*Article* **Vertical Wind Shear Modulates Particulate Matter Pollutions: A Perspective from Radar Wind Profiler Observations in Beijing, China**

#### **Ying Zhang 1, Jianping Guo 1,\*, Yuanjian Yang 2,3, Yu Wang <sup>4</sup> and Steve H.L. Yim 3,5,6**


Received: 6 January 2020; Accepted: 4 February 2020; Published: 7 February 2020

**Abstract:** Vertical wind shear (VWS) is one of the key meteorological factors in modulating ground-level particulate matter with an aerodynamic diameter of 2.5 μm or less (PM2.5). Due to the lack of high-resolution vertical wind measurements, how the VWS affects ground-level PM2.5 remains highly debated. Here we employed the wind profiling observations from the fine-time-resolution radar wind profiler (RWP), together with hourly ground-level PM2.5 measurements, to explore the wind features in the planetary boundary layer (PBL) and their association with aerosols in Beijing for the period from December 1, 2018, to February 28, 2019. Overall, southerly wind anomalies almost dominated throughout the whole PBL or even beyond the PBL under polluted conditions during the course of a day, as totally opposed to the northerly wind anomalies in the PBL under clean conditions. Besides, the ground-level PM2.5 pollution exhibited a strong dependence on the VWS. A much weaker VWS was observed in the lower part of the PBL under polluted conditions, compared with that under clean conditions, which could be due to the strong ground-level PM2.5 accumulation induced by weak vertical mixing in the PBL. Notably, weak northbound transboundary PM2.5 pollution mainly appeared within the PBL, where relatively small VWS dominated. Above the PBL, strong northerlies winds also favored the long-range transport of aerosols, which in turn deteriorated the air quality in Beijing as well. This was well corroborated by the synoptic-scale circulation and backward trajectory analysis. Therefore, we argued here that not only the wind speed in the vertical but the VWS were important for the investigation of aerosol pollution formation mechanism in Beijing. Also, our findings offer wider insights into the role of VWS from RWP in modulating the variation of PM2.5, which deserves explicit consideration in the forecast of air quality in the future.

**Keywords:** PM2.5; radar wind profiler; Beijing; wind shear

#### **1. Introduction**

Particulate particle with an aerodynamic diameter of 2.5 μm or less (PM2.5), mainly originated from industrial emissions and vehicle exhaust pollutants, and secondary aerosols forming through a series of photochemical reactions [1,2] have been shown to significantly affect the atmospheric environment [3–5], weather and climate system [6–15], and human health [16–19]. Therefore, the PM2.5 pollution and its causes have been increasingly receiving attention in recent years [20,21].

The major drivers for deteriorating or improving PM2.5 pollution are roughly twofold—aerosol emissions and meteorology—both of which are highly versatile and uncertain. In addition to high emissions accompanied with the rapid development of urbanization and industrialization, the roles of meteorological conditions, including large-scale synoptic patterns [22–26], and local meteorological conditions in the planetary boundary layer (PBL) [27–32] have been well recognized able to modulate the PM2.5 concentration. For instance, Tai et al. [27] revealed that the local meteorological conditions could explain up to 50% of the daily variability of PM2.5 in the USA from 1998 to 2008. In 68 major cities of China, ground-level PM2.5 were found to be broadly associated with local meteorological factors at seasonal, yearly, and regional scales [33]. Among various meteorological factors, the surface wind speed was one of the variables modulating ground-level PM2.5 over the Yangtze River Delta region of China, which showed that PM2.5 decreased approximately by -2.42 μg m−<sup>3</sup> for a 1 m s−<sup>1</sup> increase in wind speed [34]. In Beijing, the heavy pollution events frequently occurred under the calm wind conditions, which was generally associated with stable atmospheric stratification and shallow PBL [35–38]. The presence of high pressure in northwest parts of Beijing, linked to strong northwesterly winds, was closely associated with a significant drop in PM2.5 concentrations in Beijing [4]. Besides, the changes in circulation induced by local mountain-valley and urban heat island setting in Beijing and its surrounding areas were found to be able to modulate the diurnal variations of PM2.5 in Beijing [32].

Among others, reanalysis data and model simulations were one of the most used approaches to analyze the variation of wind with different height in the lower troposphere and its impacts on air pollution, revealing a significant role of vertical wind shear (VWS, an important indicator of dynamically vertical mixing) in modulating particulate matter pollution [37,39,40]. In recent years, there has been a surge of interest in observational investigation of VWS in connection with atmospheric pollution, most of which are based on Doppler wind lidar [26,39]. However, the Doppler wind lidar has limited capability to offer vertically resolved wind observations under pretty clean or foggy conditions. To date, the associations between vertical wind profile and surface particulate matter concentrations have yet to be fully understood in Beijing.

Fortunately, the new-generation radar wind profiler (RWP) deployed by the China Meteorological Administration (CMA) in Beijing [41] offers us the best opportunity to quantify the long-term effect of local wind vector profiles and VWS on ground-level PM2.5 pollution in Beijing. Thus, this study aims to explore the impacts of wind profiles and VWS on the wintertime PM2.5 in Beijing based on high-resolution RWP observation along with ground-level PM2.5 monitoring. The remaining contents of this work proceed as follows. In Section 2, the measurements of RWP, other related weather data, and ground-level PM2.5 are described in detail. In Section 3, the impacts of wind and VWS on PM2.5 pollution in Beijing are analyzed and discussed. Finally, the main findings are summarized in Section 4.

#### **2. Data and Methodology**

#### *2.1. Study Area*

Beijing, the capital of the People's Republic of China, is located in the north part of the North China Plain (NCP) of China and covers an area of around 16,410 square kilometers. As shown in Figure 1a, there exists a large amount of aerosol emission sources surrounding Beijing with the recent rapid economic development throughout China, especially in eastern China. Beijing is surrounded by the Yanshan Mountains to the north, and by Taihang Mountains to the west and northwestern (Figure 1b). In terms of the climate in Beijing, it typically belongs to a semi-humid continental climate in the north temperate zone, characterized by hot and humid summers due to the subsidence caused by the subtropical high, and cold, windy and dry winters which is mainly under the influence of the vast Siberian anticyclone [41–43]. Due to the huge amount of anthropogenic emission in the NCP (see Figure 1a), the atmospheric pollution (especially the PM2.5) in Beijing has been intensively analyzed in recent years [38,39,41,44].

**Figure 1.** The spatial distribution of (**a**) particulate particle with an aerodynamic diameter of 2.5 μm or less (PM2.5) emissions (mainly in the transportation, agriculture, industry, power and residual sectors) during winter in 2016, as acquired from the multi-resolution emission inventory for China (http://www.meicmodel.org/), and (**b**) monitoring stations in Beijing overlaid with topography (color shading). The red squares, blue circles, and black plus refer to the locations of radar wind profiler (RWP), PM2.5, and radiosonde (SOND) sites, respectively. The administrative boundary of Beijing is denoted by the black solid lines.

#### *2.2. Radar Wind Profiler Measurements*

The RWP is a type of remote sensing instrument that detects and processes vertical-resolved wind field information by transmitting and receiving electromagnetic beams in different directions. This instrument can provide a variety of data products, including the profiles of horizontal wind speed and direction, and vertical velocity. Specifically, the RWP data used here were collected from two sites (marked by the red squares in Figure 1b), including the Tongzhou site (116.29◦ E; 39.99◦ N) and Chaoyang site (116.47◦ E; 39.81◦ N). Both RWPs deployed in Beijing are the CFL-16 profiler, which provides 25 levels of wind speed and direction below ~3 km above ground level (AGL) with a vertical resolution of 120 m, beginning at 150 m (AGL). The measurements of wind are taken at 6 min intervals, and the detailed specifications of the RWP are given in Table 1. Prior to further analysis, the raw data have to undergo strict quality control for data consistency, continuity, and deviation [41,45,46]. To match the hourly PM2.5, the original 6-min RWP measurements were aggregated into hourly data. Our study period covers the whole boreal winter of 2018 (December of 2018 through February of 2019). To exclude the effect of wet deposition, all the measurements mentioned in this study refer to those taken on non-precipitation (i.e., rain, hail, or snow) hours, unless otherwise noted. The hourly precipitation events with precipitation amounts larger than 0.1 mm are generally defined as precipitation hours [44].

#### *2.3. Ground-level PM2.5 Concentration Measurements*

In this study, the aerosol pollution in Beijing is denoted by hourly ground-level PM2.5 concentration measurements, collected from seven air quality sites (marked by the blue circles in Figure 1b; Table 2) of the Ministry of Ecology and Environment of China. At each monitoring site, the hourly PM2.5 is measured using the tapered element oscillating microbalance method and the beta absorption method. The systematic uncertainty of ground-level PM2.5 mass concentration at these air quality monitoring

stations was controlled within 15% [47]. To avoid the uncertainties caused by aerosol heterogenous distribution, averages were taken on the PM2.5 measurements from all these seven sites.


**Table 1.** Summary of the operating and sampling characteristics of the CFL-16 Radar Wind Profiler (RWP) deployed in Beijing.

**Table 2.** Basic information for observation stations.


Note that RWP (Radar Wind Profiler), MEE (Ministry of Ecology and Environment), and SND (Radiosonde) stand for wind-profiler site, air quality site of the Ministry of Ecology and Environment, and radiosonde site, respectively. (WS: wind speed; WD: wind direction; Temperature: T; Pressure: P; Potential temperature: PT; Aerodynamic diameter smaller than 2.5 μm: PM2.5).

#### *2.4. Radiosonde and Other Meteorological Data*

The radiosonde soundings routinely measured in Beijing (116.47◦ E; 39.80◦ N, marked by the black cross in Figure 1b) were also collected to characterize the temperature inversion in association with aerosol pollution. As stated in our previous studies [29,48], the sounding balloons in China are launched twice per day at around 0800 and 2000 Beijing time (BJT = UTC + 8 h). It follows that the sounding measurements at 0800 BJT were compared with the hourly RWP and PM2.5 data at 0800 BJT. As illustrated in Figure 1b, all these meteorological stations and PM2.5 monitoring sites are evenly distributed to represent the hourly meteorological conditions in the whole urban area of Beijing well.

#### *2.5. Air Mass Back Trajectory Model*

Air masses related to regional or synoptic meteorological conditions could be responsible for the atmospheric transport of aerosol particles in the vertical and horizontal directions [26,49]. As such, we identified the main transport pathways of aerosol pollutants from surrounding regions to Beijing using the Hybrid Single-Particle Lagrangian Integrated Trajectory Model (HYSPLIT) [50]. The HYSPLIT developed by the National Ocean and Atmospheric Administration (NOAA)'s Air Resources Laboratory has been extensively used for the analyses of transboundary transport and dispersion of aerosol [51]. Based on the HYSPLIT and The GDAS (Global Data Assimilation System) reanalysis data, the frequency distribution and cluster mean of 24-h backward trajectories of each day for the period from December 1, 2018, to February 28, 2019, were calculated, respectively. The trajectory endpoint was set in the urban area of Beijing (116.37 ◦E, 40.09 ◦N) with a height of 100 m above ground level (AGL).

#### *2.6. Methodology*

In order to avoid the effects induced by the seasonal variation of aerosol at specific sites and by the spatial variation among different sites within the same region, hourly PM2.5 concentrations for a given site are normalized using the monthly mean of each site and year according to the approach by Wang et al. [52]. The PM2.5 dataset is then grouped into three subsets, each of which has the same number of samples. The lower and upper terciles of normalized PM2.5 refer to the clean (bottom 1/3) and polluted (top 1/3) conditions, respectively. In this way, comparison analysis can be performed between clean and polluted atmospheric conditions, and good sampling statistics can be maintained [11,53], even though the critical threshold of normalized PM2.5, which is used to distinguish between the clean and polluted categories, differs by various time scales. Besides, hourly wind anomalies with 120 m resolution in the vertical are calculated and then used in the subsequent analyses in diurnal association with ground-level PM2.5 concentration. This allows us to do a more detailed analysis of the wind variation for various PM2.5 pollution levels.

The VWS has been found to play important roles in the dispersion of air pollutants, and thus was calculated here to check its effects on PM2.5 variability. Therefore, the bulk shear, which refers to the magnitude of the bulk vector difference (top minus bottom) divided by height [54–56], is calculated as follows: 

$$VWS = \frac{\sqrt{(u\_{z1} - u\_{z2})^2 + (v\_{z1} - v\_{z2})^2}}{(z1 - z2)} \times 1000\tag{1}$$

where *VWS* is the vertical wind shear (units: m/(s · km)), *uz*<sup>1</sup> and *uz*<sup>2</sup> represent the zonal wind at the height of *z*1 and *z*2, respectively; and *vz*<sup>1</sup> and *vz*<sup>2</sup> represent the meridional wind at the height of *z*1 and *z*2. *z*1 is the top height and *z*2 is the bottom height.

To enhance the visual interpretation, daily 24-h period is divided into eight sub-period at 3-hour intervals, which is defined as follows [57,58]: late night (0000 – 0300 BJT), early morning (0300–0600 BJT), morning (0600–0900 BJT), late morning (0900–1200 BJT), early afternoon (1200–1500 BJT), late afternoon (1500–1800 BJT), evening (1800–2100 BJT), and night (2100–2400 BJT).

Additionally, the bivariate polar plot has been used, combining wind measurements from RWP and PM2.5 measurements in Beijing, which is expected to provide insight into the sophisticated relationship between wind and PM2.5 [59,60].

#### **3. Results and Discussion**

#### *3.1. Thermodynamic and Meteorological Variables Related To PM2.5*

Figure 2 shows the time series of observed daily PM2.5 concentrations with vertically thermodynamic (temperature) and dynamic (wind) variables in Beijing simultaneously observed for the period from December 2018 to February 2019. The heavy PM2.5 pollution tended to occur more frequently on the days with low near-surface wind speed, and warmer air at the top of PBL (Figure 2b,c), given the wintertime climatological PBL height of 1–1.5 km in Beijing [61]. This generally does not favor the vertical ventilation and horizontal dispersion of aerosols. For example, during the pollution episode from December 14, to 18 of 2018, the near-surface wind speed in Beijing was

significantly lower than those of pre- and post-periods, which was accompanied by strong thermal inversion layer. The southerly winds prevailed in the lowest 1–2 km of PBL (Figure 2a), which tended to transport aerosol particles from Hebei, a significant source region of aerosol emission to the south of Beijing (Figure 1a).

The most severe haze episode occurred during the study period persisted at least three days, starting from 12 January 2019 until 14 January 2019, during which PM2.5 exceeded 100 μg m<sup>−</sup>3. A most extremely high concentration of greater than 200 μg m−<sup>3</sup> was observed as well. Coincidently, there existed weak southwesterly winds, and strong temperature inversion in the PBL, both of which contributed to this atmospheric pollution.

**Figure 2.** (**a**) Time series of the horizontal averaged wind vectors as derived from the Radar Wind Profiler (RWP) in Beijing for the altitude ranges from the surface to 1 km (SFC–1 km), 1–2 km, and 2–3 km above ground level (AGL), which are denoted as the red, green, and blue vectors, respectively. The wind arrow is the direction towards which the wind is blowing, and the width of the wind vector is proportional to the wind speed. Time-height cross-sections of (**b**) potential temperature (PT, color shaded) from radiosonde measurements and (**c**) horizontal wind speed (color shaded) from RWP, overlaid with observed ground-level PM2.5 concentration (red lines). All these measurements were obtained at 0800 BJT during the period December 1, 2018, to February 28, 2019.

#### *3.2. Synoptic-Scale Circulation and Backward Trajectory Statistical Analysis*

In this section, we will examine the role of synoptic-scale meteorology underlying the polluted and clean episodes observed during the study period in Beijing. The first step to accomplish this is to determine the climatological wintertime winds at 925 hPa and 850 hPa pressure levels over Beijing and its surroundings. As shown in Figure 3a,c, weak westerly or southwesterly winds dominated both 850 hPa and 925 hPa pressure levels during the high aerosol-loading winter days in Beijing. By comparison, the wind fields at 850 hPa and 925 hPa were characterized by strong northwesterly winds over Beijing, which generally led to frequent intrusion of cold air mass (Figure 3b,d).

This cold advection could bring in cold and clean air from the northern regions without much anthropogenic emission sources (Figure 1a), thus resulting in low PM2.5 concentration in Beijing, such as the extremely clean atmospheric episode occurring during December 27 to 30, 2018 (Figure 2). By contrast, less lapse rate of temperature and southwesterly winds featured the synoptic conditions favoring the accumulation of PM2.5 (Figure 3a,c), well corroborating the vertical wind measurements in Beijing shown in Figure 2. Our findings were broadly consistent with the relationships between PM2.5 concentration, temperature, and wind speed in winter found in other regions of the NCP [62].

**Figure 3.** Spatial distribution of the wind field (black arrows, vector), superimposed by temperature (shaded) at 925 hPa (**a**,**b**) and 850 hPa (**c**,**d**) pressure level under polluted (left column) and clean (right column) conditions, respectively. All data are from the National Center for Environmental Prediction (NCEP) global Final (FNL) reanalysis. The areas highlighted with red lines represent the region of interest (Beijing), which is the same as the region highlighted with black lines in Figure 1a.

To further our understanding of the long-range transport to particulate matter pollution in Beijing, the 24-h backward trajectories were calculated and clustered. As illustrated in Figure 4, the prevailing northwesterly winds dominated the contribution in terms of transboundary transport (72% of all 24-h back trajectories). Interestingly, a small fraction of the trajectories (28%) came from the south, which was linked to most of the polluted episodes in Beijing during December 1, 2018, to February 28, 2019.

#### *3.3. Diurnal Variations in Vertical Winds*

Figure 5 illustrates the diurnal variations of wind speed and direction in Beijing for the heights ranging from ground-surface up to 3 km AGL under mean, polluted and clean conditions and their corresponding hodographs, and so do the anomalies of wind profile under polluted and clean conditions relative to the average wintertime winds in Beijing.

**Figure 4.** The spatial distribution of the trajectory frequency (**a**), and cluster-mean results of 24-h backward trajectories (**b**) calculated by the Hybrid Single-Particle Lagrangian Integrated Trajectory Model (HYSPLIT) ending at Beijing (116.37 ◦E; 40.09 ◦N, pentagram) at 100 m height AGL during winter from December 2018 to February 2019. The dots on the trajectories represent time node of 12h, and the percentage represents the ratio of the number of back trajectories in each cluster to the total number of back trajectories.

Overall, the mean wind speed was found to increase with height (Figure 5a). Meanwhile, the mean wind exhibited a pronounced diurnal cycle in the lower PBL (i.e., from the ground surface to 1 km AGL), and much stronger winds occurred during nighttime near the ground-surface compared with daytime (Figure 5a), which could be due to much reduced turbulence-related friction. This was also likely associated with the radiative (nocturnal) cooling in the night, which tended to reduce the eddy viscosity and momentum transfer from the upper levels, and in turn, lead to decreased wind speed [63]. However, the wind at the top of PBL or low troposphere exhibited double peaks: one is midnight and another in the later morning (1000 BJT). Under clean conditions, wind vector veered with the height between the near ground surface and 1 km AGL, followed by significant backing above 2 km (Figure 5b). Veering winds in the lowest layers of the atmosphere are most likely the result of friction-related processes while the backing winds are indicative of cold advection [64], which is mainly contributed by prevailing northwest winds (Figure 5a).

Under both polluted and clean conditions, the wind profiles showed significant diurnal variation at all heights, and its amplitude and sign differed greatly in the vertical during a daily cycle (Figure 5c,e), indicating that the lower part of PBL was characterized by prevailing southerly and northerly winds, respectively. The hodograph for polluted conditions exhibited smaller vertical shear at most heights, irrespective of the time of a day (Figure 5d). On the contrary, the curvature of the anticyclone rotation at above approximately 2 km AGL was significantly larger under clean conditions, resulting in larger wind shear (Figure 5f), which was most likely due to many more cold waves (Figures 2 and 3d).

Also, there existed significant wind anomalies under polluted and clean conditions (Figure 5g,i). Coincidently, a clockwise turning of the wind with the height was observed near the ground surface under polluted condition (Figure 5h), which was indicative of a warm air advection from the south. In contrast, the backing winds were found near the ground surface under clean conditions (Figure 5j), confirming the notion of northerlies-induced decreases in aerosol concentration. This highlights the urgency of consideration of VWS and wind direction in the aerosol pollution and its formation causes. In particular, under polluted conditions, negative wind anomalies prevailed at almost all times of the day in the PBL over Beijing, especially during the nighttime, indicating aerosol-induced changes in radiation reaching the surface could be linked to the dramatical reduction of wind speed. By comparison, positive southerly wind anomalies emerged during 1000–2000 BJT in the troposphere above 1.75 km, indicating that there were strong elevated PM2.5 transport paths, which were located at the heights above the PBL at this time period. Under clean conditions (Figure 5c), positive northerly

wind anomalies prevailed at almost all height, with the exception of negative wind anomalies above the PBL at roughly 0600, 1200, and 1900 BJT. This suggested that less aerosol tended to be associated with northerly winds, which further strengthened the role of cold advection from the northwestern parts of Beijing in reducing ground-level PM2.5.

**Figure 5.** Height-resolved diurnal variation of horizontal wind vector under all-sky mean (**a**), polluted (**c**) and clean (**e**) conditions, and their corresponding anomalies (relative to mean wind) under polluted (**g**) and clean (**i**) conditions in Beijing from December 1, 2018, to February 28, 2019. Also shown are their corresponding hodographs in the panels (**b**,**d**,**f**,**h**,**j**) on the right-hand sides. Note the vectors in panels (**g**)–(**i**) show the resultant wind anomaly direction, and the vector length and color indicate the magnitude of the resultant wind anomaly relative to the average wintertime winds in Beijing.

#### *3.4. Vertical Wind Shear Under Polluted And Clean Condition*

The vertical wind shear is known to be able to strongly influence the vertical mixing process and resultant changes in aerosol pollutants in the PBL [26]. Figures 6 and 7 show the vertical distribution of VWS under polluted and clean conditions, respectively. The leading diagonal (top right to bottom left) denotes the local VWS at each level, whereas the shading in color indicates the magnitude of VWS at least two consecutive vertical levels. Note that VWS distribution does not take into account the shear direction.

**Figure 6.** Three-hourly averaged vertical wind shear (VWS) computed between different heights under polluted conditions in Beijing for (**a**) 0000–0300 BJT, (**b**) 0300–0600 BJT, (**c**) 0600–0900 BJT, (**d**) 0900–1200 BJT, (**e**) 1200–1500 BJT, (**f**) 1500–1800 BJT, (**g**) 1800–2100 BJT and (**h**) 2100–2400 BJT, during the period from December 2018 to February 2019. The X and Y axes represent the top and bottom height of the VWS bulk, respectively.

**Figure 7.** Same as Figure 6 but under clean conditions.

Generally, the pattern of the diurnal cycle did not change much when the atmosphere evolved from clean to polluted conditions, except for the magnitude of VWS. The magnitude of VWS was found to be much smaller under polluted conditions than that under clean conditions, indicative of weaker vertical mixing in the presence of high aerosol concentration in the PBL. This, in turn, led to a stronger accumulation of ground-level PM2.5. On average, the local VWS for the polluted condition was ~8 m s−<sup>1</sup> km<sup>−</sup>1, as compared to as high as ~10.5 m s−<sup>1</sup> km−<sup>1</sup> for the clean condition. However, note that the VWS for polluted conditions between 2.5–3 km and <1.5 km was greater than that under clean conditions during the afternoon (1200–1800 BJT), which could be related with the positive southerly wind vector anomalies above 2 km during 1200–1800 BJT (Figure 5b). It implied that strong wind in the upper level and in the PBL tended to transport PM2.5 from southern Hebei province to Beijing, further deteriorating the air quality in Beijing through these strongly vertical mixing exchanges (i.e., larger VWS) in the upper level and in the PBL. This finding was in agreement with the observational evidence from Hong Kong [26]. In addition, the magnitude of VWS (> 5ms−<sup>1</sup> km<sup>−</sup>1) was small during the night to morning (1800–1200 BJT) but large during the afternoon (1200–1800 BJT) under polluted conditions. On the contrary, under clean conditions, the VWS (> 5ms−<sup>1</sup> km<sup>−</sup>1) was relatively larger during the night (2100-0600 BJT) than during the day. These differences in diurnal variations of VWS (> 5ms−<sup>1</sup> km<sup>−</sup>1) under polluted/clean conditions were probably associated with different integrated effects of the local circulation (mountain-valley and urban heat island circulations) [14,32] and synoptic patterns [65,66].

Figure 8 presents the correlation between VWS at different layers and ground-level PM2.5 under polluted and clean conditions, respectively. It generally exhibits marked differences (even with opposite signs) in the lowest part of PBL in both conditions. In particular, the correlation coefficients seemed to be positive for the altitudes from the ground surface up to 2.5 km AGL under polluted conditions (Figure 8a), which meant the weak VWS near the ground surface or lower part of PBL observed in Figure 6 favored the accumulation of aerosol. In contrast, the correlation coefficient shifted from positive to negative as the VWS occurred upward, suggesting that the stronger VWS above the PBL was linked to lower ground-level PM2.5 concentration. Interestingly, under clean conditions, a ubiquitous negative correlation was found between VWS and ground-level PM2.5 in the almost whole lower atmosphere except in the height of 1 km and that above 2.5 km and beyond (Figure 8b). The increase of VWS tended to be accompanied by enhanced vertical mixing of aerosol, leading to reduced ground-level PM2.5, which could account for this negative correlation observed for the clean condition. The exception in height above 2.5 km could be associated with air mass intrusion of long-range transported aerosol episodes [67], given the dominant height of long-range transboundary transport being generally above the PBL [68–70].

**Figure 8.** The correlation coefficient distribution between normalized ground-level PM2.5 and height-revolved VWS under (**a**) polluted and (**b**) clean conditions in Beijing. Gray dots indicate the Pearson correlation coefficient that is statistically significant at the 90% confidence level. The X and Y axes represent the top and bottom heights of the VWS bulk, respectively.

#### *3.5. The Dependency of Ground-Level PM2.5 On Vertically Resolved Winds*

The bivariate polar plots in Figure 9 showed that the normalized PM2.5 concentration (hereinafter referred to as NPM2.5) in Beijing varied by wind direction and speed at different heights, including ground-surface–1km, 1–2 km and 2–3 km. Specifically, at the height of ground surface–1km (Figure 9a), high aerosol concentration episodes (HEP, NPM2.5 > 120%) were observed when the cardinal winds occurred along the NNE–NE, and NNW–WNW directions with speeds of greater than 6 m s−<sup>1</sup> and even up to 10 m s−1. Additionally, the polluted cases were observed when the prevailing cardinal winds were in the WSW–SSE sector with wind speed 4–6 m s<sup>−</sup>1, consistent with previous findings [71]. At the heights of 1–2 km, extreme HEPs (i.e., NPM2.5 >200 %) were observed mainly when winds blew along the SW–SE sector with a speed of <6ms−<sup>1</sup> (Figure 9b). By comparison, HPEs occurred when the NE–N winds prevailed at moderate to high wind speeds (4–12 m s<sup>−</sup>1). As the atmospheric height increased to 2–3km, HEP mainly occurred when the winds were coming from NE–N sector at high wind speed (12–20 m s<sup>−</sup>1), and also occurred in the NW sector with a wind speed of 14 m s−<sup>1</sup> (Figure 9c). The surprisingly high NPM2.5 tended to occur in Beijing when the winds blew from WSW–S at weaker wind speeds (2–6 m s−1). In contrast, the NPM2.5 in the NW sector was found to be significantly reduced. In general, it was found that a smaller range of direction angle was accompanied by smaller wind shear at the lower part of the PBL, whereas stronger southerly wind prevailed with larger wind shear in the upper PBL and even above the PBL under heavy polluted conditions (Figures 5 and 7).

**Figure 9.** Bivariate polar plot of normalized PM2.5 concentration (in percent) for the altitude ranges of (**a**) ground surface to 1 km (SFC–1km), (**b**) 1–2 km and (**c**) 2–3 km AGL during the period from December 1, 2018, to February 28, 2019. The wind directions are denoted by 16 compass direction: N, NNE, NE, ENE, E, ESE, SE, SSE, S, SSW, SW, WSW, W, WNW, NW, NNW. The radial axis represents wind speed in m s<sup>−</sup>1, which increases radially outward. The concentration of PM2.5 is scaled by colors.

The region to the south of Beijing was previously recognized to be a key emission source region for the PM2.5 pollution episodes in Beijing, especially southerly or southwesterly wind prevailed [72–74]. This, in part, was attributed to transboundary PM2.5 transport [28,75]. The results presented here provided convincing evidence that regional sources largely contribute to PM2.5 below 3 km, where existed kind of main PM2.5 transport path from south to Beijing, which basically agreed with the findings from model simulation analysis [76].

Another striking feature we observed here was that a few aerosol pollution episodes occurred even as the strong northerly or northwesterly winds dominated from the ground surface up to 3 km AGL (Figure 9). This could be likely linked to the long-range transported aerosol from northwestern or northeasten China, which was verified or corroborated in previous observational and model investigations [77].

#### **4. Concluding Remarks**

Based on continuous fine-resolution radar wind profiler (RWP) observations, radiosonde measurements during the winter for the period December of 2018 to February of 2019, the height-resolved wind vectors were analyzed, along with the impact of vertical wind shear on PM2.5 pollution in Beijing. The main findings are summarized as follows:

Overall, the diurnal variations in wind profiles were found to differ greatly when classified by different ground-level PM2.5 concentrations. Specifically, the southerly wind anomalies dominated throughout the whole PBL or even beyond the PBL under polluted conditions during the course of a day, in sharp contrast to the northerly wind anomalies in the PBL under clean condition. More strikingly, under pollution conditions, the positive anomaly of southerly wind speed mainly occurred at 1.75 km AGL during 1000–2000 BJT. This favored the transboundary transport originated from significant aerosol emission source to the south of Beijing, thereby leading to high ground-level PM2.5 concentration in Beijing.

Besides, the ground-level PM2.5 pollution exhibited a strong dependence on the vertical variation of the wind direction. The VWS tended to be much weaker in the lower PBL under polluted conditions, compared with under clean conditions, which could be strong ground-level PM2.5 accumulation induced by weak vertical mixing in the PBL. Notably, the PM2.5 pollution mainly appeared within the PBL as weak southerly winds prevailed when the relatively small VWS was observed as well. Above the PBL, strong northerlies winds also favored the long-range transport of aerosols, which in turn deteriorated the air quality in Beijing as well. This was well corroborated by the results from synoptic-scale circulation and backward trajectory analysis.

In summary, not only wind profiling but also the VWS at various heights could significantly modulate the ground-level PM2.5 concentrations. Also, the present work highlighted the role that the height-resolved wind shear plays in better understanding the wintertime aerosol pollution episodes in Beijing. To increase the generalizability of the reported associations between aerosol and VWS; nevertheless, more efforts have to be made to include a much longer time series of observations at larger spatial domains in the future. More importantly, more wind and VWS measurements from RWP are desperately needed to be assimilated into the air quality model, which is expected to have great implications for improving the wintertime air quality forecast in China.

**Author Contributions:** Conceptualization, J.G.; Methodology, J.G. and Y.Z.; Validation, J.G., Y.Y. and S.H.L.Y.; Formal Analysis, Y.Z., J.G. and Y.Y.; Data Curation, J.G., Y.Z.; Writing–Original Draft Preparation, Y.Z. and J.G.; Review & Editing, J.G., Y.Y., Y.Z. and Y.W.; Visualization, Y.Z.; Supervision, J.G. and Y.W.; Resources, J.G. and S.H.L.Y.; Funding Acquisition, J.G. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by the National Natural Science Foundation of China (Grant 41771399), Ministry of Science and Technology of China (Grant 2017YFC1501401), Chinese Academy of Sciences (Grant GXDA20040502), and Chinese Academy of Meteorological Sciences (Grants 2017Z005).

**Acknowledgments:** We appreciated greatly the National Meteorological Information Center of China Meteorological Administration for providing the radar wind profiler (RWP) data (https://data.cma.cn/en/). We also sincerely appreciated the PM2.5 data made publicly accessible by the Ministry of Ecology and Environment of China (http://www.cnemc.cn/en/). Last but not least, the authors would like to thank the editor and three anonymous reviewers for their constructive comments which help improve the quality of our manuscript.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Role and Mechanisms of Black Carbon A**ff**ecting Water Vapor Transport to Tibet**

**Min Luo 1, Yuzhi Liu 1,\*, Qingzhe Zhu 1, Yuhan Tang <sup>1</sup> and Khan Alam <sup>2</sup>**


Received: 10 December 2019; Accepted: 7 January 2020; Published: 9 January 2020

**Abstract:** Although some studies reported the impact of black carbon (BC) on the climate over the Tibetan Plateau (TP), the contribution and mechanisms of BC affecting the water vapor transport to Tibet are not fully understood yet. Here, utilizing the satellite observations and reanalysis data, the effects of BC on the climate over the TP and water vapor transport to the Tibet were investigated by the Community Earth System Model (CESM 2.1.0). Due to the addition of BC, a positive net heat forcing (average is 0.39 W/m2) is exerted at the surface, which induces a pronounced warming effect over the TP and consequently intensifies the East Asian Summer monsoon (EASM). However, significant cooling effects in northern India, Pakistan, Afghanistan and Iran are induced due to the BC and related feedbacks, which reduces significantly the meridional land–sea thermal contrast and finally weakens the South Asian summer monsoon (SASM). Consequently, the water vapor transport to the south border will be decreased due to addition of BC. Moreover, through affecting the atmospheric circulation, the BC could induce an increase in the imported water vapor from the west and east borders of the TP, and an increase outflowing away from the north border of the TP. Overall, due to the BC, the annual mean net importing water vapor over TP is around 271 Gt, which could enhance the precipitation over the TP. The results show that the mean increase in the precipitation over TP is about 0.56 mm/day.

**Keywords:** black carbon; Tibetan plateau; water vapor transport; South Asian summer monsoon; East Asian summer monsoon

#### **1. Introduction**

The black carbon (BC) aerosol emitted from the combustion of some biomass and fossil fuel has a strong "Greenhouse Effect" by absorbing solar radiation and longwave radiation [1,2] in the atmosphere with a few days' lifetime [3,4]. Moreover, the BC can also affect the earth–atmosphere energy balance through indirect and semi-direct effects [5], having a profound influence on the hydrological cycle and climate [2,6,7]. However, the uncertainties in estimating the magnitude of the hydrological cycle and regional climate responses to the BC are still pronounced [8].

During recent years, the aerosol emissions, including the BC over Asia are obviously increasing, and these particles can be transported to the Tibetan Plateau (TP) by atmospheric circulations [9–11]. The BC could be deposited into snow and exert a pronounced "snow darking" effect [12], and further affect the radiation budget [13,14]. The annual mean snow albedo direct radiative forcing of BC may reach 2.9 W/m2 [15], which will reduce the snow albedo and accelerate the melting of glaciers [16–19]. Moreover, the BC over the TP can affect the properties of cloud [20–22], precipitation [22–24] and the monsoon circulations [25].

Furthermore, the BC-in-snow effect can induce an increase in surface temperature over the TP and cause the earlier onset of the South Asian summer monsoon (SASM) [26]. Besides, the "Elevated Heat Pump" (EHP) effect suggests that the BC can induce an updraft motion with a warm anticyclone circulation in the upper atmosphere over the TP in late spring or early summer [27,28], the EHP effect could reduce the SASM in summertime by its dynamic and thermal forcing [28,29]. Additionally, the decreased meridional temperature gradient from the Indian Ocean to Northern India caused by absorbing aerosols could also reduce the SASM significantly [6,30,31]. On the other hand, the BC-in-snow effect could enhance the East Asian summer monsoon (EASM) by increasing the land–sea thermal contrast [26]. The changes in EASM are closely related to the aerosols disturbing the thermal contrast between land and ocean [32]. Li et al. [33] observed that the greenhouse gases and aerosols could increase the thermal contrast between the East China and the adjacent sea, and hence the EASM. The warming effect of BC over East Asia is the main factor, which could induce the enhanced EASM [34].

Generally, the SASM and EASM are the main dynamic factors in terms of the water vapor transport. The water vapor can be carried from the ocean to the land by the circulations of SASM and EASM [35], in which the northward transport is mainly caused by the lower southerly wind [36]. The TP, which is named the "Asian water tower", is feeding several major rivers in Asia and providing fresh water for more than one third of the populations of the world [37], and has been receiving much attention [37–42]. Generally, the water vapor may be gathered in the western and southern TP, and advected to the rest of the TP [40]. The water vapor could be transported to the TP by upslope transport and up-and-over patterns [37,38]. In addition, the perturbed cyclone and anticyclone over Lake Baikal are closely related to the water situations of TP, and the warming in the northwestern Atlantic Ocean is the key factor contributing to the wetting TP [42]. However, under the global dryland expansion and warming [43,44], the potential role and mechanism of the BC affecting the water vapor transport and the burden over TP is poorly studied.

Although previous researchers have revealed that the BC has a pronounced climate effect on the SASM and EASM, which are closely related to the water vapor transport from the ocean to the TP, there are few studies focused on the effects of BC on the water vapor transport from the surrounding to the TP. In this study, the role and mechanisms of BC affecting the water vapor budget over the TP are investigated by utilizing the CESM, which is fully coupled including atmosphere, ocean, sea ice, land and land–ice components.

#### **2. Data and Methods**

#### *2.1. The Uncertainties and Applicability of Each Data*

To evaluate the model performance on the climate and water vapor transport over the TP, the product of Multi-angle Imaging Spectro Radiometer (MISR) and several reanalysis data were used to compare with simulations. The resolutions of observations and reanalysis data sets depend on the assimilation system and the accuracy of the equipment. Before the analysis, we have interpolated the results of simulations whose resolution is 0.9◦ (latitude) × 1.25◦ (longitude) to the resolutions of observations and reanalysis data sets.

#### 2.1.1. Multi-Angle Imaging Spectro Radiometer (MISR)

The MISR measures the aerosol optical depth (AOD) at a spatial resolution from 275 to 1100 m globally. Because the atmospheric path contribution from the surface-leaving radiance can be removed by taking advantage of differences in multi-angular signatures, the MISR aerosol retrieval algorithm is less sensitive to surface type especially over the bright surfaces [45]. Therefore, compared with the ground-based remote sensing, the AOD products from MISR have good accuracy over the TP. However, MISR has a limited swath coverage that is much too narrow, which has a lower frequency of observations at the given ground-based site in each orbital cycle. In this study, the MISR-3 product

derived from multiple orbits monthly with a resolution of 0.5◦ × 0.5◦ [46] was used to evaluate the simulated AOD by a model.

#### 2.1.2. Cloud and Earth's Radiant Energy System (CERES)

The CERES is used to investigate the cloud/radiation feedback, the data sets are measured by the broadband scanning radiometers [47]. The all sky surface net radiation fluxes (longwave and shortwave radiation, W/m2) were obtained from the CERES 4.0 product, whose spatial and temporal resolutions are 1.25◦ × 1.25◦ and monthly, respectively. The CERES data sets have a strong correlation with meteorological station data [48]. For the CERES products, the uncertainties in surface net radiation are attributable to the environmental parameters, including surface water vapor pressure, surface temperature, the Normalized Difference Vegetation Index (NDVI) and surface albedo. Studies show that the errors of each radiation component of the CERES product are within 20 W/m2 at the monthly scale [49].

#### 2.1.3. ERA-Interim

The ERA-interim data sets are obtained from the European Center for Medium-Range Weather Forecasts (ECMWF) covering the period from 1979 to the present. It has a good performance on describing the actual atmosphere over Asia [50]. In this study, the monthly mean skin temperatures (K) with a spatial resolution of 0.75◦ × 0.75◦ are used. Compared with the observations, the root mean square error (RMSE) of temperature is about 3.2 ◦C, and the correlation coefficient is 0.709. Generally, the ERA-interim is closer to the ground observations than many other reanalysis data over Asia [51].

#### 2.1.4. Global Precipitation Climatology Project (GPCP)

The GPCP is a merged precipitation reanalysis data which incorporates information from the low-orbit-satellite microwave, the geosynchronous-orbit-satellite infrared, and the rain observations. The GPCP can figure out the temporal and spatial features of precipitation (mm/month) quite well, having a good performance over Asia [52]. Here, the monthly mean accumulated precipitation data at the surface from GPCP-2 with a spatial resolution of 2.5◦ × 2.5◦ were used. Some studies reported that the GPCP data may overestimate the precipitation, especially when the precipitation rate increases. This observed uncertainty may be due to the shortcomings of GPCP for the retrieval of summer precipitation over land, such as mistaking higher clouds as precipitation clouds [53,54].

#### 2.1.5. Modern-Era Retrospective Analysis for Research and Applications (MERRA)

The MERRA-2 is an assimilation which includes different ground-based and space-based remote sensing information. The monthly mean surface mass concentration of BC (kg/m3) is obtained from the version 2 of MERRA (MERRA-2), which has a spatial resolution of 2.5◦ × 2.5◦. The MERRA data can describe the distribution of the BC mass concentration over Asia very well [55]. The aerosol AODs from the MERRA-2 data have a high correlation but low bias relative to the ground-based observations (e.g. sun photometer) [56]. Here, we use the surface concentration of BC from the MERRA-2 data.

#### 2.1.6. National Centers for Environment Prediction (NCEP)

The NCEP reanalysis data covers the information of satellite, and it is produced by a forecast model together with a data assimilation system. Because the NCEP data sets have a good performance on describing the winds and free atmosphere of the temperate region in the Northern Hemisphere [57], the horizontal wind (u and v component, m/s) and specific humidity (kg/kg) in NCEP data sets are used to evaluate the simulated water vapor in the atmosphere. The spatial resolution of the monthly NCEP reanalysis data used in this study is 2.5◦ × 2.5◦ [58]. For the NCEP data, due to the topographical height and systematic deviations of the assimilation model, it was a false trend on the longer time scale in the middle and lower troposphere over the TP [59].

#### 2.1.7. Emission Data of Aerosols and Greenhouse Gases

In this study, the emission data of aerosols and greenhouse gases for the year 2000 are used to be the background reference. The anthropogenic aerosol emissions from industrial production, agriculture activities and human activities are derived from the emission data of the Intergovernmental Panel on Climate Change (IPCC) Fifth Assessment Report (AR5) [60]. The emissions of BC together with sulfur dioxide are updated from Smith et al. [61]. Besides, the aerosol size distribution is classified as three lognormal modes: Aitken (ranging from 0.015 mm to 0.053 mm), accumulation (ranging from 0.058 mm to 0.27 mm), and coarse modes (ranging from 0.80 mm to 3.65 mm) [62]. The concentrations of greenhouse gases for the year 2000 are derived from the specific concentration [63].

#### *2.2. Methods of Calculating Asian Summer Monsoon and Water Vapor Transport Changes*

#### 2.2.1. Estimation of Summer Monsoon

To estimate the strength of the summer monsoon, the dynamical normalized seasonality index (DNS) was used in this paper [64], and can be calculated as using the following Equation (1):

$$DNS = \frac{\left\| \vec{V1} - V(\vec{m}, n) \right\|}{\left\| \vec{V} \right\|} - 2 \tag{1}$$

where the <sup>→</sup> *<sup>V</sup>*<sup>1</sup> is the climatological wind vector in January, the <sup>→</sup> *V* is the mean wind vector of January and July, while the <sup>→</sup> *V*(*m*, *n*) is the wind vector in the year 'n' and month 'm'.

#### 2.2.2. Changes of Water Vapor

The water vapor flux in the whole atmospheric layer (*Q*) is calculated from the land surface to 100 hPa and the water vapor budget (*N*) [42] are calculated as follows:

$$Q = -\frac{1}{\mathcal{S}} \int\_{p\_\*}^{p} q \,\vec{V} dp\tag{2}$$

$$N = \oint \mathbb{Q}dl\tag{3}$$

where *<sup>q</sup>* denotes the specific humidity (kg/kg), <sup>→</sup> *V* denotes the horizontal wind vector (u and v component, m/s), *g* denotes the constant of gravity acceleration (9.8 m/s2). Here, the atmospheric pressure at the top of atmosphere (TOA) is set as 100 hPa, and *l* denotes the border of the TP. In this study, the border of TP is considered as the east, south, west and north borders separately.

#### *2.3. Description of Model Setting*

In this study, CESM (version 2.1.0) is used to study the effects of BC on the climate over the TP. The atmospheric, land and ocean models are the Community Atmosphere Model (CAM) 5.0, Community Land Model (CLM) 4.5 and three-dimensional active Parallel Ocean Program (POP) 2.0, respectively, in CESM.

The default aerosol configuration coupled to the Modal Aerosol Model (MAM) is adopted to investigate the climate effects of BC [65]. The CAM 5.0 of CESM includes BC, dust, precursors of sulfate (SO2 and SO4), sea salt, particulate organic matter, and secondary organic aerosol in the MAM3 aerosol module [65]. The dust and sea salt modes are merged into a coarse mode in MAM3 for the online calculating of the emissions. The dry deposition process of aerosols is calculated by a parameterization which includes the information of land use and land cover [66], while the wet deposition process is calculated by the wet removal routine [67]. Meanwhile, the revised cloud macrophysics processes are considered in CESM-CAM5, in terms of the cloud fractions, and the interconversion rates between water vapor and cloud condensation.

The parameterization of cloud microphysics is utilized to calculate the droplet and ice number concentration [68]. Besides, the CAM 5.0 forms the main atmospheric components in CESM and has the capacity of simulating the aerosol–cloud interaction, which includes the cloud droplet activation by aerosols, precipitation process affected by particles and the radiative interactions of cloud particles [69–71].

In this study, two experiments are carried out to study the climatic effects of BC. The control experiment, named the 'ALL' experiment, considers the emissions of BC, particulate organic matter (POM), dust, SO2, SO4 and second organic aerosol gases (SOAG). The contrast experiment, named as the 'ALL-BC' experiment, is run with the same emissions as the ALL experiment, except for the BC emissions. The simulations started from January 2000 to December 2005 with a spin-up time in the year of 2000. The finite volume (FV) dynamical core with a resolution of 0.9◦ × 1.25◦ and a hybrid sigma-p vertical coordinate of 30 layers from the land surface to 3.64 hPa are utilized in CESM. The details of the model setup and experiment design are presented in Table 1.


**Table 1.** Physical and chemical schemes used in the Community Earth System Model (CESM) simulation.

#### *2.4. The Study Area*

Considering the typical topography of the TP and distribution of BC, we focused on the region with latitude and longitude coordinate ranges of 60◦ E–140◦ E and 5◦ N–45◦ N (see Figure 1). It covers most parts of China, the whole Indo-China Peninsula, India and Pakistan. The areas marked by the blue and the red rectangle in Figure 1 are adopted to calculate the changes in the intensity of the SASM and the EASM, respectively. Besides, the climatic effects caused by BC over the TP are analyzed for the summertime (June–July–August). Before analyzing water vapor changes due to BC, the assessment for the model ability of CESM is performed first.

**Figure 1.** The topography (km) of study area, red and blue rectangles correspond to the East Asian Summer monsoon (EASM) and the South Asian summer monsoon (SASM) regions, respectively.

#### **3. Results**

#### *3.1. Model Assessment*

Figure 2 shows the comparisons of the simulated net surface radiation budget, surface temperature (ST) and precipitation rate with observations. It shows that the "northern high and southern low" pattern of all sky surface net radiation (integrated by shortwave and longwave radiation) is simulated well (Figure 2a,d). The values of net surface radiation over South Asia, together with East Asia, the Taklimakan Desert and Pakistan range from 150–300W/m2, 300–400 W/m<sup>2</sup> and 400–450 W/m2, respectively. While the net surface radiation is underestimated over TP. Because of the bias in surface net radiation over TP, the simulated ST is also underestimated slightly (Figure 2b,e). Generally, the distributions of simulated ST are consistent with the ERA-interim. Overall, the heavy precipitation values are found over the Western India, Bay of Bengal and Indo-China Peninsula in model simulations. The areas of simulated heavy precipitation are consistent with the results from GPCP. In Figure 2c,f, over the south slopes of TP, the model overestimates the precipitation compared with the GPCP results.

**Figure 2.** Evaluations of simulated summer all sky net surface radiation (W/m2), surface temperature (ST) (K) and precipitation rate (mm/day) by the Community Earth System Model (CESM) model for the period from 2001–2005. Where, (**a**–**c**) are model simulations, while the (**d**–**f**) are derived from the data of CERES, EAR-interim and the Global Precipitation Climatology Project (GPCP), respectively. The down direction in (**a**) and (**d**) is defined as the positive value, otherwise the upward direction is defined as the negative value. The area of the Tibetan Plateau (TP) is enclosed by a black line.

Furthermore, the comparisons of simulated water vapor transports and that from the NCEP are presented in Figure 3. In general, both the simulated and observed water vapor burdens over the TP ranges from 5 to 10 kg/m2, with almost the same water vapor gradients over the borders of the TP. Over the TP, the westerly winds control the water vapor transportation. As shown in Figure 3, the water vapors are mainly imported from the south and west borders and exported from the east borders. It shows that the southwest water vapor transport channel from the Arabian Sea to the TP is important, which is consistent with the result reported by [72].

**Figure 3.** Evaluation of simulated atmospheric water vapor flux (kg/(m·s)) and column water vapor burden (kg/m2) from surface to 100 hPa by CESM model in the summer for period from 2001–2005. (**a**) Denotes the model simulations. (**b**) Derived from the NCEP. The area of TP is enclosed by a black line. The blue rectangle indicates the region of southwest vortex.

Besides, as given in Figure 3, comparing the water vapor fluxes from the model simulation and NCEP reanalysis over the Bay of Bengal, a weaker, southwest vortex (blue rectangle in Figure 3) is found in the simulation (Figure 3a). In our model simulation, the weaker southwest vortex induced more water vapor northward transport to the south border of the TP, resulting in more precipitation over the south slopes of the TP (Figure 2c).

Before analyzing the BC's impact on the Tibet climate further, an assessment of aerosol optical depth (AOD) and BC concentration distributions is needed. Figure 4 shows the comparisons of AOD from model simulations and MISR satellite observations, and BC distributions from simulations and MERRA-2 reanalysis data. Figure 4a,c, show a similar distribution feature of AOD over the East China, Northern India, Arabian Sea and Taklimakan Desert, both in model simulation and MISR data. Overall, the simulated values of AOD are higher than those from MISR, especially over the Taklimakan Desert. In contrary, the simulated AOD over other regions is lower than the reanalysis data, and that is similar to the features of surface concentration. It is found that the simulated distributions of the surface BC concentration are also consistent with the results from MERRA-2. Both the model and the reanalysis data indicated three centers with a high concentration of BC over Northern India, the Sichuan Basin and Northern China. The model simulated BC concentration over Northern India, the Sichuan Basin and Northern China ranges from 0.5–2.0 μg/m3, 2.5–3.0 μg/m3 and 1–1.5 μg/m<sup>3</sup> (Figure 4b,d). Due to the lower concentration simulated by the model, the effects of BC on the water vapor transport and climate may be underestimated over the TP. Generally, the coupled model has a good simulated ability on the precipitation, ST, water vapor flux, BC concentration and AOD distributions over the TP. Based on the model simulation, the effects of BC on water vapor transport and its mechanism are further investigated.

**Figure 4.** Distributions of aerosol optical depth (AOD) and black carbon (BC) mass concentration (μg/m3) in the summer for a period from 2001 to 2005, (**a**,**b**) simulated by the CESM model. (**c**,**d**) are derived from the data of Multi-angle Imaging Spectro Radiometer (MISR) satellite observations and MERRA-2 reanalysis data sets. The area of the TP is enclosed by a black line.

#### *3.2. The Radiative Forcing*

In this study, both the direct and indirect effects of BC are considered in the CESM model. Figure 5 shows the clear sky direct radiative forcing (DRF), latent heat (LH) at the surface and the surface sensible heat (SH) fluxes. The results revealed that the shortwave and longwave DRFs at TOA in clear sky are positive with the values about 2.05 and 0.72 W/m2, respectively (Figure 5a,b). The positive shortwave DRF at TOA indicated that the earth–atmosphere system absorbs more solar radiation, while the positive longwave DRFs are due to less outgoing longwave radiation caused by the absorption of BC. In this study, the results of the TOA DRFs are in agreement with the previous researches, in which some studies show the clear sky DRFs at TOA ranges 0.3–2.4 W/m<sup>2</sup> over east Asia [73,74] and 2–6 W/m2 over East China [75,76]. In contrast, the surface shortwave DRFs of clear sky is negative with a mean value of <sup>−</sup>0.86 W/m<sup>2</sup> (Figure 5g), while the longwave DRF is positive with an average of about 0.29 W/m<sup>2</sup> (Figure 5h). Obviously, the shortwave radiative forcing plays a dominant role. Because of the absorption of BC, the reduction of solar radiation at the surface is obvious, resulting in a cooling effect. However, the surface longwave radiative forcing is positive, the reason of which is that the cooled surface emits less longwave radiation, while the warmed atmosphere emits more longwave radiation downward to the surface. In addition, the BC-in-snow effect could reduce the surface longwave albedo, which contributes to the surface positive longwave radiative forcing also.

It is found that both the shortwave and longwave DRFs in the atmosphere are positive with the mean values of 2.9 W/m2 and 0.44 W/m2, respectively (Figure 5d,e). The positive DRFs in the atmosphere presents a pronounced warming effect of BC upon the atmosphere. In addition, due to the BC, the LH at the surface is about 0.82 W/m2 (Figure 5c), and the SH is about 0.13 W/m2 (Figure 5f). The positive heat flux at the surface could significantly offset the surface cooling effect. The net surface heat change is about 0.39 W/m<sup>2</sup> (Figure 5i), dominated by the disturbed latent heat flux (0.82 W/m2), and followed by the longwave radiative forcing (0.29 W/m2).

**Figure 5.** Shortwave and longwave direct radiative forcing (DRF) (W/m2) at the top of atmosphere (TOA) (**a**,**b**), in the atmosphere (**d**,**e**) and on the surface (**g**,**h**), latent heat flux (**c**), sensible heat flux (**f**) and the net heat flux (**i**) during the summer for period from 2001–2005. The area of the TP is enclosed by a black line. the down direction is defined as the positive value, otherwise the upward direction is defined as the negative value.

Simultaneously, the effects of BC on the cloud radiative forcing (CRF) is another important aspect for the radiation transfer process [7]. Here, we mainly analyze the changes in shortwave and longwave CRFs caused by BC. It is found that the mean change in shortwave CRF is about 1.75 W/m<sup>2</sup> (Figure 6a), while the longwave one is about <sup>−</sup>1.62 W/m2 (see Figure 6b), thus the comprehensive change in CRF is warming the earth–atmosphere system. Furthermore, the shortwave and longwave CRFs showed an opposite variation. The change in shortwave CRF is positive over the eastern TP but negative over the western TP. In contrary, the change in longwave CRF is positive over the western TP and negative over the eastern TP. Overall, the net positive changes in CRF denotes a warming effect. The changes in CRF are associated with the cloud microphysics. Due to the indirect effects, BC can affect the cloud microphysics and hence the cloud albedo. In the semi-direct effects, the BC could evaporate the cloud by absorbing more solar radiation. As a result, the increased cloud fraction over the western TP could strengthen the cloud longwave radiative forcing and reduce the shortwave radiative forcing. The decreased cloud is mainly distributed over the southeast TP, that may increase the shortwave radiative forcing while reducing the longwave radiative forcing.

**Figure 6.** Changes in cloud radiative forcing (W/m2) during summer for the period (2001–2005) for (**a**) Shortwave and (**b**) Longwave. The area of the TP is enclosed by a black line.

#### *3.3. Changes in Surface Temperature*

The anomalies in the earth–atmosphere heat equilibrium caused by BC are reflected by the perturbed ST. Based on the former studies, the changes in sea surface temperature (SST) gradients are important in terms of the monsoon dynamics [6,30], especially over the areas of tropical seas [33]. As some previous studies reported [6,30,31], the anomaly meridional SST gradients over the region (5◦–27◦N, 60◦–100◦E) tightly relates to the SASM. In this study, the region (5◦ N–27◦ N, 60◦ E–100◦ E), enclosed by the blue rectangle in Figure 7, is used to calculate the SASM changes. For EASM, the latitudinal land land–sea thermal contrast is more important compared with the meridional thermal contrast [77]. The area enclosed by the red rectangle in Figure 7 is taken as the critical area to calculate the change of EASM.

**Figure 7.** Changes in surface temperature (K) in the summer for period from 2001 to 2005. The area of the TP is enclosed by a black line. The dots denote the changes in surface temperatures that are significant above the 90% confidence level. The blue and red rectangles denote the areas to calculate the changes in the SASM and EASM indices, respectively.

Figure 7 shows that, because of the addition of BC, the surface temperature increases by 0.8–1.6 K over most parts of the TP. The warmer TP is closely related to the surface positive heat flux and the reduced surface albedo (Figure 8).

**Figure 8.** Changes in (**a**) surface shortwave albedo and (**b**) longwave albedo.

The reduced surface albedo means more solar radiation will be absorbed by the land surface. Besides, the increased SST is mainly distributed in the equatorial Arabian Sea and the Southern Bay of Bengal because of the positive heat flux at the surface. In contrast, a pronounced decrease is found over Northern India, Pakistan, Afghanistan and Iran because of the dimming effect of BC. Consequently, the surface temperature shows a decreased meridional gradient. On the other hand, the surface temperature increases over East China, but decreases over the Northwest Pacific Ocean, which could enhance the land-sea thermal contrast.

Generally, the anomalous meridional temperature gradient and the enhanced land–sea thermal contrast could affect the SASM and EASM circulations and further the water vapor transport significantly. More details would be discussed in the following sections.

#### *3.4. Changes in Atmospheric Circulation*

Figure 9 describes the altitude–latitude cross-section of zonal mean (60◦ E–120◦ E) air temperature and wind field changes. As shown in Figure 9a, over South Asia, the upper atmosphere which is marked by the red rectangle has a warming affect. Meanwhile, an anticyclonic circulation is stimulated due to the addition of BC over the area from 15◦ N to 25◦ N, which is consistent with the "Elevated Heat Pump" (EHP) effect of BC reported by Lau and Kim [28]. Due to the EHP effect induced by BC, the anomalous circulation appears opposite to the Hadley circulation, resulting in a northerly wind at the lower atmosphere, which may further reduce the water vapor transport from the Indian Ocean to the TP.

**Figure 9.** (**a**) Altitude–latitude cross-section of the changes in zonal mean air temperature (colors, K) and wind field (arrows) averaged along meridian region of 60◦ E–120◦ E in the latitude zone of 27◦ N–42◦ N in the summer for period from 2001 to 2005. (**b**) Same as (**a**) but for altitude–longitude cross-section of meridional means along latitude region of 27◦ N–42◦ N in the longitude zone of 60◦ E–120◦ E. Red rectangle indicates the area of anticyclonic circulation induced by BC.

Figure 9b describes the altitude–longitude cross-section of changes in meridional mean (27◦ N–42◦ N) air temperature and wind field over East Asia. Generally, East Asia is affected by the subtropical summer monsoon, which attributes the intensity mainly to the latitudinal thermal contrast and secondarily to the meridional thermal contrast [77]. Thus, we analyzed the latitudinal land–sea thermal contrast from 100◦ E to 140◦ E. Besides, the pattern of "western warm eastern cold" induces easterlies from the Northwest Pacific Ocean to the east border of the TP, leading to an enhanced EASM.

To estimate the changes in SASM and EASM due to the BC, the DNS indices are calculated according to Li et al. [64], as shown in Figure 10. Figure 10a shows that the SASM indices under the ALL experiment are smaller than that under the ALL-BC experiment, indicating a weak SASM. Conversely, the EASM index has the opposite variations, in which an enhanced EASM is caused by including BC in the model (Figure 10b).

**Figure 10.** Changes in the (**a**) SASM and (**b**) EASM index from 2001–2005. The red and blue bars denote the simulations under the 'ALL' and 'ALL-BC' experiments, respectively.

#### *3.5. E*ff*ects of BC on the Water Vapor Budget*

It should be noted that the EHP effect can reduce the SASM in summertime. The anticyclone in the upper atmosphere, which is caused by BC, can induce a northerly wind in the lower atmosphere from the south slope of the Tibetan Plateau to Northern India (see Figure 9a). The northerly wind is contrary to the southerly wind over the region of 0◦ N–15◦ N, where can represent the SASM. Thus, the EHP effect can reduce the SASM in summertime. In addition, the strong cooling effect over North India, Pakistan, Afghanistan and Iran, and the warming effect over South India, together with the surrounding ocean area, could reduce the meridional temperature gradient and hence weaken the SASM. Meanwhile, the warmer TP caused by BC can increase the land–sea thermal contrast and the EASM. Overall, the BC could induce a weak SASM but an enhanced EASM and further affect the water vapor transport from the ocean to the TP.

Water vapor plays a significant role in adjusting the air temperature and precipitation. As reported by previous researchers [78], the water vapor from the west and southwest borders contributed mostly to the precipitation over the TP region. Besides, the net importing water vapor to the TP is mainly attributed to the east and west borders with the values of 1991.47 kg/(m·s) and 160.13 kg/(m·s), respectively. Likewise, the exporting water vapor away from the TP is attributed mainly to the north and south borders with the values of 267.77 kg/(m·s) and 381.97 kg/(m·s), respectively [37]. Generally, the four borders of the TP have different characteristics on the water vapor transport. The importing water vapor from the west border serves as the primary contribution to the water vapor over the TP, and that from the southern border is the second contribution [37,79]. Besides, the eastern border has the main export channels, which are closely related to the atmospheric circulations [42].

Figure 11a shows that a pronounced anomalous cyclone over the Northwest Pacific Ocean and a weak cyclone system over Pakistan and Afghanistan are stimulated because of the addition of BC in the simulation. The TP, located at the north side of the cyclone over the Northwest Pacific Ocean, was affected by the perturbed east airflows extending from the Northwest Pacific Ocean to the east border of the TP. Over the western TP, the anomalous southeast flow exerted an impact on the TP because of the weak cyclone system over Pakistan and Afghanistan.

**Figure 11.** Effects of BC on (**a**) water vapor flux (kg/m/s) and (**b**) water vapor budget (Gt/year). The red, green, black and blue lines denote the west, east, north and south borders of the TP, respectively. The dots in (**a**) denote the changes in water vapor that are significant above the 90% confidence level. The purple short lines in the bar in (**b**) denote the error ranges from a negative to a positive standard deviation.

Over the south and north borders of the TP, it was controlled by the anomalous north and southeast flows, respectively. Consequently, more water vapor was exported from the TP. The mean flux of water vapor imported from the west and east borders are 6.3 <sup>×</sup> 10<sup>6</sup> kg/s and 19.5 <sup>×</sup> 106 kg/s, repsectively. The water vapors exported from the south and north border are about 6.9 <sup>×</sup> <sup>10</sup><sup>6</sup> kg/s and 10 <sup>×</sup> <sup>10</sup><sup>6</sup> kg/s, respectively. Thus, the horizontal net water vapor budget of TP is about 8.6 <sup>×</sup> <sup>10</sup><sup>6</sup> kg/s. The net annual mean water vapor budget is about 271Gt/year (see Figure 11b). This net budget is dominated by the east border of TP than others.

#### **4. Discussion**

As mentioned in Section 3.3, the pronounced increase in ST over the TP is caused by the positive surface net heat flux with the value of 0.39 W/m2. The warmer TP is dominated by the surface latent heat flux, followed by the longwave radiative forcing and sensible heat flux.

In Section 3.5, we noted that the net annual mean positive importing water vapor caused by BC over the TP is about 271 Gt/year, in which the positive feature is consistent with the previous study [72]. The anomalous water vapor could partially modify the precipitation pattern. Figure 12 shows the

anomalous precipitation rate over the TP. As given in Figure 12, precipitation was mainly increased over the northern and western TP, with a maximum value of 2 mm/day. Simultaneously, the precipitation was decreased over the southern TP with a maximum decreasing about 3 mm/day. The average value of increased precipitation due to BC over the TP is 0.56 mm/day. The increased precipitation over TP is attributed to the anomalous water vapor transport and atmospheric circulations caused by BC [2,3].

**Figure 12.** Changes in precipitation rate (mm/day) over the TP. The dots denote the changes in precipitation that are significant above the 90% confidence level.

As shown in Figure 11a, the cyclone induced by the surface cooling effect of BC over Pakistan and Afghanistan is beneficial to increase the precipitation over the Western TP. Besides, the perturbed easterly wind and the orographic lifting are in favor of increasing the precipitation over the Northern TP.

As illustrated above, the BC aerosol can induce a warmer and wetter plateau. The results are consistent with current studies [42,73,80]. Our results suggest that the heating of LH, SH and longwave radiative forcing could offset the surface cooling effect. Besides, the BC-in-snow effect could contribute to the warmer TP also [16–19].

In this study, the net surface radiative forcing has included the BC-in-snow effect already. Based on the previous studies, the "snow darkening" effect of BC could also contribute to the warmer TP [81,82], and leads to a reduced snow cover by about 10%–20%, accompanied with a decreased surface albedo. This feedback could heat the TP significantly [82]. In this study, the changes in surface albedo are shown in Figure 8. The results indicate that the mean shortwave and longwave albedo are decreased by 0.0015 and 0.0004, respectively, because of the addition of black carbon. The magnitude of the reduced shortwave albedo is about four times greater than that of the surface longwave albedo. Besides, the reduced shortwave albedo could reach to −0.03 at the south slope of the TP. It is closely related to the high concentration of BC. Generally, the reduced albedo means that the BC-in-snow could absorb more downward solar radiation and longwave radiation to heat the TP.

Moreover, the BC-in-snow effect could further significantly affect the SASM and EASM by the thermal and dynamical forcing. Based on the current study, the warmer TP could increase the land–sea thermal contrast to enhance the EASM in summertime [26], which is consistent with the result of this study. Besides, the BC-in-snow can strengthen the upward motion over the TP to advance the SASM in pre-monsoon [26], which is similar to the EHP effect of BC [27].

The premise of the EHP is that BC can stack up against the south slope of the TP in the springtime and then induce an anomalous warming anticyclone in the upper atmosphere [27]. Because of the latent heat warming effect over the TP, the meridional temperature gradient and the SASM could be enhanced in springtime. It should be noted that Lau and Kim [27] proposed the EHP by the results from the GCM model, neglecting the indirect effects by the off-line method. On the contrary, our results include the direct and indirect effects and the feedback of the ocean. The results show that EHP could induce a warm anticyclone in the upper atmosphere over BC-contaminated regions from Northern India to TP. The northerly wind of the anticyclone in the lower atmosphere and the reduced meridional ST gradient could reduce the SASM in summertime. The results are consistent with the previous studies [28,29].

In the following, a mechanism analysis is performed (Figure 13). Because of the absorption of BC, the solar radiation reaching the surface is reduced obviously, leading to a surface cooling effect. Furthermore, the ST increased over the TP because of the addition of BC. Therefore, a warmer TP can be induced by the BC.

**Figure 13.** Mechanism of BC affecting water vapor transport over the TP.

Though the thermal effects of BC on the lower atmosphere over the TP are warming, spatial discrepancies over the other regions exist. On the one hand, over the regions surrounding Pakistan, Afghanistan and Northern India, the addition of BC can decrease the ST. However, it can increase the SST of Indian Ocean. Thus, a weak thermal contrast between the land and ocean is induced, leading to a weak SASM, then to less water vapor transporting from the south border of the TP. On the other hand, the land–sea thermal contrast is intensified over East Asia, inducing an intensified EASM. The changes in SASM and EASM dramatically impact the water vapor to the TP. Furthermore, a cyclone is stimulated in the upper atmosphere due to the decreased ST over Pakistan, Afghanistan and the Northwest Pacific Ocean. Thus, the western TP is controlled by the southwest winds, leading to more water vapor being transported to the TP. Besides, the eastern TP is controlled by the east winds which is on the north side of the cyclone, resulting in more water vapor being transported to the TP also. Consequently, due to the BC, though the water vapor imported from south side is weak, and the exported water vapor from north side of TP is enhanced, more water vapor is transported from the east and west to the TP. Therefore, because of BC addition, the TP will be wetter.

The uncertainties in the results apply mostly to the aerosol–clouds interactions (ACI, indirect effects) because of the lack of an accurate resolution to represent ACI. Current understanding and classification of this ACI regime is based on the response of the droplet number concentration (Nc) to the aerosol number concentration (Na) and vertical velocity (w) [83], and that are represented by empirical parameterizations. The ACI could affect the performance of the climate model strongly [83]. The other uncertainties mainly derived from the short records of the horizontal and vertical BC distributions. An accurate distribution of BC data sets may help to interpret the transport and buildup of the BC well. Recently, using the observations which include aerosol physical and chemical properties and mixing state to constrain the simulations is an effective approach to limit the uncertainties.

#### **5. Conclusions**

The BC could significantly affect the climate over the TP especially in the summer. Here, combining satellite observations and reanalysis data, we studied the climate effect of BC over the TP by using a fully coupled model. In this study, we have estimated the results of BC affecting the water vapor transport, and we have revealed the concerning mechanism.

The simulation indicates that the BC can induce the positive radiative effects (including direct and indirect effects) and thus exert a pronounced warming over the TP. Based on a detailed analysis, the mean shortwave and longwave radiative forcing at the TOA over TP are 2.05 W/m2 and 0.72 W/m2, respectively.

Besides, considering the LH and SH, the surface net heat forcing presents positive with a value of 0.39 W/m2. Furthermore, the BC can induce an anomalous temperature and then change the atmospheric circulations. As mentioned above, because of the addition of BC, there is a pronounced decreased temperature over Pakistan and Afghanistan, but an increased surface temperature over Southern India and its surrounding ocean. Such temperature patterns can induce a weak thermal contrast between the land and sea, leading to a weak SASM. Thus, less water vapor could be transported from the Indian Ocean to the TP. Besides, over East Asia, the "western warm and eastern cold" pattern enhanced the land–sea thermal contrast over East Asia and the surrounding ocean, inducing an intensified EASM significantly. Consequently, more water vapor is transported from the east of TP. Overall, due to the BC, the net water vapor is positive over the TP, implying a net import of water vapor from the surroundings to the TP. Furthermore, the increased water vapor is closely related to the anomalies in precipitation over the TP.

**Author Contributions:** Conceptualization, Y.L.; methodology, Y.L. and M.L.; investigation, Y.L and M.L.; writing—original draft preparation, M.L.; writing—review and editing, Y.L., M.L., K.A. and Y.T.; visualization, M.L., Q.Z. and Y.T.; funding acquisition, Y.L. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was mainly supported by the Strategic Priority Research Program of Chinese Academy of Sciences (Grant No. XDA2006010301) and the National Natural Science Foundation of China (91737101 and 91744311). This work was also jointly supported by the Foundation of Key Laboratory for Semi-Arid Climate Change of the Ministry of Education in Lanzhou University.

**Acknowledgments:** We acknowledge the MISR (http://eosweb.larc.nasa.gov/PRODOCS/misr/table\_misr. html), CERES (https://ceres.larc.nasa.gov/), EAR-interim (https://www.ecmwf.int/en/forecasts/datasets/reanalysisdatasets/era-interim), MERRA-2 (https://gmao.gsfc.nasa.gov/reanalysis/MERRA-2/) and NCEP (https://www. esrl.noaa.gov/psd/data/gridded/data.ncep.reanalysis.html) science teams for providing excellent and accessible products which favor this study. We also thank the Intergovernmental Panel on Climate Change (IPCC) Fifth Assessment Report (AR5) for providing us the data sets of anthropogenic aerosol emissions (ftp: //ftp.cgd.ucar.edu/cesm/inputdata/atm/cam/chem/trop\_mozart\_aero/emis/).

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
