Long-Term Trends in Visibility and Its Relationship with Mortality , Air-Quality Index , and Meteorological Factors in Selected Areas of Korea

We investigated temporal trends in atmospheric visibility as well as their relationship to non-accident-related daily mortality, the Comprehensive Air-quality Index (CAI) in Korea, and selected meteorological factors in four cities (Seoul, Busan, Daegu, and Ulsan) and one remote island (Ullungdo) for the period 2001–2009. According to the mean values, Seoul exhibited the lowest visibility (12.2 km), followed by Ullungdo (12.6 km), Ulsan (14.2 km), Daegu (14.7 km), and Busan (17.3 km). Conversely, Ullungdo had the lowest CAI value (82.4) and no respiratory mortality (RM) occurrences at all, and Seoul had both the highest mean CAI value (122) and the highest occurrence of daily RM (5.3). There were negative correlations between visibility and two meteorological parameters (relative humidity and ambient temperature) in all five areas, whereas there were generally positive correlations between visibility and the other two meteorological parameters (precipitation and wind speed). For some lag periods, the mortality in the cities was significantly correlated with visibility and the CAI, whereas mortality in Ullungdo was not. Busan had the highest excess risk for non-accidentrelated daily mortality, followed by Daegu, Seoul, and Ulsan.


INTRODUCTION
Visibility is defined as the distance at which an observer can distinguish a black object viewed against a horizontal skyline.The reduction in atmospheric visibility in urban areas results mostly from interference with the transmission of light through the atmosphere by solid and liquid aerosols and gases.Visibility impairment in urban atmospheres is closely associated with air pollution from anthropogenic sources, including automobile exhaust, fuel combustion, solid waste incineration, and industrial emissions (Tsai et al., 2007;Deng et al., 2008;Pusheng et al., 2011;Fajardo et al., 2013;Zhuang et al., 2014).Given that air pollution results in visibility impairment, visibility has been frequently used as a qualitative measure of air quality in urban areas because visibility reduction is easily observed with the naked eye (Brumer et al., 2008).Specifically, atmospheric visibility can be 145-225 km in unpolluted areas (USEPA, 2001), whereas it is generally < 2 km in severely polluted areas (Deng et al., 2008).
Atmospheric pollution is also associated with a variety of adverse health effects, including respiratory and cardiovascular mortalities, impaired lung function, and heart disease (Anderson, 2009;Dadvand et al., 2011;Puza et al., 2011;Yu et al., 2012;Shang et al., 2013).Other studies (Huang et al., 2009;Thach et al., 2010) have attempted to link atmospheric visibility to adverse health effects.Specifically, Huang et al. (2009) reported that death rates from respiratory and cardiovascular diseases increased as atmospheric visibility decreased in Shanghai, China, where the decrease of an interquartile range of 8 km was correlated with a 3.02% increase in mortality.Similarly, Thach et al. (2010) reported that, in Hong Kong, an interquartile range of a 6.5 km decrease in visibility could result in an excess health risk of 1.92% for residents of all ages.Because a reduction in atmospheric visibility is often correlated with air pollution (Tsai et al., 2007;Deng et al., 2008;Pusheng et al., 2011;Fajardo et al., 2013;Zhuang et al., 2014), poor urban visibility is likely to be associated with adverse health effects for urban residents.However, apart from the two aforementioned studies, direct relationships between visibility and adverse health effects are rarely found in the literature.
As with other developing countries (Tsai et al., 2007;Chen et al., 2014;Zhuang et al., 2014), air pollution in many urban areas in Korea has become a critical challenge that must be overcome, because the concentrations of atmospheric pollutants often exceed international air-quality standards (Yoo et al., 2008;Son et al., 2010), suggesting poor visibility in many Korean cities. Visibility can also be influenced by meteorological factors including relative humidity, wind speed, ambient temperature, and precipitation (Chang et al., 2009;Geertsema and Schreur, 2009;Deng et al., 2011).In addition, previous studies (Tsai et al., 2007;Deng et al., 2008;Pusheng et al., 2011) have indicated that the long-term trend in visibility can be used to assess past, current, and future air pollution control measures.Accordingly, the current study was conducted to examine the long-term trend in atmospheric visibility, as well as its relationship with non-accident-related daily mortality, the Comprehensive Air-quality Index (CAI) in Korea, and meteorological factors in selected areas for the period of 2001-2009.Four major cities, each with a population of more than one million and a large number of point and nonpoint sources of anthropogenic air pollutants, were selected for this study, together with a rural remote island as a control area.The CAI, which is similar to the air-quality index established by the United States Environmental Protection Agency (USEPA, 1999) to indicate daily air quality, was determined using the daily or hourly mean concentration of five major air pollutants [particulate matter with an aerodynamic size of 10 µm or less (PM 10 ), sulfur dioxide (SO 2 ), nitrogen dioxide (NO 2 ), ozone (O 3 ), and carbon monoxide (CO)].Mortality was grouped into the categories of respiratory mortality (RM), pneumonia (PM), chronic obstructive pulmonary disease (COPD), and asthma (AM).

Survey Areas
The survey areas included four cities (Seoul, Busan, Daegu, and Ulsan) and one rural area (Ullungdo).The geographical locations of the visibility, weather, and air pollution monitoring sites in the survey areas are presented in Fig. 1.There are many residential and commercial buildings and a variety of municipal facilities and small-scale industrial factories in each of the cities.Ulsan is the location of the largest petrochemical industry complex in Korea.Seoul, the capital of Korea, has a population of ca.10.5 million, an area of 605 km 2 , and is located at 126°E and 37°N.Daegu has a population of ca.2.5 million, an area of 885 km 2 , and it is located ca. 250 km southeast of Seoul.Both Seoul and Daegu are surrounded by mountains and hills, these topographical features serve as barriers that interfere with the transport of air pollutants to other areas.Busan has a population of ca.3.6 million, an area of 767 km 2 , and is located on the southeast coast of Korea.Ulsan, which is a petrochemical industries-concentrated city, has a population of ca.1.2 million, and an area of 644 km 2 , and it is located between Daegu and Busan.Ullungdo, a remote island, has a population of ca.9,500, an area of 73 km 2 , and it is located ca. 350 km southeast of Seoul.As of December 31, 2009, the number of motor vehicles registered in Seoul, Busan, Daegu, Ulsan, and Ullungdo were 2.96 million, 1.02 million, 0.89 million, 0.42 million, and 0.003 million, respectively.

Meteorological Factors and Visibility Observations
The synoptic data for the meteorological factors and visibility were obtained from the Korean Meteorological Administration (KMA).The meteorological factors reported in this study were hourly average ambient temperature, relative humidity, wind speed, and precipitation.Eight daily visibility observations were made by well-trained technicians at 3-h intervals starting from 24:00.They measured the visual range using distinctive markers, such as tall buildings, towers, and mountain edges at established distances from the meteorological monitoring stations.These visibility observations were made on the basis of the quality control procedures adopted by the KMA.In the present study, visibility data were omitted when the relative humidity was higher than 90%, because the scattering cross sections of hygroscopic particles in humid environments are substantially higher than those for dry environments (Malm and Day, 2001).However, all relative humidity data, including those above 90%, were used to obtain summary statistics of relative humidity (Table 1).Daily-average visibility was used as a daily representative value, because our preliminary test suggested that it did not differ statistically from the visibility measured at 12:00 or 15:00 h.In addition, the correlation matrix of daily visibility and meteorological factors, and the multivariate regression results for daily visibility, meteorological factors, and the CAI obtained from the five survey areas from 2001 to 2009 were determined using the software SAS (Version 9.2).

Daily Mortality
Non-accident-related daily mortality data for residents in the five survey areas from 2001-2009 were extracted from a database held by the Korean National Statistics Agency.Causes and dates of deaths were coded based on the International Classification of Diseases (ICD-10).Excluding accidental deaths and injuries, the mortality investigated in this study was coded as follows: respiratory mortality (RM), J00-J98; pneumonia (PM), J12-J18; chronic obstructive pulmonary disease (COPD), J40-J44; and asthma (AM), J45-J46.Trends in the annual-average numbers of mortalities for the survey areas over the years 2001-2009 were determined.Multivariate regression analyses with different lag periods (0, 7, 14, 21, and 28 d) were conducted to evaluate the relationships of the mortalities to visibility and/or the CAI.The lag periods were arbitrarily chosen using periods of weeks.In addition, excess risks for daily mortality and the associated 95% confidence interval (CI) per interquartile range decrease in visibility under the different lag periods were estimated using a generalized additive Poisson regression model with penalized splines (Huang et al., 2009;Thach et al., 2010): where E(M i ) represents the expected number of mortalities

Comprehensive Air-Quality Index
To determine the CAI, air pollution data for five compounds (PM 10 , SO 2 , NO 2 , O 3 , and CO) were obtained from the Air Quality Management Bureau (AQMB) of the Korean Ministry of Environment.Twelve monitoring values per hour were nominally used to calculate a 1-h average concentration.Once the concentration data were quality assured, they were transferred to the AQMB.The air-quality data used in the present study, comprising 1-h averaged values, were extracted from the AQMB database.The air pollution data were only utilized for statistical analysis if 20 or more hourly concentrations were available during each 24-h period at any of the air-quality monitoring stations.
Daily CAIs were determined using the following equation: where CAI PM10 , CAI SO2 , CAI NO2 , CAI O3 , and CAI CO represent the partial indices of PM 10 , SO 2 , NO 2 , O 3 , and CO, respectively.The CAI value of individual air pollutants was calculated using the following equation: where CAI i represents the CAI of air pollutant i when C i,j ≤ C i ≤ C i,j + 1 ; C i denotes the daily average concentration of air pollutant i; C i,j denotes the threshold concentration of air pollutant i at air-quality grade j; C i,j + 1 denotes the threshold concentration of air pollutant i at air-quality grade j + 1; CAI i,j + 1 denotes the threshold partial index of air pollutant i at air-quality grade j; and CAI i,j denotes the threshold partial index of air pollutant i at air-quality grade j + 1.

Summary Statistics
Table 1 shows the summary statistics for visibility, the CAI, mortality counts, and the meteorological factors obtained in the four cities (Seoul, Busan, Daegu, and Ulsan) and a rural area (Ullungdo) during the period 2001-2009.Seoul had the lowest mean visibility (12.2 km), followed by Ullungdo (12.6 km), Ulsan (14.2 km), Daegu (14.7 km), and Busan (17.3 km).Similarly, Seoul had the highest mean CAI; however, Daegu had the second highest mean CAI value, followed by Busan, Ulsan, and Ullundo.Seoul also had the highest RM, followed by, Busan, Daegu, Ulsan, and Ullungdo.
The daily values of the meteorological factors varied depending on area (Table 1).Specifically, the mean temperature values in Seoul and Ullungdo were similar, and were both lower than those of the other cities.The relative humidity in the four cities was not significantly was similar, and was higher than the other cities.Likewise, the precipitation was similar in Busan and Ullungdo, where it was higher than in the other cities.Seoul had the lowest visibility, which was attributed to the fact that it had the most severe air pollution.Seoul had the highest mean CAI    122), indicating the worst air pollution levels of all the sites investigated.Seoul also had the highest mean occurrence of daily RM (5.3), pneumonia (1.5), COPD (1.6), and asthma (0.9).However, Busan had the second highest mean occurrence of RM and individual respiratory diseases, followed by Daegu and Ulsan, while there was no RM in Ullungdo.
Interestingly, Ullungdo, a remote island, had the second lowest visibility, with a mean of 12.6 km and a range of 0.5-30.0km, although the population and the number of motor vehicles on the island are much lower than in the cities, and there was no mortality recorded.These results are likely to be associated with a consequence of the frequent formation of fog above the ocean due to evaporation of sea water, rather than air pollution.It is well known that high levels of humidity have an impact on atmospheric visibility.Specifically, Malm and Day (2001) demonstrated strong correlations between a reduction in visibility and the relative humidity in several areas of the United States.As shown in Table 1, Ullungdo had the highest relative humidity with a mean value of 72.2% and a range of 14.5-99.3%,while this area had the lowest CAI values with a mean value of 82.4 and a range of 17.0-637.Fine particulate matter, such as PM 2.5 (particulate matter with an aerodynamic diameter of 2.5 µm or less), which is a major contributor to visibility impairment due to its tendency to scatter and absorb sunlight (Lee and Sequeira, 2002;Zhao et al., 2011;Chen et al., 2014), could also result in a decrease in visibility on the island of Ullungdo.Fine particles on the remote island may be formed by photochemical reactions of precursor pollutants such as hydrocarbons and nitric oxide transported from the mainland of Korea, under strong sunlight conditions above the ocean.

Trends in Annual-Average Visibility
Trends in annual average visibility in Seoul, Busan, Daegu, Ulsan, and Ullungdo over the period of 2001-2009 were examined (Fig. 2).The annual-average visibility decreased generally in all areas except Ulsan over the nine-year period.The decreasing trend in the four areas was attributed to annual variation in air pollution levels.A number of earlier studies conducted in other countries have also reported the presence of decreasing trends in annual visibility in urban areas, mostly due to air pollution (Tsai et al., 2007;Molnár et al., 2008;Chang et al., 2009;Zhao et al., 2011;Deng et al., 2012).In particular, Deng et al. (2012) reported that the long-term visibility in southeastern China declined at a rate of 3.1 km/10-y over the period of 1973-2010.The decreasing trend in annual average visibility in the three urban areas (Seoul, Busan, and Ulsan) was primarily attributed to the gradual increase in the number of motor vehicles, which are known to be a major emission source of air pollutants in most cities in Korea (Park and Lee, 2011).For example, the number of motor vehicles registered in Seoul, increased gradually from 2.6 to 3.0 million over the nine-year study period.However, in Ulsan, a city based on the petrochemicalindustry, there was no ascending or descending trend in annual average visibility, although the number of motor vehicles registered in the city increased gradually from 0.32 to 0.41 million over the nine-year study period.These results were attributed to both industrial and automotive emissions of air pollutants in Ulsan, with industrial emissions accounting for ca.60% of the total emissions of atmospheric pollutants in the city (Hieu and Lee, 2010).In addition, the decrease in visibility on the remote island location of Ullungdowas attributed to the long-range transportation of air pollutants from the mainland of Korea.

Relationship of Visibility to Meteorological Factors
The relationship of visibility to the meteorological factors measured in the survey areas was examined.Table 2 shows the correlation matrix of daily visibility and meteorological factors obtained from Seoul, Busan, Daegu, Ulsan, and Ullungdo from 2001 to 2009.Atmospheric visibility in the survey areas was either negatively or positively correlated with the four meteorological parameters (relative humidity, ambient temperature, wind speed, and precipitation).There were negative correlations between visibility and two meteorological parameters (relative humidity and ambient temperature) in all five areas.Although relative humidity itself does not significantly influence atmospheric visibility, a high humidity level can result in significant water absorption by hygroscopic particles such as sulfates, thereby increasing the scattering cross section of particles and reducing visibility (Deng et al., 2011).In addition, high ambient temperatures accelerate photochemical reactions to form secondary fine  1.00 a VIS, visibility (km); T, temperature (°K); RH, relative humidity (%); WS, wind speed (m/s); PC, precipitation (mm); *, p < 0.05.aerosols, which can scatter sun light effectively (Chang et al., 2009).In contrast, visibility was generally positively correlated with the other two meteorological parameters (precipitation and wind speed).Precipitation decreases the concentration of air pollutants through precipitation scavenging, thereby increasing atmospheric visibility (Geertsema and Schreur, 2009).In addition, there were positive correlations between visibility and wind speed in the four cities, whereas there was a negative correlation in the remote area, Ullungdo.High wind speeds in urban areas could increase atmospheric visibility, because the wind-derived atmospheric mixing results in lower particle concentrations (Deng et al., 2011).For Ullungdo, other parameters such as sea fog and fine particle formation might have confounded the effects of wind on atmospheric visibility.

Relationship of Mortality to Visibility and CAI
The relationship between mortality, visibility, and the CAI was examined using a multivariate analysis method.The correlation between visibility and the CAI was not significant (data not shown).The relationship of mortality to visibility and the CAI varied between the study areas and with different lag periods.Table 3 reveals the regression equations for mortality as a dependent variable and visibility and the CAI as independent variables for Seoul, Busan, Daegu, Ulsan, and Ullungdo from 2001 to 2009 according to the lag period.For some lag period conditions, the mortalities for the cities were significantly correlated with visibility and the CAI, whereas for Ullungdo they were not.In all four cities, there was a significant relationship between mortality, and visibility and the CAI with a lag period of 14 d, suggesting that this lag period is the optimal value when determining the relations of mortality with visibility and the CAI in urban areas.Similarly, other studies have reported significant correlations between mortality and atmospheric visibility in Hong Kong and Shanghai, China (Huang et al., 2009;Thach et al., 2010).
In the four cities, where a significant relationship was found between mortality, and visibility and the CAI, the relationship between total RM and specific diseases were further investigated by a Spearman correlation analysis.Table 4 displays a correlation matrix of the occurrence of RM, PM, COPD, and AM for a lag period of 14 d in Seoul, Busan, Daegu, and Ulsan, over the period of 2001-2009.RM was significantly correlated with all three individual diseases (PM, COPD, and AM), while the three diseases themselves did not show any significant correlations.In all the cities, COPD was the largest contributor to RM, followed by PM and AM (Table 1).Specifically, the daily average occurrences of COPD in Seoul, Busan, Daegu, and Ulsan were 1.6, 1.0, 0.6, and 0.3, respectively, while the respective values for RM occurrences were 5.3, 2.8, 1.8, and 0.7.
Excess risks that might result from visibility reduction in the four cities were also determined using a generalized  5 displays the estimated excess risks for non-accident-related daily mortality and an associated 95% confidence interval per interquartile range decrease in visibility for a lag period of 14 d in Seoul, Busan, Daegu, and Ulsan from 2001 to 2009.The inter-quartile ranges of Seoul, Busan, Daegu, and Ulsan were 7.4, 10.6, 6.6, and 7.8 km, respectively.Moreover, Busan had the highest excess risk for non-accident-related daily mortality, which was associated with the decrease in visibility, followed by Daegu, Seoul, and Ulsan.It is noteworthy that although the confidence levels include zero values, which indicate zero excess risk, the positive confidence levels can still be used for the conservative management of air quality.Consistently, several previous studies (Abbey et al., 1995;Vajanapoom et al., 2001;Thach et al., 2010) have reported that a decrease in urban visibility is strongly correlated with the increased mortality from respiratory illness and other diseases.

CONCLUSIONS
In this study, long-term trends in atmospheric visibility and the relationship with daily RM, the CAI, and meteorological factors in four cities and one remote island were examined.The cities were the most severely polluted and displayed both the lowest visibility and the highest occurrences of RM.Interestingly, Ullungdo, a remote island, had the second lowest visibility, although the population and the number of motor vehicles were much lower compared to the cities and there was no mortality recorded.The results in Ullungdo were attributed to the frequent formation of fog above the ocean due to evaporation of sea water rather than the effect of air pollution.There were negative correlations between visibility and two meteorological parameters (relative humidity and ambient temperature), in both the cities and the remote island location, while there were generally positive correlations between visibility and the other two meteorological parameters (precipitation and wind speed).For some lag period conditions, the mortalities in the cities were significantly correlated with visibility and the CAI, whereas in Ullungdo there was no correlation were not.

Fig. 1 .
Fig. 1.Geographical map of visibility, weather, and air pollution monitoring sites.

Table 2 .
Correlation matrix of daily visibility and meteorological factors obtained from Seoul, Busan, Daegu, Ulsan, and Ullungdo during 2001−2009 a .

Table 4 .
Correlation additive Poisson regression model.Table

Table 5 .
Estimated excess risks (%) for non-accident-related daily mortality and associated 95% confidence interval (%) per inter-quartile range (km) decrease in visibility for a lag period of 14 d in Seoul, Busan, Daegu, and Ulsan during 2001−2009.