Tải bản đầy đủ - 0 (trang)
1 Multi-City, Multi-County and Site-Specific Studies

1 Multi-City, Multi-County and Site-Specific Studies

Tải bản đầy đủ - 0trang

6 Eastern and



2. Peng et al. 2009

119 counties

Multi-county studies

1. Bell et al. 2009

106 counties

2. ACS Study (Pope 151 U.S. localities

et al. 1995)

Multi-city studies

1. Six Cities Study

(Dockery et al.


PM2.5, sulfate

PM2.5, sulfate, acidity

Significant mortality associations found for

all-cause, cardiopulmonary mortality for

locality with highest PM2.5 vs. lowest

(RR = 1.26, 1.37); acidity not associated.

Sulfate and PM2.5 were very highly

correlated (0.98), sulfate associations

nearly identical to PM2.5 associations.

Significant mortality risks found for allcause, cardiopulmonary mortality with

both PM2.5 (RR = 1.17, 1.31) and sulfate

(RR = 1.15, 1.26)


Time series study (same

20 PM2.5 components (those in

Significant associations in cardiovascular

Peng et al. (2009) below, plus

day hospital admissions)

hospital admissions found for

13 elements which are minor

interquartile range increase in vanadium,

components of PM2.5, mostly

nickel, and elemental carbon components


of PM2.5, but not for other PM2.5


Time series study (daily

7 largest PM2.5 components

Significant associations in cardiovascular

emergency hospital

(sulfate, nitrate, silicon,

hospital admissions found only for


elemental carbon, organic

elemental carbon, not for other PM2.5

carbon, sodium and ammonium

components (0.80% increase in


hospitalizations per one IQR

(interquartile range increase) in EC

Prospective cohort study,

survival since


Prospective cohort study,

survival since


Table 1 Multi-city studies, multi-county studies, studies in Atlanta area: varying conclusions regarding which particle components are associated with allcause mortality, cardiopulmonary mortality, or emergency department visits for cardio-vascular disease


Geographic basis

Temporal basis

Particle components studied

Associations found

Distinguishing Health Effects Among Different PM2.5 Components


31 hospitals in

Atlanta area

41 hospitals in

Atlanta area

2. Tolbert et al.


Time series study (daily

emergency department


Time series study (daily

emergency department


206 rural and urban Prospective cohort study,


survival since


Atlanta area studies

1. Metzger et al.


4. Lipfert et al.


Table 1 (continued)


Geographic basis

Temporal basis

3. Lipfert et al.

206 rural and urban Prospective cohort study,



survival since



Emergency department admissions for

PM10, PM2.5, 10 to 100 nm

cardiovascular disease associated with

particle count, water soluble

EC, OC, and PM2.5 in single pollutant

metals, sulfate, acidity, OC, EC

models, but not with other PM2.5


Emergency department admissions for

PM2.5, sulfate, OC, EC, total

cardiovascular disease associated with

carbon, water soluble metals

EC, OC, total carbon, but not with other

(1998-2004 period of study)

PM2.5 components

Particle components studied

Associations found

PM2.5, sulfate, traffic density

Consistent significant mortality risk

(surrogate for traffic emissions)

associations with traffic density in 19

single and multi-pollutant models;

significant PM2.5 association in one of

two single pollutant models, but not in

any of four multi-pollutant tests; sulfate

not associated with elevated mortality


15 elements (mostly metals), EC, Significant mortality risk associations with

OC, nitrate, sulfate, PM2.5,

traffic density in 9 of 12 single and multitraffic density (surrogate for

pollutant models1; nitrate, EC, V, and Ni

traffic emissions)

associated with elevated mortality risk in

single pollutant, occasional multipollutant models; sulfate and PM2.5 not

associated with elevated mortality risks


T. J. Grahame

Temporal basis

Time series study (daily

emergency department


Particle components studied

In factor analysis, following

factors identified: gasoline,

diesel, wood smoke, soil,

secondary sulfate 1 and 2,

cement kiln, railroad, and

metal processing. Particles

include EC, OC, Se, nitrate,

sulfate, K and Zn

Associations found

In factor analyses, emergency department

admissions for cardiovascular disease

associated with sources of carbonaceous

emissions (gasoline, diesel, wood smoke).

PM components associated with CVD

hospital admissions were EC, OC, and K

(a tracer for wood smoke). Sulfate,

nitrate, Se, and Zn not associated.

In the three models of 12 in which traffic density was not significantly associated with elevated mortality risks, EC was also a variable. Like traffic density,

EC is a proxy for vehicular (mainly diesel) emissions


Table 1 (continued)


Geographic basis

3. Sarnat et al.

27 hospitals in


Atlanta area

Distinguishing Health Effects Among Different PM2.5 Components



T. J. Grahame

geographic scale; (3) find health associations with components of vehicular

emissions such as EC (and with a proxy for traffic emissions, traffic density),

which could not have occurred in the multi-city studies because such emissions

were unmonitored; and (4) do not find associations with sulfate. Studies which

include metals such as V and Ni are uncommon, but these emissions are often

significantly associated with health endpoints when included (as with two of the

multi-county studies, Bell et al. (2009), and Lipfert et al. (2006b)). In general,

most researchers would expect that a majority of PM2.5 components are likely to

cause health effects, although effects may differ in type and severity.

Studies in the Atlanta area also monitor for many pollutants, including components of traffic emissions such as EC, and report similar findings as the multicounty studies.

Because these are population-based epidemiology studies, covering broad

geographic areas and millions of people, exposure misclassification for pollutants

with spatial variability across a metropolitan area will occur to some degree in

each. For pollutants whose concentrations vary significantly over a locality, such

exposure misclassification is expected to reduce the size and significance of effect

estimates, as explained below. Therefore health risk associations found for locally

variable emissions such as BC/EC may be larger than reported in these studies.

Yet, because of exposure issues, results of these studies cannot be definitive

regarding causality, without additional evidence from toxicology and from studies

with better exposure information.

1.2 Progress in Identifying Health Relevant Sources

and Components of PM2.5

The Clean Air Scientific Advisory Committee (CASAC) is the scientific body

charged with advising EPA on proposed regulation of the six ‘‘criteria’’ air pollutants, including PM. In their June 13, 1996 letter to the EPA Administrator,

CASAC members reflected diverse opinions about whether to regulate PM2.5, and

if so, at what level to set the standard (CASAC 1996). Among the issues raised in

the letter were exposure misclassification (the first issue identified in the Abstract

above), the influence of confounding pollution variables (the 3rd issue in the

Abstract), and the lack of understanding of toxicological mechanisms (also discussed in the Abstract).

In 1998 and in later years, the NRC emphasized the need to identify which components of PM2.5 are most, or least, toxic (National Research Council 1998, 2004).

Despite these concerns, progress was slow with regard to differentiating health

effects among PM2.5 constituents until 2002, when several studies of different

aspects of highway pollution came forward.

Zhu et al. (2002a, b) demonstrated that concentrations of ultrafine emissions,

black carbon, and carbon monoxide (CO) were highly elevated adjacent to freeways in Los Angeles, but dropped by up to an order of magnitude 150 m away.

Distinguishing Health Effects Among Different PM2.5 Components


This study caused many in the air pollution health research community to focus

more clearly on near-roadway emissions.

Concomitantly, the small cohort study of Hoek et al. (2002) was the first of many

‘‘highway proximity’’ studies to show elevated health risks from close exposure to

vehicular emissions. Hoek et al. (2002) found highly elevated risks of cardiopulmonary mortality for those living within 100 m of a freeway or within 50 m of a

major urban arterial, versus those living further away, after taking personal and

sociological factors into account. Risks were significantly elevated for near-highway

BC concentrations, but not for background BC concentrations. Another cohort

study, the Canadian study of Finkelstein et al. (2004), found a ‘‘mortality rate

advancement period’’ of 2.5 years for those living in such close proximity to major

roads, a health risk comparable to having chronic ischemic heart disease.

The combination of the results of the Zhu et al. (2002a,b) studies, showing

elevated vehicular emissions in close proximity to major roads, with the results of

the early ‘‘highway proximity’’ studies, showing elevated mortality and morbidity

risks for those living within this zone of high pollution, sparked increased research

of how vehicular emissions might cause observed harm. Extant studies associating

various cardiac and cardiovascular outcomes with close proximity to major roads

are reviewed by Grahame and Schlesinger (2009).

Several studies by Li et al. (2002a, b, 2003) found oxidative stress to be caused

by diesel emissions in human bronchial epithelial lung cells in vitro, with the

smallest (ultrafine) particles, highly correlated with polycyclic aromatic hydrocarbon (PAH) content, causing the most damage. Due to their tiny size, these

ultrafine particles easily penetrate walls of lung epithelial cells. Once inside, the

chemical toxicity of the ultrafine PM causes oxidative stress. These findings were

later replicated in vivo by McDonald et al. (2004), who also found that removing

most carbonaceous matter in diesel exhaust (including 100% of BC) abrogated the

oxidative stress as well as inflammation. Carbonaceous matter was removed with a

catalyzing filter of the type required by EPA in 2007 for new on-road diesel


Salvi et al. (1999) found in healthy human volunteers an increase in neutrophils

in airway lavage after 1 h of exposure to diluted diesel exhaust. This same effect

was also found by Huang et al. (2003), in a study which examined effects of two

subsets of soluble components and took place near major roads, but which did not

measure for diesel emissions or organic PM2.5 species.

These and ensuing studies led many researchers to focus more on vehicular

emissions in analyzing health effects of air pollutants. Just as importantly, they came

to understand the importance of research methodologies which emphasized relatively

accurate subject exposure to pollutants—in large part because the highway proximity

studies showed that only with better exposure to vehicular emissions, were elevated

risks consistently found with such emissions. With this understanding, researchers

were motivated to develop new methodologies, such as those which enabled the

mapping of a pollutant ‘‘surface’’ over a geographical area, such as Los Angeles, and

then associate health outcomes with the more precise measure of a pollutant at a

subject’s home (Jerrett et al. 2005; Kuenzli et al. 2005; Maynard et al. 2007).


T. J. Grahame

Increased emphasis on exposure was not limited to vehicular emissions,

although vehicular emissions are usually the dominant source of local emissions

where the majority of people live, in large urban areas. NYU researchers found

associations between V and Ni in residual oil emissions and inflammatory effects

in vitro (Maciejczyk and Chen 2005), and between Ni from a distant smelter and

cardiovascular effects in vivo (Lippmann et al. 2006). Both studies utilized

strategies, including the relatively new step of backcasting wind trajectories, to

better characterize the sources of the metals which were associated with each

health effect.

Increasingly sophisticated research has taken place, particularly in the last five

years, suggesting risk of harm from specific components or sources of PM2.5. For

example, vehicular emissions—including ultrafine carbonaceous emissions, various organic gases, and PM2.5 black carbon—have been suggested as presenting

cardiovascular health risks in reviews (Samet 2007; Adar and Kaufman 2007;

Grahame and Schlesinger 2009). Emissions containing nickel, such as those of Ni

smelters and of residual oil, in studies with good exposure information, also appear

likely to have cardiovascular consequences (Lippmann et al. 2005, 2006; Bell et al.


Distinguishing characteristics of newer research include:

• More accurate exposure information in population, based and human panel

epidemiological studies.

• Examination of a wide variety of emissions, both particle phase and gaseous, in

the same study.

• Combining information from toxicological studies with information from epidemiological (including panel) studies with the above characteristics.

We will now turn to the first of these characteristics.

2 Importance of Accurate Exposure Information

in Epidemiological Studies

A fundamental epidemiological principle is that random exposure misclassification

biases results towards the null, i.e., results in an underestimation of an input

variable’s true relative risk (Hennekens and Buring 1987). In an air pollution

setting, more accurate measurement of subjects’ exposure to a harmful pollutant in

an epidemiology study produces effect estimates which are larger and more often

statistically significant. For example, Zeger et al. (2000) used both daily ambient

and personal exposure measurements in the PTEAM study, and found a smaller

coefficient for mortality risk in regressions using measured ambient levels, such as

from a central monitor, than when using personal exposure data. In Van Roosbroeck et al. (2008), effect estimates were two to three times higher for schoolchildren when personal exposure measures of soot and NO2 were used, than when

measurements taken outside the school were used. These examples show that the

Distinguishing Health Effects Among Different PM2.5 Components


better the estimate of a subject’s exposure to a pollutant, the larger the risk estimate of a damaging pollutant is likely to be.

For a single pollutant, these observations are straightforward. But what about

studies examining effects of two or more pollutants, which have different measurement errors? Goldberg and Burnett (2003) suggest that exposure misclassification of consequential input variables could actually lead to ‘‘transference’’ of

causal effects ‘‘from less precisely to more precisely measured variables.’’ If in

studies of two or more pollutants, associations are not only reduced for the poorly

measured pollutants, but also transferred from the poorly measured to the better

measured pollutants, then this exposure misclassification might explain why some

studies find associations with emission X but not emission Y, while other studies find

the opposite. Further, if such transference occurs, then the importance of relying

more heavily on studies with reasonably accurate exposure information is magnified.

In the world of published journal articles, can we demonstrate both that associations are stronger when subject exposure is well characterized, and that there

may be transference of associations from the poorly measured to the better measured emissions? Grahame (2009) addressed this question for heart rate variability

(HRV), an easily measured cardiac health endpoint. Perhaps because of the noninvasive nature of this measure, there appear to be more human panel studies of

this endpoint than of other cardiac and cardiovascular endpoints. HRV changes

appear not only to have medical relevance in their own right for those with a

history of heart disease, but also to cause and be a marker for oxidative stress, e.g.

in the heart, a chronic risk factor for cardiovascular disease (Rhoden et al. 2005;

Chahine et al. 2007).

Grahame (2009) found that in studies where subject exposure to BC was reasonably well characterized, it was almost always strongly and significantly associated with changes in HRV. In such studies, regional air masses containing large

amounts of sulfate but few urban emissions were not associated with changes in

HRV. However, BC was significantly associated with HRV changes only infrequently in studies where subject exposure across a city was characterized by

central monitor readings. In the study with the poorest subject exposure information, there appeared to be transference of associations from BC (with no

associations) to sulfate (large and significant associations). Grahame and Schlesinger (2009) made similar but preliminary findings in studies of ST-segment

depression and arrhythmias, e.g., the better the subject exposure to traffic emissions, the stronger the associations with either endpoint. However, insufficient

empirical work has been done on such transference of associations, when considering groups of studies; thus further investigation is a research priority.

3 Examining Wide Variety of Pollutants in the Same Study

Concentrations of various pollutants are often well correlated in epidemiological

studies. This can mean that taken individually, each pollutant might show


T. J. Grahame

a significant positive association with a given health endpoint. On a stagnant day,

most pollutant concentrations will be elevated, and thus many might show associations with a daily health endpoint.

For example, in Tolbert et al. (2007), four of the five pollutants examined for

the 1993–2004 time frame (CO, NO2, PM10, and ozone) were significantly associated with daily respiratory disease emergency room visits. In two-pollutant

models, ozone and PM10 remained significant in all tests, and in three-pollutant

models, only ozone retained significance. Tolbert et al. (2007) point out that there

are important issues with multi-pollutant models, some due to measurement error

for one or more of the pollutants. Tolbert et al. (2007) point out that when two

pollutants are thought to be independent risk factors for a given health endpoint,

and when they are correlated with each other, then it may be ‘‘appropriate to use a

two-pollutant model to adjust the effect estimate of each pollutant for confounding

by the other pollutant.’’ Tolbert et al. (2007) also note that ‘‘while it should be kept

in mind that there may be residual confounding as a result of model misspecification or measurement error, the risk estimates from this model are likely to be

more valid than those from each pollutant’s single-pollutant model.’’

In general, use of single pollutant models will be suspected of finding associations which may not be causal, unless it can be shown that the pollutant in

question is not correlated with other pollutants thought to be risk factors, and that

there is toxicological evidence suggesting biological mechanisms by which the

pollutant could cause the observed harm. Multi-pollutant models also have issues,

such as properly dealing with collinearity of various pollutants. Despite these

difficulties (and taking into account exposure misclassification and the possibility

of transferring associations ‘‘from less precisely to more precisely measured

variables’’), multi-pollutant models (especially two pollutant models) may allow

us to better identify the most harmful emissions (or the least harmful ones)

compared to single pollutant models. Again, however, these results by themselves

are insufficient to confirm causality of a given PM2.5 species, absent additional

evidence, in particular when exposure misclassification is present.

3.1 Factor Analyses

Factor analyses have been developed as a method of reducing large amounts of

correlated data to a smaller number of ‘‘factors.’’ In air pollution research, such

factors are often thought to represent sources of emissions. Yet in practice such

factors often represent a mix of sources, especially in earlier, less complete

analyses. Factor analysis is a multivariate receptor modeling approach, using

pollutant concentrations obtained from a monitoring site. For a description of a

popular method of factor analysis (Positive Matrix Factorization, or PMF), see

Paatero and Tapper (1993, 1994) and Paatero (1997). Unmix is another type of

factor analysis. These methods of identifying factors, and potentially sources,

possibly may be helpful in distinguishing health effects among sources of

Distinguishing Health Effects Among Different PM2.5 Components


pollutants. A similar method, not a type of factor analysis, is chemical mass

balance (CMB), which can be used to determine source contributions but requires

a priori knowledge of major sources and characteristics of their emissions.

Like the other epidemiology models in Table 1, exposure misclassification is a

given, because central monitor readings are used to develop the ‘‘factors.’’ That

said, it is useful to review the findings of different factor analyses and to see how

factor analyses have developed over time. The next two subsections briefly

illustrate some of the characteristics and issues with contemporary factor analyses.

3.2 Factor Analyses in Phoenix Area

The many factor analyses done in the Phoenix area can be distinguished from each

other by the differences in their results, which in turn depend upon which emissions are monitored as well as upon differing methodologies.

Mar et al. (2000) found five factors, which they identified with apparent sources:

vehicular emissions (enriched in Fe, Zn, OC, EC, NO2, and CO); soil dust (enriched

in Si, Al, and Fe); vegetative burning (highly enriched with K); a local SO2 source;

and sulfate enriched factor, thought to represent regional coal, fired power plants.

Ramadan et al. (2000) explored sources of PM2.5 in and nearby Phoenix during

the same time period, also using PMF. By utilizing more PM components,

Ramadan et al. (2000) found eight factors. Three large smelters were operating

approximately 100–170 km from Phoenix in the mid to late 1990s. In contrast to

Mar et al. (2000), which didn’t utilize copper PM data, Ramadan et al. (2000)

found a factor associated with ‘‘nonferrous smelting.’’

Using one data set (DFPSS), Ramadan et al. (2000) found the level of S in the

smelter factor to be about half the level of S in the ‘‘secondary sulfate or (coal fired

power plant)’’ factor (Fig. 1 in Ramadan et al.), suggesting that perhaps one-third of

the sulfate in the Mar et al. (2000) analysis could come from regional copper

smelters. Using a second data set (DICHOT), Fig. 8 in Ramadan et al. (2000)

suggests that the average amount of fine PM in the smelter factor and the coal power

plant factor may have been about equal. Thus the inclusion of an additional metal—

copper—appears to have added to understanding of regional sources of sulfate.

Lewis et al. (2003) used Unmix to analyze sources of PM2.5 in Phoenix and

surroundings. Lewis et al. (2003) did not identify a factor associated with smelter

emissions, but found a factor associated with diesel emissions substantially larger

than did Ramadan et al. (2000). Whether the divergence in findings between

Ramadan et al. (2000) and Lewis et al. (2003) is the result of different factor

analysis models (PMF and Unmix) or some other reason is not clear.

Mar et al. (2006) reported on results of a workshop in which factor analyses

from nine different practitioners were generated for the Phoenix area. Six groups

identified six factors (soil/crustal material; traffic [gasoline and/or diesel]; secondary sulfate; biomass/wood combustion; sea salt; and copper smelters). Two of

the factor analyses identified four factors, and one identified but two factors.


T. J. Grahame

A later study, Brown et al. (2007), also using PMF, differed from the preceding studies in several ways. Probably the two most important differences are

(1) use of emissions data from both sides of the US–Mexico border, and (2) an

incremental probability analysis linking factor strengths and receptors using

wind direction.

Brown et al. (2007) found a total of nine factors. A new factor is the ‘‘secondary

transport’’ factor, enriched moderately with sulfate but mostly with bromine (Br),

and accounting for 7% of PM2.5 mass. This factor is associated with sources west

of Phoenix with strength in southern California and northern Mexico, possibly

suggesting the use of methyl bromide in agriculture as well as miscellaneous urban

emissions. This factor is in addition to a regional power generation factor with

higher sulfate loadings.

The regional power factor in Brown et al. (2007) suggests that the two largest

sources of sulfate from power plants in the Phoenix area, based upon the combination of source emissions and the use of incremental probability analysis, are

(1) a large coal plant in Nevada near the California/Nevada/Arizona border, and

(2) a large residual oil, fired power plant in Puerto Libertad, Sonora, Mexico, on

the Gulf of California.

Thus, among the many factor analyses for Phoenix we now find four distinct

potential sources of sulfate: (1) coal fired power plants; (2) copper smelters; (3) a

‘‘secondary transport’’ factor from west of Phoenix, with sources on the California/

Mexico border; and (4) residual oil emissions from the Puerto Libertad plant in

Sonora. All of these factors will have different co-emissions co-mingled with the

sulfate (most of which is not directly emitted, but is the product of atmospheric

transformation of SO2 to sulfuric acid to neutral ammonium sulfates). This

diversity of findings indicates the need not merely to monitor for a widespread

emission such as sulfate, but also to employ sophisticated methods to distinguish

among sulfate sources, before examining health effects. These issues are discussed

at greater length in Grahame and Hidy (2007).

3.3 Factor Analyses in the Southeast US

This same sensitivity to nuance in the types and number of pollutants measured

can be seen in a pair of studies in the southeastern US. The first study (Liu et al.

2005) used 19 particle components (mostly metals) to identify factors at four

locations (2 urban/rural pairs). Eight factors were resolved at the urban sites, and

seven at the rural sites. A motor vehicle factor was identified which had high

concentrations of both EC and OC.

The second study (Liu et al. 2006) utilized several more pollutant components,

in addition to the components in the first study, to aid in determining factors.

Concentrations of 4 gas components (CO, SO2, HNO3, and NOy) were used, and

both EC and OC were split into temperature-resolved fractions (4 OC fractions,

3 of EC).

Distinguishing Health Effects Among Different PM2.5 Components


Major differences in factors between the studies occurred with the addition of

added pollutant information in the second study. First, instead of a single motor

vehicle factor, separate diesel and gasoline emission factors were identified.

Secondly, several factors changed in their percentage contribution to PM2.5 mass

between the two studies. At a downtown Atlanta site, there was a decrease of 8.6%

in secondary sulfate contribution (from 37.0 to 28.4%); an increase in wood

smoke contribution from 13.0 to 22.2% (prescribed agricultural burning in

spring and residential wood burning in winter are the main contributors); and a

decrease in the non-sulfate coal combustion factor from 3.0 to 1.9%. In a rural

area (Yorkville), the secondary sulfate contribution changed little, but the secondary nitrate factor increased from 7.0 to 15.3%; wood smoke dropped from

20.0 to 16.6% contribution; and non-secondary sulfate coal combustion dropped

from 6.0 to 0.7%.

Apparently the addition of SO2 to the analysis enabled refinement of the coal

combustion factor. The newer non-secondary sulfate coal plant emission numbers

comport better with known percentages of coal fly ash in ambient air. Coal fly ash

(CFA), mainly composed of aluminosilicates with calcium or sodium, iron, and

trace materials, is the major primary PM2.5 coal plant stack emission (CFA

atmospheric concentrations typically are in the tens of ng/m3, up to *200 ng/m3,

depending on the number of coal plants in the region). The use of temperature,

differentiated EC and OC fractions allowed differentiation of diesel and gasoline

emissions, and also caused the contributions of other factors to change (e.g., wood


Sarnat et al. (2008), one of the Atlanta area studies and the only factor analysis

in Table 1, utilizes a factor analysis most similar to that of Liu et al. (2006), one of

the two most sophisticated of the factor analyses reviewed herein (Brown et al.

2007 being the other). Sarnat et al. (2008) compares the results of this factor

analysis (e.g., health effect attributions) with results of other analytical methodologies, including CMB as well as using representative PM components as tracers

of sources, and finds consistent results. There may be several reasons why Sarnat

et al. (2008) were able to find consistent results among different methodologies,

where others (e.g., Mar et al. 2000) did not. First, the factor analysis utilized was

more advanced in the number and choice of input components. Secondly, there

weren’t significant unforeseen emission sources (e.g., large cross-border sources in

the Phoenix area found only by Brown et al. (2007), in addition to copper smelter

emissions identified by some analysts). Third, co-mingling of emissions from

disparate sources was less likely because steel mills, coke ovens, smelters, and

residual oil combustion sources are absent in the Atlanta area.

4 Different Sources of Common Pollutants

While many epidemiology studies today find associations with pollutants such as

OC, EC or BC (which are virtually the same), or sulfate, it is important to

Tài liệu bạn tìm kiếm đã sẵn sàng tải về

1 Multi-City, Multi-County and Site-Specific Studies

Tải bản đầy đủ ngay(0 tr)