Sources of polycyclic aromatic hydrocarbons to Lake Michigan

Multimedia Model for Polycyclic Aromatic Hydrocarbons (PAHs) and Nitro-PAHs in Lake Michigan. Lei Huang and Stuart A. Batterman. Environmental Science...
0 downloads 0 Views 1MB Size
Environ. Sci. Technol. 1993, 27, 139-146 Dries, Do;De Corte, B.; Liessens, J.; Steurbaut, W.; Dejonckheere, w.; Verstraete, w. Biotechnol. Lett. 1987, 9, 811-816. Jones, T. W.; Winchell, L. J . Environ. Qual. 1984, 13, 243-247. Valentine, J. P.; Bingham, S. W. Can. J . Bot. 1976, 54, 2100-2107. Butler, G. L.; Deason, T. R.; O'Kelley, J. C. Bull. Environ. Contam. Toxicol. 1975,13, 149-152. Pape, B. E.; Zabik, M. J. J . Agri. Food Chem. 1970, 18, 202-207. Pelizzetti, E.; Tosato, M. L. Environ. Sci. Technol. 1990, 24, 1559-1565. Frimmel, F. H.; Hessler, D. P. Photochemical Degradation of Triazine and Anilide Pesticides in Natural Waters. Abstracts of Papers, 203rd National Meeting of the Am-

erican Chemical Society, San Francisco, CA, April 5-10, 1992; American Chemical Society: Washington, DC, 1992; AGRO 87. (36) Armstrong, D.E.; Chesters, G. Environ. Sci. Technol. 1968, 2, 683-689.

Received for review June 4, 1992. Revised manuscript received September 24,1992. Accepted September 28,1992. This study was supported, in part, by the Iowa Department of Natural Resources Geological Survey Bureau, through the Big Spring Basin Demonstration Project, with funds provided from the Iowa Groundwater Protection Act. Roger Koster, the Big Spring Project Coordinator for the Iowa State University Cooperative Extension Service, provided assistance in obtaining landowner cooperation in the study.

Sources of Polycyclic Aromatic Hydrocarbons to Lake Michigan Determined from Sedimentary Records Erik R. Chrlstensen" and Xiaochun Zhangt

Department of Civil Engineering and Mechanics and Center for Great Lakes Studies, University of Wisconsin-Milwaukee, Milwaukee, Wisconsin 53201 Four sediment cores from Lake Michigan were 210Pband 13'Cs dated and analyzed for 12 polycyclic aromatic hydrocarbons (PAHs). Sources of PAHs were determined based on multiple linear regression of US. energy consumption figures for coal, petroleum, and wood. Profiles of PAH fluxes in sediments with mixing were reconstructed with the inverse Berger-Heath model. PAHs generated from wood burning amount to 4.6-9.3 ng/cm2 per 1015Btu independent of location. For coal, the significance as an input source decreases slowly with increased distance from shore, and quickly for petroleum. Maxima in the PAH records were observed corresponding to 1985, reflecting petroleum-derived PAHs as shown by the BghiPIIP ratio of 2.18 f 0.37 and the similarity of the historical record to petroleum consumption data, and to the mid-l950s, which may be caused by a maximum in coke production in addition to a switch in home heating from coal to oil and gas. Introduction Twelve polycyclic aromatic hydrocarbons (PAHs) of primary concern in terms of environmental significance (1)were chosen for this sediment investigation. Because some PAH compounds such as benzo[a]pyrene are proven carcinogens, it is important to examine the levels of these compounds in sediments and to identify their sources. The most important PAH formation processes are burning of coal, coke production in the iron and steel industry, and open burning of refuse and other materials (2). PAH pollution from vehicles is important on a local basis (3). According to Bjarseth and Ramdahl(4), -36% of the input of PAHs to the atmosphere in the United States comes from motor vehicle combustion. The residence time of PAHs in the atmosphere is mostly governed by particle size and atmospheric conditions. For small particles of l-pm diameter it varies between a few days and 4-6 weeks, but is only a few days or less than 1

-

+Presentaddress: Wisconsin Department of Natural Resources, WR-2, P.O.Box 7921, Madison, WI 53707. 0013-936X/93/0927-0139$04.00/0

day for particles with l-10-pm diameter (5). Most studies show that atmospheric input is the primary source of PAHs to sediments and soils, especially for remote sites (6-14). There have been many qualitative studies of PAHs in sediments including those of Lake Michigan (15,16).Heit et al. (17)examined correlations between PAH fluxes into the sediments of Cayuga Lake, NY, and fossil fuel usage for the states of New York, Ohio, and Pennsylvania and the United States as a whole. However, no detailed quantitative analyses were conducted to estimate the proportion of these sources. Multiple linear regression was applied by Smith and Levy (18) to estimate the PAH input to the sediment near aluminum smelter plants in Saguenay Fjord. They only distinguished between two pathways of PAHs to the sediments. Sexton et al. (19) used stepwise analysis to determine the sources of PAHs in the air of Waterbury, VT. This study is an attempt to quantitatively resolve the significance of the proposed energy consumption inputs to Lake Michigan, based on accurately dated sediment records. Further sedimentological and geochemical data for the cores considered here and for other cores will be published elsewhere (20). Materials and Methods PAH Compounds and Dating Methodology. Details of sampling and PAH and radionuclide analysis are given elsewhere (20). Sampling locations are shown in Figure 1,and core coordinates and other parameters are listed in Table I. The following 12 PAH compounds were determined: phenanthrene (PhA), anthracene (An), fluoranthene (FlA), pyrene (Py), chrysene (Chy), benz[a]anthracene (BaA), benzo[b]fluoranthene (BbFlA), benzo[klfluoranthene (BkFlA), benzo[a]pyrene (BaP), ideno[1,2,3-cd]pyrene(IF)dibenz[a,h]anthracene , (dBahA), and benzo[ghi]perylene (BghiP). Mass sedimentation rates and zloPbfluxes were calculated as in ref 21. The date based on sedimentation rate was calculated by dividing the average cumulative mass for an interval with the mass sedimentation rate. Focusing

0 1092 American Chemical Society

Environ. Sci. Technol., Vol. 27, No. 1, 1993

139

Table 1. Sampling Site Coordinates. Water Depths, Core Lengths,and Sediment Core Parameters (22) sediment core CLM88E CLM88F SLM88B NLM88D

coordinates latitude longitude north west 43'28'50'' 43'21'21'' 42'38'49'' 44'40'46''

water depth (m)

87°29'17" 87O35'08" W56'52" 86O44'07"

gravity core length (em)

mass sedimentation rate (g cm-l yi')

72 93

0.0136 f 0.0013 0.0320 0.0021

150

115 162 270

100

93

0.01C6

steady-state 210Pbflux focusing factor (dpm cur2 F-') 210Pb '"Cs 0.8950 1.1554 0.4752 0.988

O.MX)8

0.0144 f O.WO6

factors were determined from a correlation between 210Pb flux [i.e., decay constant (yr-') times 210Pb inventory (dpm/cm2)] and focusing factor (22),and from '"Cs data as in 21. Multiple Linear Regression. The records of total PAHs were reconstructed for the two cores affected by mixing in the top layers, CLM88E and NLM88D (23). The depths and times of deposition of the mixing layers were estimated as 0.70 cm and 8.35 yr (CLM88E) and 0.60 em and 5.31 yr (NLM88D); see Zhang (22). The principle of linear addition of source functions can be written in matrix form

89'

68'

870

0.926 1.196 0.492 1.022 86'

1.57 2.91 2.46 0.977

85'

46'

450

440

c : 1 : } +{ : }

43'

02

(1)

or

42'

880

870

86'

850

Flgure 1. Lake Michigan sampling snes.

where [F(t)]are the measured PAH fluxes, [F'(t)]are the PAH fluxes calculated from n source functions, the [&(ti)] matrix contains the time-dependent source functions, and [e] is an error vector. The objective is to fmd the constants [a] (al,a2,...,a,,) such that Fit,), F l t J , ...,F'(tJ is the best possible approximation to the measured vector [ F ( t ) ]i.e., , to minimize S m

S = EWi(F(t,)- F'(ti))' i=1

(3)

where mi is an appropriate weighting factor. In order to solve for the a's, reflecting the importance of the sources and changes of PAHs during transport, the rn X n matrix (rn 2 n) is converted to a n n X n matrix (22). The fractional contribution pij,of source number i, 15 i 5 n, to the PAH flux at the time tj, 1 5 j 5 rn, is then

p.. LJ =

ai$&,)

al+'(tj) + Q

Z M j )

+ ...+ Q"+"(tj)

(4)

Results and Discussion Dating of Sediment Cores. The results of the 210Pb and 'Ws counting are shown in Table II for core CLM88F. Mass sedimentationrates, 21Tbfluxes,and focusingfactors for all four cores are listed in Table I. Measured total PAH 140 Envlrm. Sol. Technol.. Vol. 27, No. 1. 1993

concentrations for CLM88F are also included in Table 11. Note that 137Csan$ PAH data are from a box core and the " T b data are from a gravity core. The relatively small diameter gravity cores can occasionally undergo some compaction during sampling. However, the dating of sediment layers will not be affected by this since we are expressing sedimentation rates on a mass basis (g cm-2 yr-9.

The dates determined from the sedimentation rates are in reasonable agreement with the available 'Ws data (a), except for CLM88F and SLM88B. In the case of CLM88F. an adjustment of the rate from 0.0320 to 0.046 g cm-2yr-' places the ' T s maximum correctly around 196-1966,and a PAH maximum at 1951 in accordance with the time of the maximum of NLM88D and results from other studies (17, 25). For SLM88B. the 13'Cs data exhibit an irregular, significant tail which indicates downward migration of this radionuclide. This is also reflected in the significant discrepancy between focusing factors based on 210Pband I3Ts (Table I). However, the 1963 W s peak is fairly well confirmed. The mass sedimentation rates, " T b fluxes, and focusing factors (Table I) are within the range of values found for a different set of cores collected from Lake Michigan in 1984 (21). Focusing factors based on 'ITb were considered the more accurate ones and are therefore used below.

Table 11. *l0Pb and lS7CsActivities, Total P A H Concentrations, and Dates Determined from the Sedimentation Rates for Core

CLM88F

a

depth (cm)

cumulative mass (g/cm2)

net zloPb acti? (dpm/g)

0.0-0.5 0.5-1.0 1.0-1.5 1.5-2.0 2.0-3.0 3.0-4.0 4.0-5.0 5.0-6.0 6.0-7.0 7.0-8.0 8.0-9.0 9.0-10.0 10.0-12.0 12.0-14.0 14.0-16.0

0.060 0.120 0.207 0.293 0.486 0.698 0.915 1.142 1.369 1.602 1.833 2.063 2.547 3.121 3.634

31.7 f 0.7b 31.7 f 0.7b 36.1 f 0.5b 36.1 f 0.5* 24.3 f 0.5 21.9 f 0.4 16.1 f 0.4 15.8 f 0.4 11.5 & 0.4 7.51 f 0.40 7.33 f 0.29 6.82 f 0.30 4.10 f 0.23 1.18 f 0.62 0

137cs activ (dpm/g)

PAH concn'

0.0320

0.046

(ng/g)

g cm-2 yr-l

g cm-2 yr-'

30.2 f 0.3

4190 6099 4459 5041 5514 4003 3716 4347 3838 3979 4496 2625 1438 2535 538

1987 1985 1983 1980 1976 1970 1963 1956 1949 1942 1934 1927 1916 1899 1882

1987 1985 1984 1983 1980 1975 1970 1966 1961 1956 1951 1946 1938 1926 1915

*

30.7 0.3 29.7 f 0.3 28.0 f 0.2 16.7 f 0.2 16.1 f 0.2 17.4 f 0.3 7.28 f 0.16 13.1 f 0.2 12.9 k 0.3 2.88 f 0.07 1.41 f 0.05 0.94 f 0.04

sedimentation rate date

Supported zloPbactivity, 3.15 dpm/g. bAverage. 'Coefficient of variation, 17%. 40 I

I

i

-

Coal

Coal (Illinois) Coal used for coke production ( U S )

...... Petroleum Wood

30 -

20

3

-

I

10

/

t

0

1900

1950

2000

isso

1900

1950

2000

Year

Year

Flgure 2. Energy consumptlon In the United States.

Source Functions. Figure 2 shows the energy consumption in the United States from 1850 to 1988when the sediment samples were collected. The energy consumption data for coal and petroleum in the time period 1960-1988 are from the Energy and Information Administration (EIA) (26). The amount of wood consumed was obtained from EIA (27) for the time period 1949-1983, from EIA by personal communication for 1988, and by linear interpolation for the time period 1983-1987. All data before 1960 for coal, petroleum, and wood are from Hottel and Howard (28). The coal consumption data for Illinois (26,29),shown in Figure 3, follow the national trend fairly well, except that the Illinois numbers are 13 times lower and show a flattening after the 1950s. The contribution of coal used for coke production to the generation of PAHs may be important (5,16). Coke is used

-

Flgure 3. Consumptlon of coal in Illlnols and of coal used for coke production in the Unlted States.

not only in the steel industry but also for domestic heating purposes. Coal used for coke production contributes to the PAH formation first during the generation of coke and second during the combustion of coke. Coal used for coke production was calculated using a conversion factor of 0.0268 X 1016Btu per million short tons of coke produced (28). The amount of coal used for coke production is from the Minerals Yearbook (30) and EIA (31). The coke production before 1880 was not available. The resulting source function for coal used for coke production in the United States has a maximum value of -3 X 1015Btu during the mid-1950s (Figure 3). Measured and Calculated PAH Fluxes. Figure 4 shows a comparison of the measured and calculated total PAH fluxes to CLM88E. With three proposed sources (Figure 4a), there is a large discrepancy between the calEnviron. Sci. Technol., Vol. 27, No. l , 1993 141

I03

CLMBBE 0 Measured

e

as

(a)

i

80

1

0 Measured

I

Calculated

(b)

!

A

60

1

40

-

d a

20 20

1

1 1

1 0 1850

. 1

Calculated

I

I

CLMBBE

O ,9oc

1300

195'3

Year

Year

1950

2'

L

,

,

1 a50

,

, '900

,

Year

,

195c

-

I 2:03

Figure 4. Comparlson of measured total PAH flux for core CLM88E wlth that calculated from the U.S. energy consumption of (a) coal, petroleum, and wood, (b) coal, petroleum, and wood, IBH, and (c) coal, petroleum, wood, and coal used for coke production, IBH. IBH means that the Inverse Berger-Heath model was applied. Actual measured data are Indicated with open triangles, while open circles reflect measured points obtained .. through llnear lnterpolatlon. -

CLM 88F

340

-

320

-

0 Measured

CLM88F C Measured

33C -

0 Calculated

1850

1930

Calculated

P

Year

1950

200c

190Q

Year

1950

io03

Figure 5. Comparison of measured total PAH f l u for core CLM88F wlth that calculated from the US. energy consumption of (a) coal, petroleum, and wood and (b) coal, petroleum, wood, and coal used for coke productlon.

culated and measured fluxes during the mid-1950s. After the inverse Berger-Heath model reconstruction, the discrepancy is enhanced and the time of the peak is shifted to the early 1960s instead of the mid-l950s, but the overall agreement between measured and calculated data is improved (Figure 4b). If four sources are considered, the calculated total PAH fluxes are even closer to the measured data (Figure 44. For CLM88F, the estimated total PAH fluxes generally coincide very well with the trends of the measured fluxes (Figure 5a). Note in particular that the consumption peaks during the late 1920s (coal) and 1978 (petroleum) are reflected in the measured data, as are the valleys during the Depression, 1930-1940 (coal), and around 1983 (petroleum). With four sources, the calculated flux profile 142 Envlron. Scl. Technol., Vol. 27, No. 1, 1993

provides a better match during the 1940s-1960s (Figure 5b). For NLM88D, there is a discrepancy observed between the calculated and the measured total PAH fluxes when three sources are used, as shown in Figure 6a. Improved results for four sources are shown in Figure 6b. The inverse Berger-Heath model does not improve the fit. However, it excludes petroleum as a source (Table 111). In general, a significant discrepancy is demonstrated during the 1950s-1960s for all four cores including SLM88B (22)if three and, to a lesser extent, if four sources were considered. This peak was also observed by Gschwend and Hites (8), Ohta et al. (32),Eadie et al. (33),McVeety and Hites (W), and Heit et al. (17). These authors attributed the peak

100

100

I

NLM88D

NLM88D

0 Measured

0 Measured

0 Calculated

0 Calculated

(b)

60

1

60

1

h L

\ \

c"

40

40

-

20

20

Q

1850

1900

Year

1950

2000

0

'

1850

1900

Year

1950

'

'

2000

Figure 8. Comparison of measured total PAH flux for core NLM88D with thet calculated from the US. energy consumptlon of (a) coal, petroleum, and wood and (b) coal, petroleum, wood, and coal used for coke production.

Table 111. Coefficients (a's) of Proposed Source Functions for PAH Fluxes to Sediments and for the Fluxes after Reconstruction with the Inverse Berger-Heath Model (Focus Corrected)

three four

applying inverse Berger-Heath model

factors

CLM88E

1. coal 2. petroleum 3. wood 1. coal (other) 2. petroleum 3. wood 4. coal (coke) 1. coal 2. petroleum 3. wood 1. coal (other) 2. petroleum 3. wood 4. coal (coke)

1.71 0.20 4.55

sediment core NLM88D CLM88F" 1.96 5.23 0.66 5.09 10.8

1.37 0.39 6.95 0.54 0.23 6.67 6.93

1.66 0.22 8.65 0.09 0.06 9.32 11.3 0.33O

22.6b 0.26b 8.64b

SLM88B 0.05 0.25 5.47 0.2P 5.00" 1.22'

9.28' 5.42' 7.330

Factors after removal of negative sources. Using Illinois coal consumption data.

to a switch in home heating from coal to oil and gas. Coal used for coal-fired power plants produces, in fact, less PAHs than coal for residential heating (4,34). Goldberg et al. (35) found similarly a peak of charcoal in southern Lake Michigan sediments around 1955 and linked this peak to the installation of improved control devices to remove fly ash from the stack gases after 1955. The use of pulverized coal for large commercial power plants further decreases the formation of PAHs (4, 17). In addition to the above reasons for the PAH peak observed during the 1950s, we suggest that emissions from coke production and consumption may have contributed to ita formation. This is based on the previous discussion of the importance of coke for PAH generation, the time of the peak of the coke source function (Figure 3), and the fact that ita inclusion as a fourth source function improves the fit between measured and calculated PAH records, as can be seen from Figures 4-6 and later discussion.

Coefficients (a's) of the Source Functions. Table I11 summarizes the conversion factors (a's) of the source functions for the four cores before and after application of the inverse Berger-Heath model. For the purpose of comparison between sediment cores, these factors have been corrected with the 210Pb-basedsediment focusing factors (Table I). The coefficients are expressed in ng of PAH/cm2 per 1015Btu. Wood is excluded as a positive source for CLM88F, probably because the measured data (after 1915) do not go long enough back in time to make wood burning a significant source (Figure 2). Use of Illinois coal consumption data for core NLM88D improves the fit somewhat as will be discussed later, but leaves the coefficients (a's) virtually unchanged except for coal, where a = 1.66 has been multiplied with a scaling factor of 13.6 corresponding to the average ratio between the United States and the Illinois consumption data (Figures 2 and 3). Environ. Scl. Technol., Vol. 27, No. 1, 1993

143

Table IV. Measured and Calculated Fluxes of PAHs (ng yr-I) and Estimated Contribution of Sources to PAH Fluxes after the Inverse Berger-Heath Model Reconstruction for CLM88E contribution (%) year

calcd F’(t)

1986 1980 1973 1968 1958 1956 1942 1926 1911 1895 1890

36 37 36 31 27 33 33 31 26 20 20

measd F(t) 42 44 42 35 72 68 34 24 34 19 19

coal (other) 23 19

14 16 15 14 19 22 21 11

9

wood

coal (coke)

19 20

40 30 21 16 21 20 24 30 43 70 76

18 31 44 50 50 54 52 45 35 19 16

19 15 12 5 3 1 0 0

From Figures 2 and 3 and Table I11 we see that the relative contribution of sources (eq 4) is fairly unaffected by this change. For example, for 1955 we find 59 (coal), 12 (petroleum), and 29% (wood) based on U.S. consumption data compared with 61, 13, and 26%, respectively, using the Illinois data. Following eq 4,the coal percentage, p o is calculated as

However, as Figures 2 and 3 suggest, the difference is more significant if more recent data are used. A case in point is the 1985 contributions, which are 53,12, and 35% (US.coal consumption) vs 40, 18, and 42% (Illinois coal consumption) in the given order. For wood, the a’s are 4.6-9.3 ng of PAH/cm2 per 1015 Btu independent of location. For coal used for coke production, coal for other uses, and petroleum, the a’s decline largely in the following sequence: CLM88F, CLM88E, NLM88D, and SLM88B. As an example, the coefficients for coke used for coke production are 10.8,6.93,7.33-11.3, and 1.22, respectively. This sequence is in the order of increasing distance from land-based sources considering the prevailing wind direction from west to east. Thus, as expected, the greatest PAH load is generally occurring closest to the sources. The ratio between the a’s for coal used for coke production and for coal for other uses ranges between 13 (CLM88E) and (SLM88B). Considering that the coal for coke produdion curve (Figure 3) is higher than the U.S. coal consumption curve (Figure 2) divided by 13, this means that coal used for coke production generally contributes more than 50% of the totalcoal-derivedPAH flux. An example of source apportionment based on eq 4 is shown for core CLM88E in Table IV. The relevance of the numbers in this table may be checked by comparing with the relative percentages of wood, oil, and coal carbon particles identified in a dated Lake Michigan core by Griffin and Goldberg (36). For example, for the time period 1953-1978, they found 10 (wood), 76 (coal), and 14% (oil) vs our 1968 numbers, in percent of total PAH flux, of 16 (wood),66 (coal), and 19% (oil). Similarly, for petroleum in the time periods of 1953-1978,1928-1953, 1903-1928, and 1878-1903, they found 14,8,3, and 0%, in the given order, to be compared with our corresponding 1968,1942,1911, and 1890 numbers of 19,5,1, and 0%. Tables similar to Table IV are easily generated for the other cores. 144

Environ. Sci. Technol., Vol. 27, No. 1, 1993

min sediment core”

petroleum

21

Table V. Statistics of the Multiple Linear Regression Model

CLM88E (3) (IBH-3) (IBH-4) SLM88B (3) (4) NLM88D (3) NLM88D (3Y (4) (IBH-3) (IBH-3)d CLM88F (3) (4)

x2

error allowed for x2 = df

(W)

dfb m-n

corr coeff

R2

28

52 49 45

0.3468 0.4110 0.3805

163 157

31 31

48 48

0.3000 0.2452

126

26

54

0.3860

103 96 793 139

23 23 64 28

54 53 56 53

0.4084 0.4330 -0.083 0.4020

126 120

28 28

45 44

0.8168 0.8212

249 146 128

37 29

” The notes following the core identification indicate different number of sources and data reconstruction by the inverse BergerHeath model. Number 3 represents three sources: coal, petroleum, and wood. Number 4 represents four sources: coal used for coke production, petroleum, wood, and coal used for other sectors. * m , number of data points of the time series representing the source function 4; n, number of source functions considered. cIllinois coal consumption data. dThe lowest three points were deleted. Core CLM88F is remarkable because it has more than an 80% contribution to PAHs from petroleum since about 1970. While we have observed similar recent PAH fluxes in two cores collected from Green Bay in 1988 (20),such PAH profiles do not generally appear to have been reported (17,%), except perhaps by Furlong et al. (37). We suggest that the reasons for this are (a) most other measurements have most recent data from about 1981, (b) core CLM88F receives a large petroleum-derived PAH contribution because of its proximity to the shore (Figure l), and (c) the relatively high sedimentation rate and low water depth allows for a high resolution of the PAH record. The high petroleum contribution to PAHs in core CLM88F is supported by the high BghiPIIP ratio of 2.18 f 0.37 in this core over the last 50 years. The corresponding BghzF/IP ratios for the other three cores are 1.41 f 0.10 (CLM88E), 1.39 f 0.17 (NLM88D),and 0.71 f 0.22 (SLM88B), indicating increasing relative contribution of wood burning to the PAH load (4, 38). Also, Benner et al. (39) found a value of 1.74-2.0 for this ratio based on a study of PAHs in Baltimore Harbor Tunnel, confirming that the high value in core CLM88F reflects petroleum dominance. The relationship of the a’s to yield may be seen from the CLM88F data (ng of PAH/cm2 per 1015Btu): coal for other uses -0.7, coal for coke production -11, and petroleum -5. Let us further assume that the factor for wood is -5 ng of PAH/cm2 per 1015Btu. These numbers may be compared with emission factors (mg of PAH/kg) from Bjprrseth and Ramdahl (4): coal for other uses 60 (residential), 0.02-0.8 (power generation); coal for coke production 15; petroleum for mobile sources 5-10; wood 29-40 (residential), 2 (power generation). Although the latter numbers should be expressed on a Btu basis, it appears that the a’s are largely proportional to the emission factors. Statistical Analysis. Table V is a summary of statistical results of the multiple linear regression. x2 is very much error dependent, as shown below:

where u is the relative standard deviation, which was in the range of 9-48% for individual PAHs based on results from gas chromatographic analyses. If the PAH concentration in the sediment is high, then the relative analytical error will be reduced (22). The addition of individual errors gives relative errors which are less than the individual ones. We calculated the relative standard error for total PAH to be 1770,which is used in the calculations. If x2 approximately equals the number of degrees of freedom, then the model fits the experimental data. In order to reach an optimal fit of the multiple linear regression in terms to x2 using three or four source functions, 28% relative standard deviation should be allowed for CLM88F, 23% for NLM88D, 31% for SLM88B, and 37% for CLM88E. If the inverse Berger-Heath model is applied, these numbers change to 64 and 28% for NLM88D and CLM88E, respectively (Table V). That the relative error in most cases is so close to 17% for x2 = df shows that the model is quite realistic. Model Limitations and Potential. Although geochemical and sedimentological factors can play a role in shaping the sedimentary PAH records, we believe that the main features of the four records considered here are driven by the stated energy inputs. Furlong et al. (37) found a significant amount of variability on a regional basis which may in part be related to different energy sources. The high correlation found by these authors between sulfur emission in the state of New York and the total PAH flux in a Big Moose Lake core may reflect that coal rather than petroleum is the major source of PAHs to this core. Other cores with PAH profiles resembling that of CLM88F may receive a larger proportion of petroleumderived PAHs. A possible sedimentological artifact of the measured PAH records is that they all, even CLM88F (Figure 5), show a decline in PAHs in the most recent, Le., the uppermost, layer. This may be an indication that the uppermost fine particles of high PAH content are resuspended into the water column (15). The nepheloid layer of NLM88D had in fact a relatively high PAH concentration of 6.54 pg/g. Total organic carbon as measured by loss on ignition (LOI) showed some variation between the four cores, with the lowest values (4-8%, dry sediment) obtained for NLM88D and the highest (7-12%) for CLM88F. The higher PAH flux for the latter core (Figure 5) may in part be related to the higher LOI. However, the shape of the PAH records seems to be unrelated to LOI. Our model assumes that all PAHs are transported, e.g., in the atmosphere, with equal efficiency. Thus, selective photooxidation of compounds such as anthracene, which is known to occur over long distances (8),can cause some error (510%) in the results. As stated here, fuel Btu’s are analogues for PAHs. Each fuel type can have a separate but constant PAH yield vs time. In reality, however, the PAH yield may change as is especially the case when coal for power generation is substituted for coal for residential heating. With known emission factors, this aspect could be taken into account. Making this adjustment may not be as important as the above emission factors (mg of PAH/kg) would suggest, since coal for other uses generally produces less than 50% of the coal-derived PAHs, and coal for coke production has a fairly constant PAH yield. Substituting Illinois for US. coal consumption data improved the fit of the model enough to show that use of local

-

consumption data in fact may be preferable, as one would expect, but not sufficiently to demonstrate that this would make a major difference. Nevertheless, future improvementa in the model may well include the use of carefully weighted energy consumption data for the surrounding states.

Conclusions A simple multiple linear regression model based on a least squares method is applicable to estimate the sources of PAHs to lake sediments. The model fits historical consumption data for coal, petroleum, and wood to dated sedimentary PAH records. Coal has been a significant source of PAHs since the beginning of this century. Ita relative contribution ranges from 20 to 30% in a remote core (SLM88B), 30 to 70% in intermediate cores (CLM88E and NLM88D), to 10 to 80% in a near-shore core (CLM88F). The contribution of petroleum-derived PAHs to the total PAH load was virtually zero before 1900, is less than 40% in all cores except CLM88F where the share has been higher than 80% since 1970, and still is rising in 1988. The fraction of PAHs derived from wood burning is greatest (>40%) for the remote core (SLM88B) and less (>20%) for the intermediate cores. PAHs from wood burning are distributed fairly evenly over the lake, whereas PAHs from coal combustion decrease from near shore to remote sediments. PAHs from petroleum drop off dramatically from the coast to the remote lake sites. Peak PAH fluxes are observed not only from the early 1950s-1960s but also in the most recent layers (1985) of CLM88F. On the basis of the high BghiPIIP ratio (2.18 f 0.37) during the last 50 years in this core, and the similarity of the sedimentary PAH record to the US.petroleum consumption data, we conclude that the 1985 maximum in CLM88F comes from petroleum-derived PAHs. Regarding the 1950-1960’s peak we find that maximum U S . coke production during that time period may have contributed to its formation in addition to the switch in residential heating from coal to gas and oil. The inverse Berger-Heath model was successfully used to reconstruct the PAH input to CLM88E affected by mixing as shown by an improved fit of the calculated PAH fluxes to the source functions, but it was not successful for NLM88D, possibly due to magnification of errors caused by the unmixing process. Acknowledgments We thank Robert Harley for assistance with sampling and Mark Hermanson and L.-Y. Yan for radioanalytical work.

Literature Cited (1) Sullivan, J. R.; Delfino, J. J. A Select Inventory of Chernicals Used in Wisconsin’s Lower Fox River Basin; University of Wisconsin Sea Grant Institute: Madison, WI,1982. (2) Neff, J. M. Polycyclic Aromatic Hydrocarbons in the

Aquatic Environment, Sources, Fates and Biological Effects; Applied Science Publishers: London, U.K., 1979. (3) Particulate Polycyclic Aromatic Matter; National Academy of Sciences: Washington, DC, 1972. (4) Bjarseth, A., Ramdahl, T., Eds. Handbook of Polycyclic

Aromatic Hydrocarbons: Emission Sources and Recent Progress in Analytical Chemistry; Marcel Dekker: New York, 1985; pp 1-85. (5) Suess, M. J. Sci. Total Enuiron. 1976, 6, 239-250. (6) Hites, R. A.; Laflamme, R. E.; Farrington, J. W. Science 1977, 198,829-831. (7) Cranwell, P. A.; Koul, V. K. Water Res. 1989,23,275-283. Environ. Scl. Technol., Vol. 27, No. 1, 1993

145

Environ. Sci. Technol. 1993, 27, 146-152

Gschwend, P. M.; Hites, R. A. Geochim. Cosmochim. Acta 1981,45, 2359-2367. Giger, W.; Schaffner, C. Anal. Chem. 1978, 50, 243-249. Jones, K. C.; Stratford, J. A.; Tidridge, P.; Waterhouse, K. S.; Johnston, A. E. Environ. Pollut. 1989, 56, 337-351. Jones, K. C.; Stratford, J. A.; Waterhouse, K. S.; Furlong, E. T.; Giger, W.; Hites, R. A.; Schaffner, C.; Johnston, A. E. Environ. Sci. Technol. 1989, 23, 95-101. Muller, G.; Grimmer, G.; Bohnke, H. Naturwissenschaften 1977,64, 427-431. Wakeham, S. G.; Schaffner, C.; Giger, W. Geochim. Cosmochim. Acta 1980,44, 403-413. Tan,Y. L.; Heit, M. Geochim. Cosmochim.Acta 1981,45, 2267-2279. Helfrich, J.; Armstrong, D. E. J. Great Lakes Res. 1986, 12, 192-199. Eadie, B. J. In Toxic Contaminants in the Great Lakes; Nriagu, J. O., Simmons, M. S., Eds.; John Wiley & Sons Publishers: New York, 1984; pp 195-211. Heit, M.; Tan, Y. L.; Miller, K. M. Water,Air, Soil Pollut. 1988, 37, 85-110. Smith, J. N.; Levy, E. M. Enuiron. Sci. Technol. 1990,24, 874-879. Sexton, K.; Liu, K. S.; Hayward, S. B.; Spengler, J. D. Atmos. Environ. 1985, 19, 1225-1236. Zhang, X.; Christensen, E. R.; Yan, L.-Y. Submitted for publication in J. Great Lakes Res. Hermanson, M.; Christensen, E. R. J . Great Lakes Res. 1991, 17, 33-50. Zhang, X. Ph.D. Dissertation, Dept. of Civil Engineering and Mechanics, University of Wisconsin-Milwaukee, 1991. Christensen, E. R.; Klein, R. J. Enuiron. Sci. Technol. 1991, 25, 1627-1637. Yan, L.-Y. M.S. Thesis, Dept. of Civil Engineering and Mechanics, University of Wisconsin-Milwaukee, 1991. McVeety, B. D.; Hites, R. A. Atmos. Environ. 1988, 22, 511-536.

State Energy Data Report, Consumption Estimates

1960-1988; DOE/EIA-0214 (88); Energy Information Administration (EIA): Washington, DC, 1988a. (27) Estimation of Wood Energy Consumption, 1949-1983; Energy Information Administration (EIA): Washington, DC, 1981,1983. (28) Hottel, H. C.; Howard, J. B. New Energy Technology;MIT Press: Cambridge, MA, 1971. (29) Hidy, G. M.; Henry, R. C.; Hansen, D. A.; Ganesan, K.; Collins, J. Analysis of Trends in Historical Acid Precursor

Emissions and Their Airborne and PrecipitationProducts; (30) (31) (32) (33)

ERT Document P-B538; Environmental Research and Technology, Inc., Westlake Village, CA, 1983. Minerals Yearbook; U.S. Department of the Interior: Washington, DC, 1950. Annual Review; DOE/EIA-0214(88); Energy Information Aministration (EIA): Washington, DC, 1988b. Ohta, K.; Handa, N.; Matsumoto, E. Geochim. Cosmochim. Acta 1983, 47, 1651-1654. Eadie, B. J.; Robbins, J. A.; Faust, W. R.; Landrum, P. F.

In Organic Substances and Sediments in Water: Processes and Analytical; Baker, R., Ed.; Lewis Publishers: Ann Arbor, MI, 1991; Vol. 2, pp 171-189. (34) Dasch, J. M. Enuiron. Sci. Technol. 1982, 16, 639-645. (35) Goldberg, E. D.; Hodge, V. F.; Griffin, J. J.; Koide, M.; Edgington, D. N. Environ. Sci. Technol. 1981,15,466-471. (36) Griffin, J. J.; Goldberg, E. D. Science 1979,206, 563-565. (37) Furlong, E. T.; Cessar, L. R.; Hites, R. A. Geochim. Cosmochim. Acta 1987, 51, 2965-2975. (38) Freeman, D. J.; Cattell, F. C. R. Environ. Sci. Technol.1990, 24, 1581-1585. (39) Benner, B. A., Jr.; Gordon, G. E.; Wise, S. A. Enuiron. Sci. Technol. 1989,23, 1269-1278.

Received for review February 25, 1992. Revised manuscript received July 6,1992. Accepted September 24,1992. This work was sponsored by US.National Science Foundation Grants CES-8701184 and BCS-8921000.

Atmospheric Chemistry of Hydrofluorocarbon 134a. Fate of the Alkoxy Radical CF30 Jens Sehested

Section for Chemical Reactivity, Environmental Science and Technology Department, Riser National Laboratory, DK-4000 Roskiide, Denmark Timothy J. Wallington"

Research Staff, SRL-3083, Ford Motor Company, P.O. Box 2053, Dearborn, Michigan 48 121-2053

rn The atmospheric chemistry of the alkoxy radical CF30 produced in the photooxidation of hydrofluorocarbon (HFC) 134a has been investigated wing Fourier transform infrared spectroscopy. CFBOradicals are shown to react with methane and CF3CFH2to give CF30H. CF30H decomposes to give COFz and HF. The rate constant for the CF30H + reaction CF30 CF3CFHz (HFC-134a) CF CFH was determined to be kI9 = (1.1f 0.7) X cmd molecule-' s-' at 297 K. The implications of ow results for the atmospheric chemistry of CF30 radicals and HFC-134a are discussed.

+

-

Introduction Recognition of the adverse effect of chlorofluorocarbon (CFC) release into the atmosphere has led to an international effort to replace CFCs with environmentally acceptable alternatives (1-3). Hydrofluorocarbon 134a (1,1,1,2-tetrafluoroethane) is a viable substitute for CFC-12 148

Environ. Sci. Technol., Voi. 27, No. 1, 1993

in automotive air-conditioning systems. Prior to large-scale industrial use of HFC-l34a, the environmental consequences of i b release into the atmosphere should be considered. To define the environmental impact of HFC-134a release, the atmospheric photooxidation products of HFC-134a need to be determined. The main atmospheric loss mechanism for HFC-134a is reaction with the OH radical, reaction 1. Studies of the CF3CFH2 OH CF3CFH + HzO (1)

+

-

HFC-134a kinetics of this reaction ( 4 ) have shown that the atmospheric lifetime of HFC-134a is approximately 15 years. The alkyl radical formed in reaction 1 reacts rapidly (within 1p under tropospheric conditions) with molecular oxygen to give the peroxy radical CF3CFHO2(reaction 2). CF3CFH + 0 2 + M CF3CFHOz f M (2) Studies in our laboratories have shown that reaction with

0013-936X/93/0927-0146$04.00/0

-+

0 1992 American Chemical Society