Quantifying the Inﬂuences of PM 2.5 and Relative Humidity on Change of Atmospheric Visibility over Recent Winters in an Urban Area of East China

: Fine particulate matters (PM 2.5 ) and relative humidity (RH) in the ambient atmosphere are the leading anthropogenic and natural factors changing atmospheric horizontal visibility. Based on the analysis of environmental and meteorological data observed over 2013–2019 in Nanjing, an urban area in East China, this study investigated the inﬂuences of PM 2.5 and RH on atmospheric visibility changes over recent years. The visibility had signiﬁcantly negative correlations with the PM 2.5 concentrations and RH changes. The nonlinear relationships existed between PM 2.5 concentrations and visibility, as well as between RH and visibility, with the inﬂection points in the atmospheric visibility changes. The PM 2.5 inﬂection concentrations were 81.0 µ g m − 3 , 76.0 µ g m − 3 , 49.0 µ g m − 3 , and 33.0 µ g m − 3 , respectively, for the RH ranges of RH < 60%, 60% ≤ RH < 80%, 80% ≤ RH < 90%, and RH ≥ 90%, indicating that the improvement of visibility with reducing PM 2.5 concentrations could be more di ﬃ cult under the humid meteorological condition. The visibility changes were most sensitive to PM 2.5 concentrations in the RH range of 60–80% in this urban area of East China. The relative contributions of natural factor RH and anthropogenic factor PM 2.5 to variations of wintertime atmospheric visibility were quantiﬁed with 54.3% and 45.7%, respectively, revealing an important role of natural factor RH in the change of atmospheric visibility in the urban area of East Asian monsoon region.


Introduction
Atmospheric horizontal visibility (hereinafter, visibility) is meteorologically defined as the distance at which a normal observer can perceive a black object viewed against the horizon [1]. For the air environment, visibility can express the ambient atmospheric opacity reflecting the levels of air pollution, especially haze pollution resulting from high aerosol concentrations, which could deteriorate atmospheric visual range [1,2]. With the adverse influence of low visibility on traffic, human health, climate change, and other significant aspects, the spatial and temporal changes of visibility have attracted considerable attention worldwide as part of scientific studies of atmospheric zones, as well as one suburban site over the national air quality monitoring network in China ( Figure 1). The PM 2.5 concentrations were averaged over nine observational sites in Nanjing to characterize the variations of fine aerosol particles over this urban area. The meteorological data of surface observation in Nanjing were obtained from the Meteorological Data Sharing Network of China Meteorological Administration [51]. The meteorological data of surface observation included visibility and RH with temporal resolutions of 3 h in order to analyze the meteorological variations and the influence of RH on visibility for this study. Daily mean visibility, RH, and PM 2.5 concentrations were computed from the hourly observation data of meteorology and environment.
Atmosphere 2020, 11,461 3 of 12 the variations of fine aerosol particles over this urban area. The meteorological data of surface observation in Nanjing were obtained from the Meteorological Data Sharing Network of China Meteorological Administration [51]. The meteorological data of surface observation included visibility and RH with temporal resolutions of 3 h in order to analyze the meteorological variations and the influence of RH on visibility for this study. Daily mean visibility, RH, and PM2.5 concentrations were computed from the hourly observation data of meteorology and environment.

Categorizations of Atmospheric Visibility and RH
In this study, we categorized daily visibility into the good, low, and poor visibility levels, respectively, with the lower limit of 10 km and upper limits of 10 km and 5 km [35,52]. To more quantitatively analyze the variation of visibility, the visibility values were divided into three levels: > 10.0 km, 5.0-10.0 km, and < 5.0 km. Aerosol light scattering enhancement factor grows smoothly with RH < 60% and exhibits a rapid enhancement at high RH > 60% [45]. The aerosol hygroscopic effect becomes significant when RH exceeds 90% [45]. Thus, we divided RH into four ranges: RH < 60%, 60% ≤ RH < 80%, 80% ≤ RH < 90%, and RH ≥ 90%.

A Measure of Relative Importance in Multiple Regression
Multiple regression analysis has two distinct applications: Prediction and explanation [53]. When multiple regression is used for explanatory purpose, we are interested in the extent of each variable contributing to the response variable [54]. Standardized regression coefficients are the most common measure of relative importance when multiple regression is used [55,56].
We constructed a stepwise multiple linear regression model to quantify the effect of RH and PM2.5 on visibility variability. The model fitted the daily visibility with daily RH and PM2.5. The daily values of visibility, PM2.5 and RH satisfied approximately normal distribution. The fit has the form:

Categorizations of Atmospheric Visibility and RH
In this study, we categorized daily visibility into the good, low, and poor visibility levels, respectively, with the lower limit of 10 km and upper limits of 10 km and 5 km [35,52]. To more quantitatively analyze the variation of visibility, the visibility values were divided into three levels: > 10.0 km, 5.0-10.0 km, and < 5.0 km. Aerosol light scattering enhancement factor grows smoothly with RH < 60% and exhibits a rapid enhancement at high RH > 60% [45]. The aerosol hygroscopic effect becomes significant when RH exceeds 90% [45]. Thus, we divided RH into four ranges: RH < 60%, 60% ≤ RH < 80%, 80% ≤ RH < 90%, and RH ≥ 90%.

A Measure of Relative Importance in Multiple Regression
Multiple regression analysis has two distinct applications: Prediction and explanation [53]. When multiple regression is used for explanatory purpose, we are interested in the extent of each variable contributing to the response variable [54]. Standardized regression coefficients are the most common measure of relative importance when multiple regression is used [55,56].
We constructed a stepwise multiple linear regression model to quantify the effect of RH and PM 2.5 on visibility variability. The model fitted the daily visibility with daily RH and PM 2.5 . The daily values of visibility, PM 2.5 and RH satisfied approximately normal distribution. The fit has the form: Atmosphere 2020, 11, 461 where Y(t) is the standardized time series of daily visibility for all winters over 2013-2018 in Nanjing, and X k (t) is the corresponding time series for the standardized affecting factors (PM 2.5 and RH) k ∈ [1, n]. β k is the standardized regression coefficient. The contribution (C x k ) of a given predictor (x k ) to response variable (Y) is: Figure 2 showed the monthly variations of visibility in Nanjing averaged over 2013-2019, with the error bars indicating the standard deviations for the mean visibility based on the daily values. Similar to most urban areas in China, visibility in Nanjing varied seasonally from the high values in summer (July and August) to the low values in winter (December, January, and February) during the past seven years, which could be in association with a combination of air pollutant emissions and seasonal variance in meteorology [57]. Hence, we focused this study on the wintertime visibility variations and their relationship with wintertime PM 2.5 and RH. is the standardized regression coefficient. The contribution ( ) of a given predictor ( ) to response variable (Y) is: Figure 2 showed the monthly variations of visibility in Nanjing averaged over 2013-2019, with the error bars indicating the standard deviations for the mean visibility based on the daily values. Similar to most urban areas in China, visibility in Nanjing varied seasonally from the high values in summer (July and August) to the low values in winter (December, January, and February) during the past seven years, which could be in association with a combination of air pollutant emissions and seasonal variance in meteorology [57]. Hence, we focused this study on the wintertime visibility variations and their relationship with wintertime PM2.5 and RH.  58,59]. Differently, the interannual variations of wintertime RH presented a weak humid trend in the urban environment from 2014 to 2018, reflecting the change of natural factor RH in the East Asian monsoon region [60,61]. The slight increase in RH during these winter periods could intensify the hygroscopic growth of aerosol particles, thus increasing light extinctions and reducing visibility. Although the number of days with the level of "good visibility" had an obvious enhancement over the past winters, especially since 2016, the level of "poor visibility" dominated the wintertime ambient air with 42-63% days, with a weak interannual change over 2013-  58,59]. Differently, the interannual variations of wintertime RH presented a weak humid trend in the urban environment from 2014 to 2018, reflecting the change of natural factor RH in the East Asian monsoon region [60,61]. The slight increase in RH during these winter periods could intensify the hygroscopic growth of aerosol particles, thus increasing light extinctions and reducing visibility. Although the number of days with the level of "good visibility" had an obvious enhancement over the past winters, especially since 2016, the level of "poor visibility" dominated the wintertime ambient air with 42-63% days, with a weak interannual change over 2013-2018 in Nanjing ( Figure 3b). All of the mean visibility values averaged over each winter ranged between 5-10 km in the level of "low visibility" (Figure 3a). were given in Table 1. The visibility had significantly negative correlations with the PM2.5 concentrations and RH. The light extinction efficiency of the coarse particles was substantially lower than that of fine particles, and aerosols with diameters of 0.5-2 μm were most efficient for scattering visible light [49,62]. In terms of particles, we could expect that the visibility degradation was mainly caused by the increment of the PM2.5 concentrations during winter in Nanjing. However, the better correlation between visibility and RH was found than that between visibility and PM2.5 concentrations for all winters (Table 1), which could indicate that the natural factor RH might play the more important role in visibility change compared with the anthropogenic factor PM2.5 concentrations in this urban region. The impact of RH on the size and chemical composition of aerosol particles could alter aerosol optical property for changing visibility in the ambient atmosphere [32,35].  The correlation coefficients of daily visibility with PM 2.5 and RH during wintertime of 2013-2018 were given in Table 1. The visibility had significantly negative correlations with the PM 2.5 concentrations and RH. The light extinction efficiency of the coarse particles was substantially lower than that of fine particles, and aerosols with diameters of 0.5-2 µm were most efficient for scattering visible light [49,62]. In terms of particles, we could expect that the visibility degradation was mainly caused by the increment of the PM 2.5 concentrations during winter in Nanjing. However, the better correlation between visibility and RH was found than that between visibility and PM 2.5 concentrations for all winters (Table 1), which could indicate that the natural factor RH might play the more important role in visibility change compared with the anthropogenic factor PM 2.5 concentrations in this urban region. The impact of RH on the size and chemical composition of aerosol particles could alter aerosol optical property for changing visibility in the ambient atmosphere [32,35].  in changing visibility, further decreasing PM 2.5 concentrations could enhance the visibility sharply (Figure 4a). Similarly, an inflection point existed in the nonlinear relationship between atmospheric visibility and RH changes (Figure 4b). In humid air environment with high RH, the visibility dropped slowly with increasing RH, while RH changed visibility significantly in dry air with low RH.

Variations in Atmospheric Visibility
Atmosphere 2020, 11, 461 6 of 12 visibility dropped slowly with increasing RH, while RH changed visibility significantly in dry air with low RH. Based on the observation analysis, it was found that the nonlinear relationship between PM2.5 concentrations and visibility depended on different RH values in the ambient atmosphere. With the same PM2.5 concentrations, the higher the RH was, the lower the visibility was ( Figure 5). The increases of RH can promote the moisture absorption of the particles and strengthen the scattering and absorption capacities to the visible light, leading to the impairment of atmospheric visibility [32,35]. Power functions were used to fit the nonlinear relationships between visibility and PM2.5 concentrations in different RH ranges ( Table 2). The visibility changes in this urban area of East China were most sensitive to PM2.5 concentrations in the RH range of 60-80% with a power exponent of −0.77, and the visibility changes showed the least sensitivity to PM2.5 concentrations when RH ≥ 90% with a power exponent of −0.55 (Table 2). Under the high humid condition (RH ≥ 90%), the visibility values were mostly lower than 5 km for the "poor visibility," even with low PM2.5 concentrations ( Figure 5), which could have resulted from the formation of fog droplets in ambient air with water vapor saturation. Under the dry air condition (RH < 60%), a large part of the observed visibility was longer than 5 km in the good and low visibility levels with the enhancement of PM2.5 concentrations ( Figure 5), which might have been caused by the low aerosol water contents and weak water vapor uptake by fine aerosol matters [49]. Based on the observation analysis, it was found that the nonlinear relationship between PM 2.5 concentrations and visibility depended on different RH values in the ambient atmosphere. With the same PM 2.5 concentrations, the higher the RH was, the lower the visibility was ( Figure 5). The increases of RH can promote the moisture absorption of the particles and strengthen the scattering and absorption capacities to the visible light, leading to the impairment of atmospheric visibility [32,35]. Power functions were used to fit the nonlinear relationships between visibility and PM 2.5 concentrations in different RH ranges ( Table 2). The visibility changes in this urban area of East China were most sensitive to PM 2.5 concentrations in the RH range of 60-80% with a power exponent of −0.77, and the visibility changes showed the least sensitivity to PM 2.5 concentrations when RH ≥ 90% with a power exponent of −0.55 (Table 2). Under the high humid condition (RH ≥ 90%), the visibility values were mostly lower than 5 km for the "poor visibility," even with low PM 2.5 concentrations ( Figure 5), which could have resulted from the formation of fog droplets in ambient air with water vapor saturation. Under the dry air condition (RH < 60%), a large part of the observed visibility was longer than 5 km in the good and low visibility levels with the enhancement of PM 2.5 concentrations (Figure 5), which might have been caused by the low aerosol water contents and weak water vapor uptake by fine aerosol matters [49]. Figure 5 also showed the inflection points on the fitting curves of visibility and PM 2.5 changes. The inflection point of PM 2.5 concentrations was a turning point in the change of visibility induced by PM 2.5 concentrations. Atmospheric visibility showed low and high sensitivity with PM 2.5 concentration changes, respectively, above and below the inflection points ( Figure 5). Under four RH ranges of RH < 60%, 60% ≤ RH < 80%, 80% ≤ RH < 90%, and RH ≥ 90%, the inflection points of PM 2.5 concentrations in changing visibility reduced to 81.0 µg m −3 , 76.0 µg m −3 , 49.0 µg m −3 , and 33.0 µg m −3 ( Figure 5, Table 3), which indicated that improving atmospheric visibility with reducing PM 2.5 concentrations could be more difficult in more humid air environment. An important task of atmospheric environment control is reducing the PM 2.5 concentrations below the inflection points in order to achieve a significant improvement in visibility [36]. Atmosphere 2020, 11, 461 7 of 12   Figure 5 also showed the inflection points on the fitting curves of visibility and PM2.5 changes. The inflection point of PM2.5 concentrations was a turning point in the change of visibility induced by PM2.5 concentrations. Atmospheric visibility showed low and high sensitivity with PM2.5 concentration changes, respectively, above and below the inflection points ( Figure 5). Under four RH ranges of RH < 60%, 60% ≤ RH < 80%, 80% ≤ RH < 90%, and RH ≥ 90%, the inflection points of PM2.5 concentrations in changing visibility reduced to 81.0 μg m −3 , 76.0 μg m −3 , 49.0 μg m −3 , and 33.0 μg m −3 ( Figure 5, Table 3), which indicated that improving atmospheric visibility with reducing PM2.5 concentrations could be more difficult in more humid air environment. An important task of atmospheric environment control is reducing the PM2.5 concentrations below the inflection points in order to achieve a significant improvement in visibility [36].
According to the National Ambient Air Quality Standards of China released by the Ministry of Ecology and Environment of China [63] (http://www.mee.gov.cn/), light air pollution level of PM2.5 is categorized by the daily average PM2.5 concentration exceeding 75 μg m −3 in ambient air. There is a widely used haze definition with visibility of less than 10 km and RH of less than 90% [64,65]. Based on the fitting curves of nonlinear relationships between visibility and PM2.5 concentrations in four RH ranges of RH < 60%, 60% ≤ RH < 80%, 80% ≤RH < 90%, and RH ≥ 90%, Table 3 presented the critical values of PM2.5 concentrations on the fitting curves at visibility of 10km in low visibility level. For the visibility at 10 km, the critical PM2.5 concentrations varied significantly between 43.0 μg m −3 , 30.7 μg m −3 , 11.4 μg m −3 , and 3.8 μg m −3 in dry and humid air with four RH ranges of RH < 60%, 60% ≤ RH <  According to the National Ambient Air Quality Standards of China released by the Ministry of Ecology and Environment of China, light air pollution level of PM 2.5 is categorized by the daily average PM 2.5 concentration exceeding 75 µg m −3 in ambient air. There is a widely used haze definition with visibility of less than 10 km and RH of less than 90% [63,64]. Based on the fitting curves of nonlinear relationships between visibility and PM 2.5 concentrations in four RH ranges of RH < 60%, 60% ≤ RH < 80%, 80% ≤RH < 90%, and RH ≥ 90%, Table 3 presented the critical values of PM 2.5 concentrations on the fitting curves at visibility of 10km in low visibility level. For the visibility at 10 km, the critical Atmosphere 2020, 11, 461 8 of 12 PM 2.5 concentrations varied significantly between 43.0 µg m −3 , 30.7 µg m −3 , 11.4 µg m −3 , and 3.8 µg m −3 in dry and humid air with four RH ranges of RH < 60%, 60% ≤ RH < 80%, 80% ≤ RH < 90%, and RH ≥ 90% (Table 3), and all the critical PM 2.5 concentrations with much lower than the light PM 2.5 pollution level at 75 µg m −3 implied that the standard of clean air environment with PM 2.5 < 75 µg m −3 had apparent inconformity with the threshold of visibility at 10 km for haze pollution, reflecting the different influences of anthropogenic and natural factors on atmospheric environment changes. Table 3 also gave the critical values of visibility on the fitting curves at PM 2.5 concentrations of 75 µg m −3 in light air pollution level. The critical visibility values at PM 2.5 of 75 µg m −3 were 7.2 km, 5.0 km, 3.0 km, and 1.9 km, respectively, with different RH ranges ( Table 3). The increases of RH could significantly deteriorate atmospheric visibility, providing a larger contribution to the visibility change. The critical values of visibility were below the visibility threshold of 10 km for haze pollution under all RH conditions, indicating a discrepancy of haze pollution defined by visual ranges and PM 2.5 concentrations. PM 2.5 concentrations and RH, as the leading anthropogenic and natural factors changing atmospheric visibility, exerted an integrate influence on wintertime visibility impairment in this urban area of East China.

Relative Contribution of RH and PM 2.5 to Wintertime Visibility Variations
The impairment of wintertime visibility is generally linked with rapid increases in anthropogenic pollutant emissions in conjunction with unfavorable meteorological conditions for air pollutant dispersion [65,66]. This has led to a hot debate concerning whether the notable changes in visibility were mainly attributed to the meteorological conditions and anthropogenic emission control. Based on the discussion above, the anthropogenic factor PM 2.5 led to the visibility impairment by light extinction, and natural factor RH promoted aerosols' hygroscopic growth to a certain extent. We quantified their relative contribution of the leading anthropogenic and natural factors, PM 2. 5 The goodness of fitting was 0.58, passing the confidence level of 99%, which could mean the inclusion of PM 2.5 and RH accounted for the major contribution to wintertime visibility variations from 2013 to 2018 in Nanjing.
We estimated the relative importance of RH and PM 2.5 by the method of comparing the standardized regression coefficients mentioned in Section 2.2.2. The daily variations of wintertime visibility represented by natural factor RH was 54.3%, indicating a higher contribution as compared with the anthropogenic factor PM 2.5 contribution of 45.7%. That is, in spite of the effective PM 2.5 reduction in this urban area, natural factor RH was a dominant driver for the changes of wintertime visibility over recent years. The change of natural factor RH was closely connected with climate change in this East Asian monsoon region, implicating an importance of the regional climate change for variation of atmospheric visibility and environment.

Conclusions
Anthropogenic and natural factors affecting visibility change were investigated based on data of environmental and meteorological observations over 2013-2019 in Nanjing, an urban area of East China. We recognized PM 2.5 and RH changing atmospheric visibility and quantified the relative contribution of leading anthropogenic and natural factors to the variations of atmospheric visibility over recent winters in this urban area of East China. The application of these results will improve the representativeness and reliability of the use of visibility in the field of air pollution attribution and the determination of its climatic effects.
In spite of the effective PM 2.5 reductions after 2013, the level of "poor visibility" dominated the ambient atmosphere over winter in this urban area of East China. Although the visibility had significantly negative correlations with the PM 2.5 concentrations and RH, the inflection points existed in the nonlinear relationships between visibility and PM 2.5 concentrations, as well as between visibility and RH. Under the RH ranges of RH < 60%, 60% ≤ RH < 80%, 80% ≤ RH < 90%, and RH ≥ 90%, the inflection points of PM 2.5 concentrations were 81.0 µg m −3 , 76.0 µg m −3 , 49.0 µg m −3 , and 33.0 µg m −3 , respectively, reflecting that reducing the PM 2.5 concentrations to the inflection points for improving the visibility would be more difficult in more humid air environment in context of climate change. The visibility changes were most and least sensitive to PM 2.5 concentrations, respectively, in the RH range of 60-80% and RH≥90% in this urban area of East China. The relative contributions of natural factor RH and anthropogenic factor PM 2.5 to the changes of visibility were quantified with 54.3% and 45.7%, respectively, in this urban area, indicating an important role of natural factor RH in the changes of atmospheric visibility in the urban area of East Asian monsoon region.
This study revealed the influences of PM 2.5 and RH on changing wintertime visibility with the seven-year data of environmental and meteorological observations over an urban area in East China. It should be emphasized that the particle size distribution, optical properties, and the chemical composition of aerosols exert the impacts on atmospheric visibility. The influences of atmospheric particles and meteorological conditions to atmospheric visibility are more complicated, which could be further investigated based on the long-term observations on physical and chemical properties of particle and fine meteorology. The influences of long-range transport of PM 2.5 , weather patterns, and human activities on visibility change could be a further study with a more comprehensive numerical model of air quality and fine observations of meteorology and environment.

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