RELATIVE CONTRIBUTIONS OF DIFFERENT SOURCES OF URBAN AEROSOLS: APPLICATION OF A NEW ESTIMATION METHOD TO MULTIPLE SITES IN CHICAGO F. GATZ

DONALD Atmospheric

Sciences Section, Illinois State Water (First

Survey, P.O. Box 232, Urbana.

received 8 June 1973 and injinal,/brm

21 February

Illinois 61801, U.S.A.

1974)

Abstract-Information on the relative contributions of various types of pollution sources is required by pollution control strategists in planning for cleaner air. A new method for estimating the percentage contributions (source coefficients) of various sources to the total atmospheric aerosol content has recently been published. It requires detailed data on the chemical composition of (1) the urban aerosol and (2) emissions from the various sources considered. Enough data are now available for Chicago to permit a tentative estimation of source coefficients for the following aerosol sources: automobiles, fuel oil burning, cement manufacturing, iron and steel manufacturing, coal burning, and wind-raised soil dust. Comparison of estimated source coefficients for Chicago and published results for the Los Angeles area indicates that the method gives reasonable results, This paper also examines the. possibility of applying the method to multiple stations within a single city or region. Results show significant variations in the contributions af the various sources to aerosol concentrations within the city. They also suggest that the method may provide help in locating unknown sources.

INTRODUCTION

Ultimately, the price that we are willing to pay for environmental quality will be decided by the general public. This decision should be based on the best possible information about the detrimental effects of pollution and the costs of its abatement. Some degree of emission control is probably necessary to protect the health of people living in urban areas, as well as to protect the natural environment. However, controls impose direct costs on industry and indirect costs on consumers, and can have severe side effects on individual workers if businesses close or move. Because the consequences of imposing controls, or not imposing them, can be severe, decisions on these matters should be based on adequate information. Likewise, decisions about the specific sources that must install control devices should be based on accurate information. For the greatest improvement of air quality, emission controls should generally be applied first to the sources that contribute most to the particular problem. This means that the control official must have information on the relative contributions of possible sources. A broad range of specific problems can occur. The pollutant can vary from total suspended particulate matter to a specific toxic gas or metal. The space scale of the problem can range from a whole city to the neighborhood of a specific polluter. The time scale can range from long-term control to meet Federally established air quality criteria to shortterm episode control. 4.1 0 I *

1

Episode control requires air quality measurements in real time or near real time. Automatic equipment is available for continuously monitoring many gaseous pollutants. Such real time monitoring is not presently feasible for aerosols, although recent developments in sample analysis can reduce the total time needed for sampling. sample transport. and analysis to a few hours or less. The techniques and results discussed in this paper apply to aerosols only and thus apply primarily to long-term controls at the present time. As monitoring techniques improve. applications to episode control will increase. A technique for estimating the relative contributions of a number of different types ot solid particulate sources has been published recently by Miller LV~1. (1973) and applied to Pasadena, California. This method requires two kinds of information: (I) detailed chemical compositions of emissions from major sources: and (2) detailed information on the chemical composition of the urban aerosol. Both types of information arc just bcginning to become available. A limited amount.of both kinds of information is now available for C’hicago, and It appears timely to see whether some preliminary estimates of relative source contributions can be made for that city. Winchester and Nifong (1971) have provided many of the emission compositions needed, and a number of authors have published chemical compositions for the Chicago-Northwest Indiana aerosol. One of these papers (Brar ct trl.. 1970) contains concentrations of total particulate matter and about 20 individual elements measured over one 24-h period at 22 sampling stations in Chicago. The existence of these data makes it possible to apply the method of Miller ct (I/. (1972) at each station. Such a calculation should show the intra-tit) variation of the percentage contribution of the various sources to ( 1) the total aerosol burden and (2) the atmospheric burden of any of the individual elements. Because the measurements were made over on14 24 h. the results should not be taken as representing any more than that period. Nevcrthcless. we should be able to set if the /~rc~lhodis useful. The purposes of this paper. then, arc: (I) to investigate the possibility of extending the method of Miller t~‘tul. to a city with a wider variety of source types than Pasadena. and (2) to apply the method to multiple stations within the city, thus showing possible intracity variations in the relative contributions of the various sources to the total aerosol and to concentrations of specific chemical components.

The method is that given by Miller et al. ( 1972). and is described age Pi of any element i in the aerosol is given by: I-‘, =

5

x;,

/I,,

C‘,

as follows. The percent-

(1)

where pij is the percentage of element i in the particulate matter emitted from source .i. Cj is the fraction of the aerosol sample contributed by source ,i and xi,i is a coefficient of fractionation. For continuity. the sum of all fractional contributions must equal unity.

cc,= but in most cases it is not practical

1

to treat all possible

sources

Different

sources

of urban

3

aerosols

According to Miller et al. C(ijis the fraction of species i in source j that appears at the sampling site. Fractionation, a systematic change in elemental relative abundances, is known or suspected to occur, for example, during the formation of sea spray aerosols and in the wind-caused suspension of dust from soils. Once particles are airborne, fractionation can occur from the gravitational settling of larger particles if they contain a disproportionate fraction of any elements. These effects have, for the most part, not been expressed quantitatively, however and will be ignored in this paper. CHICAGO

AEROSOL

The major anthropogenic Nifong (1971) as:

SOURCES

AND

THEIR

COMPOSITIONS

aerosol sources for Chicago were listed by Winchester and

(1) coal burning; (2) coke production; (3) (4) (5) (6)

fuel oil burning; automotive fuel burning; iron and steel manufacturing; cement manufacturing. Table.

1. Compositions

Auto* Al As Br Ca Cd Cl co Cr cu Fe

of emissions

Cementt 2.5

pollution

sources,

per cent

Coal and cokef

Fuel oil:

Iron and steel8

14.0 0.016

5.0

2.4

5

4.0 0.004

0.4

5.4

0.8

0.009 0.03 0.04 7.0 0.0000~

0.15 0.12 0.16 2.5

Soill

7.9 44 6.8

0.4

2.7

Hg K La Mg Mn Na Ni Pb SC Ti V Zn

for some Chicago

1.6 38.7

II

0.10 1.2

40

0.14

0.8

0.3

0.024 0.4 0.04 0.13

0.03 1.5 6.0 0.18

0.9 0.08 0.09

0.03 2.5 0.05

0.002 0.005 0.003 3 0.00004**

1.6 2.4

1.8

i.004 0.7 0.03 0.6 0.005 0.005 0.00 15 0.3 0.007 0.01

* Miller et a/. (1972). t Emission assumed to have the same composition as the finished product (Winchester and Nifong, quoting Schueneman et al., 1963). $ Winchester and Nifong (1971). 5 From Winchester and Nifong (1971), Table 10, and their assumptions concerning relative emissions various types of processes, i.e. sinter plants, open hearth, etc. /I Shacklette et al. (1971). ![ Billings and Matson (1972). ** Shacklette, Boerngen and Turner (1971).

1971,

from

Most of these sources were identified as important contributors to the variance of metal concentration in 30 U.S. cities by BlifTord and Meeker ( 1967). In addition, at least one major natural source, namely soil dust, should be considcrcd. Table 1 gives the elemental composition of emissions of each major source used in this work. as found in the literature. The literature also contains measured concentrations of elements in Chicago air. To be useful in this work, the total concentration of suspended particulate matter must also have been measured. so the concentration of each element in air GIII lx converted to a per cent of the total aeros,ol. A number of literature sources of Chicago aerosol composition have beenassembled for comparison in Table 2. In gcncral. the studies that report mcasurcmcnts averaged over months, seasons. or years have data on onl) a limited number of elements and those reporting many elements are limited to sample.\ taken on one day or a few days. To provide a single aerosol composition for the city for use in the calculations. a “model” aerosol composition was compiled (Table 2). The long-term measurements were used for the model whenever they wcrc available for ;I given clement. but when onI>, short-term

Harrison and Winchester1

19w Al AS Br Ca C‘d

C‘I (‘0 Cr (‘11

_

Mg Mn Na Nl Pb SC TI \ Zn

(

I ‘)6-I

ompo\itc model

30 0.01 : II

0.0

I7

IO

. 4’Approximate values, computed from Harrison VI tri. ( IO7I I nvrthne,t Indiana data. assuming a mean uspended particulate loading of I SO log m ‘. ** Consistent with two samples collected and analv~ctl h\ Loucks ( 1969). havmg approximate values of 2.4 and 7.3 per cent for an assumed total particulate load+of I50 /(g m ‘.

Different

sources

of urban

aerosols

5

measurements were available, they were used. For a few elements, measurements were not available for Chicago and those from nearby northwest Indiana were used as estimates. TRACER

ELEMENTS

The first step in the calculation of source coefficients (Cj) was the selection of tracer elements for each source. A tracer element should make up a large fraction (2 10 per cent) of the total emissions from its source, but should not be present in large amounts in the .emissions from other sources. The tracer elements used for the sources considered in this work are listed in Table 3. The obvious choices for the auto, cement, and fuel oil tracers are Pb, Ca and V, respectively. Pb and Ca make up 40 and 44 per cent of the emissions from their respective sources. In instances where Pb values are not available, Br is an acceptable substitute, as will be shown later in this paper; V makes up only 2.5 per cent of its source emissions, but it has the advantage that fuel oil is its only major source. Likewise Mn is only 2.4 per cent of iron and steel emissions, but they are its primary source. Iron is also listed as a second tracer for iron and steel (hereafter referred to simply as “steel”) manufacturing sources. Despite its 39 per cent abundance, Fe is only a second choice because it is a constituent, sometimes a major one, of the emissions from all the other sources considered. A major difficulty was encountered in assigning tracer elements to the coal and coke (hereafter referred to simply as “coal”) and soil sources. Among the elements available for consideration (i.e. those that have been measured in air), abundance is the first criterion. However, Al is an abundant element in both sources and it is also present at the level of a few per cent in cement, fuel oil and steel sources. In coal emissions, Ca and Fe are the only other available elements present in excess of 1 per cent and both of these have other major sources. For soil, other elements such as K, La and SC were investigated as possible tracers, although only K makes up more than 1 per cent of the soil source. The investigation of K, La, SC and Al as possible tracers for soil was based on the Chicago model,aerosol composition and the assumption that soil was the only source of the element in aerosols. For each potential tracer clement, a maximum soil contribution was computed from

derived from equation (1). The soil contribution is an upper limit because if other sources contribute to Pi, that portion due to soil will be less than the total Pi of the aerosol, and

Table

3. Tracer

elements for emissions sources

from

Source

Tracer

Automobiles Cement Fuel oil Iron and steel Coal and coke Soil

Pb, Br Ca V Mn, Fe Al Al

various

a reduction in Pi would cause a reduction of Cltli,. The results of these calculations are shown in Table 4. The soil contribution cannot be any larger than the smallest Cz,j; calculated, namely 0.40 for Al. The larger values found indicate additional non-soil sources of those elements. Al is known to have additional sources also. including some strong ones. Nevertheless it was concluded that the best infor~tioll on the SOUKXcoefficient for soil (and coal) would come from the use of Al as a tracer for both, with appropriate consideration of upper limits and a reasonable apportioning of Al between the two sources, Becuase all of the tracer elements. even the best ones. have small contributions from secondary sources. it is necessary to use an iterative procedure. where coefficients of secondary sources may have to be gucsscd at, or ignored. on the first calculation. The coefficients computed on the first try are then used on the second. and so on. untit the source coefficients reach constant values. Since this is best explained by example, the Appendix shows the details of the calculation for Chicago, based on the observed data (Pi) in Table 2 and the emission compositions (I?,;) given in Table 1.

Source coeficients for Chicago were computed using the general procedure, outlined in the Appendix. To obtain information on the possible range of values that could occur. source coefficients were computed separately for each of the following assumptions: (1) each source is the onl~ ~ontrib~ltor of its trace element:

(2) all Al comes from coal: (3) all Al comes for soil: (4) Al not contributed bq. cement. fuel oil, or steel is attributed and soil.

equally to coal

Results are given in the next section

The existence of data on multi-element aerosol concentrations at 22 stations in the Chicago arca on 4 April 1968 (Brar er trl.. 1970) offered the possibility of applying this same technique to each station. The results should show what variations occur across the city in the relative contributions of the various sources to the total aerosol burden and to the burden of individual elements. Also. differences between observed and computed data might point out source areas for each element treated. The procedure illustrated above was repeated for each station in the Brar ct al. data, with the use of the emission compositions in Table 1 and the aerosol compositions

Differentsources

of urban

aerosols

reported by Brar et al. A few changes were necessary because of certain characteristics of the data. No data for Ca were reported. Thus, the cement source coefficient could not be calculated. No Pb data were reported, but this was not a serious matter, because Br is a valid substitute. Initial computation with Mn as the tracer for steel manufacturing emissions yielded values between 0.13 and 0.51 for Cstee,.When these values were used to compute expected percentages of other elements in the atmospheric aerosol, the computed Fe values far exceeded observed values. This suggested that the original values for Cstee,were excessive. Comparison of the Mn concentrations reported by Brar et al. (see mean value in Table 2) with CASN and NASN measurements in Chicago and with those of Harrison ef al. (1971) in northwest Indiana indicated that the Mn concentrations reported by Brar et al. were anomalously high. Thus Fe was used as the tracer for steel manufacturing emissions, after appropriate correction of PFe for contributions from all other Fe sources. Other information available

by this method

In addition to the possibility of computing the area1 distribution of source coefficients from the Miller et al. method, it is also possible to compute a number of parameters pertaining to the distribution of individual elements in the sampling network. One parameter of obvious interest is the concentration of the element in air (pg m-3), one of the basic pieces of data used in this technique. In the case of a toxic heavy metal, a map showing the distribution of concentration in a city would show the location of any areas where the concentration was above permissible limits. This information by itself does not suggest any possible control strategies, since information on the relative contributions of the various sources is still needed. Such information can be supplied in the following way. From equation (l), still ignoring aij, it is clear that the percentage Picj, of any element i in the aerosol, due to source j only, is: Pi(j)

Thus, the concentration

=

pij

Cj.

in air xicj, of element i, due to source j only, is: Xicj,

=

Pi(j,

S = pij Cj S

where S is the concentration (pg me3) of the atmospheric aerosol. The area1 distribution of Xicj,for the various sources would show, by comparison with the total concentration, the relative contribution of each source to the atmospheric concentrations of the element in question. Two additional derived parameters pertain to the problem of locating unknown sources of particular elements. By computing the sum1 j xicj)of the concentration of species i from the various known sources and subtracting it from the observed concentration, one is left with the concentration due to unknown sources. Plotting the area1 distribution of such concentrations may help to find the location of unknown sources. A search for unknown sources may be aided by a plot of the area1 distribution of the relative amount (percentage) of the concentration at each receptor that is due to unknown sources. Examples of the area1 distribution of these derived parameters will be presented in the next section.

s

DONALI)

F GA r/

The source coefficients for Chicago. computed for the four different assumptions concerning sourceetracer relationships described earlier. are given in Table 5. (Note that cocflicients are now expressed in per cent). Assumption 1. that each source has a unique tracer. results in maximum values, or upper limits. on the contribution from each source. Assumptions 2 and 3 assign Al exclusively to coal and soil. respectively. but allow for the esistcnce of secondary sources of the other tracers, and account for their contribution when computing source coefficients. Assumption 4 apportions AI not due to other Sources ecluall~ between coal and soil. and is considered the best estimate of actual conditions. Thus. the coefficients computed under Assumption 4 will be used in the comparison of the present results with those of others. A useful method of checking whether the computed source coctlicients arc reasonable is to compute the percentage of the total aerosol due to each element for which data are available. on the basis of known emission compositions and the computed source coetficients. The computed compositions for the source cocficients found under each of the four assumptions are given in Table 6. The Chicago composite model composition is also shown, to facilitate comparison of the observed and computed compositions. The use of upper-limit values ofa11the source coefficients (Assumption I ) results in computed contributions for the tracer elements that exceed observed values b> factors of up to about 2. The Chicago composite model aerosol is not considered so accur’ate that fktor-of-? discrepancies automatically signal large errors in the source coefficients. However. because the discrepancy occurred in more than one clement, and because the upper-limit source coefficients would account for more than 65 per cent of the total aerosol despite the fact that a number of large sources have not been considered. one suspects that the upper-limit coefficients are indeed somewhat escossive. The upper limit coelticients yielded even larger (10 x ) excesses of computed over observed compositions for Ni and Ti. b11t this ivas also true for the cocflicicnts computed under the other assumptiona. Possible rcasons for this arc discussed later. Assumptions 2~ 4 represent Lariations in the apportioning 01’Al between coal and soil. It was hoped that these comparisons of computed and obsericd compositions would bc of help in deciding which assumption was more nearI> corrccl. For most elements. the various assumptions make littlc difference in the computed compositions. The rcsult4 l’ot elements whose chief source is soil. such as K. La and SC, indicate. honovcr, that C’,,,,, cannot be near 0. and could possibly bc as high as 40 per cent. For the present. then. WC’shall

* Assumptions

explained in text

Different Table 6. Comparison Chicago composite model (observed) Ali AS Br Cat Cd Cl co Cr CU Fe Hg K La Mg Mnt Na Ni PbP SC Ti V-t Zn

2 0.02 0.2 2$ 0.01 3 0.004 0.006 0. I 3 0.004 O.S$ 0.002:: 0.9:: 0.1 0.4 0.025 I.1 0.00072 0.01 0.04 0.6

of observed

sources

of urban

and computed

aerosol

I 4.2 0.0022 0.22 3.1 0.~056 0.19 0.0045 0.008 1 0.076 4.0 0.000019 0.80 0.0016 0.51 0.12 0.32 0. IO 1.1 0.00060 0.25 0.054 0.097

9

aerosols compositions

Assumption 2

(per cent of total aerosol)

used in computation* 3

2.2 0.0022 0.22 2.0 0.0#56 0.19 0.003 I

0.0056 0.072 2.6 0.0000028 0.0012 0 0.21 0.10 0.074 0.078 1.1 0 0.13 0.041 0.089

* Assumptions defined in text. t Elements used as tracers for the various sources. : Approximate values, computed from Harrison rl al. (1971) northwest pended particulate loading of 150 pg m- ‘.

2.2 0 0.22 2.0 0 0.19 0.0029 0.0037 0.062 2.8 0.0000 16 o.so 0.0016 0.38 0.10 0.26 0.092 1.1 0.00060 0.12 0.040 0.075

Indiana

data.

4 2.0 0.0010 0.22 2.0 0.~026 0.19 0.0030 0.0045 0.068 2.6 0.0000085 0.36 0.~072 0.28 0.10 0.15 0.087 1.1 0.00027 0.11 0.041 0.082

assuming

mean

sus-

take the coefficients computed under Assumption 4 as the best estimates available for the strengths of the various sources.

Maps showing the distributions of source coefficients for automobile, fuel oil, steel, coal, and soil sources at 22 locations in the Chicago area are presented in Fig. 1. The map (Fig. la) showing the distribution of C.,,, indicates that the contribution of this source to the total aerosol ranges from less than I per cent (station Q) to more than 5 per cent (station H). Most of the stations show auto contributions in the l-2 per cent range. The area1 variation of the fuel oil source coefficient is shown in Fig. 1b, and indicates that Cr,,, oil is generally less than 1 per cent. Only four stations had values greater than 1 per cent. Two of these (stations F and V) had values over 4 per cent and suggest that a sizeable source may be located somewhere in the central part of the city. The variation of Cstecl(Fig. lc) has a range from about 3 per cent (station C) to about 8 per cent (stations B, I/ and G). and overall a rather uniform distribution over the city. This may be surprising in view of the known strong source in the southeastern sections of Chicago and in neighboring northwest Indiana, but the existing wind conditions during sample collection will have to be examined to see whether this source would have been upwind or downwind of the sampling network when the samples were collected.

IO

a

C

DO\AI

I>

F.