the effects of pm2.5 concentrations and relative humidity ... · the effects of pm2.5...

The Effects of PM2.5 Concentrations and RelativeHumidity on Atmospheric Visibility in BeijingXiaoyan Wang1 , Renhe Zhang1 , and Wei Yu1

1Department of Atmospheric and Oceanic Sciences & Institute of Atmospheric Sciences, Fudan University, Shanghai,China

Abstract Atmospheric visibility is often used as a proxy for ambient air quality. However, in additionto the concentrations of particulate matter, visibility is also affected by meteorological conditions. Therelative contributions of PM2.5 concentrations and meteorological conditions to visibility are not yet clear.In this study, the individual contributions of PM2.5 concentrations and the relative humidity (RH) to thevisibility are measured based on observations at 12 stations in Beijing. We find that visibility decreasesquickly as PM2.5 concentrations increase at clean condition and decreases slowly when PM2.5

concentrations exceed 100 μg/m3. The visibility decreases linearly as PM2.5 concentrations increase whenPM2.5 concentrations lower than 50 μg/m3; however, there tends to be an exponential relationship asPM2.5 concentrations increase. The PM2.5 concentrations can explain 50% of the variance in visibility inclean and high‐humidity environments, whereas this fraction decreases to 10–15% when the PM2.5

concentrations exceed 200 μg/m3. In contrast, RH has little effect on visibility under dry conditions. Whenthe RH exceeds 40%, atmospheric visibility tends to display an inversely proportional and exponentialrelationship with RH under polluted and clean conditions, respectively. Under high‐humidity and highlypolluted conditions, up to 40% of the variance in visibility is associated with RH. The PM2.5

concentrations dominate the variations in visibility under dry or low PM2.5 concentrations, and thecontributions of RH become increasingly important as PM2.5 concentrations and humidity increase. Thedifference of aerosol types and weather conditions increase the uncertainties of correlation coefficientsbetween visibility and PM2.5/RH. This study emphasizes the atmospheric conditions when employingvisibility as a proxy for air pollution.

Plain Language Summary Atmospheric horizontal visibility provides the public with anintuitive grasp of air quality. People commonly think that the lower the atmospheric visibility is, themore severe is the ambient air pollution. However, in addition to the particulate matter concentrations,visibility is also affected by other meteorological factors, especially the relative humidity. Our study foundthat the variation in visibility is dominated by the PM2.5 concentrations only under dry or low PM2.5

concentrations conditions, in which the visibility can be considered as a proxy of air pollution. However, asthe PM2.5 concentrations and humidity increase, visibility decreases gradually. Compared to PM2.5

concentrations, the contributions of the RH to visibility become increasingly important in low visibilityconditions, and visibility does not exactly reflect the variation in ambient air quality.

1. Introduction

During the last decade, the severe and frequent air pollution events in China have received widespreadattention (Shen et al., 2017; Tao et al., 2017; Watts et al., 2015; R. Zhang, 2017). A good metric for this pollu-tion is the concentration of ambient particulate matter smaller than 2.5 μm (i.e., PM2.5). Considerable atten-tion has been paid to these particles, given their negative effects on human health, society, and publicwelfare (Han & Tong, 2015; Inoue & Takano, 2011; Kan et al., 2004; Silva et al., 2017); moreover, these par-ticles play a critical role in driving climate change because they scatter and absorb solar radiation (Dinget al., 2016; C. Fu et al., 2017; Jacobson, 2001; L. Lin et al., 2016; J. Li et al., 2011; Xia, 2015). Aerosol particlesare efficient in absorbing and scattering solar radiation, and changing the intensity and direction of sunlightand resulting in a reduction in the atmospheric horizontal visibility (Barber, 1984; Husar & Jd, 2000; Panet al., 2009; Singh et al., 2017). Therefore, atmospheric visibility provides the public with an intuitive graspof air quality (Jacob, 1999). Atmospheric visibility is a conventional meteorological parameter in China, andvalues of this parameter are available from as long as the 1960s (H. Chen & Wang, 2015).

©2019. American Geophysical Union.All Rights Reserved.

RESEARCH ARTICLE10.1029/2018JD029269

Key Points:• The relative contributions of PM2.5

concentration and relative humidityto the atmospheric visibility weremeasured

• PM2.5 concentrations dominate thevariations in atmospheric visibilityunder dry or low PM2.5

concentrations• The contribution of relative

humidity to the visibility becomesincreasingly important as the PM2.5

concentrations and RH increase

Correspondence to:R. Zhang,[email protected]

Citation:Wang, X., Zhang, R., & Yu, W. (2019).The effects of PM2.5 concentrations andrelative humidity on atmosphericvisibility in Beijing. Journal ofGeophysical Research: Atmospheres,124, 2235–2259. https://doi.org/10.1029/2018JD029269

Received 28 JUN 2018Accepted 19 JAN 2019Accepted article online 28 JAN 2019Published online 26 FEB 2019

Author Contributions:Conceptualization: Xiaoyan Wang,Renhe ZhangData curation: Xiaoyan Wang, Wei YuFunding acquisition: Renhe ZhangInvestigation: Xiaoyan WangMethodology: Wei YuSupervision: Renhe ZhangValidation: Xiaoyan WangVisualization: Xiaoyan WangWriting ‐ original draft: XiaoyanWangWriting – review & editing: XiaoyanWang, Renhe Zhang

WANG ET AL. 2235

https://orcid.org/0000-0002-4353-3363

https://orcid.org/0000-0001-7750-8679

http://dx.doi.org/10.1029/2018JD029269

http://dx.doi.org/10.1029/2018JD029269

mailto:[email protected]

https://doi.org/10.1029/2018JD029269

https://doi.org/10.1029/2018JD029269

http://publications.agu.org/journals/

http://crossmark.crossref.org/dialog/?doi=10.1029%2F2018JD029269&domain=pdf&date_stamp=2019-02-26

The Chinese government has gradually built a national air quality monitoring network since the end of 2012,but the duration of the available PM2.5 data set is insufficient to support studies of the attribution of long‐term air quality variations and climatic effects of aerosols. Due to the long period over which visibility hasbeen observed and the convenience of access to the data, many previous studies have employed visibilitydata sets as a proxy for air quality to attribute to the historical variations in air quality. For example, R.Zhang et al. (2014) investigated themeteorological conditions that affect the severe pollution event over east-ern China that occurred in January 2013 by using visibility observations. By employing a visibility of 10 kmand a relative humidity (RH) of 85% to represent the thresholds of haze, the annual number of wintertimehaze days over central and eastern China is closely related to the strength of the East Asian winter monsoon(Q. Li, Zhang, & Wang, 2016; R. Zhang et al., 1996). Most previous studies have employed mean visibilityvalues or the number of haze days based on the visibility and RH thresholds as a metric of air quality andattribute the long‐term variations in these parameters.

However, the PM2.5 concentration that corresponds to a visibility of 10 km varies from region to region.For example, this quantity is 110 μg/m3 in Beijing but 65 μg/m3 in Shanghai (K. Huang et al., 2012; P.Zhao et al., 2011). This property reduces the reliability and consistency of air quality studies when a fixedvisibility threshold is employed to reflect the occurrence of haze. In addition, if air quality is measuredonly in terms of whether or not haze occurs, according to the atmospheric visibility threshold, theintensity of air pollution will be ignored. Thus, using visibility does not provide an accurate assessmentof variations in air quality, especially for changes that occur on days with severely polluted or excellentair quality. Therefore, determining the quantitative relationship between observed visibility and PM2.5

concentrations is important for improving the accuracy and representativeness of visibility data sets asa proxy for air quality.

The relationship between PM2.5 concentrations and visibility is complex and nonlinear because of the varia-bility in the size distribution, mixing state, and chemical composition of aerosol particles (Covert et al., 2010;J. Lin et al., 2014; Malm, 2000; Seinfeld & Pandis, 2006; Q. Wang, Sun, et al., 2015). However, due to thecostly instruments and intractable data sets involved, most routine field sampling does not measure themicrophysical and chemical parameters of particles. Alternatively, meteorological conditions affect theemissions of primary particles, the formation of secondary aerosols (S. Guo et al., 2014; R. Huang et al.,2014; G. Wang et al., 2016), the regional transportation of air pollutants, and the hygroscopic growth ofparticles (J. Chen, Zhao, et al., 2014; Y. C. Liu et al., 2016; Sun et al., 2016; Tang et al., 2015), resulting inthe observed variations in the microphysical characteristics and chemical compositions of particles.

Extensive studies have been carried out to establish the quantitative relationship between PM2.5 concentra-tions and atmospheric visibility, accounting for the ambient meteorological conditions. The relationship

between PM2.5 concentrations and visibility has the form of a power function in the JingJinJi and PearlRiver Delta regions (X. Fu et al., 2016; P. Zhao et al., 2011), whereas this relationship can be described byan exponential function in Xi'an and at Tai Mountain (Cao et al., 2012; T. Zhao et al., 2017).

The optical characteristics of aerosols are determined by their chemical compositions, particle size, andshape, which are closely conditioned by atmospheric humidity (Petters & Kreidenweis, 2007; Pilinis et al.,1995; Yan et al., 2009). Humidity affects the aerosol extinction coefficient, because of the hygroscopicchemical components in aerosols, such as sulfate, nitrate, and some organic matters, which can take upwater vapor to grow in size and thus enhance the scattering ability and the single scattering albedo (SSA)of aerosols (Yan et al., 2009). Therefore, the visibility would sharply decrease when the ambient RHwas highat the same level of aerosol mass concentration (Hodkinson, 1966; X. Liu et al., 2013; Malm & Day, 2001).Aerosols can be divided into hydrophilic and hydrophobic species according to its hydroscopicity.Diameter of hydrophilic salts (e.g., ammonium sulfate and ammonium nitrate) will grow many times thanthat at dry condition and scatter more light, but the diameter of hydrophobic aerosols (e.g., black carbon)does not change even at high RH (X. Liu et al., 2013). In addition, most atmospheric aerosols are externallymixed with respect to hygroscopicity and consist of more and less hygroscopic subfractions (Meier et al.,2009; Swietlicki et al., 2008). The ratio between these fractions as well as their content of soluble materialdetermines the hygroscopic growth of the overall aerosol. Hence, the value of the aerosol extinctionefficiency is depended on the chemical composition and its size distribution and not a constant (H. Denget al., 2016; X. Liu et al., 2012, 2013). Therefore, in addition to particle concentrations, the ambient RH

10.1029/2018JD029269Journal of Geophysical Research: Atmospheres

WANG ET AL. 2236

also has a substantial impact on visibility, and the quantitative relationships between PM2.5 concentrationsand visibility vary according to the RH (X. Deng et al., 2008; Yang et al., 2015).

The mass concentration of PM2.5 is measured with moisture removed, but people's intuitive on air quality isdirectly to the range of visibility, which is also highly depended on the ambient humidity. To eliminate theeffects of hygroscopic growth on visibility, semiempirical correlations between visibility and RH are used toinvestigate long‐term variations in air quality (Che et al., 2007; Kotchenruther et al., 1999; J. Li, Li, et al.,2016; K. Wang et al., 2009, 2012). In addition, the RH correction based on the aerosol hygroscopic growthfactor is widely adopted in retrieving satellite‐observed aerosol optical depth to determine surface PM2.5 con-centrations (Q. He et al., 2016; J. Guo et al., 2009; Kong et al., 2017; J. Lin & Li, 2016; Ma et al., 2016; C.Zheng, Zhao, et al., 2017). Moreover, the visibility may display either an exponential or power law relation-ship with the PM2.5 concentrations at the same location, depending on the conditions of wind speed, winddirection, and precipitation intensity (J. Wang et al., 2006).

Both the PM2.5 concentrations and meteorological parameters affect variations in visibility. In some cases,variations in visibility are determined primarily by the particle concentrations. In such cases, the visibilitycan represent the level of air quality very well. However, in some other cases, variations in visibility aredominated by changes in meteorological conditions, and the visibility may not be a good proxy for localair quality at that time. RH is one of the most important meteorological parameters that affect variationsin visibility. However, it is not clear whether any threshold values of PM2.5 concentrations and RH exist thatdefine the conditions under which visibility can be applied in assessing air quality.

The Chinese national air quality monitoring network has been built since 2013, and a data set that describesthe hourly PM2.5 mass concentrations measured by this network is available. Moreover, China has graduallyexchanged manual visibility observations for automatic observations beginning in 2014 (Yin et al., 2017).The use of automatic observation methods avoids the inhomogeneity of manual observations and improvesthe temporal resolution from four times per day to hourly. The synchronous and high temporal resolutionPM2.5 concentration measurements and automatic visibility observations represent a rare opportunity toinvestigate the sensitivity of visibility to PM2.5 concentrations and ambient RH. As the political, cultural,and educational center of China, Beijing is one of the most populous cities in the world. Therefore, in thisstudy, we will explore the effects of PM2.5 mass concentrations and RH on visibility in Beijing since 2014using the hourly PM2.5 concentrations and conventional meteorological data sets. This study assesses theindividual effects of PM2.5 concentrations and RH on variations in visibility and determines the dominantfactors that affect visibility.

2. Data and Methods2.1. Data

Hourly observations of the concentrations of PM2.5, PM10, and other air pollutants released since 2013 wereobtained from the Ministry of Ecology and Environment of the People's Republic of China (X. Wang et al.,2018). Because the observational system is under development and is gradually improving, the duration overwhich data are available differs among the stations. Measurements of the hourly PM2.5 mass concentrationsmade at the 35 state or provincial control stations in Beijing are available from 2013 to 2016.

Visibility is the maximum distance at which an observer can discern the outline of an object against the hor-izon sky (Singh et al., 2017; K. Wang et al., 2012). Historically, routine meteorological measurements of vis-ibility were observed manually four times per day. The accuracy of manual visibility observations dependsstrongly on the experience and skill of the observer, the characteristics of targets, and space interval betweenreference targets. The methods used to observe visibility have gradually changed from manual to automaticsince January 2014. This change helps to avoid the effects of man‐made inhomogeneity of observed visibilityand results in observations with a much higher temporal resolution. Hourly visibility data collected by 12automatic weather stations (AWSs) located in Beijing are used in this study. The data from three of these sta-tions are available from January 2014, whereas the data series of the other stations begin in January 2016(Table 1). Each of the 12 AWSs is matched with a PM2.5 station, and the nearest PM2.5 station is paired witheach AWS, as shown in Figure 1. Other hourly routine meteorological measurements, including RH, preci-pitation, and wind speed, are also collected at the 12 AWSs.


WANG ET AL. 2237

Due to the limited range of the instruments, the maximum observed visibility is 35 km. However, given thatthe actual visibilitymay exceed 35 km,we delete all of themaximumvisibility values (i.e., 35 km) in this studyto avoid the effects of instrument range. Some weather phenomena (e.g., fog, dust, and precipitation) thatalso reduce the atmospheric horizontal visibility are excluded. Hourly observations of low visibility (less than750 m) accompanied by high RH (>90%) are deleted to eliminate the effects of fog. Data collected within 3 hrof the occurrence of precipitation are removed to eliminate the effects of precipitation. Dust events are alsoexcluded because of the negative effects of dust on visibility. A low PM2.5/PM10 ratio is typically used toexclude high PM events caused by long‐range dust transport (Y. Chen et al., 2015; Q. Fu et al., 2010; Tonget al., 2017; Y. Q. Wang, Zhang, et al., 2015), for which we take PM2.5/PM10 ratios of <30% to represent dustevents. The data during the Spring Festival are removed to avoid the effects of human activities. To ensure thedata quality, we deleted the hourly PM2.5 concentration data if they were higher than the synchronous PM10

concentration. For the visibility and RH data set, the hourly data were deleted if the hour‐to‐hour variationexceeded the ±3 standard deviations. We also conducted an interstation comparison, in which the data dif-ference should be lower than 50% percent with respect to the hourly mean value.

2.2. Methods

Aerosol particles (especially fine particles) have considerable effects on visibility due to their ability to absorband scatter light. This behavior is determined mainly by the number size distribution, relative refractive

Table 1Information on the AWSs and Their Paired PM2.5 Stations and the Periods for Which Visibility Data Are Available

AWS/PM2.5 station Visibility data duration AWS/PM2.5 station Visibility data duration

ShunYi/ShunYi January–December 2016 TongZhou/TongZhou January–December 2016HaiDian/WanLiu January–February 2016 ChaoYang/NongZhanGuan January–December 2016YanQing/YanQing January 2014 to December 2016 ChangPing/ChangPing January–December 2016MiYun/MiYun January 2014 to December 2016 MenTouGou/MenTouGou January–December 2016HuaiRou/HuaiRou January–December 2016 Beijng/YiZhuang January 2014 to December 2016PingGu/PingGu January–December 2016 ShiJingShan/GuCheng January–December 2016

Note. The locations of the automatic weather stations (AWSs) and the PM2.5 stations are shown in Figure 1.

Figure 1. Locations of automatic weather stations (AWSs) and their nearest neighboring PM2.5 stations in Beijing.


WANG ET AL. 2238

index and shapes of the particles (H. Deng et al., 2016; Yu et al., 2016). Meanwhile, in addition to thecharacteristics of the particles, RH is another important factor; it impacts visibility by changing the opticalproperties of aerosol particles (Kong et al., 2017; Qu et al., 2015). Therefore, in this study visibility can beconsidered to be a function of PM2.5 concentrations and the RH, as shown in equation (1).

f Vis;PM;RHð Þ ¼ F f Vis; PMð ÞRH; f Vis;RHð ÞPM� �

(1)

Here, f(Vis, PM)RH indicates the effects of PM2.5 concentrations on variations in visibility when the RHremains constant, whereas f (Vis, RH)PM shows the contribution of the RH to visibility when the PM2.5

remains constant. Due to the effects of RH on the hygroscopic growth of aerosols, both the particle mass con-centration and extinction property increase as ambient RH increase. Because of the observation principle ofinstruments, the PM2.5 mass concentrations are measured at a relative constant RH level, which indicatesthe observed PM2.5 concentrations are dry mass concentrations rather than actual concentrations at ambientRH. Therefore, the visibility reduction that results from the aerosol hygroscopic growth is considered as theeffects of RH on the particle extinction (i.e., f (Vis, RH)PM).

The running correlationmethod is used to analyze the relative contributions of PM2.5 concentrations and theRH to variations in visibility. The correlation is first calculated in a window of the first N (the window size)observations. The window is then moved by n (the step size) to a new position, and the correlation is recal-culated. This procedure is repeated until correlations are calculated for the whole data series. We adopt atwo‐way running correlation to assess the PM2.5 concentrations and the RH, respectively. First, the rangeof the PM2.5 concentrations is fixed (the window of the PM2.5 concentrations is written as PMwin), and wethen move the RH from the lowest to the highest values using a fixed step size (RHstp) and window size(RHwin). The correlation coefficients between visibility and the RH and the PM2.5 concentrations are calcu-lated within each PM‐RH window. The range of the PM2.5 concentrations is then moved to the next windowwith a step of PMstp, and the RH is moved as the last step. This process is repeated until the maximum valueof PM2.5 concentrations is reached.

To determine the contributions of the PM2.5 concentrations to the variations in visibility (f(Vis, PM)RH), therange of RH values is set to be a narrow range (i.e., a small RHwin is chosen, meaning that the RH remainsapproximately constant), and the correlation between the PM2.5 concentrations and visibility is calculatedbased on a wide range of PM2.5 concentrations (PMwin). By analogy, the effects of RH on visibility( f (Vis, RH)PM) are determined using a wide RHwin and a narrow PMwin. Many previous studies haveshown that different nonlinear relationships (e.g., inversely proportional or exponential functions) may existbetween visibility and PM2.5 concentrations, depending on the ambient RH (Cao et al., 2012; X. Deng et al.,2008; Yang et al., 2015; P. Zhao et al., 2011).We use linear, inverse, and logarithmic functions to calculate thecorrelation and regression relationships between the visibility data and the PM2.5 concentrations and theRH; the relevant equations are shown in equations (2)–(4), respectively.

Vis ¼ aX þ b (2)

1Vis ¼ aX þ b= (3)

lnVis ¼ aX þ b (4)

where X indicates the PM2.5 concentrations or RH and the parameters a and b are the slope and intercept ofthe linear regression model, respectively. The linear regression equations of 1/Vis and In (Vis) with RH andthe PM2.5 concentrations are then converted to nonlinear equations (i.e., inversely proportional and expo-nential functions) that describe Vis as a function of RH and PM2.5.

3. Results3.1. Relationship of Visibility With PM2.5 Concentrations and RH

Figure 2 shows the seasonal relationship between hourly PM2.5 concentrations and atmospheric horizontalvisibility in Beijing. This figure shows that the visibility decreases as PM2.5 concentrations increase. The rela-tionship between the PM2.5 concentrations and visibility is nonlinear. The visibility decreases quickly at low


WANG ET AL. 2239

PM2.5 concentrations (less than approximately 100 μg/m3) and then decreases slowly. The relationshipsbetween PM2.5 concentrations and visibility depend strongly on the ambient RH. The second column inFigure 2 shows that, for a given PM2.5 concentration, a lower horizontal visibility results from conditionswith relatively high RH due to the hygroscopic behavior of aerosol particles. The standard deviation ofvisibility within specific ranges of PM2.5 concentrations and RH can be used to express the uncertainty ofthe relationships between PM2.5 concentrations and the RH with visibility. The results indicate a largeabsolute value of the standard deviation of visibility at relatively low PM2.5 concentrations, and thisquantity decreases as the PM2.5 concentrations increase. However, with respect to the mean visibility, therelative standard deviation of visibility is relatively small at the lower PM2.5 concentrations in the fourthcolumn of Figure 2. The relative standard deviation grows to its maximum value and then stays flat whenthe PM2.5 concentrations exceed 100 μg/m3. The relative standard deviation of visibility tends to increasewith increasing humidity, which indicates that the contributions of PM2.5 concentrations and the RH tovariations in visibility are more uncertain under conditions of high humidity and in highlypolluted environments.

Atmospheric visibility is the most intuitive measure of air pollution due to the light extinction of particles.The ambient air quality standards of China's Ministry of Environmental Protection (2012) classifies air qual-ity into six categories that reflect the effects of air quality on human health. Taking PM2.5 as the primary pol-lutant, the threshold of PM2.5 concentrations between the excellent, good, lightly polluted, moderatelypolluted, heavily polluted and severely polluted categories are 35, 75, 115, 150, and 250 μg/m3, respectively.Figure 3 shows three correlation coefficients of visibility with the PM2.5 concentrations and RH for differentair quality levels that are based on linear, inverse, and logarithmic visibility, respectively. Taken together,the available data indicate a strong negative correlation between the visibility and PM2.5 concentrations

Figure 2. Seasonal relationship between the PM2.5 concentrations and atmospheric horizontal visibility according to the relative humidity (RH; i.e., the hourlyobserved visibility, the mean visibility, the standard deviation of visibility and the ratio of the standard deviation to the mean value). The mean and standarddeviation of the visibility values within running windows of the PM2.5 concentrations of 20 μg/m

3 determined using a step size of 5 μg/m3 for different RH rangesare shown. Each average visibility is the mean value of at least 50 samples.


WANG ET AL. 2240

according to all three of the correlation functions. For cases associated with excellent and good air quality(i.e., PM2.5 concentrations less than 75 μg/m3), the visibility shows a significant linear or exponentialcorrelation with the PM2.5 concentrations, and relatively high correlation coefficients of approximately−0.8 are obtained. However, the correlation weakens as the PM2.5 concentrations increase. Given greateramounts of pollution, the absolute values of the correlation coefficients drop below 0.5, especially forlightly to heavily polluted conditions. The inversely proportional or exponential relationships betweenvisibility and the PM2.5 concentrations are stronger than the linear correlation in highly pollutedenvironments. In contrast, the correlation between visibility and RH increases as the PM2.5

concentrations increase. The strongest correlation between RH and visibility can exceed −0.6, based onthe inversely proportional or exponential functions in severely polluted conditions.

Figure 3. Correlation coefficients of visibility with the PM2.5 concentrations (left column) and the relative humidity (RH;right column) under different air pollution conditions. The original, inverse, and logarithmic values of visibility areused to calculate the correlation coefficients with PM2.5 concentrations and the RH. The negative correlation coefficientsof the inverse visibility with the PM2.5 concentrations and the RH are shown here to match the y axis with the other twokinds of correlation; that is, positive correlation coefficients are shown. All of the correlation coefficients pass asignificance test at the 0.01 level. Hourly observations that feature low visibility (less than 750 m) accompanied by highrelative humidity (RH > 90%) are deleted to eliminate the effects of fog. All hourly data without fog and dust are used tocalculate the All correlation coefficient. In addition, the hourly data are divided into several segments based on PM2.5concentrations of 0–75 μg/m3, 75–150 μg/m3, 150–250 μg/m3, and above 250 μg/m3. These ranges correspond to excellentand good air quality, lightly and moderately polluted, heavily polluted and severely polluted conditions, respectively,according the ambient air quality standards of China's Ministry of Environmental Protection (GB 3095‐2012).


WANG ET AL. 2241

3.2. Running Correlation of Visibility With PM2.5 Concentrations and RH

As shown in section 3.1, both RH and the PM2.5 concentrations contribute to the variations in atmosphericvisibility. In this section, we discuss their individual contributions to the variations in visibility based on therunning correlation method. The individual contribution of PM2.5 concentrations to variations in visibility isevaluated by keeping the RH nearly unchanged. The size of the sliding window of RH (RHwin) is set to a nar-row value of 10%, and the step size of RH (RHstp) is set to 5%. Accordingly, a wide PMwin of 60 μg/m

3 and astep size of 10 μg/m3 are assigned to PM2.5. In this case, the variations in visibility within each RH‐PM win-dow can be considered to be the results of variations in the ambient PM2.5 concentrations, rather than RH.Figure 4 shows the linear, inversely proportional, and exponential correlation coefficients between the PM2.5

concentrations and visibility (i.e., R(Vis, PM)RH). The linear and exponential correlations are much stron-ger than that of the inversely proportional relationship. The linear correlation coefficients between the PM2.5

concentrations and visibility show the same pattern as the exponential correlations; the correlation coeffi-cients decrease as the PM2.5 concentrations increase. These results indicate strong negative correlationsbetween the PM2.5 concentrations and visibility. Statistically significant absolute correlation coefficients thatexceed 0.5 are obtained when the PM2.5 concentrations are below 80 μg/m3, and these correlation coeffi-cients are even larger (i.e., −0.7) under high‐humidity conditions. For conditions under which the PM2.5

concentrations fall within the range of 80 to 200 μg/m3, the correlation coefficients between visibility andthe PM2.5 concentrations tend to be sensitive to the ambient humidity. The correlation coefficients are

Figure 4. Running correlation coefficient of visibility with PM2.5 concentrations (left column) and relative humidity (RH; right column). The original (upper row),inverse (middle row), and logarithmic (bottom row) values of visibility are used to calculate the correlation coefficients with the PM2.5 concentrations andthe RH, respectively. The negative correlation coefficients of the inverse visibility with the PM2.5 concentrations and the RH are shown here for comparison with theother two. In terms of the correlation coefficient between visibility and the PM2.5, the step size and window size of the PM2.5 concentrations are set to 10 and60 μg/m3, respectively; the step size and window size of RH are set to 5% and 10%, respectively. In terms of the correlation coefficient between visibility and the RH,the step size and window size of the RH are set to 5% and 40%, respectively; the step size and window size of the PM2.5 concentrations are set to 10 and 20 μg/m3,respectively. The x axis and y axis show the medians of the PM2.5 concentrations and the RH windows, respectively. The correlation coefficients marked withdots indicate correlations that are statistically significant at the 0.01 level. Because the size of the slidingwindow of the PM2.5 concentrations is 60 μg/m

3 (left column),there is no correlation coefficient between visibility and the PM2.5 when the PM2.5 concentrations are less than 30 μg/m3. By analogy, the correlation coefficientsbetween visibility and RH are missing when RH is below 20% in the right column. Sliding windows that contain more than 100 observations each are shown.


WANG ET AL. 2242

lower under very dry or very humid conditions, and they are typically higher when the humidity is not toohigh or too low. When the PM2.5 concentrations exceed 200 μg/m3, both the linear and exponentialcorrelation coefficients are not significant and decrease to −0.2 or less.

We set PMwin to a small range and the RHwin to a relatively large range to address the individual contribu-tion of RH to variations in visibility. In contrast to R(Vis, PM)RH, RHwin and PMwin are set to 40% and20 μg/m3, respectively, to obtain the correlation coefficient between RH and visibility (R(Vis, RH)PM).RHstp and PMstp are consistent with R(Vis, PM)RH, that is, 5% and 10 μg/m3, respectively. The right columnof Figure 4 shows that the correlation coefficients of the inversely proportional and exponential relationshipsexceed those of the linear relationship. RH has little effect on variations in visibility under dry conditionswhen RH is less than 40%, which is consistent with the relationship between RH and the aerosol hygroscopicgrowth factor that was identified by many previous studies. That is, PM2.5 concentrations dominate varia-tions in visibility under dry conditions. When RH exceeds 40%, the correlation coefficient between visibilityand the RH increases as RH and the PM2.5 concentrations increase. Under high‐humidity and highly pol-luted conditions, the effects of RH on decreases in visibility are remarkable and are associated with signifi-cant correlation coefficients of approximately −0.6.

The magnitudes of the running correlation coefficients that describe the strength of the relationshipsbetween the PM2.5 concentrations and the RH with visibility depend strongly on the size of the sliding win-dow. The larger the size of the sliding window is, the greater the absolute values of the correlation coeffi-cients will be. We perform a sensitivity analysis by changing PMwin and the RHwin to determine theeffects of the sizes of the sliding windows on the magnitudes of R(Vis, PM)RH and R(Vis, RH)PM, respec-tively. Figure 5 shows the values of R(Vis, PM)RH based on 50 μg/m3 PMwin (compared to 60 μg/m3 inFigure 4) and R(Vis, RH)PM according to an RHwin value of 30% (40% in Figure 4). The correlation

Figure 5. Running correlation coefficient of visibility with PM2.5 concentrations (left column) and the RH (relative humidity; right column). This figure is similarto Figure 4; however, different sizes of sliding windows are used. In the case of the correlation coefficient between visibility and the PM2.5 concentrations, thesize of the sliding window of the PM2.5 concentrations is reduced to 50 μg/m3 (compared to 60 μg/m3 used in Figure 4). The sliding window of the RH is setto 30% (compared to 40% used in Figure 4) to calculate the correlation coefficient between visibility and the RH.


WANG ET AL. 2243

coefficients shown in Figure 5 are lower than those shown in Figure 4; the values in Figure 4 are calculatedusing larger PM and RH sliding windows. However, the correlation coefficients show the same PM‐RHpattern as that shown in Figure 4. Thus, we hereafter show results based on a PMwin of 60 μg/m3 and aRHwin of 40%.

3.3. Regression Analysis of Visibility With PM2.5 Concentrations and RH

Three kinds of correlation coefficients are applied in Figure 4 to present the relationships of visibility withthe PM2.5 concentrations and the RH. We find that the visibility and the PM2.5 concentrations display linearor exponential correlations, and the inversely proportional and exponential relationships between visibilityand the RH are more robust than the linear correlation. Therefore, we use the running linear and exponen-tial functions to establish regression models that describe the relationship between visibility and the PM2.5

concentrations, respectively. For the case of RH and visibility, regression analyses are conducted using theinversely proportional and exponential regressions. The same sliding windows and step sizes for RH andthe PM2.5 concentrations used to produce Figure 4 are applied here.

Figure 6 shows the seasonal coefficients of determination (R2) of the regression equations. The R2 values ofthe two kinds of regressionmodels show the same seasonal PM‐RH pattern. In terms of the regressionmodelthat describes visibility as a function of the PM2.5 concentrations, both the linear and exponential modelsshow higher R2 values during summer and spring and lower R2 values during winter. The fraction oforganic/inorganic and secondary/primary compositions varies day to day due to the more frequent air pol-lution events in winter, which may lead to a more inhomogeneous and diverse mixtures of aerosols andreduce the R2 in winter. The R2 values of the regression models decrease as the PM2.5 concentrations

Figure 6. R2 values of the running regression equations of visibility and PM2.5 concentrations (first and second columns) and RH (relative humidity; third andfourth columns). The original and logarithmic visibility are used to establish the regression equations with PM2.5, respectively (first and second columns,respectively). The inverse and logarithmic visibility are used to establish the regression equations with the RH (third and fourth columns, respectively). R2 valuesmarked with dots indicate regression equations that are statistically significant at the 0.01 level. Sliding windows that contain more than 100 observationseach are shown. MAM = March–May; JJA = June–August; SON = September–November; DJF = December–February.


WANG ET AL. 2244

increase. The PM2.5 concentrations can explain approximately one half of the variance in visibility when thePM2.5 concentrations are below 60 μg/m3 in summer. However, in other seasons, the R2 values are sensitiveto the ambient RH at low PM2.5 concentrations. During spring and autumn, approximately 50% of the var-iance in visibility can be attributed to the variations in PM2.5 concentrations in high‐humidity environments;however, the contribution of PM2.5 concentrations decreases to 30% in dry environments. The effects of thePM2.5 concentrations on visibility are complex in wintertime; this season yields the highest R2 value of 0.4under dry conditions and low PM2.5 concentrations. With regard to the light andmoderate air pollution cate-gories, which feature PM2.5 concentrations of approximately 80~150 μg/m3, 20–30% of the variance in visi-bility in spring is associated with PM2.5 concentrations; however, this proportion is only 10–15% for the otherthree seasons. The R2 values tend to be below 0.1 when the PM2.5 concentrations exceed 200 μg/m3 for allof seasons.

The inversely proportional and exponential running regressions between visibility and RH are evaluated toobtain the individual effects of RH on visibility (i.e., f(Vis, RH)PM). As shown in the third and fourth col-umns of Figure 6, there is good agreement between the R2 values of the two kinds of regression equations.Variations in the atmospheric visibility are independent of ambient RH until the RH reaches 40% and thePM2.5 concentration concentrations exceed 80 μg/m3. The R2 values of the regression models increase totheir maximum values under high‐humidity and highly polluted conditions. These values are highest inthe spring, which features R2 values that exceed 0.4.

The R2 values based on two different regression functions, that is, linear and exponential correlation forf (Vis, PM)RH and inversely proportional and exponential regression for f(Vis, RH)PM, display the samedistribution as the variations in PM2.5 concentrations and the RH, as shown in Figure 6. The ratio of theR2 of the linear regression of f (Vis, PM)RH to the R2 of the exponential regression is shown in Figure 7 tocompare the robustness of the two types of regression functions. The R2 of the linear regression is higherthan that of the exponential regression when the PM2.5 concentrations are below 50 μg/m3. These resultsindicate that the relationship between visibility and the PM2.5 concentrations is linear for low PM2.5 concen-trations. Nevertheless, as PM2.5 concentrations increase, they tend to display an exponential correlation. Forvisibility and RH, visibility exhibits an exponential relationship with RH in dry or clean environments. Thisrelationship becomes inversely proportional under high‐humidity and polluted conditions. According to theabove analysis, both the PM2.5 concentrations and RH contribute to variations in atmospheric visibility.PM2.5 concentrations dominate the variations in visibility under dry conditions or at low PM2.5 concentra-tions, while the contribution of RH becomes increasingly important as PM2.5 concentrations and ambientRH increase.

3.4. Sensitivity of Atmospheric Visibility to PM2.5 Concentrations and RH

The linear regression functions that describe the relationship between inverse and logarithmic visibility withPM2.5 concentrations and the RH based on equations (3) and (4) are converted to the inversely proportionaland exponential regression models as follows:

Vis ¼ 1aX þ b

and Vis ¼ eaXþb

where X indicates the PM2.5 concentrations or RH and parameters a and b are the slope and intercept inequations (3) and (4), respectively. In addition, we then take the relative derivatives of visibility with respectto the RH and the PM2.5 concentrations to evaluate the sensitivity of visibility to the two variables. The thirdand fourth columns in Figure 8 show the relative derivatives of visibility with respect to the PM2.5 concen-trations according to the linear and exponential regression functions of f (Vis, PM)RH, respectively. The sen-sitivities of visibility to the PM2.5 concentrations based on the two regression models are strongly consistentwith each other. At low PM2.5 concentrations, the visibility is very sensitive to the variations in PM2.5 con-centrations, and the visibility decreases quickly as PM2.5 concentrations increase. Regarding the sensitivityto the RH, visibility shows relatively strong sensitivities to the RH at the highest RH levels when the ambientPM2.5 concentrations are below 100 μg/m3. The threshold values of the insensitive PM2.5 concentrationsremain about the same when the RH is less than 60% and then decrease rapidly. The relative sensitivitiesof visibility to the RH highlight the effects variation in the high‐humidity variation to the reduction of


WANG ET AL. 2245

visibility, especially during summertime. Visibilities under the low PM2.5 concentrations and humidenvironment are most sensitive to the changes in RH.

3.5. Effects of Aerosol Source on the Relationship of Visibility With PM2.5 Concentrations and RH

The particle extinction property is a strong function of particle size, particle shape, chemical composition,and hygroscopic properties of aerosols (X. Liu et al., 2008). Size‐dependent hygroscopic properties are oftenobserved, corresponding to distinct composition for Aitken and accumulation mode aerosols. The hygro-scopic properties of inorganic species are well characterized, but organics can exhibit hygroscopic behaviorranging from hydrophobic to somewhat hydrophilic (Cubison et al., 2008; Tijjani & Uba, 2013; Yan et al.,2009). Atmospheric aerosols consist of a complicated and different mixture of various chemical compositionswith diverse aerosol types depending on aerosol source (J. Chen, Qiu, et al., 2014; Cheng, Wiedensohler,

Figure 7. The ratio of two kinds of coefficients of determination (R2) based on the different visibility regression equations shown in Figure 6. RH = relative humid-ity; MAM = March–May; JJA = June–August; SON = September–November; DJF = December–February.


WANG ET AL. 2246

et al., 2008). Therefore, we analyze the effects of aerosol type, air mass backward trajectory, and synopticprocess on the particle extinction property in Beijing.

The aerosol optical data used in this section were originally from Aerosol Robotic Network (AERONET). Toensure the data quality, only the level 2.0 data set at Beijing and Beijing_CAMS are involved to classify theaerosol types. The AERONET spectral products are available at the wavelengths of 440, 675, 870, and1020 nm. The SSA spectral behavior and Ångström exponent (AE440–870) parameter were applied to identifyaerosol mixture. Dust aerosol tend to have increasing SSA values with wavelength, while carbonaceousaerosol, such as those from fossil fuel and biomass burning, show decreasing SSA valves with increasingwavelength (J. Li et al., 2015). The absorption of aerosol mixtures is contributed by both dust and anthropo-genic aerosol, which leading to a nonmonotonic SSA spectral behavior. Tian et al. (2018) employed thewidely used parameter, that is, AE440_870, as another criterion to identify aerosol types, in which the 0.7and 1.4 of AE are taken as the threshold of anthropogenic and dust aerosols. We summarized the aerosolclassification criteria in Table 2. The aerosol is identified as anthropogenic, dust, or mix‐type aerosol onlyif both the SSA spectral and AEmeet the criteria in Table 2. For those whomeet none of the SSA‐AE criteria,the samples are deleted in this study, which indicates aerosols with significant characteristics of types areinvolved. Due to the high missing rate of the SSA data in the two AERONET stations, the daily aerosol typesare available only at 315 days over the year of 2014 to 2016. The occurrence of each aerosol types is shown inFigure 9. The visibilities are more relevant to PM2.5 concentrations for the specific aerosol types compared tothe overall samples, especially for the anthropogenic aerosol conditions. The significant effects of RH on vis-ibility exhibit under polluted and high‐humidity conditions. But there is no sample during the high PM2.5

concentrations for the anthropogenic and dust aerosol types because of the limitation occurrence of these

Figure 8. Proportional sensitivity of visibility to PM2.5 concentrations (first and second columns) and RH (relative humidity; third and fourth columns) determinedusing different regression models (unit: %). The relative derivatives of visibility with respect to PM2.5 concentrations and the RH are used to measure theproportional sensitivity of visibility to these two variables. The relative derivative is defined as dVis=Vis

dX=X 100% ¼ dVisdX

XVis 100%; here, X indicates the PM2.5 concentra-

tions or the RH. dVisdX is the derivative of visibility to RH or PM2.5 concentrations. MAM = March–May; JJA = June–August; SON = September–November;DJF = December–February.


WANG ET AL. 2247

two aerosol types. There is no significant improvement on the correlationsbetween visibility and RH for the mix‐type aerosol compared to thosebased on overall samples.

To analyze the effects of air mass origins on particle extinction properties,the Trajectory Statistics (TrajStat) model (http://ready.arl.noaa.gov/HYSPLIT.php) is employed for backward trajectory clustering in Beijing.The input meteorological data were downloaded from the NationalOceanic and Atmospheric Administration global reanalysis data set. Thestarting height was set to 10 m, and the endpoint was Beijing. The back-

ward time was 24 hr. Figure 10 shows the annual five clustering results, that is, northwest long‐distanceair mass (T1), southwest short‐distance air mass (T2), west short‐distance air mass (T3), north‐northwestlong‐distance air mass (T4), and east short‐distance air mass (T5). Except for T2, the other backward trajec-tories show nearly the same frequency during the whole year. The correlation coefficients between visibilitywith PM2.5 concentration and RH corresponding to each backward trajectory type are also shown inFigure 10. It shows relative good air quality for Beijing when the air mass comes from northwest (T1) ornorth‐northwest (T4) directions, while the southwesterly airstream (T2) is easier to aggravate the severeair pollution. Compared to the correlation coefficient based on the all samples that is shown in Figure 4,the visibility is more relevant to the PM2.5 concentration for T2, T3, and T5 trajectory types. When PM2.5

Table 2Characteristics of the SSA and AE Values Under the Control of DifferentAerosol Types

Aerosol types SSA spectral AE

Anthropogenic Decreasing >1.4Dust Increasing <0.7Mixed type Nonmontonic 0.8 < AE < 1.3

Note. SSA = single scattering albedo.

Figure 9. Occurrence of each aerosol type during 2014 to 2016 (top panel) and the annual mean correlation coefficient between visibility and PM2.5/relative humid-ity (RH) based on different aerosol types. The aerosol types are identified based on the criteria in Table 2. The logarithmic and inverse values of visibility are used tocalculate the correlation coefficients PM2.5 concentrations and RH, respectively. The step size and window size of PM2.5 and RH are the same as those in Figure 4.


WANG ET AL. 2248

http://ready.arl.noaa.gov/HYSPLIT.php

http://ready.arl.noaa.gov/HYSPLIT.php

concentrations are lower than 100 μg/m3, the correlation coefficients can reach up to−0.7 to−0.6 and do notvary with the ambient RH. Except for the east short‐distance air mass, the relationship between visibility andRH under each specific trajectory type shows almost the same pattern as that based on the all samples.

In addition to the aerosol types and air mass backward trajectory, the local meteorological conditions alsoaffect the ambient aerosol property. The T‐mode principal component analysis has been successfully appliedto the studies of general circulation models, the climate change, and air pollution (Valverde et al., 2015; Xuet al., 2016). The obliquely rotated T‐mode principal component analysis method developed by COST action733 (http://www.cost733.org) has been used to operate the circulation classification over Beijing. The

Figure 10. Annual mean backward trajectory cluster in Beijing for the period of 2014 to 2016 (top panel) and the annual mean running correlation coefficient ofvisibility with PM2.5 concentrations (left column) and RH (relative humidity; right column) based on each backward trajectory. The logarithmic and inverse valuesof visibility are used to calculate the correlation coefficients PM2.5 concentrations and RH, respectively. The step size and window size of PM2.5 and RH are the sameas those in Figure 4.


WANG ET AL. 2249

http://www.cost733.org

anomalies of the daily mean ERA‐interim mean surface air pressure and wind field data set from 2014 to2016 over 100–130°E and 25–55°N were applied to identify the circulation classification. Six typicalcirculation types (CTs) are revealed and illustrated in Figure 11, that is, high‐pressure center (CT1), highpressure in west and low in east (CT2), high pressure in north and low in south (CT3), low pressure innorthwest (CT4), low pressure in northeast (CT5), and low‐pressure center (CT6). The frequencies of thesix circulation patterns are comparable in spring (figure is not shown here). CT6 and CT1 dominate thecirculation pattern in summer and autumn, respectively. The occurrences of CT1 and CT2 are morefrequent than other CTs. Figure 12 exhibits the comparison of the correlation coefficients betweenvisibility and PM2.5/RH corresponding to different circulation patterns. It shows noticeable improvementof the correlation between visibility and PM2.5 concentrations for the CT3, CT4, CT5, and CT6, especiallyfor the clear and high‐humidity conditions. In terms of relationship between visibility and RH, there is nomeasurable increase in correlation coefficients except for the CT3 conditions. Therefore, all of the aerosoltype, air mass trajectory, and circulation patterns can affect to some extent the relationship betweenvisibility and PM2.5 concentrations and RH. Correlation between visibility and PM2.5 is more sensitive tothe aerosol or weather characters than the correlation between visibility and RH.

3.6. Effects of Other Factors on Variations in Visibility

We have described the contributions of ambient RH and PM2.5 concentrations to variations in visibility.However, the visibility at a fixed RH and PM2.5 concentration level is highly uncertain, as reflected by thestandard deviation of visibility that is shown in Figure 2. We select the top and bottom 10% of the visibilityvalues within each interval of PM2.5 concentrations to compare their differences. The left column ofFigure 11 demonstrates that the top and bottom 10% of the visibility values change with PM2.5 concentra-tions and RH. In general, the RH values of the top 10% of the visibility values are much lower than thoseof the bottom 10% due to the hygroscopic behavior of aerosol particles. However, there is no obvious differ-ence in the RH between the top and bottom 10% of the visibility values at low PM2.5 concentrations duringthe spring and winter. The curves that describe the relationship between the mean RH and the mean PM2.5

concentration for the top and bottom 10% of the visibility values are shown in the middle column ofFigure 13. This column shows great differences in the RH between the top and bottom 10% of visibility

Figure 11. Typical circulation patterns based on the anomaly of mean surface pressure and 10‐m U‐V wind during the period of 2014 to 2016. Number in thebracket indicates the frequency of the specific circulation type. CT = circulation type.


WANG ET AL. 2250

values in summer and autumn, and the RH discrepancy between the two conditions remains nearlyunchanged as the PM2.5 concentrations increase. That is, in summer and autumn, the difference in thetop and bottom 10% of the visibility values can be attributed to the difference in RH. Nonetheless, forspring and winter, there is a small difference in the RH between the top and bottom 10% of visibilityvalues at low PM2.5 concentrations. In addition, the differences increase and then change only slightlywhen PM2.5 concentrations exceed 100 μg/m3. Even though the fine particles dominate the aerosol

Figure 12. Annual mean running correlation coefficient of visibility with PM2.5 concentrations (left column) and RH (relative humidity; right column) based oneach circulation type (CT).


WANG ET AL. 2251

extinction, some components in coarse particle can still absorb or scatter solar radiation resulting in thereduction of atmospheric visibility. The right column of Figure 13 shows the ratio of PM2.5 to PM10.Percentage of PM2.5 in bottom 10% of visibility conditions is higher than that in top 10% conditions by 5–10% in summer and autumn, which is attributed to the aggravated air pollution caused by fine particles inhigh‐humidity conditions. However, for the spring and winter, the percentage of PM2.5 in PM10 showsalmost the same magnitude between top and bottom 10% visibility under clear condition, and thepercentage for bottom 10% visibility is much lower than that in summer and autumn. The relative higherpercentage of coarse particles (i.e., low ratio of PM2.5 to PM10) in bottom 10% visibility conditions absorbsand scatters more solar radiation and contributes to the low visibility under low PM2.5 concentrations.Therefore, the high proportion of coarse particle may be one reason for the bottom 10% visibility in winterand spring.

The atmospheric total optical extinction coefficient is defined as the sum of the extinction due to the gasphase and the particle phase (Cheng, Zhang, et al., 2008; Malm, 2000; Singh et al., 2017). The impacts ofgas‐phase molecules on visibility include Rayleigh scattering of air molecules and the absorption of solarradiation by NO2 (X. Deng et al., 2008; Pitchford et al., 2007; Pui et al., 2014; Sati & Mohan, 2014). Thus,except for the high RH and high proportion of coarse particles, which influence the particle extinction,the NO2 content dominates themagnitude of gas‐phase extinction. Figure 14 exhibits the frequency distribu-tion of the occurrence time of the bottom and top 10% visibility conditions. If the RH exceeds 70%, low vis-ibility is considered to be the result of hygroscopic growth of aerosol particles. Employing an RH value of 70%

Figure 13. Relationship between the PM2.5 concentrations and the top or bottom 10% of the visibility values according to the relative humidity (RH; left column),the mean RH variation as the PM2.5 concentrations increase (middle column), and the ratio of PM2.5 to PM10 under top and bottom 10% visibility (right column).The top and bottom 10% of visibility values within each 5‐μg/m3 interval of the PM2.5 concentrations are shown in the left column. MAM = March–May;JJA = June–August; SON = September–November; DJF = December–February.


WANG ET AL. 2252

as a threshold, the frequency distributions of the occurrence time are constructed separately. Figure 14reveals that the RH values of most of the top 10% of visibility values are below 70%, which indicates thatthe high visibility can be attributed to the occurrence of dry conditions. However, only in summer andautumn, the RH values that correspond to the majority of the bottom 10% of the visibility values exceed70%. In addition, most low visibility values that are accompanied by high humidity occur at night due tothe diurnal cycle of RH. During spring and winter, the RH values associated with approximately twothirds of the bottom 10% of the visibility values are below 70%. Thus, we conclude that factors other thanhigh humidity result in the low visibility. Moreover, for the bottom 10% of the visibility values in springand winter, the times at which RH is below 70% tend to occur during the morning rush hour (i.e., 7:00–9:00 a.m.) and evening rush hour (i.e., 18:00–20:00).

In general, the gas extinction coefficient can be considered as constant. However, the high vehicle emissionsduring rush hour raise the level of NOx in the atmosphere, which may increase the gas extinction coefficientdue to the absorption of solar radiation. We compare the NO2 concentrations associated with the top andbottom 10% of the visibility values under low‐ and high‐humidity conditions. Figure 15 indicates that, underdry conditions with low PM2.5 concentrations during spring and winter, the NO2 concentrations associatedwith the bottom 10% of visibility values exceed those associated with the top 10% of the visibility values. Thedifference in NO2 concentrations between the top and bottom 10% of visibility values decreases gradually as

Figure 14. The frequency of the occurrence time of the top and bottom 10% of visibility values when the relative humidity (RH) is less than 70% (left column) andgreater than 70% (right column). The blue and orange numbers in each subplot represent the sample sizes of the top and bottom 10% of visibility values, respectively.MAM = March–May; JJA = June–August; SON = September–November; DJF = December–February.


WANG ET AL. 2253

the PM2.5 concentrations increase. In other words, the effects of gas extinction cannot be ignored at lowPM2.5 concentrations in spring and winter; this conclusion is consistent with the results of X. Deng et al.(2008). However, a majority of the bottom 10% of visibility values that occur in summer and autumn canbe attributed to high humidity.

4. Conclusions and Discussion

The relationships of atmospheric visibility with PM2.5 mass concentrations and RH in Beijing are investi-gated using hourly observations of these quantities. Visibility decreases quickly as PM2.5 concentrationsincrease at low PM2.5 concentrations and decreases slowly when ambient PM2.5 concentrations exceed100 μg/m3. The relationships between the PM2.5 concentrations and visibility depend strongly on the ambi-ent RH. At a given PM2.5 concentration, higher RH values are associated with lower horizontal visibility.Thus, both PM2.5 concentrations and the RH affect the variations in atmospheric visibility.

Running correlation and regressionmethods are used to reveal the individual contributions of PM2.5 concen-trations and the RH to the variations in visibility and to determine the dominant factors that affect visibility.

Figure 15. The variations in mean NO2 concentrations with the PM2.5 concentrations given the top and bottom 10% of the visibility values when the relativehumidity (RH) is less than 70% (left column) and greater than 70% (right column). MAM = March–May; JJA = June–August; SON = September–November;DJF = December–February.


WANG ET AL. 2254

Three alternative functions, that is, linear, inversely proportional, and exponential functions, are examinedto determine the best fit model that describes the relationship between RH and PM2.5 concentrations to vis-ibility. In general, the linear and exponential correlations between visibility and the PM2.5 concentrationsare much stronger than that of the inversely proportional relationship. Both linear and exponential correla-tion coefficients between PM2.5 concentrations and visibility decrease as the PM2.5 concentrations increase.The results reflect strong negative correlations between the PM2.5 concentrations and visibility; statisticallysignificant correlation coefficients that exceed 0.5 occur when the PM2.5 concentrations are below 80 μg/m3.The correlation coefficients are even higher under high‐humidity conditions, for which PM2.5 concentra-tions can explain up to 50% of the variance in visibility. When the PM2.5 concentrations exceed200 μg/m3, both the linear and exponential correlation coefficients are not significant and decrease to−0.2 or less, and only 10–15% of the variations in visibility can be explained by the PM2.5 concentrations.Although both the linear and exponential correlations are more robust than the inversely proportional rela-tionship, the R2 value of the linear regression exceeds that of the exponential regression when the PM2.5 con-centrations are below 50 μg/m3. This result indicates that visibility and PM2.5 concentrations obey a linearrelationship in low PM2.5 concentrations. Nevertheless, as the PM2.5 concentrations increase, the relation-ship tends to become exponential.

In contrast, RH has a small effect on the variations in visibility under dry conditions (RH< 40%), which indi-

cates that variations in visibility are dominated by the PM2.5 concentrations. When the RH exceeds 40%, theinversely proportional and exponential correlations between RH and visibility are more robust than the lin-ear relationship; in addition, they tend to be exponential in clean environments and inversely proportionalunder polluted conditions. The correlation coefficient between visibility and the RH increases as RH and the

PM2.5 concentrations increase. This correlation coefficient reaches its negative maximum of 0.6 under high‐humidity and highly polluted conditions, under which the RH can explain approximately 40% of the var-iance in visibility.

Both the RH and the PM2.5 concentrations affect the variations in visibility. Comparing their relative contri-butions, we find that the PM2.5 concentrations dominate the variations in visibility under dry conditions orlow PM2.5 concentrations, and the contribution of RH becomes increasingly important as the PM2.5 concen-trations and the ambient humidity increase. In addition to the ambient PM2.5 concentrations and RH, theeffects of coarse particle and gas extinction cannot be ignored at low PM2.5 concentrations. The effects ofaerosol type, air mass backward trajectory, and circulation patterns on the correlation between visibilityand PM2.5/RH were discussed. Most of the correlation coefficients will increase in the specific aerosol orweather type compared to the overall cases. The correlation coefficients between visibilities and PM2.5 aremore sensitive to the aerosol type and weather conditions than those between visibilities and RH.

The observed PM2.5 concentrations are drymass concentrations at a constant RH level due to the principle ofinstruments. We considered the visibility reduction according to the hygroscopic growth of aerosols as the

effect of RH in this study. Both the increase in the dry mass concentration of PM2.5 and the hygroscopicgrowth of aerosols reduce the visibility. Therefore, when taking the long‐term observed visibilities as a proxyfor air pollutants, the RH effect on extinction should be eliminated rather than directly using the haze‐dayoccurrence only based on the visibility value.

The tapered element oscillating microbalance (TEOM) and Beta Attenuation Monitor have been standar-dized and recommended by several organizations (e.g., Environmental Protection Agency and EuropeanMonitoring and Evaluation Programme) and used worldwide to obtain PM2.5 concentrations. HourlyPM2.5 dry mass concentrations are publicly available with the assumption that water is effectively removedprior to the measurement. However, as both the filter material of the monitor and the aerosol particles arecapable of retaining a significant amount of water even at low RH, the basic assumption may not be valid,resulting in significant bias in reported PM2.5 concentrations (Chow, 1995; Perrino et al., 2013). Adsorbedwater may be partly retained on the filter and particulates even at low RHs (called hysteresis; Kiss et al.,2017). The magnitude of water retention strongly depends on the filter material. Tierney and Conner(1967) concluded that the effect of RH on glass fiber filters was insignificant, but at RH > 55% the effecton collected particulates was significant. Demuynck (1975) demonstrated that the amount of water irrever-sibly retained by cellulose filters is comparable to or even higher than the mass of particulate matter


WANG ET AL. 2255

collected during the foggy day. In addition, the particulate mass collected in 1 hr is small; thus, the biascaused by water may be excessive.

Except for the hysteresis effects of filters, the loss of semivolatile components is another factor for theobserved PM2.5 concentration bias. The TEOM instrument has the advantage over conventional gravimetricmethods of particle mass monitoring to provide data on an almost real‐time basis and to be a cost and labor‐effective method, which is widely used in China (K. He et al., 2001; Yin et al., 2017; J. Zhang et al., 2017; Y,Zheng, Che, et al., 2017). To minimize interferences from the evaporation and condensation of water ontothe filter, there is a heating system at the inlet of the TEOM with temperature of about 50 °C. However,the higher temperature can also lead to significant volatilization losses of the some semivolatile componentsduring collection. Particulate ammonium nitrate and semivolatile organic compounds are thought to be themain compounds lost in the TEOM system (Aurelie Charron et al., 2004; Chow, 1995). The amount of semi-volatile compounds associated with the particles is expected to depend on the temperature and the RH(Stelson & Seinfeld, 1982). Additionally, the RH has a significant impact on the growth of particulate matter.Comparisons between TEOMs and reference gravimetric methods in different countries have shown that forwarmer and dryer regions the agreement is better than for colder and damper regions (Williams &Bruckmann, 2002). Therefore, the measurement bias of the hourly PM2.5 mass concentrations will increasethe uncertainty of our results, especially at high‐humidity conditions.

This study indicates the conditions under which atmospheric visibility can be used as a proxy for air pollu-tion. The application of these results will improve the representativeness and reliability of the use of visibilityin the field of air pollution attribution and determining its climatic effects. However, all of the results arebased on new, automatically collected visibility observations. When the results are extended to long‐term,manual observations of visibility, homogenization and systematic deviations between manual andautomatic observations should be considered due to the differences in the principles used in making theobservations and the uncertainties associated with manual observations.

ReferencesAurelie Charron, M., Harrison, R., Moorcroft, S., & Booker, J. (2004). Quantitative interpretation of divergence between PM10 and PM2.5

mass measurement by TEOM and gravimetric (Partisol) instruments. Atmospheric Environment, 38(3), 415–423. https://doi.org/10.1016/j.atmosenv.2003.09.072

Barber, P. (1984). Absorption and scattering of light by small particles. Journal of Colloid and Interface Science, 98, 290–291.Cao, J., Wang, Q., Chow, J., Watson, J., Tie, X., Shen, Z., et al. (2012). Impacts of aerosol compositions on visibility impairment in Xi'an,

China. Atmospheric Environment, 59, 559–566. https://doi.org/10.1016/j.atmosenv.2012.05.036Che, H., Zhang, X., Li, Y., Zhou, Z., & Qu, J. J. (2007). Horizontal visibility trends in China 1981–2005. Geophysical Research Letters, 34,

L24706. https://doi.org/10.1029/2007GL031450Chen, H., &Wang, H. (2015). Haze days in North China and the associated atmospheric circulations based on daily visibility data from 1960

to 2012. Journal of Geophysical Research: Atmospheres, 120, 5895–5909. https://doi.org/10.1002/2015JD023225Chen, J., Qiu, S., Shang, J., Wilfrid, O. M., Liu, X., Tian, H., & Boman, J. (2014). Impact of relative humidity and water soluble constituents

of PM2. 5 on visibility impairment in Beijing, China. Aerosol and Air Quality Research, 14, 260–268.Chen, J., Zhao, C., Ma, N., & Yan, P. (2014). Aerosol hygroscopicity parameter derived from the light scattering enhancement factor

measurements in the North China Plain. Atmospheric Chemistry and Physics, 14, 3459–3497.Chen, Y., Luo, B., & Xie, S. (2015). Characteristics of the long‐range transport dust events in Chengdu, Southwest China. Atmospheric

Environment, 122, 713–722. https://doi.org/10.1016/j.atmosenv.2015.10.045Cheng, Y., Wiedensohler, A., Eichler, H., Heintzenberg, J., Tesche, M., Ansmann, A., et al. (2008). Relative humidity dependence of aerosol

optical properties and direct radiative forcing in the surface boundary layer at Xinken in Pearl River Delta of China: An observationbased numerical study. Atmospheric Environment, 42(25), 6373–6397. https://doi.org/10.1016/j.atmosenv.2008.04.009

Cheng, Y., Zhang, Y., Hu, M., & Wiedensohler, A. (2008). An observation‐based method for investigating the atmospheric aerosol radiativeproperties in Pearl River delta of China, in. Beijing: Science Press.

China's Ministry of Environmental Protection (2012). Ambient air quality standards. Retrieved from http://210.72.1.216:8080/gzaqi/Document/gjzlbz.pdf, (last access 15 February 2019).

Chow, J. C. (1995). Measurement methods to determine compliance with ambient air quality standards for suspended particles. Journal ofthe Air & Waste Management Association, 45(5), 320–382. https://doi.org/10.1080/10473289.1995.10467369

Covert, S., Charlson, J., & Ahlquist, C. (2010). A study of the relationship of chemical composition and humidity to light scattering byaerosols. Journal of Applied Meteorology, 11, 968–976.

Cubison, M., Ervens, B., Feingold, G., Docherty, K., Ulbrich, I., Shields, L., et al. (2008). The influence of chemical composition and mixingstate of Los Angeles urban aerosol on CCN number and cloud properties. Atmospheric Chemistry and Physics, 8(18), 5649–5667. https://doi.org/10.5194/acp‐8‐5649‐2008

Demuynck, M. (1975). Determination of irreversible absorption of water by cellulose filters. Atmospheric Environment, 9, 523–528.Deng, H., Tan, H., Li, F., Cai, M., Chan, P., Xu, H., et al. (2016). Impact of relative humidity on visibility degradation during a haze event: A

case study. Science of the Total Environment, s569–570, 1149–1158.Deng, X., Tie, X., Wu, D., Zhou, X., Bi, X., Tan, H., & Li, F. (2008). Long‐term trend of visibility and its characterizations in the Pearl River

Delta (PRD) region, China. Atmospheric Environment, 42(7), 1424–1435. https://doi.org/10.1016/j.atmosenv.2007.11.025


WANG ET AL. 2256

AcknowledgmentsThis study was supported by theNational Basic Research Program ofChina (2015CB453203), National KeyResearch and Development Program ofChina (2016YFA0600602), and theNational Natural Science Foundation ofChina (41790470 and 41805117). Thevisibility data and the other conven-tional meteorological observations wereobtained from the website (http://data.cma.cn). In addition, the hourly PM2.5

data set is available from the website(http://www.pm25.in).

https://doi.org/10.1016/j.atmosenv.2003.09.072



https://doi.org/10.1029/2007GL031450

https://doi.org/10.1002/2015JD023225



http://210.72.1.216:8080/gzaqi/Document/gjzlbz.pdf

http://210.72.1.216:8080/gzaqi/Document/gjzlbz.pdf

https://doi.org/10.1080/10473289.1995.10467369

https://doi.org/10.5194/acp-8-5649-2008

https://doi.org/10.5194/acp-8-5649-2008


http://data.cma.cn

http://data.cma.cn

http://www.pm25.in

Ding, A., Nie, W., Huang, X., Chi, X., Sun, J., Kerminen, V. M., et al. (2016). Long‐term observation of air pollution‐weather/climateinteractions at the SORPES station: A review and outlook. Frontiers of Environmental Science & Engineering, 10, 15. https://doi.org/10.1007/s11783‐016‐0877‐3

Fu, C., Ding, A., & Wu, J. (2017). Review on studies of air pollution and climate change interactions in monsoon Asia: Research andforecast. In C. Chang, H. Kuo, N. Lau, et al. (Eds.), The Global Monsoon System Research and Forecast (Vol. 25, pp. 315–326). Hangzhou,China: World Meteorological Organization.

Fu, Q., Zhuang, G., Li, J., Huang, K., Wang, Q., Zhang, R., et al. (2010). Source, long‐range transport, and characteristics of a heavy dustpollution event in Shanghai. Journal of Geophysical Research, 115, D00K29. https://doi.org/10.1029/2009JD013208

Fu, X., Wang, X., Hu, Q., Li, G., Xiang, D., Zhang, Y., et al. (2016). Changes in visibility with PM(2.5) composition and relative humidity at abackground site in the Pearl River Delta region. Journal of Environmental Sciences, 40, 10–19.

Guo, J., Zhang, X., Che, H., Gong, S., An, X., Cao, C., et al. (2009). Correlation between PM concentrations and aerosol optical depth ineastern China. Atmospheric Environment, 43(37), 5876–5886. https://doi.org/10.1016/j.atmosenv.2009.08.026

Guo, S., Hu, M., Zamora, M. L., Peng, J., Shang, D., Zheng, J., et al. (2014). Elucidating severe urban haze formation in China. Proceedings ofthe National Academy of Sciences of the United States of America, 111(49), 17,373–17,378. https://doi.org/10.1073/pnas.1419604111

Han, Y. Q., & Tong, Z. (2015). Health effects of fine particles (PM 2.5) in ambient air. Science China. Life Sciences, 58, 1–3.He, K., Yang, F., Ma, Y., Zhang, Q., Yao, X., Chan, C. K., et al. (2001). The characteristics of PM2. 5 in Beijing, China. Atmospheric

Environment, 35, 4959–4970.He, Q., Li, C., Geng, F., Zhou, G., Gao, W., Yu, W., et al. (2016). A parameterization scheme of aerosol vertical distribution for surface‐

level visibility retrieval from satellite remote sensing. Remote Sensing of Environment, 181, 1–13. https://doi.org/10.1016/j.rse.2016.03.016

Hodkinson, J. (1966). Calculations of color and visibility in urban atmospheres polluted by gaseous N [O. Sub. 2]. International Journal ofAir and Water Pollution, 10, 137–144.

Huang, K., Zhuang, G., Lin, Y., Wang, Q., Fu, J. S., Zhang, R., et al. (2012). Impact of anthropogenic emission on air quality over amegacity‐revealed from an intensive atmospheric campaign during the Chinese Spring Festival. Atmospheric Chemistry and Physics, 12,17,151–17,185.

Huang, R., Zhang, Y., Bozzetti, C., Ho, K., Cao, J., Han, Y., et al. (2014). High secondary aerosol contribution to particulate pollution duringhaze events in China. Nature, 514(7521), 218–222. https://doi.org/10.1038/nature13774

Husar, R., & Jd, M. (2000). Distribution of continental surface aerosol extinction based on visual range data. Atmospheric Environment,34(29‐30), 5067–5078. https://doi.org/10.1016/S1352‐2310(00)00324‐1

Inoue, K., & Takano, H. (2011). Particulate matter–induced health effects: Who is susceptible? Environmental Health Perspectives, 119,446–454.

Jacob, D. (1999). Introduction to atmospheric chemistry. New Jersey: Princeton University Press.Jacobson, M. (2001). Strong radiative heating due to the mixing state of black carbon in atmospheric aerosols. Nature, 409(6821), 695–697.

https://doi.org/10.1038/35055518Kan, H., Chen, B., Chen, C., Fu, Q., & Chen, M. (2004). An evaluation of public health impact of ambient air pollution under various energy

scenarios in Shanghai, China. Atmospheric Environment, 38(1), 95–102. https://doi.org/10.1016/j.atmosenv.2003.09.038Kiss, G., Imre, K., Molnár, Á., & Gelencsér, A. (2017). Bias caused by water adsorption in hourly PM measurements. Atmospheric

Measurement Techniques, 10(7), 2477–2484. https://doi.org/10.5194/amt‐10‐2477‐2017Kong, L., Xin, J., Liu, Z., Zhang, K., Tang, G., Zhang, W., & Wang, Y. (2017). The PM2.5 threshold for aerosol extinction in the Beijing

megacity. Atmospheric Environment, 167.Kotchenruther, R. A., Hobbs, P. V., & Hegg, D. A. (1999). Humidification factors for atmospheric aerosols off the mid‐Atlantic coast of the

United States. Journal of Geophysical Research, 104(D2), 2239–2251. https://doi.org/10.1029/98JD01751Li, J., Carlson, B. E., & Lacis, A. A. (2015). Using single‐scattering albedo spectral curvature to characterize East Asian aerosol mixtures.

Journal of Geophysical Research: Atmospheres, 120, 2037–2052. https://doi.org/10.1002/2014JD022433Li, J., Li, C., Zhao, C., & Su, T. (2016). Changes in surface aerosol extinction trends over China during 1980‐2013 inferred from quality‐

controlled visibility data. Geophysical Research Letters, 43, 8713–8719. https://doi.org/10.1002/2016GL070201Li, J., Yi, Y., Minnis, P., Huang, J., Yan, H., Ma, Y., et al. (2011). Radiative effect differences between multi‐layered and single‐layer clouds

derived from CERES, CALIPSO, and CloudSat data. Journal of Quantitative Spectroscopy and Radiative, 112(2), 361–375. https://doi.org/10.1016/j.jqsrt.2010.10.006

Li, Q., Zhang, R., & Wang, Y. (2016). Interannual variation of the wintertime fog–haze days across central and eastern China and itsrelation with East Asian winter monsoon. International Journal of Climatology, 36(1), 346–354. https://doi.org/10.1002/joc.4350

Lin, J., Donkelaar, A. V., Xin, J., Che, H., & Wang, Y. (2014). Clear‐sky aerosol optical depth over East China estimated from visibilitymeasurements and chemical transport modeling. Atmospheric Environment, 95, 258–267. https://doi.org/10.1016/j.atmosenv.2014.06.044

Lin, J., & Li, J. (2016). Spatio‐temporal variability of aerosols over East China inferred by merged visibility‐GEOS‐Chem aerosol opticaldepth. Atmospheric Environment, 132, 111–122. https://doi.org/10.1016/j.atmosenv.2016.02.037

Lin, L., Gettelman, A., Fu, Q., & Xu, Y. (2016). Simulated differences in 21st century aridity due to different scenarios of greenhouse gasesand aerosols. Climatic Change, 1–16.

Liu, X., Cheng, Y., Zhang, Y., Jung, J., Sugimoto, N., Chang, S.‐Y., et al. (2008). Influences of relative humidity and particle chemicalcomposition on aerosol scattering properties during the 2006 PRD campaign. Atmospheric Environment, 42, 1525–1536. https://doi.org/10.1016/j.atmosenv.2007.10.077

Liu, X., Gu, J., Li, Y., Cheng, Y., Qu, Y., Han, T., et al. (2013). Increase of aerosol scattering by hygroscopic growth: Observation, modeling,and implications on visibility. Atmospheric Research, 132, 91–101.

Liu, X., Zhang, Y., Cheng, Y., Hu, M., & Han, T. (2012). Aerosol hygroscopicity and its impact on atmospheric visibility and radiativeforcing in Guangzhou during the 2006 PRIDE‐PRD campaign. Atmospheric Environment, 60, 59–67. https://doi.org/10.1016/j.atmosenv.2012.06.016

Liu, Y. C., Wu, Z. J., Tan, T. Y., Wang, Y. J., Qin, Y. H., Zheng, J., et al. (2016). Estimation of the PM 2.5 effective hygroscopic parameter andwater content based on particle chemical composition: Methodology and case study. Science China Earth Sciences, 59, 1683–1691.

Ma, Z., Hu, X., Sayer, A. M., Levy, R., Zhang, Q., Xue, Y., et al. (2016). Satellite‐Based Spatiotemporal Trends in PM2.5 Concentrations:China, 2004–2013. Environmental Health Perspectives, 124, 184.

Malm, W. (2000). Spatial and seasonal patterns and temporal variability of haze and its constituents in the United States. CooperativeInstitute for Research in the Atmosphere, Colorado State University.


WANG ET AL. 2257

https://doi.org/10.1007/s11783-016-0877-3

https://doi.org/10.1007/s11783-016-0877-3

https://doi.org/10.1029/2009JD013208


https://doi.org/10.1073/pnas.1419604111

https://doi.org/10.1016/j.rse.2016.03.016

https://doi.org/10.1016/j.rse.2016.03.016

https://doi.org/10.1038/nature13774

https://doi.org/10.1016/S1352-2310(00)00324-1

https://doi.org/10.1038/35055518


https://doi.org/10.5194/amt-10-2477-2017

https://doi.org/10.1029/98JD01751

https://doi.org/10.1002/2014JD022433

https://doi.org/10.1002/2016GL070201

https://doi.org/10.1016/j.jqsrt.2010.10.006

https://doi.org/10.1016/j.jqsrt.2010.10.006

https://doi.org/10.1002/joc.4350








Malm, W. C., & Day, D. E. (2001). Estimates of aerosol species scattering characteristics as a function of relative humidity. AtmosphericEnvironment, 35(16), 2845–2860. https://doi.org/10.1016/S1352‐2310(01)00077‐2

Meier, J., Wehner, B., Massling, A., Birmili, W., Nowak, A., Gnauk, T., et al. (2009). Hygroscopic growth of urban aerosol particles inBeijing (China) during wintertime: A comparison of three experimental methods. Atmospheric Chemistry and Physics, 9, 6865–6880.https://doi.org/10.5194/acp‐9‐6865‐2009

Pan, X., Yan, P., Tang, J., Ma, J., Wang, Z., Gbaguidi, A., & Sun, Y. (2009). Observational study of influence of aerosol hygroscopic growthon scattering coefficient over rural area near Beijing mega‐city. Atmospheric Chemistry and Physics, 9(19), 7519–7530. https://doi.org/10.5194/acp‐9‐7519‐2009

Perrino, C., Canepari, S., & Catrambone, M. (2013). Comparing the performance of Teflon and quartz membrane filters collecting atmo-spheric PM: Influence of atmospheric water. Aerosol and Air Quality Research, 13, 137–147. https://doi.org/10.4209/aaqr.2012.07.0167

Petters, M., & Kreidenweis, S. (2007). A single parameter representation of hygroscopic growth and cloud condensation nucleus activity.Atmospheric Chemistry and Physics, 7(8), 1961–1971. https://doi.org/10.5194/acp‐7‐1961‐2007

Pilinis, C., Pandis, S. N., & Seinfeld, J. H. (1995). Sensitivity of direct climate forcing by atmospheric aerosols to aerosol size and compo-sition. Journal of Geophysical Research, 100(D9), 18,739–18,754. https://doi.org/10.1029/95JD02119

Pitchford, M., Malm, W., Schichtel, B., Kumar, N., Lowenthal, D., & Hand, J. (2007). Revised algorithm for estimating light extinction fromIMPROVE particle speciation data. Journal of the Air & Waste Management Association, 57(11), 1326–1336. https://doi.org/10.3155/1047‐3289.57.11.1326

Pui, D., Chen, S., & Zuo, Z. (2014). PM2.5 in China: Measurements, sources, visibility and health effects,and mitigation. Particuology, 13,1–26. https://doi.org/10.1016/j.partic.2013.11.001

Qu, W., Wang, J., Zhang, X., Wang, D., & Sheng, L. (2015). Influence of relative humidity on aerosol composition: Impacts on lightextinction and visibility impairment at two sites in coastal area of China. Atmospheric Research, 153, 500–511. https://doi.org/10.1016/j.atmosres.2014.10.009

Sati, A. P., & Mohan, M. (2014). Analysis of air pollution during a severe smog episode of November 2012 and the Diwali Festival overDelhi, India. International Journal of Remote Sensing, 35(19), 6940–6954. https://doi.org/10.1080/01431161.2014.960618

Seinfeld, J., & Pandis, S. (2006). From air pollution to climate change. Atmospheric Chemistry and Physics, 429–443.Shen, H., Shu, T., Chen, Y., Ciais, P., Güneralp, B., Ru, M., et al. (2017). Urbanization‐induced population migration has reduced ambient

PM2.5 concentrations in China. Science Advances, 3, e1700300.Silva, R. A., West, J. J., Lamarque, J. F., Shindell, D. T., Collins, W. J., Faluvegi, G., et al. (2017). Corrigendum: Future global mortality from

changes in air pollution attributable to climate change. Nature Climate Change, 7.Singh, A., Bloss, W. J., & Pope, F. D. (2017). 60 years of UK visibility measurements: Impact of meteorology and atmospheric pollutants on

visibility. Atmospheric Chemistry and Physics, 17, 1–32.Stelson, A. W., & Seinfeld, J. H. (1982). Relative humidity and pH dependence of the vapor pressure of ammonium nitrate‐nitric acid

solutions at 25 °C. Atmospheric Environment, 16, 993–1000.Sun, Y., Chen, C., Zhang, Y., Xu, W., Zhou, L., Cheng, X., et al. (2016). Rapid formation and evolution of an extreme haze episode in

Northern China during winter 2015. Scientific Reports, 6(1), 27151. https://doi.org/10.1038/srep27151Swietlicki, E., Hansson, H.‐C., Hämeri, K., Svenningsson, B., Massling, A., McFiggans, G., et al. (2008). Hygroscopic properties of sub-

micrometer atmospheric aerosol particles measured with H‐TDMA instruments in various environments—A review. Tellus Series B:Chemical and Physical Meteorology, 60(3), 432–469. https://doi.org/10.1111/j.1600‐0889.2008.00350.x

Tang, G., Zhu, X., Hu, B., Xin, J., Wang, L., Münkel, C., et al. (2015). Impact of emission controls on air quality in Beijing during APEC2014: Lidar ceilometer observations. Atmospheric Chemistry and Physics, 15, 12,667–12,680. https://doi.org/10.5194/acp‐15‐12667‐2015

Tao, J., Zhang, L., Zhang, R., & Cao, J. (2017). A review of current knowledge concerning PM2.5 chemical composition, aerosol opticalproperties, and their relationships across China. Atmospheric Chemistry and Physics, 1–70.

Tian, P., Zhang, L., Ma, J., Tang, K., Xu, L., Wang, Y., et al. (2018). Radiative absorption enhancement of dust mixed with anthropogenicpollution over East Asia. Atmospheric Chemistry and Physics, 18, 7815–7825. https://doi.org/10.5194/acp‐18‐7815‐2018

Tierney, G. P., & Conner, W. D. (1967). Hygroscopic effects on weight determinations of particulates collected on glass‐fiber filters.American Industrial Hygiene Association Journal, 28(4), 363–365. https://doi.org/10.1080/00028896709342537

Tijjani, B., & Uba, S. (2013). The effect of hygroscopic growth on urban aerosols, the International Institute for Science. Technology andEducation (IISTE), 25, 58–75.

Tong, D. Q., Wang, J. L., Gill, T. E., Lei, H., &Wang, B. (2017). Intensified dust storm activity and valley fever infection in the southwesternUnited States. Geophysical Research Letters, 44, 4304–4312. https://doi.org/10.1002/2017GL073524

Valverde, V., Pay, M. T., & Baldasano, J. M. (2015). Circulation‐type classification derived on a climatic basis to study air quality dynamicsover the Iberian Peninsula. International Journal of Climatology, 35, 2877–2897. https://doi.org/10.1002/joc.4179

Wang, G., Zhang, R., Gomez, M. E., Yang, L., Zamora, M. L., Hu, M., et al. (2016). Persistent sulfate formation from London fog to Chinesehaze. Proceedings of the National Academy of Sciences of the United States of America, 113, 13,630–13,635. https://doi.org/10.1073/pnas.1616540113

Wang, J., Zhang, Y., Min, S., Liu, X., Zeng, L., Cheng, C., & Xiao‐Feng, X. (2006). Quantitative relationship between visibility and massconcentration of PM2.5 in Beijing. Journal of Environmental Sciences, 18, 475–481.

Wang, K., Dickinson, R., & Liang, S. (2009). Clear sky visibility has decreased over land globally from 1973 to 2007. Science, 323(5920),1468–1470. https://doi.org/10.1126/science.1167549

Wang, K., Dickinson, R., Su, L., & Trenberth, K. (2012). Contrasting trends of mass and optical properties of aerosols over the NorthernHemisphere from 1992 to 2011. Atmospheric Chemistry and Physics, 12(19), 9387–9398. https://doi.org/10.5194/acp‐12‐9387‐2012

Wang, Q., Sun, Y., Jiang, Q., Du, W., Sun, C., Fu, P., & Wang, Z. (2015). Chemical composition of aerosol particles and light extinctionapportionment before and during heating season in Beijing, China. Journal of Geophysical Research: Atmospheres, 120, 12,708–12,722.https://doi.org/10.1002/2015JD023871

Wang, X., Dickinson, R., Su, L., Zhou, C., & Wang, K. (2018). PM 2.5 pollution in China and how it has been exacerbated by terrain andmeteorological conditions. Bulletin of the American Meteorological Society, 99, 105–119.

Wang, Y. Q., Zhang, X. Y., Sun, J. Y., Zhang, X. C., Che, H. Z., & Li, Y. (2015). Spatial and temporal variations of the concentrations of PM10,PM2.5 and PM1 in China. Atmospheric Chemistry and Physics, 15, 13,585–13,598. https://doi.org/10.5194/acp-15-13585-2015

Watts, N., Adger, P. W. N., Agnolucci, P., Blackstock, J., Byass, P. P., Cai, W., et al. (2015). Health and climate change: Policy responses toprotect public health. British Dental Journal, 219, 1861–1914.

Williams, M., & Bruckmann, P. (2002). Guidance to member states on PM10 monitoring and intercomparisons with the reference method,Final Report der EC Working Group on Particulate Matter.


WANG ET AL. 2258

https://doi.org/10.1016/S1352-2310(01)00077-2

https://doi.org/10.5194/acp-9-6865-2009

https://doi.org/10.5194/acp-9-7519-2009

https://doi.org/10.5194/acp-9-7519-2009

https://doi.org/10.4209/aaqr.2012.07.0167

https://doi.org/10.5194/acp-7-1961-2007

https://doi.org/10.1029/95JD02119

https://doi.org/10.3155/1047-3289.57.11.1326

https://doi.org/10.3155/1047-3289.57.11.1326

https://doi.org/10.1016/j.partic.2013.11.001

https://doi.org/10.1016/j.atmosres.2014.10.009


https://doi.org/10.1080/01431161.2014.960618

https://doi.org/10.1038/srep27151

https://doi.org/10.1111/j.1600-0889.2008.00350.x

https://doi.org/10.5194/acp-15-12667-2015

https://doi.org/10.5194/acp-18-7815-2018

https://doi.org/10.1080/00028896709342537

https://doi.org/10.1002/2017GL073524

https://doi.org/10.1002/joc.4179



https://doi.org/10.1126/science.1167549

https://doi.org/10.5194/acp-12-9387-2012

https://doi.org/10.1002/2015JD023871

https://doi.org/10.5194/acp-15-13585-2015

Xia, X. (2015). Parameterization of clear‐sky surface irradiance and its implications for estimation of aerosol direct radiative effect andaerosol optical depth. Scientific Reports, 5(1), 14376. https://doi.org/10.1038/srep14376

Xu, J., Chang, L., Qu, Y., Yan, F., Wang, F., & Fu, Q. (2016). The meteorological modulation on PM2. 5 interannual oscillation during 2013to 2015 in Shanghai, China. Science of the Total Environment, 572, 1138–1149. https://doi.org/10.1016/j.scitotenv.2016.08.024

Yan, P., Pan, X., Tang, J., Zhou, X., Zhang, R., & Zeng, L. (2009). Hygroscopic growth of aerosol scattering coefficient: A comparativeanalysis between urban and suburban sites at winter in Beijing. Particuology, 7(1), 52–60. https://doi.org/10.1016/j.partic.2008.11.009

Yang, Y., Liu, X., Qu, Y., An, J. L., Jiang, R., Zhang, Y., et al. (2015). Characteristics and formation mechanism of continuous hazes inChina: A case study during the autumn of 2014 in the North China Plain. Atmospheric Chemistry and Physics Discussions, 15(7),10,987–11,029. https://doi.org/10.5194/acpd‐15‐10987‐2015

Yin, Z., Wang, H., & Chen, H. (2017). Understanding severe winter haze events in the North China Plain in 2014: Roles of climateanomalies. Atmospheric Chemistry and Physics, 17, 1–27.

Yu, X., Ma, J., An, J., Yuan, L., Zhu, B., Liu, D., et al. (2016). Impacts of meteorological condition and aerosol chemical compositions onvisibility impairment in Nanjing, China. Journal of Cleaner Production, 131, 112–120. https://doi.org/10.1016/j.jclepro.2016.05.067

Zhang, J., Chen, Z., Lu, Y., Gui, H., Liu, J., Liu, W., et al. (2017). Characteristics of aerosol size distribution and vertical backscatteringcoefficient profile during 2014 APEC in Beijing. Atmospheric Environment, 148, 30–41. https://doi.org/10.1016/j.atmosenv.2016.10.020

Zhang, R. (2017). Atmospheric science: Warming boosts air pollution. Nature Climate Change, 7(4), 238–239. https://doi.org/10.1038/nclimate3257

Zhang, R., Qiang, L., & Zhang, R. (2014). Meteorological conditions for the persistent severe fog and haze event over eastern China inJanuary 2013. Science China Earth Sciences, 57, 26–35.

Zhang, R., Sumi, A., & Kimoto, M. (1996). Impact of El Nino on the East Asianmonsoon: A diagnostic study of the '86/87 and '91/92 events.Journal of the Meteorological Society of Japan, 74(1), 49–62. https://doi.org/10.2151/jmsj1965.74.1_49

Zhao, P., Zhang, X., Xu, X., & Zhao, X. (2011). Long‐term visibility trends and characteristics in the region of Beijing, Tianjin, and Hebei,China. Atmospheric Research, 101(3), 711–718. https://doi.org/10.1016/j.atmosres.2011.04.019

Zhao, T., Yang, L., Yan, W., Zhang, J., Lu, W., Yang, Y., et al. (2017). Chemical characteristics of PM1/PM2. 5 and influence on visual rangeat the summit of Mount Tai, North China. Science of the Total Environment, 575, 458–466.

Zheng, C., Zhao, C., Zhu, Y., Wang, Y., Shi, X., &Wu, X. (2017). Analysis of influential factors for the relationship between PM2.5 and AODin Beijing. Atmospheric Chemistry and Physics, 19, 1–57.

Zheng, Y., Che, H., Zhao, T., Zhao, H., Gui, K., Sun, T., et al. (2017). Aerosol optical properties observation and its relationship tometeorological conditions and emission during the Chinese National Day and Spring Festival holiday in Beijing. Atmospheric Research,197, 188–200. https://doi.org/10.1016/j.atmosres.2017.07.003


WANG ET AL. 2259

https://doi.org/10.1038/srep14376

https://doi.org/10.1016/j.scitotenv.2016.08.024

https://doi.org/10.1016/j.partic.2008.11.009

https://doi.org/10.5194/acpd-15-10987-2015

https://doi.org/10.1016/j.jclepro.2016.05.067


https://doi.org/10.1038/nclimate3257

https://doi.org/10.1038/nclimate3257

https://doi.org/10.2151/jmsj1965.74.1_49



the effects of pm2.5 concentrations and relative humidity ... · the effects of pm2.5...

Documents