Environ. Scl. Technol. 1002, 26, 1444-1457
Murray, J. W.; Gill, G. Geochim. Cosmochim. Acta 1978,
42,9. Whitlow, S.L.;Rice, D. L. Water Res. 1985,19,619. Stumm, W.; Morgan, J. J. In Aquatic Chemistry; Wiley Interscience: New York, 1970;Chapter 6. Wedepohl, K. H. In Geochemistry; Holt Rinehart and Winston: New York, 1971;Chapter 2. Tsumura, A.; Yamasaki, S.; Kihou, N. Radioisotopes 1991,
40,279. Sholkovitz, E. R.; Elderfield, H. Global Geochem. Cycles
1988,2, 157. Goldstein, S.J.; Jacobsen, S. B. Earth Planet. Sci. Lett. 1988,89,35. Elderfield, R.; Goddard, R. U.; Sholkovitz,E. R. Geochim.
Cosmochim. Acta 1990,54,971. M.0.;Asmode, J. F.; Foster, P.; Vant'dack, L. Anal. Chem. 1981,53,1766. Andreae,
(36) Mok, W. M.; Wai, C. M. Anal. Chem. 1987,59,233. (37) Tipping, E.; Woof, C.; Backes, C. A.; Ohnstad, M. Water Res. 1988,22,321. (38) Vlassopoulos, D.;Wood, S. A.; Mucci, A. Geochim. Cosmochim. Acta 1990,54,1575. (39) Andreae, M.0.Anal. Chem. 1977,49,820. (40) Tanaka, S.;N h u r a , M.; Kishi, Y.; Hashimoto, Y. Nippon Kagaku Kaishi 1986,727. (41) Orvini, E.; Gallorini, M. J.Radioanal. Chem. 1982,71,75. (42) Takayanagi, K.; Wong, G. T.M. Geochim. Cosmochim.Acta 1985,49,539. (43) Li, W. C.; Victor, D. M.; Chakrabarti, C. L. Anal. Chem. 1980,52, 520.
Received for review November 15, 1991. Revised manuscript received March 17, 1992. Accepted March 19,1992.
Flux Estimates and Pattern Recognition of Particulate Polycyclic Aromatic Hydrocarbons, Polychlorinated Dibenzo-p -dioxins, and Dibenzofurans in the Waters outside Various Emission Sources on the Swedish Baltic Coast Carina Naf,* Dag Broman, Harald Pettersen, Carl Rolff, and Yngve Zebiihr
Aquatic Chemical Ecotoxicology, Department of Zoology, Stockholm University, S-I0691 Stockholm, Sweden Sediment trap fluxes of polycyclic aromatic hydrocarbons (PAHs) and polychlorinated dibenzo-p-dioxins and dibenzofurans (PCDD/Fs) were estimated outside various industrial emission sources and a large city (Stockholm) and at remote water areas. The direct influence area of all emission sources was limited to the inner 10-50 km2,indicating a fast settling of particulate PAHs and PCDD/Fs. The flux of 15 PAHs within the 10 km2 closest to the aluminum smelter was 560 kg y-l, which was 56 times the background flux. The other emission sources caused fluxes between 300 and 10 kg y-l in the corresponding areas. The PAH flux to the whole Baltic was estimated to 400 ton y-l. A pulp and paper mill was identified as a major PCDD/F source generating a gross flux of 68 g y-l (1.1g y-l toxic equivalents) in the 10 km2 closest to the source. Corresponding fluxes of the other industrial sources were 19-7 g y-l (0.22-0.05g y-l as toxic equivalents). The flux of PCDD/F to the Baltic was estimated to be 300 kg y-l (4 kg y-l toxic equivalents). Multivariate pattern recognition significantly separated all sampling areas (except one), indicating a unique PAH and PCDD/F emission profile of all the sources. The compounds responsible for the separations were identified. With fossil fuel consumption data, the total PAH emission from all Swedish Baltic municipalities to the coastal waters was estimated to be 4.5 ton y-l (i.e., 1% of the total Baltic flux). ~~
~
~
Introduction
Research on polycyclic aromatic hydrocarbons (PAHs) in the aquatic environment has been extensive during the last decades. The major reason for this is the high load of these compounds in many areas in combination with their toxic, mutagenic, and carcinogenic potential ( I , 2 ) . Several PAHs are also found on the US.EPA's list of priority pollutants (3). The other group of organic contaminants that is focused on in this study is the chlorinated dibenzo-p-dioxins and dibenzofurans (PCDD/Fs). They are also of environmental concern due to their toxicity, 1444
Environ. Sci. Technol., Vol. 26, No. 7, 1992
carcinogenic potential, and potential to damage or interfere with immunologicaland reproductive systems (4). Some PCDD/F isomers also have a potential to biomagnify in aquatic food chains. Both PAHs and PCDD/Fs are formed primarily from incomplete combustion of different types of organic matter. In the investigated areas presented here, PAHs are primarily derived from the combustion of fossil fuels in mobile sources (e.g., automobiles, trucks, ships, airplanes, etc.), from residential heating and power generation, and also from municipal solid-waste incineration. Specific industrial processes such as metal smelting (e.g., steel and aluminum), petroleum catalytic cracking, bitumen production, and graphite electrode production also emit PAHs (5). A minor portion also originates from dicharges of noncombusted fossil fuel products. PCDD/Fs are formed in flames, engines, and high-temperature melting processes of organic matter in the presence of organic chlorine (6). Examples of some of the sources for PCDD/Fs in the study areas within this investigation are gasoline engines (with chlorine additives), various solid-waste incinerators, and specific industrial processes such as some types of metal smelters (7). PCDD/Fs are also formed during intentional or unintentional combustion of various commercial and technical chlorinated products, such as PCBs and chlorinated phenols. Further, emissions associated with the chlorine bleaching of pulp and paper is an example of a noncombustion process which is of great significance to the Baltic (8). Submicron particles generated by combustion seem to be of crucial importance for the long-range input and particulate distribution of compounds such as PAHs and PCDD/Fs in remote aquatic areas (9). Supermicron particles (>1pm), on the other hand, undergo significant gravitational settling and therefore become more important in the vicinity of an emission source (IO). The diffusive exchange of vapor-phase PAHs and PCDD/Fs over the air-water interface is also of importance at remote loca-
0013-936X/92/0926-1444$03.Q0/0
0 1992 American Chemical Society
tions (11)and probably also in the vicinity of an emission source. This process can be described as a slow dynamic exchange whereas the submicron particles have to be washed out of the atmosphere during precipitation events. In the water, compounds like PAHs and PCDD/Fs settle out primarily in association with large rapidly settling particles (12).This settling particulate matter (SPM) often consists of aggregates of smaller particles which are formed by biotic processes, such as fecal pellet formation, and by abiotic processes, such as flocculation and agglomeration. Floccules are inorganic particle aggregates that are held together by electrostatic forces and which often get colonized by bacteria and protozoans. Agglomerations are comprised of organic and inorganic matter weakly held together by surface tension and organic cohesion due to various biological activities. These mechanisms probably also facilitate relatively fast settling of the combustionderived particles discussed above (13). Further, the comparatively high organic content of SPM results in a highabsorbing capacity for the dissolved PAHs and PCDD/Fs due to the high organic carbon-water partition coefficients displayed by these compounds. SPM collected in sediment traps during a longer period of time therefore reflects an integrated picture of the PAHs and PCDD/Fs found in the water mass due to the efficient scavenging capacity of SPM for both dissolved and suspended particulate forms of these compounds. This is also confirmed in an earlier study from the Baltic (14)in which the PAH and PCDD/F chemical profiles of SPM collected in sediment traps in remote areas were compared with the corresponding profiles of the sum of particulate and dissolved matter in the waters. Results from the study by Broman et al. (14)do not indicate any selective loss of PAHs and PCDD/Fs via sedimentation, such as has been described in other studies (12,15). The study presented here discusses results of estimated gross fluxes of PAHs and PCDD/Fs in water areas outside various potential emission sources as well as at remote background locations along the Swedish Baltic coast. Estimates of the total load in the Baltic Sea are also given. To investigate possible similarities and/or differences in the chemical pattern of the various sediment trap samples and to point out emission source characteristic PAH and PCDD/F compounds, chemical profile comparisons and a multivariate pattern-recognition method called SIMCA (Soft Independent Modeling of Class Analogy) have been used (16).The gross fluxes of PAHs outside all Swedish Baltic municipalities are further calculated on the basis of fossil fuel consumptions. Materials and Methods The sediment traps used were of a self-suspended buoyant type with cylindrical collection vessels of glass contained in gimbal-mounted PVC cylinders. The trap design and anchorage technique used are described in detail elsewhere (17). Chloroform was used as a preservative in the collection vessel as described by Broman et al. (18),where a description of the workup procedure of the collected SPM material can also be found. The wet SPM samples were extracted for 24 h with toluene in a Soxhlet apparatus equipped with a Dean-Stark trap for the collection of water (19).Before extraction the following internal standards were added: 2-methylanthracene, picene, perylene-d12and dibenz[a,i]pyrene for the PAH analyses and eight 13C-labeled PCDD/F standards (2,3,7,8-TCDF, 2,3,7,8-TCDD, 2,3,4,7,8-PnCDF,1,2,3,7,8PnCDD, 1,2,3,4,7,8-HxCDF, 1,2,3,6,7,8-HxCDD, 1,2,3,4,6,7,8-HpCDD, and OCDD) for the PCDD/F analyses. The extracts were volume reduced and eluted
through a silica column, after which the samples were fractionated on HPLC according to the method described by Zebiihr et al. (20)and Broman et al. (14). The HPLC fractions containing the PAHs were treated with a dimethylformamide cleanup procedure followed by elution through a silica column. A detailed description of the cleanup procedure used is given elsewhere (18,21). The quantitative analysis was done by gas chromatography using a flame ionization detector GC/FID (Varian 3300) and a 25-m CP-Sil-8 capillary column (Chrompack). Identification of the compounds was done by means of GC retention times of known standards and mass spectra obtained with a Hewlett-Packard (HP) 5890 Series I1 GC coupled to a H P 5971A mass selective detector. A PAH standard mixture containing all compounds which were analyzed, including the internal standards, was extracted and further cleaned up in the same manner as the SPM samples. The standard mixture was then used to establish the response factors for all compounds analyzed. The 15 PAHs analyzed (PAH15)were fluoranthene (Flu), pyrene (Pyr), 2-methylpyrene (2MePyr), l-methylpyrene (lMePyr), benzo[ghi]fluoranthene (B[ghi]F), cyclopenta[cd]pyrene (Cp[cd]P), benz [a]anthracene (B[a]A), chrysene/triphenylene (Chr/Tri), benzo[k]fluoranthene (B[k]F), benzo[e]pyrene (B[e]P), benzo[a]pyrene (B[alp), perylene (Per), indeno[1,2,3-cd]pyrene (Ind), benzo[ghi]perylene (B[ghi]P), and coronene (Cor). The HPLC fraction containing the PCDD/Fs were cleaned up using a number of open chromatography columns in several steps according to Smith et al. (22). GC/MS analyses were performed using a VG 70E mass spectrometer and a Hewlett-Packard 5790 GC with a 40-m SP-2331 capillary column (Supelco). The PCDD/Fs analyzed were all 2,3,7,8-substituted isomers [the 2,3,4,8-/ 2,3,7,8-TCDF, 1,2,3,4,8-/ 1,2,3,7,8-PnCDF, and 1,2,3,4,7,9-/1,2,3,4,7,8-HxCDF were unresolved and quantified together] and the s u m of all isomers in the eight congener groups. The 2,3,7,8-substituted isomers were given toxic equivalent factors according to the Nordic model (23),and the concentrations are expressed either as the total of all PCDD/Fs analyzed or as TEQ (i.e., the concentrations of the 2,3,7,&substituted isomers multiplied by their factors). Result and Discussion Factors of Importance for the Fluxes and Concentration in the Various Investigation Areas. The investigated water areas outside the seven different emission sources and their position on the Swedish Baltic coast are illustrated in Figure 1A-C. The figure also shows where the sediment traps used to collect SPM (settling particulate matter) were located during the l-year-long sampling period. The characteristics of the seven emission sources are summarized in Table I (i.e., the various types of industries or sources, their production, and the size of the cities or municipalities in which they are situated). The number of sediment traps deployed in each area is also listed in the table together with the number of PAH1, and PCDD/F analyses made on the collected SPM. Both the magnitude of the fluxes of PAHs and PCDD/Fs and the size of the affected area are likely to be influenced by water exchange with the open sea. The magnitude of the water exchange is, however, difficult to evaluate. Water exchange caused by the tide, which is of crucial importance in many other coastal regions, is of no significance since the Baltic is essentially a nontidal area. On the other hand the larger freshwater surface currents induced by river inflow, and the proportion of each study area confronting the open Baltic is of considerable imEnvlron. Sci. Technol., Vol. 26, No. 7, 1992
1445
I\
.....................................
.
4
B;
. .
I
................,
SKELLEFTEHAMN
0
I
2,"'
L/
HARNOSAND
SUNDSVALL
.
I
. :
:1p.
..........2 STOCKHOLM
L
................
. A
NYNASHAMN
OXELOSUND
ngwe 1. (Panel A) Map over the Swedish Bank ccast showing the positions of the Merent emfssion sourca areas and the rerote background
areas. (Panel E) Map over me emission souce areas at Skelleftehamn (metal smeiter). W m h n d (graphite elactrodegroducingindustry). and Sundsvall (aluminum mer)The . beck dots indicate Um sediment trap locations. the armws indicate larger freshwater surface currents induced by river WMOw. the broken lines indicate the stretch at each water area which faces the open Baltic. the black areas indicate me localization of the industries. and the areas marked with a thin line represent urban areas. (Panel C ) Map over Um emission source areas at lggesund (pulp and paper mill). Stockholm (large city). Nynishamn (petroleum refinery). and Oxel6sund (steal plant). For markings in the maps see panel B.
portance (Figure 1B and C). Water meas with a freshwater surface c u r r e n t and a stretch that faces t h e open sea p r o b a b l y represent areas w h i c h have a very good water I446 Envkon. Sci. Technol., Vol. 26. NO. 7. 1992
exchange. All sediment t r a p locations in t h e area at t h e m e t a l smelter in Skelleftehamn represent t h i s t y p e of location (except perhaps t h e innermost location). T h e outer
Table I. Names of Different Emission Areas and Two Remote Background Areas, Emission Source Characteristics, Number of Sediment Traps Deployed, and Number of PAH and PCDD/F Analyses no. of emission area
emission source characteristics
Skelleftehamn metal smelter with annual production of 100000 ton of Cu, 40000 ton of Pb, 270 ton of Ag, 8 ton of Au, 160000 ton of H2S04,and 70000 ton of SOz and a small city (3000 inhabitants) Hiimosand graphite electrode-producing industry with annual production of 11000 ton of synthetic graphite and a minor city (20 000 inhabitants) Sundsvall aluminum smelter with annual production of 100000 ton of aluminum, city (60000 inhabitants), and possible secondary influence from pulp and paper mills pulp and paper mill with annual production of 315000 ton of pulp and paper and a Iggesund small city (4000 inhabitants) larger nonindustrialized city (1.5 million inhabitants) with an annual petroleum Stockholm product consumption of -2.5 million m3 petroleum refinery with annual turnover of 700 000 ton of crude oils and a minor city Nynishamn (15000 inhabitants) steel plant with annual production of 1.2 million ton of steel and minor city (13000 Oxelosund inhabitants) Gaviksfjiirden remote background area in the Bothnian Sea remote background area in the middle Baltic DroDer Siisfiiirden
sediment trap locations in the area outside the aluminum smelter in Sundsvall, the pulp and paper mill in Iggesund, and the nonindustrialized city of Stockholm also represent areas with a very good water exchange. The same is likely to be valid for the outer sediment trap locations outside the petroleum refinery in Nyniishamn and the outermost sediment trap location at the steel plant in Oxelosund due to their exposure to the open Baltic, even though they lack freshwater-induced surface currents. All other sediment trap locations in this investigation, including the remote background areas, can be considered as intermediates between those with a high degree of exposure described above and the innermost area at the pulp and paper mill in Iggesund, and the innermost area of the archipelago of Stockholm which have a moderate exchange with the open sea. In areas with a good water exchange, the size of the area affected by the emissions from the source will be greater, and thus the relative particulate PAH and PCDD/F flux will decrease. However, due to flocculation and aggregation processes, the particulate flux can in certain situations increase when freshwater water surface currents mix with the saltier seawater (13). This mechanism can increase the fluxes at the areas which have a contribution of freshwater surface currents, i.e., outside Skelleftehamn, Harnosand, Sundsvall, Iggesund, and Stockholm. The salinity differences between the open Baltic and the fresh river water are, however, small in all the water areas of this investigation (3-5 9’60). Another factor which is important for the particulate flux is the characteristics of the particulate emissions and whether the source outlets were to the water, to the air, or both. For example, if the particulate PAH and PCDD/F are emitted bound to supermicron particles, the flux of these contaminants will increase in the receiving water area to a much higher extent compared to an emission source which emits a significant part of PAHs and PCDD/Fs bound to submicron particles (10). This is probably true for particulate outlets to both air and water. All emission sources have significant outlets to both air and water except for the graphite electrode-producing industry, which has no water outlets. The proportion between air and water outlets, at the various industries, is difficult to evaluate because of the lack of information. It can be assumed, however, that the air outlets dominate the emissions of PAHs at all industries except perhaps the aluminum smelter where the air and water outlets were
no. of traps
PAH anal.
no. of PCDD/F anal.
5
5
4
5
5
2
6
6
4
5
5
4
11
21
6
12
5
5
5
3 3
3 3
1 1
more comparable. The dominance of air outlets is probably also valid for the PCDD/Fs emissions with an exception for the pulp and paper mill at Iggesund. Another factor, previously mentioned in the introduction, which increases the particulate flux of PAH and PCDD/F, is the variation in the flux of organic matter. A high dissolved organic content in the water facilitates an increase in size of the particles and thereby the flux. An outgoing freshwater current with high concentrations of dissolved organic matter (e.g., humic materials) might on the other hand adsorb dissolved PAHs and PCDD/Fs and facilitate the redistribution of the compounds (24,25), and thereby reducing the particulate flux of PAHs and PCDD/Fs. The variation in biological production between the various water areas is also difficult to evaluate. In general, the Particulate flux of organic carbon decreases from south to north in the Baltic (26). Still, local variations in the coastal nutrient load are most likely to reduce this effect. Meteorological constraints that can be of importance for the distribution of PAHs and PCDD/Fs in the near-source water areas are precipitation and the intensity and direction of the winds. The annual variations of these constraints are, however, very small between the various areas. Flux Estimates of PAHISand PCDD/F. Gross fluxes of PAH and PCDD/F have been measured in each emission source area (described above) and considered as results of the different source emissions. The fluxes were then compared to the background gross flux of the corresponding areas. The general background flux was calculated from the sediment traps which had been located in remote background areas of the Baltic proper and the Bothnian Sea (Table I). The total gross flux outside each emission source was calculated as follows. For each sediment trap a corresponding subarea was identified with the assumption that any point that was closer to a given sediment trap than any other belongs to the same subarea. These subareas were thus limited by the shore lines and the intersections of lines of equidistant points to all next-nearest neighbors of a sediment trap. The flux in each subarea was assumed to be equal to that of the corresponding sediment trap. The delimitations of these subareas were adjusted to avoid inclusion of areas that were likely to be unaffected due to physical obstacles, powerful river outlets, or main surface streams. The outer limit of the investigated area at each emission source was given by a circular periphery, equidistant with the outEnviron. Scl. Technol., Vol. 26, No. 7, 1992
1447
Table 11. Total Accumulated Yearly Fluxes (Including Background F l u x ) of PAHI,, Total P C D D / F , and P C D D / F Expressed as T E Q within 10 km2 Closest to Emission Sources"
total PCDD/F SBR, %
PAHE source area
0-10 km2,kg
SBR, %
0-10 km*, g
Skelleftehamn Hiirnosand Sundsvall Iggesund Stockholm Nynashamn Oxelosund background
14 200 560 16 150 31 290
100 2000 5600 200 1500
9.5 1.3 19
300 2900
TEQ 0-10 km2,g
SBR, %
68 na na
100 100 200 900
0.14 0.053 0.22 1.1
100 100 200 1100
na na
na na
na na
18
200
0.23
200
1.7
10
0.10
"The relation between the total flux in the 10 km2 nearest an emission source and the deposition in an area of equal size exposed to background flux only is indicated (SBR). SBRs given in even hundreds. Samales not analvzed are indicated bv na. 1100
70
- 1
1200
200
1000 800
900 400
/
800 0
700
-h
600
-
0
10 20 30 40
-401
1
0)
h
500
0)
30
-
20
-
0
10
20
30
400
300
200 100
0 0
5
10
15
20
25
30
0
2
4
6
8
10
km2 Flgure 2. Accumulated total flux of PAH,, in the accumulated 30 km2 closest to the emission source. Background flux Is indicated by a dashed line. A return to background levels of flux is indicated by slopes similar to that of background flux. The inset includes a larger area of 50 km2. Symbols identical to those shown in Figure 7.
km* Figure 3. Accumulated total flux of total PCDDIF In the accumulated 10 km2 closest to the emisslon source. Background flux Is indlcated by a dashed Ilne. A return to background levels of flux Is indlcated by slopes slmllar to that of background flux. The Inset Includes a larger area of 30 km2. Symbols Identical to those shown in Flgure 7.
ermost sediment trap station, and with center in the emission source and end points at the intersection with the shore lines. The term flux used in the text below will consistently refer to the gross flux, i.e., the total amount of matter or compounds settling per unit of time. The flux in the water area outside Stockholm has been calculated by a somewhat different method, which is described in the section headed Total Flux of PAH15from Swedish Baltic Coastal Municipalities Calculated from the Consumption of Fossil Fuel. The flux in the 10 km2 closest to an emission source (Table 11) was calculated both as mass per year and as an emission source to background flux ratio (SBR). The SBR was calculated as the ratio between the flux in the 10 km2closest to the source and the flux in an area of equal size exposed to background flux only. The accumulated fluxes of PAH15,total PCDD/F, and TEQ in the water areas surrounding each emission source have been calculated (Figures 2-4). The accumulated flux was calculated by adding the estimated flux in each subarea to that of the next subarea in order of increasing distance to the main outlet of the emission source. The accumulated flux is expressed as a function of increasing area. A flux function that becomes parallel with the background
flux line (dotted lines in Figures 2-4) indicates a return to background flux levels. The flux of PAHI5 at the different emission sources (Figure 2) can be described as belonging to one of three major groups depending on the magnitude of the flux and the size of the affected area: (Group A) Low or Moderate Influence Compared to Background Flux. In this group the water areas outside the metal smelter (Skelleftehamn), the pulp and paper mill (Iggesund), and the oil refinery (Nynlishamn) were found. The fluxes were close to background levels for all of these sources. Only the oil refinery caused a slightly elevated flux level, but the influence was only seen in the first 5 km2. The estimated SBRs were in the range of 100-300% (14, 16, and 31 kg y-' of PAHIS were deposited in the closest 10 km2 of Skelleftehamn, Iggesund, and Nynlishamn, respectively). As mentioned above both the metal smelter and the oil refinery are situated in water areas which have a good exchange with the open sea. Nevertheless, it is notable that the fluxes are so low. The outer sediment trap locations outside the pulp and paper mill also have a good water exchange with the open sea in addition to the as-
1448
Environ. Sci. Technol., Vol. 26, No. 7, 1992
1.2
r
I
0
I
O
lo 2o
2
30
4
6
8
10
kmz Figure 4. Accumulated total flux of TEQ In the accumulated 10 km2 closest to the emission source. Background flux is Indicated by a dashed line. A retwn to background levels of flux Is indicated by slopes similar to that of background flux. The Inset includes a larger area of 30 km2. Symbols identical to those shown in Figure 7.
sumed low emissions of PAH for this type of industry. (Group B) Elevated PAH15Fluxes but a Comparatively Rapid Return to Background Levels. In this group were found the areas outside the graphite electrode-producing industry (Hiirnosand), the steel plant (Oxelosund), and the large city (Stockholm). As mentioned above the flux for Stockholm was calculated according to a somewhat different method described below (eq 5). The reason for this was that the area is a complex estaurine-like archipelago area whereas the others are comparatively open coastal areas or a short bay (the inner part outside Iggesund). The emission source influence area of Hiirnosand was found to be around 10 km2. It is worth pointing out that the graphite electrode-producing industry should release its significant emissions through a 90-m-high chimney; still the closest area is rather affected. The steel plant at Oxelosund might have a slightly more far reaching influence than what has been shown by the present data, but due to the loss of sediment traps a t Oxelosund the investigated area was too small to detect any return to background flux levels. SBRs were in the range of 1500-3000%. The estimated flux in the inner 10 km2 of the steel plant and the graphite electrode industry were both in the same order of magnitude (0.3 and 0.2 ton y-l) as the estimated emission source flux from Stockholm city for the corresponding area (0.15 ton y-l; further discussed below). (GroupC) Drastically Elevated PAH15Flux Levels and Long-Range Influence. This category consisted of only the water area outside the aluminum smelter at Sundsvall with an emission source to background relation of 5600% (560 kg y-l). The influence was far-reaching and no evident return to background levels could be seen within the first 50 km2 (inset in Figure 2). This industry has outlets both to air and water which, together with the assumed high PAH emissions, lead to a far-reaching particulate load. The aluminum smelter was found to be a major emission source of PAH, and it is worth noting that the flux (including background flux) in the 30 km2closest
to the source was more than 1ton of PAH16 y-'. In general, the estimated direct influence of all the industrial emission sources appeared to be limited to the inner 10-50 km2. Also, from the background flux values presented in Table II,the totalannual flux of PAH15on the whole Baltic (375000 km2)was estimated. The extrapolation resulted in a total gross flux in the order of 400 ton y-l. This figure can be compared with the fluxes given above from outside the various emission sources and with the total annual flux from the Swedish Baltic coastal municipalities that was estimated to be 4.5 ton y-l (for a more detailed description, see the section headed Total Flux of PAH16from Swedish Baltic Coastal Municipalities Calculated from the Consumption of Fossil Fuel). The flux of total PCDD/F (Table I1 and Figure 3) and PCDD/F expressed as toxic equivalents, TEQ (Figure 4), has been calculated by the same method as described above. The subareas correspondingto sediment traps are, however, not the same as for PAH16since not all traps at each emission source were analyzed for PCDD/F. New subareas were calculated with the same method as used for the PAHs, and the whole area was delimited by the outermost sediment traps analyzed for PCDD/F. In this study, no PCDD/F analyses were done on the samples from the sediment traps outside the oil refinery at Nynashamn or on the samples from Stockholm. There was no evident trend of return to background levels within the 10 first km2 for any of the investigated emission source areas, but two main groups could be identified on a flux-level basis: (Group A) Low or moderate influence compared to background flux was found outside the metal smelter (Skelleftehamn), the graphite electrode industry (Hiirnosand), the aluminum smelter (Sundsvall), and the steel plant (Oxelosund), SBRs were in the range of 100-200% for both total PCDD/F and TEQ. In the nearest 10 km2 9.5, 7.3, 19, and 18 g PCDD/F y-l were deposited respectively, which is equivalent to 0.14, 0.05, 0.22, and 0.23 g of TEQ y-l, respectively. (Group B) Drastically Elevated Flux Levels. The emissions from the pulp and paper mill at Iggesund had a profound and far-reaching influence on the flux in the surrounding water area. It deposited 900% more than the background flux of total PCDD/F (68 g y-l was deposited in the nearest 10 km2) and 1100% more than the background flux of TEQ (1.1g y-l was deposited in the nearest 10 km2). No evident return to background levels could be seen in the nearest 30 km2 (insets in Figures 3 and 4). A possible cause of the far-reaching influence of the pulp and paper mill could be that the water from this industry contains high amounts of dissolved organic matter that might increase the water solubility of the emitted PCDD/Fs and thereby facilitate longer transport before settling. The total flux on the Baltic was estimated to approximately 300 kg y-l of total PCDD/F and about 4 kg y-l of TEQ using the data of the background flux (Table 11). This can be compared to the flux within the nearest 30 km2 at Iggesund which was found to be approximately 200 g y-' of total PCDD/F and 3 g y-l of TEQ. Concentrations and Chemical Profiles of PAHI, and PCDD/F. The PAH15chemical profiles of the various samples are illustrated in Figure 5, and the corresponding concentrations can be found in Table 111. All chemical profiles in the figure are normalized average values, and the concentration range in the table is given on an organic carbon weight basis. The profiles and the concentrations are based on samples from the four to nine Environ. Scl. Technol., Vol. 26, No. 7, 1992
1449
Figure 5. Normalized average PAH profiles based on samples from the sediment traps (n = 4-9) which were located closest to the emission sources respectively.
Table 111. Range of Gross Fluxes and Concentrations of Total PCDD/F, PCDD/F Expressed as TEQ, and PAHISat Various Emission Source Areas and Remote Background Areasa fluxes
source area Skelleftehamn Hiirnosand Sundsvall Iggesund Stockholm Nyniishamn Oxelosund backgrounds a
concentrations
total PCDD/F, pg cm-2 y-' rnax min
pg cm-2 y-' max min
140 53 200 1100
92 36 83 500
2.2 0.50 2.3 16
na na
na na
180
180
2.5
120
24
1.4
TEQ,
na na
1.3 0.4 1.4 9.9
PAHm ng y-l max min
2.3
140 2600 6100 220 2900 660 4000
90 160 600 130 130 70 1200
0.30
180
28
na na
total PCDD/F, ng g-' C max min
TEQ,
Pg g-I c -____ max
min
24 16 29 90
14 10 14 30
420 170 400 1300
200 150 190 590
na na
na na
na na
na na
20
18
250
11
9
120
PAHm rg g-l c max min
240
20 830 670 27 160 75 400
14 50 160 10 20 4 130
120
18
11
SamDles not analvzed are indicated bv na.
sediment traps which were located closest to the source in each water area (see Figure lB,C). A few PAHs, most frequently Flu, Pyr, or Chr/Tri (for abbreviations see Materials and Methods), predominate the PAH15profiles of the various emission sources. However, in the samples collected in the water area by the metal smelter in Skelleftehamn, Ind was found in the highest amounts. The high molecular weight compounds B[ghi]P and Cor were also found in comparatively high amounts in these samples. The PAH15 concentrations found in the SPM samples collected outside this industry were the lowest (14-20 mg/g C) among the source areas studied. This can be explained by the possible low or moderate PAH emissions discussed above in combination with a good water exchange in the area. The exposed positions of these sediment traps to the open sea are illustrated in Figure 1B. The average PAH15 profile of the samples collected outside the graphite electrode-producing industry (Harnosand) was dominated by Flu and Chr/Tri. B[a]P was also found in comparativelyhigh amounts outside this emission source. The samples from this area showed the highest maximum PAHIBconcentration (830mg/g C) of all areas of this investigation whereas the minimum concentration (50 mg/g C) was lower than the corresponding concentration at the aluminum smelter (Sundsvall) and the steel plant (Oxelosund) (discussed below). 1450
Envlron. Sci. Technol., Vol. 20, No. 7, 1992
Chr/Tri was also predominating in the chemical profile of the samples from the sediment traps outside the aluminum smelter (Sundsvall). In this area the second highest PAH15.maXim~mconcentrations (670 mg/g C) of the whole investigation was found. Further, the high PAH,, minimum concentration (160 mg/g C) found indicates that the local influence area of this industry is rather large despite the fact that the water area probably has a very good exchange with the open sea (Figure 1B). The PAH15 profile outside the pulp and paper mill (Iggesund) was dominated by Flu. As for the water area at Skelleftehamn (the metal smelter) the maximum concentration outside this industry was low (27 mg/g C) and comparable to the concentrations at the background areas (see Table 111). The PAH15 profile of the samples from the traps located in the estuarine-like archipelago outside the large city of Stockholm showed relatively high amounts of Pyr, Flu, and Chr/Tri. The maximum concentration in the area was lower (160 mg/g C) than outside the graphite electrodeproducing industry, the aluminum smelter, and the steel plant despite the size of the city and the high fossil fuel consumption in this area. In the sediment trap samples from outside the petroleum refinery (Nynlishamn), the same PAH compounds were found in the highest concentrations as outside Stockholm. The relative amounts of the higher molecular
Skelleftehamn
HB r ndsand
Sundsvall
lggesund
Oxelesund
Flgure 8. Normalized average PCDDlF congener profiles based on samples from those sediment traps (n = 2-4) which were located closest to the emission sources respectively.
weight PAHs, including B[a]P were, however, higher in the samples outside the refinery. The PAHI5concentrations found in this water area were fairly low (4-75 mg/g C). The PAH15profile of the sediment trap samples collected outside the steel plant at Oxelosund was dominated by Flu and Chr/Tri followed by Pyr and B[a]P, which occurred in relatively high amounts. The maximum PAHI6 concentration in this area (400 mg/g C) was the third highest in this study. The PCDD/F congener profiles which are shown in Figure 6 are also normalized average values, and the concentration ranges found in Table I11 are also given on an organic carbon weight basis. The congener profiles and the concentrations are based on samples taken from the two to four sediment traps which had been located closest to the sources respectively (see Figure 1B and C). In the congener profile of the metal smelter (Skelleftehamn) the OCDD was found in the highest concentrations, but the lower chlorinated dibenzofurans were also found in comparatively high amounts. The maximum PCDD/F concentration outside the metal smelter, expressed as TEQ (420 pg/g C), was slightly less than 4 times the corresponding concentrations at the background areas (120 pg/g C). The profiles representing the graphite electrode-producing industry in Hiirnosand and the aluminum smelter in Sundsvall showed a high degree of similarity. Both were dominated by the HpCDFs followed by the OCDD. The concentrations in both areas were relatively low even though the highest values at Sundsvall were clearly enhanced (400 pg TEQ/g C) in comparison to the background areas. The sediment trap samples from outside the pulp and paper mill (Iggesund) showed a quite different profiie with the HxCDDs as the major congener group. This was also the water area which exhibited the highest concentrations compared to the other emission source areas and the background areas, for both total PCDD/F and TEQ. The minimum concentration of total PCDD/Fs (30 ng/g C) outside the pulp and paper mill in Iggesund was, however, rather low compared to the background concentrations (9 ng/g C). This might be explained by a very high degree of water exchange with the open sea at the outer sediment trap stations and/or that a large portion of PCDD/Fs were associated with dissolved organic material which can be transported far from the source without settling. In the light of the enhanced PCDD/F concentrations outside the pulp and paper mill in Iggesund, the somewhat enhanced concentration of PCDD/F in Sundsvall might just as well be explained by the influence from the pulp and paper mills in the area as well as that of the aluminum smelter. However, the congener profiles of the two areas showed very little similarity. The city of Sundsvall itself probably also contributed to the enhanced PCDD/F concentration.
OCDD was found in highest concentrations in the samples from the steel plant in Oxelosund. As for the metal smelter water area in Skelleftehamn, the maximum total PCDD/F concentration and the concentration expressed in TEQ were just slightly enhanced (20 ng/g C and 250 pg/g C, respectively) compared to the background areas. No PCDD/F analyses have been made in this study on the sediment trap samples from the city of Stockholm or on the samples collected outside the petroleum refinery at Nynashamn. Multivariate Pattern Recognition Analysis on the Emission Source Samples In order to evaluate possible similarities and/or differences in the chemical patterns of the various sediment trap samples, a multivariate statistical method was used. The objective was to investigate the possibility of pinpointing the various emission sources, and if so which PAH compounds and PCDD/F isomers would be of greatest importance for this possible separation. The multivariate method used, called SIMCA (Soft Independent Modeling combines pattern recognition of Claw Analogy) (16,27,28), with principal component analysis and has been successfully used for classification problems in, for example, environmental chemistry (29,30).With a sufficient amount of variables it is possible to investigate similarities between objects (samples) even though a limited number of objects are present in the data set. The variables used in the SIMCA analysis of the PAH samples were the concentrations of all analyzed compounds plus the sum of all PAHs (i.e., 16 variables), and for the analysis of the PCDD/F samples the concentrations of all 2,3,7,8-substituted isomers plus the s u m of these, the eight congener group concentrations, the total sum of all PCDD/Fs, and the total sum of all TEQ concentrations (Le., 28 variables) were used. All analytical data used in the SIMCA analysis in this study were log transformed and scaled ( x + l),and some of the variables were also weighted. The PAH variables were weighted on a 3-deg scale. Maximum weights were given to all PAHs which separated clearly in all the GC analyses and to the sum of all PAHs. In some samples, B[ghi]F, B[a]P, Per, and Cor coeluted partly with other compounds and were thus given weights corresponding to two-thirds of the maximum weight. 2MePyr, lMePyr, and Cp[cd]P often partly coeluted with other compounds and were also found on, or close to, the detection limit. These were given one-third of the maximum weight. For the PCDD/F variables a 5-deg scale was used. The isomers or congener groups which had no 13C-labeled standard or which were frequently on, or close to, the detection limit were given a weight of one-fifth of the maximum weight. The rest of the isomer concentrations and congener group concentrations which were clearly above the detection limit were weighted to two-, three-, or four-fifths of the maximum weight, or to the maximum Environ. Sci. Technol., Vol. 26,
No. 7, 1992 1451
weight, depending on the recovery of the internal standard in their congener groups, respectively. Finally, the total concentration value of all 2,3,7,8-substituted isomers and the total concentration value on a TEQ basis were weighted to two-fifths, whereas the concentration value of total PCDD/Fs were weighted to three-fifths of the maximum weight. The PAH and PCDD/F variables were tested separately in the SIMCA analysis primarily because of the few PCDD/F analyses made relative to the number of PAH analyses. Further, the above-described flux and concentration data indicate that the emissions of PAHs and PCDD/Fs were not coupled together. For example, in the samples taken outside the pulp and paper mill at Iggesund, low PAH concentrations were found whereas the PCDD/F concentrations were high. A combined SIMCA analysis with the 16 PAH and the 28 PCDD/F variables gave a result which was more complex to interpret. At the first level of the SIMCA analysis, the objective was to get an overview of the whole data set, i.e., all samples (see Table I) which have been analyzed for their content of PAHs. (The PCDD/F samples are discussed later in the text). This was done to find possible object clustering and to identify potential outliers. In the SIMCA analysis this is made by constructing a multidimensional space (M-space)where each variable defies one coordinate axis, and consequently, each object (sample) will describe a point in this M-space. The first principal component (PC1) can be described as a vector located in M-space and pointing in the direction depicting the largest spread in the data. The second principal component (PC2) is located in an orthogonal position to PC1 and explains the largest spread in the data in this direction. The third (PC3) is then located in orthogonal position to PC1 and PC2, etc. This distances in M-space from each object to the PC models (the residuals) are also calculated. The standard deviations of these residuals are used later when estimating the tolerance intervals for the class models (see below). To visualize the localization of the objects in the multidimensional M-space, linear projections of the objects in a two-dimensional plane can be made. The projection of PC1 versus PC2 of the PAH data set showed that the various objects separated clearly with regard to emission source clustering. Overlaps were, however, registered, and some more or less obvious outliers were found. In general, the outliers were identified as the samples from the sediment traps that had been located at the longest and/or the shortest distance from the emission sources, respectively. Two principal components were significant, and PC1 explained 94% of the variance in the data set and PC2 a little more than 2%. At this first level it was also possible to calculate the loading of each variable, i.e., the PAHs that contributed most to the separation of the objects in PC1 and PC2. In this case Flu, Pyr, B[a]A, Chr/Tri, Ind, and total PAH were the most important variables for the separation. Primarily B[a]A and Chr/Tri but also Flu and total PAH were the most important variables for the separation of the aluminum smelter (Sundsvall) and the steel plant samples (Oxelosund) from the other samples. Pyr was important for the separation of the petroleum refinery samples (Nynashamn) and Ind for the separation of the metal smelter samples (Skelleftehamn) from the other samples, respectively. A t the second level of the SIMCA analysis, the differences or similarities between chosen groups of objects can be tested. For the PAHs this was done by making separate class models of the sediment trap samples from the various 1452
Environ. Sci. Technol., Vol. 26, No. 7, 1992
emission source areas. Seven models were developed, each one describing an area outside a specific emission source. The samples from the sediment traps which were located at the longest distance from the emission sources, respectively (often an outlier in the above-discussed PC1/ PC2 plot), were used as test objects to examine the range of the local influence area. Unfortunately, this resulted in four of the emission source class models being based only on four objecta, which is rather few for this type of analysis. As additional background samples in the analyses, the six samples from the two background areas, listed in Table I, were also used as test object. Two class models at a time were used to classify all other test objects, i.e., all background samples and all other samples from the rest of the emissions source classes. The results are illustrated in the Coomans plots in Figure 7. The Coomans plots graphically demonstrate if the test objects are significantly described by any of the two emission source class models or if they belonged to neither of them. The tolerance intervals were calculated from the residual variances of the class models by using an F-test (16). The lines in the plots of Figure 7 are equivalent to 95% confidence limits. In Coomans plot 1 (Figure 7 ) the belongings of all remaining PAH test objects were examined against one class model made of the samples from the petroleum refinery (class model 1; Nynbhamn) and one class model made of the samples from the large city (class model 2; Stockholm). The result showed that several of the test objects did not significantly differ from the petroleum refinery model, and these were two background samples of the refinery, two background samples of the large city, and two samples from the pulp and paper mill. Also, some of the objects of class model 2 were not separable from class model 1. Several other samples were also close to class model 1,for example, one of the background samples from the middle Baltic proper, one of the background samples from the Bothnian Sea, and one of the background samples from the metal smelter. All test objects were, however, significantly separated from class model 2, i.e., the sediment trap samples from outside the city of Stockholm. One background sample from Stockholm was, however, close to the model. The sample from the sediment trap station located nearest to the graphite electrode-producing industry showed the least similarity to class model 1, and the sample collected nearest the metal smelter showed the least similarity to class model 2. Coomans plot 2 (Figure 7 ) illustrates that all the test objects were significantly separated from both class model 1 (the pulp and paper mill; Iggesund) and class model 2 (the metal smelter; Skelleftehamn). A cluster of samples were located relatively close to these class models however, including the background sediment trap station of the class model areas, the samples from the middle Baltic proper background area, the samples from the Bothnian Sea background area, and some of the background sediment trap stations of the large city. The least similarity to the two class models was shown by three sediment trap stations at the aluminum smelter and the stations located closest to the graphite electrode-producing industry and the steel plant. Class model 1 in Coomans plot 3 was made up of the four sediment trap samples which were located closest to the graphite electrode-producing industry (Harnosand), and class model 2 was made up of the four samples which were located closest to the steel plant (Oxelosund). Also in this Coomans plot, all test objects were significantly separated from the two class models. The object which
.
ii
2
'
'
1
*
' :' A A( A + A 0
B O
'
0.
A
8
A
1' +-
+
6' +
+
b x
+
o
' *'
.
8 .
ff
n
'' *
X
V
*TT
Vv
Distance from class model 1 Melal Smelter etc. OBackground . I I .
+Pulp and Paper M i l l etc. X Background (3
+Graphite Electrode Prod. Ind. etc. *Background II
0 Large City elc. OBackground - I$
8
. .
Aluminium Smelter etc. Background 31
. .
. .
A Petroleum
.
Refinery elc.
ABackground
. . (3
s p (1)
V S t e e l Plant etc. VBackground . I,
.
Background Archipelago Area in lhe Middle Baltic Proper
Background Archipelago Area In the Bothnian Sea
Figure 7. Six Coomans plots of the PAH (plots 1-4) and PCDD/F (plots 5-6) data sets. The axes represent the distances from the class models respectively, and the thin lines around the class models indicate the tolerance level at 5 % probability (p = 0.05) around the classes.
was located closest to class model 1 was the background sample from the same water area. The same was valid for class model 2, i.e., the object located closest was the background sample from the steel plant. One sample from the metal smelter and one from the petroleum refinery showed the least similarity to the two class models in Coomans plot 3. The class model of the remaining emission source sample, the aluminum smelter (Sundsvall), is shown in Coomans plot 4 in Figure 7. All test objects were located at relatively large distances from the significance limit of this model, and the object which was located closest was the background sample from the aluminum smelter. The
samples which showed the highest degree of dissimilarity were the ones from the petroleum refinery. The first level SIMCA analysis which was done on the PAH data was also done on all PCDD/F analyses listed in Table I. The purpose here was also to get an overview of the total PCDD/F data set. Four principal components were significant and had a satisfyingly high degree of variance explanation. PC1 explained 76% of the variance in the data set, PC2 7%, PC3 4%, and PC4 a little less than 2%. The various objects in the PC1 versus PC2 plot separated also in a emission sourcewise clustering. The three samples which had been collected closest to the pulp and Envlron. Sci. Technol., Vol. 26, No. 7, 1992
1453
.
ng cm-2. y” 3500
3000
2500
2000
1500
1000
500
0
Figure 8. Flux of PAH,, in the Stockholm archipelago. The map section of the figure indicates the distances of the sediment trap locations to the center of Stockholm city. The numeric distance to the city center of each sediment trap is Indicated by arrows to the corresponding area 0. The delimitation of the areas (Ql-9) discussed in the text are indicated by the concentric rings and the angle C#J. The estimated point source induced flux for each sediment trap and year is indicated by symbols.
paper mill separated especially well from the others. The variable loading plot showed that primarily 1,2,3,6,7,8HxCDD and 2,3,4,8-/2,3,7,8-TCDF accounted for most of the separation of the pulp and paper mill samples. 1,2,3,7,8,9-HxCDD, total HxCDD, 1,2,3,4,7,9/1,2,3,4,7,8HxCDF, and 2,3,7,8-TCDD also had a relatively high separation capacity; especially the last two compounds mentioned for the separation of the metal smelter and the aluminum smelter from the others. A t the second level of the SIMCA analysis on the PCDD/F samples, the differences between the samples outside the various emission sources were tested by making separate class models of the emission source sediment trap samples. Because of the relatively few analyses results, compared to the PAHs (see Table I), only four emission source class models could be made. The first two of these were the aluminum smelter samples (Sundsvall; class model 1in plot 5 in Figure 7) and the steel plant samples (Oxelosund; class model 2 in plot 5 in Figure 7). In this plot all test objects differed significantly from the two claw models, with two pulp and paper mill samples showing the highest discrepancy. The background sample of the steel plant was located closest to the steel plant class model. In the last two class models constructed (the pulp and paper mill in Iggesund and the metal smelter in Skelleftehamn), the background samples outside the respective emission sources were included so that each class model was based on a minimum of four objects. Nevertheless, all test objects in Coomans plot 6 (in Figure 7) differed significantly from both class model 1, which was constructed from the pulp and paper mill samples (Iggesund), and class model 2, which was the metal smelter samples (Skelleftehamn).
Total Flux of PAH,, from Swedish Baltic Coastal Municipalities Calculated from the Consumption of Fossil Fuel Parallel to the measured fluxes of PAH15in the water areas outside the different industrial emission sources, the correspondingfluxes outside all Swedish municipalities on the Baltic coast were also calculated based on their con1454
Environ. Sci. Technol., Voi. 26, No. 7, 1992
sumption of fossil fuels. For these calculations it was assumed that the flux of PAH15measured in the Stockholm archipelago was a result of the fossil fuel combustion of various sources within the Stockholm area. A short description of the possible sources within the Stockholm area is given as a background to the following calculations. Stockholm is a large nonindustrialized city inhabited by approximately 1.5 million people. The import of petroleum products is about 2.5 million m3 per year and there are about 0.6 million gasoline-powered automobiles and 20 thousand lighbduty diesel-powered trucks and automobiles in the area. Diesel oil is the main energy source for domestic heating, and there is one large municipal incinerator in the region. The ferry lines from Stockholm to Finland are intensively trafficked, and the archipelago area is also utilized by more than 150 thousand leisure boats during the summer. Most of the small industries and all households are connected to three-step municipal wastewater treatment. As a basis for the flux calculation in the Stockholm archipelago, the data from the study by Broman et al. (18) and Brunstrom et al. (31)was used, i.e., the same data referred to and discussed in the previous sections (data from ref 31 only the outermost station). The number of traps used in the calculations and their location in the Stockholm recipient can be found in Table I and Figure 1C. The distances from the Stockholm city center to the different sediment traps are shown as concentric rings in Figure 8. The fluxes of PAH15were found to decrease in an exponential manner, and the flux at each trap location is indicated by symbols in the left part of Figure 8. In the waters directly outside Stockholm ( Q = 1)the measured flux was 2700 ng cm-2 y-’, which decreased to a background level of approximately 100 ng cm-2y-l at 30 km distance ( Q = 7-9). The flux appeared to be fairly unaffected by increasing distance at distances over 30 km. The observed flux is therefore assumed to consist of two components, an emission source induced flux and a background flux component. This can be expressed formally by a twocomponent theoretical flux function (1) where F&r) is the flux per unit area at the distance r to central Stockholm.
Table IV. Estimated Source and Background Fluxes in Archipelago of Stockholm" 3
1 2 3 4 5
6 7 8 total
r, km
to
surface, km2
1 6.2 12 17 22.5 27 33.5 38
0.37 0.20 0.19 0.23 0.24 0.20 0.45 0.46
0.6 12.3 32.9 54.7 85.5 73.1 290.3 242.8 790
F k ) , kg km-2 y-l
PD", kg y-'
BD", kg y-l
TD,, kg y-'
PD"/TD,, %
PD"/TD,, %
17.5 154.9 117.6 52.2 22.6 5.5 5.9 1.2 380
0.6 12.2 32.6 54.3 84.9 72.5 288.1 240.9 790
18.1 167.1 150.2 106.5 107.5 78.0 293.9 242.1 1200
97 93 78 49 21 7 2 0.5 32
3 7 22 51 79 93 98 99 68
27.4 8.3 2.7 1.5 1.1 1.0 1.0 1.0
The areas 31-8 refer to the areas indicated in Figure 8. For symbols in table head see Glossary.
The term ae-8' represents the emission source-induced component of the flux at the distance r from Stockholm city ( a and p are constants). B is the background flux in the Baltic estimated by the asymptote (B) of the regression (eq 1). The function1, indicated in Figure 8, has been fitted by a nonlogarithmic iterative least-squares method ( a = 3389 ng cm-2y-l, /3 = 0.248, B = 99.2 ng cm-2 y-l, 9 = 0.99).
+
F,(r) = a e-@' B
(1)
The estimated asymptotic value of background flux (B) agrees very well with the results from the previous section, which discussed the industrial source emissions, where the average background flux at the reference stations was estimated to 100 ng cm-2 y-l (Table 11). If the fraction of the flux that is locally induced is assumed to decrease in a symmetrically concentric manner around the emission source (Stockholm), the flux at any point in that plane can be predicted by eq 1 if r is substituted by x 2 + y2 where x and y are coordinates in a two-dimensional system of coordinates with origin in the center of Stockholm. The rectangular coordinates x and y can be substituted by the polar co-ordinates r and 6' (where x = r cos(6')and y = r sin@)). In the areas Q (Figure 8)) limited by two concentric rings a t inner and outer distances, ria, ron,and the angles 0 and 4, the proportion of water to total surface is en. If a and B are expressed in, for example, kilogram kilometer-2 year-l, and the water surface is assumed to be evenly distributed in Q along the gradient, the emission source-induced flux in Q can be estimated by the integral2 and the background flux by3
+
@ria l)e-orin -
+ l)e-@On
P2 BDn =
ea
jr0'&'rB d6' dr = ea4B[rOn2- ria2/2]
(3)
rill
The total flux was calculated for Q1-8 each (Table IV) according to eqs 1and 2. The total flux of PAH15in the archipelago can then be calculated by summing the Qs (eq 4). E
TD, = C (PDn + BDn) 0=1
(4)
TD, is equivalent to the surface-corrected rotation volume of eq 1 around a vertical axis through the city center of Stockholm. The rotation is limited by the areas Q1-8. Area 9 has been excluded, since according to eq 2 more than 99% of the total emission source flux occurs within a radius of 27 km from the city center at the found
values of (Y and 8. The rapidly decreasing proportion of locally-induced flux (PDn/TD,) is indicated in the last two columns of Table IV. The emission source flux of PAH15 in the inner 38 km of the Stockholm archipelago was by eq 2 estimated to be approximately 400 kg y-l, and the background flux in the same area was estimated by eq 3 to be approximately 800 kg y-l (Table IV). The total flux of PAH15 in the areas 01-8 was thus approximately 1.2 ton y-1. The estimate of total flux in the Stockholm archipelago has then been used to estimate the order of magnitude of the locally-induced flux caused by all Swedish Baltic coast municipalities. Approximately 2.8 million people live in the 30 Swedish Baltic coast municipalities with more than 10000 inhabitants. The total municipal and industrial consumption of fossil fuels in these municipalities (32,33)in 1985 was (on a weight basis) approximately 6.3 million ton y-l, of which approximately 70 ?% was gasoline and light-weight diesel oils and approximately 30% was coal. If it is assumed that the PAH15flux induced by diffuse nonpoint sources as cities is qualitatively similar and quantitatively proportional to the combustion of fossil fuels in the cities, an approximate estimate of the total local flux induced by Swedish east-coast municipalities can be calculated by eq 5. TPD,, = PD,
E, + c=l CE,
lmlr e-@' o o
ra
d6' dr
(5)
Ec denotes the consumption of fossil fuels in the municipality C, and E8 is the corresponding value for Stockholm. For the Stockholm area the calculated flux value, PDs (Table IV) was used. The deposition area for the other municipalities in eq 5 was assumed to be an open half circle, and no correction for water to total surface was made. This leads to a slight overestimation of total flux. The differences in emissions per unit of fuel that is caused by different combustion techniques or types of fuel are not accounted for. The calculation (eq 5) is based on the flux caused by the emissions from the city of Stockholm. In the summed combustion of the smaller Baltic coastal municipalities, a larger relative proportion of the combustion occurs in various industrial facilities and central municipal heating. The large-scale combustion facilities have considerably higher efficiency than the smaller ones which can lead to an overestimate of the total flux induced by the coastal municipalities. In spite of the fact that the legislation in Sweden promotes a fairly uniform standard of combustion facilities, such differences in combustion can cause insecurity in the estimate (eq 5). However, the insecurity in the different emissions caused by the composition of various fuels, e.g., oil or coal, are also of significance. Environ. Sci. Technol., Vol. 26, No. 7, 1992
1455
According to the calculations above, the total emission source flux (not including the background flux) of PAHI5 from Swedish east-coast municipalities (TPDsw) was calculated (eq 5) to be 4.5 ton y-l, of which the Stockholm area contributes with 0.4 ton y-l (PDs). As mentioned above, the total background flux of PAH15on the Baltic (375000 km2) can be estimated to be in the order of 400 ton y-l using the approximate value of the background flux (B). The total emission source flux of PAHIS thereby amounts to approximately 1% of the background (4.5 of 400 ton y-9. The method used to calculate the relative importance of locally-induced flux and diffuse background flux thus indicates that diffuse background flux is of capital importance for the total transport of PAH15to the Baltic. The result that the local flux was small in relation to the total flux of the Baltic cannot, however, be interpreted as a measure of the total importance of the coastal municipalities in the transport of €A ' H6l to the Baltic. The importance of the contribution to the diffuse background flux from these local sources in relation to other more distant sources remains unknown. It must be emphasized that the calculations (eqs 1-5) are attempts to estimate orders of magnitude rather than attaining exact quantitative measurements. At the emission sources of Oxelosund (steel plant) and Hiirnosand (graphite electrode-producing industry), the total fluxes of PAH15in the 30 km2 closest to the source were found to be in the same order of magnitude as the largest city in Sweden (Stockholm). The aluminum smelter of Sundsvall alone caused a flux of PAH15 in the 30 km2 closest to the emission source that was approximately 4 times as large as that of Stockholm. All the industrial emission sources studied caused a flux of PAH15, in the 30 km2 (including background flux) closest to the source which was equivalent to one-fourth of the total estimated locally induced flux from the Swedish east-coast municipalities. Acknowledgments
We are most grateful to Ron Johnstone for reviewing the manuscript.
B
P
estimated flux in Stockholm city center, kg km-2 Y-l nonpoint source background flux, kg km+ y-l exponential constant of flux decrease with increasing r total background flux in area 9, kg y-' consumption fossil fuels in coastal city C (industrial municipal), ton y-l consumption fossil fuels in Stockholm (industrial municipal), ton y-l proportion of water to land surface in the area 9 limited by ria,rea, and the angles 0 and 4 angle of coastline in the archipelago of Stockholm, 3.23 radians flux at distance r to Stockholm city center, ng cm-' y-l or kg km-2 y-l no. of coastal municipalities (>loo00 inhabitants) along the Swedish east coast (29) total source induced flux in area 9 , kg y-l distance to Stockholm city center, km inner limit of integral in 9, km outer limit of integral in Q, km total flux in Stockholm archipelago areas (1-8), kg
+ +
-
m
Y-l
total source induced flux from Swedish coastal municipalities, kg y-' 1456
Literature Cited (1) Mix, M. M. Mar. Environ. Res. 1986,20 (1 and 2), 141 pp. (2) Varanasi, U., Ed.; Metabolism of Polycyclic Aromatic Hydrocarbons in the Aquatic Environment; CRC Press,
Inc.: Boca Raton, FL, 1989; p 341. (3) Keith, L. H.; Telliard, W. A. Environ. Sci. Technol. 1979, 13, 416-423. (4) Poland, A.; Greenlee,W. F.; Kende, A. S. Proc. Ann. N.Y. Acad. Sci. 1979,320, 214-230. (5) Bjorseth, A., Ramdahl, T., Eds.; Handbook of Polycyclic Aromatic Hydrocarbons; Marcel Dekker Inc.: New York and Basel, 1985; Vol. 11, p 407. (6) Lustenhouwer, J. W. A.; Olie, K.; Hutzinger, 0. Chemo-
sphere 1980,9,501-522. (7) Marklund, S. Dioxin emissions and environmentalimissions. A study of polychlorinated dibenzodioxins and dibenzofurans in combustion proteases. Thesis,University of Umei, Sweden, 1990, p 63. (8) Swanson, S. E. Dioxin in the bleach plant. Thesis, University of Umei, Sweden, 1988, p 401. (9) Bidleman, T. F. Environ. Sci. Technol. 1988,22,361-367. (10) Broman, D.; Niif, C.; Wik, M.; Renberg, I. Chemosphere 1990,21, 69-77. (11) Baker, J. E.; Eisenreich, S. J. Environ. Sci. Technol. 1990, 24,342-352. (12) Baker, J. E.; Eisenreich, S. J.; Eadie, B. J. Environ. Sci. Techno!. 1991,25,500-509. (13) Broman, D.; Niif, C. Dynamics and distributionof polycyclic
aromatic hydrocarbons (PAHs) and polychlorinated dibenzo-p-dioxins and dibenzofurans (PCDD/Fs) in the aquatic environment-A review with focus on the Baltic Sea. Submitted for publication in Ambio. (14) Broman, D.; Niif, C.; Rolff, C.; Zebiihr, Y. Environ. Sci. Technol. 1991,25,1850-1864. (15) Baker, J. E.; Eisenreich, S. J.; Johnson, T. C.; Halfman, B. M. Enuiron. Sci. Technol. 1985, 19, 854-861. (16) Wold, S.; Albano, C.; Dunn, W. J., 111; Esbensen, K.;
Hellberg,S.; Johansson,E.; SjGstrom, M. In Food Research and Data Analysis; Martens, H., Russwurm, H., Jr., Eds.;
Glossary CY
area limited by ria,rOaand the angles 0 and 4, km2 Registry No. Flu, 206-44-0;Pyr, 129-00-0;2mePyr, 3442-78-2; ImePyr, 2381-21-7;B(ghi)F, 203-12-3; B(a)A, 56-55-3;Chr, 21801-9; Tri, 217-59-4; B(k)F, 207-08-9;B(b)P, 192-97-2;B(a)P, 5032-8; Per, 198-55-0; Ind, 193-39-5;B(ghi)P, 191-242; Cor, 191-07-1; TCDF, 30402-14-3; PnCDF, 30402-15-4; HxCDF, 55684-94-1; HpCDF, 38998-75-3; OCDF, 39001-02-0; TCDD, 41903-57-5; PnCDD, 36088-22-9;H,CDD, 34465-46-8;HpCDD, 37871-00-4; OCDD, 3268-87-9. 9
Environ. Sci. Technol., Vol. 26, No. 7, 1992
Applied Science: New York, 1983; pp 111-147. (17) Broman, D.; Kugelberg, J.; Niif, C. Estuarine, Coastal Shelf Sci. 1990, 30, 429-436. (18) Broman, D.; Colmsjo, A.; Ganning, B.; Niif, C.; Zebuhr, Y. Environ. Sci. Technol. 1988, 22, 1219-1228. (19) Lamparski, L. L.; Nestrick, T. J. Chemosphere 1989,19, 27-31. (20) Zebiihr, Y.; Niir, C.; Broman, D.; Lexh, K.; Colmsjo, A,; Ostman, C. Chemosphere 1989,19,39-44. (21) Niif, C.; Broman, D.; Brunstrom, B. Distribution and metabolism of polycyclic aromatic hydrocarbons (PAHs) inje&d into eggs of chicken (Gallus domesticus) and common
eider duck (Somateria mollissima). Environ. Toxicol. Chem., in press. (22) Smith, L. M.; Stalling, D. L.; Johnson, J. L. Anal. Chem.
1984,56,1830-1842. (23) Ahlborg, U. G. Chemosphere 1989, 19, 603-608. (24) Hassett, J. P.; Anderson, M. A. Environ. Sci. Technol. 1979, 13, 1526-1529. (25) Webster, G. R. B.; Muldrew, D. H.; Graham, J. J.; Sarna, L. P.; Muir, D. C. G. Chemosphere 1986,15, 1379-1386. (26) Elmgren, R. Rapp. P.-v. RBun. Cons. int. Explor. Mer 1984, 183,152-169. (27) Wold, S.;Albano, C.; Dunn, W. J., III; Edlund, U.; Esbensen,
K.; Geladi, P.; Hellberg, S.; Johansson, E.; Lindberg, W.; Sjhtrom, M. In Proceedings of the NATO Advanced Study
Environ. Sci. Technol. 1902,26,1457-1460
Institute on Chemometrics; Cosenza, Italy; Kowalski, B. R., Ed.; Reidel: Dordrecht, 1983; pp 17-97. Wold, S.;Esbensen, K.; Geladi, P. Chemom. Zntell. Lab. Syst. 1987,2, 37-52. Dunn, W. J., 111; Stalling, D. L.; Schwartz, T. R.; Hogan, J. W.; Petty, J. D.; Johansson, E.; Wold, S. Anal. Chem. 1984,56, 1308-1313. Stalling, D. L.; Peterman, P. H.; Smith, L. M.; Norstrom, R. J.; Simon, M.Chemosphere 1986,15, 1435-1443. Brunstrom, B.; Broman, D.; Dencker, L.; Nhf, C.; Vejlens, E.; Zebuhr, Y. Extracts from settling particulate matter collected in the Stockholm archipelago waters: Embryolethality, immunotoxicity, and EROD-inducing potency of fractions containing aliphatica/monoaromatics,diaromatics, or polyaromatics. Environ. Toxicol. Chem., in press.
(32) Deliveries and consumption of gasoline and fuel oil divided into municipalities in 1985. Statistiska meddelanden; E 13 SM 8601; Statistiska Centralbyrln, Enheten for industristatistik, Distributionen, s-701 89 Orebro, Sweden, 1986. (33) Industrial statistics. Consumption of purchased energy in 1985. Data by regions and branch. Statistiska meddelanden; E 14 SM 8701; Statistiska Centralbyrh, Enheten for industristatistik, Distributionen, S-701 89 Orebro, Sweden, 1987.
Received for review September 3,1991. Accepted February 25, 1992. The study was financially supported by the Swedish Environmental Protection Agency.
Methylmercury Determination as Volatlle Methylmercury Hydride by Purge and Trap Gas Chromatography in Line with Fourier Transform Infrared Spectroscopy Marco Filippeiii,t Franco Baldi,* Frederick E. Brlnckman,* and Gregory J. Oison*v*sli Laboratorio Chimico d’Igiene e Profilassi, 1-19100 La Spezla, Italy, Dipartlmento di Biologia Ambientale, Universita di Slena, via P.A. Mattioli 4, 1-53100 Siena, Italy, and National Institute of Standards and Technology, Gaithersburg, Maryland 20899
H We report a novel aqueous derivatization of methylmercury chloride by N&H4 to methylmercury hydride (CH,HgH), a volatile and unexpectedly stable species which we determined has a half-life of approximately 2 h. The analytical apparatus consisted of a purge and trap (PT) unit linked to a gas chromatograph (GC), in line with a Fourier transform infrared spectrometer (FTIR). The sample was purged with nitrogen, and volatile compounds were concentrated in a cold trap. The trap was heated, and the methylmercury hydride was separated from other gaseous compounds by a wide-bore GC column linked to the FTIR spectrometer. The detection limit of the method was 0.15 pg and was linear up to 40 pg of MeHgH. The concentration of methylmercury in a reference tuna fish as determined by PT/GC/FTIR was the same as that determined by conventional analyses in an 11-laboratory intercalibration. Analytical methods employed previously have not permitted identification of the methylmercury hydride species. Our results also suggest the need to investigate further the occurrence in nature of this highly volatile and unexpectedly stable species, especially in view of previous reports of biohydridization of other metal and metalloid species. Introduction
The biogeochemical cycle of mercury has been widely studied. Mercury is reduced, oxidized, methylated, and demethylated by a number of biological and abiotic processes. Monomethylmercury is of special concern because of ita enhanced toxicity, lipophilicity, bioaccumulation, and volatility compared to inorganic mercuric ions. Inorganic mercury in seawater occurs as chloro complexes with HgC142-as the dominant species (1). The hydride species, MeHgH, is commonly believed to be unstable in water, decomposing rapidly to elemental mercury and methane. ~~
t Laboratorio Chimico
* Universita di Siena.
~~
d’Igiene e Profilassi.
*NationalInstitute of Standards and Technology. It Present address: Pittsburgh Energy Technology Center, P.O. Box 10940, Pittsburgh, PA 15236. 0013-936X/92/0926-1457$03.00/0
Many of the methods for the ultratrace detection of methylmercury in a variety of matrices are not capable of determining the true speciation of the molecule. For example, the measurement of methylmercury by the “classical” Westoo method (2) involves extraction of methylmercury species from water with toluene and subsequent analysis by thin-layer chromatography, gas chromatography with electron capture detection, or atomic absorption determination of total mercury. These analyses do not provide direct speciation and are subject to positive interferences (3-5). Other organomercurials such as dimethylmercury can be produced naturally in the environment (6-8). However, with conventional analytical methods, dimethylmercury is transformed to methylmercury species by addition of 0.1 N HC1 in the f i s t step of extraction (9). Jackson et al. (10)suggested the use of purge and trap (PT) sampling in line with a gas chromatograph (GC) equipped with a flame photometer detector for organotin speciation in Chesapeake Bay waters. The addition of NaBH4 to the sample resulted in formation of volatile organotin hydrides. However, methylmercury is generally thought to be reduced by N&H4 to elemental mercury (11). Nonetheless, we previously detected formation of a volatile organic mercury species upon treatment of an aqueous solution of methylmercuric chloride with N&H4 (12).
The aim of this paper was to identify the volatile Hg derivative formed by the reaction of methylmercury chloride with aqueous solutions of NaBH, and to determine if this novel derivatization technique could be used with a GC coupled with Fourier transform infrared spectrometer (FTIR) detection to provide a new method for reliable speciation of methylmercury. We identified the volatile Hg species as MeHgH by PT/GC/FTIR and GC/mass spectroscopy (MS) analysis and isotopically confirmed these results with sodium borodeuteride. Experimental Section
Apparatus. The GC/FTIR system was described previously (13). A 50-m-length X 0.53-mrn4.d. wide-bore
0 1992 American Chemical Society
Environ. Scl. Technol., Vol. 26, No. 7, 1992 1457
