Relationships among Particle Fractions of Urban and Non-urban Aerosols

Data on mass concentrations of PM10 and PM2.5, obtained at four urban sites in western Taiwan from 1998 to 2000, were analyzed to examine the pattern of seasonal and yearly variations in the PM2.5/PM10-2.5 ratio, the relationship among particle fractions and the variability of each particle fraction. The results were compared with those reported in the literature for urban and non-urban areas in several countries. Even though the annual mean of the PM2.5/PM10-2.5 ratio at a site might fall within a relatively narrow range over several years, the seasonal mean of the ratio could still vary considerably within a year. These results imply that there is no long-term characteristic value of the ratio for a community. Furthermore, different urban areas did not necessarily have similar ranges of the ratio. Results for the relationship among particle fractions and the variability of each particle fraction also indicated significant differences between communities. For the areas where both PM2.5 and PM10-2.5 are moderately correlated with PM10, separate measurements of PM2.5 and PM10-2.5 are needed for a better assessment of the underlying causes for the health effects of particulate matter.


Methods
The four monitoring stations were Kuting (KT) May), summer (June to August), and fall (September to November).Accordingly, one year referred to the 12 months from December of the preceding year to November of the current year.
The seasonal and annual mean concentrations were calculated from the daily average concentrations, and the seasonal and annual means of the PM 2.5 /PM 10 ratio were computed from the daily values of the ratio.
The coefficients of determination (R 2 ) for the relationships among particle fractions at each station were calculated for each season and each year.The coefficient of variation (CV, the standard deviation divided by the mean) was used to examine the variability of PM 2.5 , PM 10-2.5 , and PM 10 concentrations.To determine which particle fraction had a greater influence on the variation of PM 10 concentrations, the contribution of each of the two components, PM 2.5 and PM 10-2.5 , to the variability of PM 10 was calculated by dividing the standard deviation of each component by the mean of PM 10 concentrations.The results of this analysis indicate that rainfall and traffic volume did not significantly influence the seasonal variations of the PM 2.5 /PM 10-2.5 ratios at the four sites.Comparing weekday and weekend PM 2.5 / PM 10-2.5 ratios in the daytime of sunny days showed that the weekend ratios were, in general, only slightly higher than the weekday ratios at all four stations.It suggests that vehicular emissions contributed only slightly more than other sources during the weekend.Similarly, the means of the daytime values of the ratio were only slightly higher than those of the nighttime values.The analysis did not yield a clear trend in the effects of rainfall.Heavy rainfall wets road surfaces, thereby suppressing the resuspension of coarse particles.

PM 2.5 and PM 10 Concentrations
On the other hand, raindrops preferentially remove fine particles that consist of more soluble components such as sulfate and nitrate.
The pattern of seasonal variations of the PM 2.5 / PM 10-2.5 ratio differed to some extent across stations from December 1999 to November 2000 (Fig. 4).
Kuting station had a relatively narrow range of seasonal mean PM 2.5 /PM 10-2.5 ratios (0.93-1.06), indicating that the contribution from motor vehicles, the main source, to ambient aerosol remained relatively constant throughout the year.The range over which the seasonal mean ratio varied at the other three stations was somewhat wider but still varied within a factor of 2. In the absence of PM 2.5 data, it is tempting to estimate PM 2.5 concentrations from the PM 2.5 /PM 10 ratio (calculated from existing data), on the assumption that each community has a characteristic annual mean of the ratio.Such an assumption needs close examination.As shown above, the seasonal mean ratio can vary considerably over time at a site.The pattern of year-to-year change in the ratio also differs from site to site (Fig. 5).
Neither the seasonal variation of the PM 2.5 / PM 10-2.5 ratio at urban sites nor that at the non-urban sites exhibits a clear pattern.Data obtained in Birmingham, U.K., show that the mean of the ratio was 4. For these primarily non-urban sites, the ratio generally varied within a factor of 2 at each site.
The only major exception was Rocky Mountain, for which the ratio was 0.33 in winter and 1.04 in spring.Kim et al. (2000b)  observed at downtown Los Angeles on a day under stagnation conditions.The second type was a low ratio due to blowing dust, such as 0.28 observed at Rubidoux on a Santa Ana wind day.The ratio may exhibit small spatial variations in a metropolitan area such as Philadelphia, where PM 2.5 is the predominant contributor to particulate pollution.
The aerosol concentrations measured at eight sites located within metropolitan Philadelphia during the summers of 1992 and 1993 (Burton et al., 1996) showed that the mean values of the ratio at the eight sites were similar, ranging only from 2.45 to 3.35.
The relatively high values for the ratio in metropolitan Philadelphia are not necessarily typical of urban areas.For instance, downtown Los Angeles had a value of 1.46 (Kim et al., 2000a) and the ratios at the urban sites in Canada ranged from 0.56 to 1.44.The results shown in Table 1 also indicate that the ratios at non-urban sites do not fall within a narrow range.

3.3Relationship among Particle Fractions
The temporal variation of aerosol concentrations is an important parameter used in time-series epidemiological studies to test the relationship between particulate matter indicators and health outcomes.If an association between PM 10 concentrations and human mortality rates exists, PM 2.5 is a better indicator than PM 10-2.5 , if PM 2.5 and PM 10 are highly correlated but PM 10-2.5 and PM 10 are poorly correlated.On the contrary, PM 10-2.5 is a better indicator if PM 2.5 and PM 10-2.5 are highly correlated but PM 2.5 and PM 10 are poorly correlated.If both PM 2.5 and PM 10-2.5 are moderately correlated with PM 10 , then both can serve as indicators and separate measurements of PM 2.5 and PM 10-2.5 are needed for a better assessment of the underlying causes of health effects.and PM 10-2.5 (Wilson and Suh, 1997).The difference in relationships among particle fractions between Philadelphia and western Taiwan arises mainly because the Philadelphia aerosols had much higher PM 2.5 /PM 10-2.5 ratios (2.45-3.35)than the aerosols in western Taiwan (0.78-1.99).
The pattern of year-to-year variations in R 2 for the relationships among particle fractions differed across stations from 1998 to 2000 (Fig. 7).The value of R 2 for the relationship between PM 10 and PM 2.5 decreased considerably at the FS station over the period, while that between PM 10 and PM 10-2.5 did not change much.In contrast, the values of R 2 remained relatively constant at the KT station.
Figure 8 shows the seasonal variations of the coefficient of determination for the relationships among particle fractions at the FS station from December 1999 to November 2000.The coefficient of determination showed a moderate seasonal variation for the relationship between PM 10 and PM 2.5 , but a higher variation between PM 10 and PM 10-2.5 .The marked variation in the coefficient of determination for the relationship between PM 2.5 and PM 10-2.5 suggested that the driving forces for fluctuations in concentration of these two particle fractions changed independently of each other from season to season.3.4Variability of PM 10 , PM 2.5 , and PM 10-2.5 The underlying causes of the association between PM 10 concentrations and human mortality rates can also be examined by comparing the variability of various particle fractions.Figure 9 shows that, for all four stations from December 1999 to November 2000, PM 2.5 and PM 10-2.5 contributed almost equally to the coefficient of variation of PM 10 concentrations.However, further calculations indicated that the relative contributions of PM 2.5 and PM 10-2.5 to the variability of PM 10 varied considerably from season to season at each site, from year to year at each site, and from site to site for each year.

Conclusions
For both urban and non-urban sites, the seasonal mean of the PM 2.5 /PM 10-2.5 ratio could vary considerably within a year.Different urban areas have ratios with markedly different ranges.Care must therefore be taken when estimating PM 2.5 from the ratio by assuming that the ratio has a characteristic value for a community.
The relationships among particle fractions and the variability of each particle fraction could differ are highly or moderately correlated in some communities.For these areas, data on both PM 10 and PM 2.5 are still needed for epidemiological studies.If resources are limited, PM 2.5 can be monitored only in urban centers where PM 2.5 is the predominant contributor to PM 10 .Before biologically active chemical species that actually cause the adverse health effects are clearly identified, extensive monitoring of PM 2.5 is not well justified in areas with low PM 2.5 /PM 10-2.5 ratios.
Complexity in physical and chemical characteristics of ambient aerosols has led to considerable difficulty in identifying the determinants in particulate pollutants that are responsible for adverse health effects.As a consequence, air quality standards for particulate pollution have made use of indicators such as TSP, PM 10 , and PM 2.5 , instead of concentrations of specific chemical species.Over past decades, the PM indicator has progressively narrowed in particle size range, as the scientific understanding of the association between particulate pollutants and health effects has advanced.Although the recent *Corresponding author: Tel: +1-310-394-4089 E-mail address: cswang@ntu.edu.twestablishment of PM 2.5 standards in the U.S. was based on the results of definitive epidemiological studies, a lack of concentration data on this particle fraction poses a problem for other countries to consider adopting PM 2.5 as an indicator in air quality standards.Measurements of PM 2.5 concentrations are therefore being made in many countries both as part of scientific studies and as test runs of routine monitoring.Relationships between different particle fractions have also been studied, partly with the objective of examining the possibility of using PM 10 as a surrogate for PM 2.5 or PM 10 .Taiwan, as in many newly industrialized countries, has been tackling serious particulate pollution problems over the past three decades.The sources of particulate pollutants in urban areas include vehicular exhausts, fugitive dust from construction sites, resuspended road dust, products of gas-to-particle conversion by chemical reactions that involve gaseous pollutants, and emissions from smokestacks in nearby industrial parks.The relative contributions of these sources have changed over the years.Construction activities during the early stages of urbanization contributed a significant amount of coarse particles, while the contribution of vehicular emissions to fine particles has markedly increased in urban centers in the 1990s.Consequently, the concentration of fine particles has remained at relatively high levels, even though the PM 10 concentration has been leveling off owing to the implementation of various pollution control programs.The Taiwan Environmental Protection Administration began to set up air quality monitoring stations in 1982.As of 2001, the number of monitoring stations had increased to 72.PM 10 has been one of the criteria pollutants monitored at these stations.The monitoring of PM 2.5 began in 1997 at four selected stations in urban areas and one station near a major highway in western Taiwan.This study analyzes the data on mass concentrations of PM 2.5 and PM 10 obtained at the four urban stations from December 1997 to November 2000.The seasonal and year-to-year variations in the PM 2.5 /PM 10-2.5 ratio, the relationship among PM 2.5 , PM 10-2.5 , and PM 10 , and the variability of each particle fraction are calculated.The calculated results are compared with those for urban and non-urban areas in Australia, Canada, U.K., and U.S.A. to examine whether any general pattern exists in the relationship between particle fractions.

Figure 2 .
Figure 2. Variations in mass concentrations of PM 2.5 and PM 10 at the four stations from 1998 to 2000.The U.S. PM 2.5 standard (annual average) is 15 µg m -3 .

Figure 1 Figure 3 .Figure 3
Figure 1 shows the mass concentrations of PM 10 and PM 2.5 at the four stations from December 1999 to November 2000.The mean concentrations of both PM 10 and PM 2.5 differed considerably among various stations.For instance, the annual mean PM 10 concentrations ranged from 54 to 103 µg m -3 and the annual mean PM 2.5 concentrations ranged from 25 to 55 µg m -3 .Metropolitan Kaohsiung, located in southern Taiwan, is surrounded by several industrial parks and therefore has a serious particulate pollution problem, as indicated by the

c
photochemical reactions or fewer resuspended coarse particles from road surfaces.Other meteorological conditions, such as rain, solar radiation and atmospheric stability, may also affect the ratio.

Figure 4 .
Figure 4. Seasonal variations of PM 2.5 /PM 10-2.5 at the four stations from December 1999 to November 2000.

Figure 5 .
Figure 5. Variations in PM 2.5 /PM 10-2.5 for the four stations from 1998 to 2000.
88 from October 1994 to March 1995 and 0.72 from May to September 1995.In an analysis of particulate concentrations measured at 42 sites in the Interagency Monitoring of Protected Visual Environments (IMPROVE) network in Class I visibility areas throughout the United States during the 1993 seasonal year, Eldred et al. (1997) reported the seasonal mean of the ratio for 12 sites.

Figure 6 .
Figure 6.Coefficient of determination (R 2 ) for the relationships among particle fractions at the four stations from December 1999 to November 2000.

Figure 7 .
Figure 7. Variations in the coefficient of determination (R 2 ) for the relationships among particle fractions at KT and FS from 1998 to 2000.

Figure 6 Figure 8 .
Figure6shows the relationship among particle fractions as indicated by the coefficient of determination (R 2 ) for the four monitoring stations from December 1999 to November 2000.The correlation between PM 10 and PM 2.5 and the correlation between PM 10 and PM 10-2.5 were both relatively high (R 2 in the ranges of 0.56-0.79and 0.75-0.96,respectively).However, the correlation between PM 2.5 and PM 10-2.5 was moderate (R 2 in the range of 0.25-0.50).In contrast, the Philadelphia aerosols during the summers of 1992 and 1993 exhibited a high correlation (average R 2 =0.90) between PM 2.5 and PM 10 , a moderate correlation (average R 2 =0.35) between PM 10 and PM 10-2.5 , and a low correlation (average R 2 =0.11) between PM 2.5

Figure 9 .
Figure 9. Coefficient of variation (CV) for PM 10 concentrations and the contributions of PM 2.5 and PM 10-2.5 to PM 10 CV at the four stations from December 1999 to November 2000.
that increases in human morbidity and mortality rates are associated with the mass concentration of ambient particles in a certain size fraction, the properties of the particles that cause adverse health effects remain unknown.The mass concentration of particles in a certain size fraction is unquestionably a simple and convenient index for monitoring and regulatory purposes.The current rationale for using PM 2.5 and PM 10 as separate indices follows mainly from the observation that the chemical compositions of fine and coarse particles in many communities differ greatly.Exception may arise in regions where fine particles contain a substantial amount of crustal material.Additional questions arise if the mass concentration of particles in a certain size -quantify, but biochemically active compounds.Friedlander and Yeh (1998) provided evidence of the involvement of peroxides in particulate pollutants in causing adverse health effects.This study indicates that PM 10 and PM 10-2.5

Table 1
those reported for various types of sites and regions in the United States, Canada, and Australia.Interestingly, the ratios in western Taiwan, southern California, and Canada vary over similar ranges.The ratios for the primarily non-urban sites across the United States were in a much wider range

Table 1 .
Ranges of the annual mean of the PM 2.5 /PM 10-2.5 ratio for various regions.
a Four urban sites during the period from December 1997 to November 2000.bMeans of seasonal means for the 1993 seasonal year at 12 non-urban sites.