Environ. Sci. Technol. 2005, 39, 2711-2718
Five-Stage Environmental Exposure Assessment Strategy for Mixtures: Gasoline as a Case Study K A R E N L . F O S T E R , † D O N M A C K A Y , * ,† THOMAS F. PARKERTON,‡ EVA WEBSTER,† AND LYNNE MILFORD† Canadian Environmental Modelling Centre, Trent University, 1600 West Bank Drive, Peterborough, ON, Canada, K9J 7B8, and Toxicological and Environmental Sciences Division, ExxonMobil Biomedical Sciences Inc., 1545 Route 22, East Annandale, New Jersey 08801-0971
A five-stage strategy is suggested for conducting an exposure assessment of mixtures that may contain numerous chemical components. The stages are: (1) determination of mixture composition and variability, (2) selection of component groups within the mixture and documentation of criteria used for this selection, (3) compilation of relevant property data for each group, (4) assessment of environmental fate of each group, and (5) assessment of environmental and human exposure to each group and to the mixture as a whole. A subsequent step is the assessment of environmental and/or human risk associated with the individual and aggregate exposure to each group. The approach is illustrated by application to gasoline, which is treated as 24 component groups or hydrocarbon blocks. Focusing on stages 2-4, the illustration shows that the groups display widely different environmental fates as a result of their different physicochemical properties, degradation half-lives, and mode-of-entry into the environment. As a result, the relative proportions of groups in each environmental medium (such as air and water) differ greatly from that of the original mixture. It is thus important to treat gasoline and similar mixtures as a number of component groups instead of as a single substance. A generic procedure is suggested in which the model is run for unit emissions of each component group to air, water, and soil. These results are compiled into matrices that can then be conveniently scaled to actual emission rates without rerunning the model. Methods for determining subsequent exposure and risk are also briefly outlined.
Introduction There is a recognized need to categorize and appropriately regulate all chemicals of commerce with respect to human and environmental risk. Initiatives have been taken internationally, such as those of Canada (Canadian Environmental Protection Act 1999, Sections 73 and 74) and the European Union (Existing Chemicals Regulation 793/93). Risk assessments conducted for these and other initiatives are commonly done for individual chemicals, for example, acetaldehyde (1) and various hydrocarbons (2). In many cases, however, chemicals are introduced into the environment as mixtures, * Corresponding author phone: (705)748-1011 x1489; fax: (705)748-1080; e-mail:
[email protected]. † Trent University. ‡ ExxonMobil Biomedical Sciences Inc. 10.1021/es048734p CCC: $30.25 Published on Web 02/26/2005
2005 American Chemical Society
examples being PCBs, dioxins, toxaphenes, and petroleum distillates such as gasoline. Regulatory and scientific communities differ on the nomenclature used to describe mixtures. Regulatory agencies often consider a “mixture” to be a deliberate preparation of two or more ingredient chemicals. Substances such as those above that are derived from a common source are termed “complex substances”. In the scientific arena, both complex substances and preparations are usually referred to as “mixtures”, while a high molar mass compound such as a protein may be regarded as a “complex substance”. We use the scientific terminology here. The task of performing risk assessments on mixtures is daunting because they may contain hundreds of different and unique chemicals that cannot practically be assessed individually. Further, the physicochemical properties of these component chemicals and their relative proportions in the mixture may not be adequately known. This task can be simplified by grouping chemical components of the mixture into a manageable number of “pseudo-components” (3-8), “hydrocarbon blocks” (9), “fractions” (10), or “groups”, a practice that has been widely applied in oil spill modeling. For example, PCBs, which have 209 possible congeners and were formulated for use as commercial mixtures (Aroclors in North America), have been treated as homologue groups based on the number of chlorine atoms present in the compound, for example, mono, di, tri, etc. (11-14). In this study, we describe and illustrate a systematic strategy that can be used to assess the exposure associated with the environmental release of a mixture. Five stages are suggested: (1) determination of mixture composition and variability, (2) selection of component groups within the mixture and documentation of criteria used for this selection, (3) compilation of relevant property data for each group, (4) assessment of environmental fate of each group, and (5) assessment of environmental and human exposure to each group and to the mixture as a whole. This exposure information can reveal which component groups cause the highest exposure and, with appropriate toxicological information, can form the basis of an assessment of the resulting environmental and/or human risk associated with each group and the mixture as a whole. Gasoline is used to illustrate this strategy, with the primary focus being on the second, third, and fourth stages. The use of unit emission-model response matrices to facilitate both present and future exposure and risk assessments is also illustrated by a simple example. The first, fifth, and final risk stages are briefly discussed.
1. Determination of Mixture Composition The composition of the mixture and associated variability must be established first because it is the foundation for subsequent stages in the assessment. The composition may be variable spatially (e.g., country to country) and temporally (e.g., seasonally). In the case of gasoline and other petroleum fuels, potential sources of variability include the season of use, the crude oil source and refining practices, as well as local or regional regulatory specifications that dictate gasoline quality. Comprehensive analyses by gas chromatography were obtained for over 200 hydrocarbons present in a survey of Western European gasoline samples from CONCAWE (15). Typically, the volumetric composition was 1.6-32% normal (n-) alkanes (primarily with 4-7 carbon (C) atoms), 0.9248% branched (iso-) alkanes (primarily C5-C8), 0-6.4% n-alkenes (primarily C4-C8), 0.13-89% iso-alkenes (primarily C4-C8), 0.29-26% cyclic-alkanes (primarily C6 and VOL. 39, NO. 8, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2711
FIGURE 1. The 24 component groups for gasoline. C7), and 1.86-54% aromatics (primarily C7-C9). The extent of variability in composition between samples should be evaluated and in this case is reflected in the reported ranges for the various hydrocarbon classes presented later.
2. Selection of Component Groups Grouping chemical components of mixtures is ultimately a matter of judgment and will be mixture-specific, but it can be informed by the combined consideration of five criteria, (i) physicochemical properties, (ii) physical separation of the mixture by analytical methods, (iii) indicators of environmental hazard, (iv) proportion in the mixture, and (v) molecular structural similarities. Physicochemical properties determine the overall environmental partitioning and fate of chemicals. The key properties of vapor pressure, water solubility, and octanolwater partition coefficient (KOW) determine the partitioning between air, water, and hydrophobic solids such as soil or sediment (16). The organic carbon in soil and sediment is generally considered to be approximated as octanol. Thus, grouping chemicals on the basis of physicochemical properties yields component groups with common environmental sinks. For example, it may be beneficial to group volatile chemicals with similarly high vapor pressures that tend to partition into air. Chemicals may also be grouped by the physical separation of the mixture using analytical fractionation methods, for example, the retention times of groups eluted from a gas chromatographic column. Analytical fractionation groups also reflect the gradient in a particular property such as KOW or octanol-air partition coefficient (KOA). A chemical’s hazardous properties are generally taken to refer to a chemical’s “inherent” or “intrinsic” persistence (P), bioaccumulation (B), long-range transport potential (LRTP), and inherent toxicity (iT) and are independent of the quantity of chemical present. Risk properties depend on the quantity of chemical and thus on the dose or exposure concentrations. Hazardous properties and mixture composition can be considered together when defining component groups. For example, a chemical of high inherent toxicity that makes up only a small fraction of the overall mixture may be considered as a component group on its own, justifying more effort to obtain accurate physicochemical properties and estimates of fate and health effects. In gasoline, an example is naphthalene. Conversely, another chemical that is considered to contribute little to health risk may be combined with other similar chemicals even if it makes up a greater fraction of the mixture. In gasoline, examples are higher carbon number alkanes. Chemicals can also be grouped based on toxic modeof-action considerations. For gasoline, it can be assumed that all component chemicals have a narcotic mode-of-action and pose an additive risk to the exposed organism (17). However, because benzene is also a human carcinogen (18), it is best treated as an individual group. 2712
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 8, 2005
Molecular structure can be used to indicate similarities in inherent properties, such as physicochemical properties and indicators of hazard, for example, aromatics, alkanes, and alkenes. These properties tend to change systematically with structure, enabling the selection of component groups based on a gradient in particular properties. This is the principle underlying quantitative structure-property relationships (QSPRs). Applying these five criteria, 24 component groups were established for gasoline as shown in Figure 1, thus capturing the variation in properties, fate, and effects between chemical components. For other mixtures, a different number of groups will apply. Hydrocarbons were grouped primarily into structural classifications including alkanes and alkenes with normal, branched, and cyclic structures, and aromatics with one or two rings, and also by the number of carbon atoms in the compounds ranging from 3 to 12. Some groups contain multiple structural classes and carbon numbers; for example, group 1 contains C3 and C4 normal and branched alkanes, while others such as groups 16 and 17 contain a single unique hydrocarbon, benzene and toluene. Non-hydrocarbons and gasoline additives were not considered but in principle could be addressed by inclusion of additional groups. Having established the groups, the next step is to define the mixture composition and its variability, in this case for the liquid state, but also of interest is the equilibrium vapor. In some cases, an equilibrium aqueous phase composition may also be useful. Figure 2 shows a comparison of the average relative liquid-state group compositions of Western European gasoline by mass obtained by Peterson and Parkerton (19) from an analysis of CONCAWE (15) data. The vapor state composition was calculated as the equilibrium vapor composition above liquid gasoline using Raoult’s law. The variability in each group ranges from 7% to 40% of the average. The liquid and vapor compositions are very different, with 89% of the vapor composition composed of the more volatile groups, notably numbers 1, 2, 5, 8, and 10. Liquid and vapor state volume, mole, and mass compositions of the groups are given in the Supporting Information.
3. Compilation of Property Data for Groups The properties required for an exposure assessment include physicochemical properties (molar mass, boiling point, density, vapor pressure, solubility in water, and Henry’s law constant), partition coefficients (octanol-water, air-water, and octanol-air, KOW, KAW, and KOA respectively), and degradation half-lives (in air, water, soil, sediment, and sewage treatment plants). Bioaccumulation factors in fish may also be required. These values were compiled for gasoline groups using judgment informed by both measured data and estimation methods as well as the relative proportions of specific chemical structures within each group as described in this section. Physicochemical Properties and Partition Coefficients. Obtaining appropriate property data for groups 12, 16, 17, 18, and 23 was straightforward because cyclohexane, ben-
FIGURE 2. Liquid and vapor gasoline compositions. zene, toluene, ethylbenzene, and naphthalene uniquely define these groups. However, the other groups contain multiple hydrocarbon components. For example, group 7 contains all C9, 10, and 11 normal and branched alkanes. To obtain properties for each group that adequately represent all of the hydrocarbon components, a subset of hydrocarbons were selected for each structural class and carbon number within a group, and these are given in the Supporting Information. Group 7 is thus represented by six compounds C9, 10, and 11 “n-alkanes” (nonane, decane, and undecane) and C9, 10, and 11 “iso-alkanes” (2,2,5-trimethylhexane, 2-methylnonane, and 2-methyldecane). Representative hydrocarbon components were selected on the basis of molecular structure, relative volumetric composition in gasoline, and property data availability. The overall group property data sets were calculated as the composition weighted average of the representative component values within each group. The values of molar mass, density, vapor pressure, solubility in water, and KOW for each of the representative components were compiled primarily from Mackay et al. (20) and Verbruggen et al. (21), which are themselves summaries of published physicochemical property data. Boiling points were compiled from Ferris (22). To ensure consistency of the compiled data, comparisons were made with data from other sources. The predictive models EPIWIN (version 3.10) (Syracuse Research Corp., http://esc.syrres.com/interkow/estsoft.htm) and SPARC (accessed September 2002) (University of Georgia, http:// ibmlc2.chem.uga.edu/sparc/) were used for comparisons of density, vapor pressure, solubility, and KOW. The Antoine equation was used to calculate vapor pressures using correlations from Reid et al. (23) and Stephenson and Malanowski (24). Values used in applicable European Union
risk assessment reports for specific hydrocarbons (2) were also compiled. A parallel investigation into the group physicochemical properties in gasoline at 25 and 12 °C was also kindly provided by Bokis and Heidman (25) of ExxonMobil. A high level of consistency was observed in the gathered data, which is not surprising because these hydrocarbons are well studied. Physicochemical and partitioning properties, typically measured under laboratory conditions at room temperature and atmospheric pressure, 25 °C (298 K) and 1 atm, were corrected to a reference temperature of 12 °C (285 K) where possible. This temperature is recommended in the European Union Technical Guidance Document on Risk Assessment (TGD) (9). Details of these corrections are given in the Supporting Information. Table 1 summarizes the temperature corrected data set. It is useful to compile a chemical space plot in which chemicals are positioned depending on their KOW, KAW, and KOA partition coefficients. The position on the plot can be used as an indication of the likely partitioning of chemicals in the environment at equilibrium. Typically, two of the partition coefficients are selected for the x and y axis and diagonal lines can be drawn to estimate values of the third partition coefficient because it can be calculated from the other two. Figure 3 shows a log KAW versus log KOW plot of the type used by Gouin et al. (26), with diagonal lines showing the corresponding log KOA values. The imposed dark lines show the regions of the space that result in 1% and 99% distribution to air, water, and octanol as a result of Level I equilibrium calculations. The 24 gasoline groups are shown as dots with sizes dependent on the proportion in average Western European liquid gasoline by mass. It is evident that the predicted fates of these groups will be quite different VOL. 39, NO. 8, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2713
TABLE 1. Physicochemical Properties and Partition Coefficients for Gasoline Component Groups at 12 °C group no.
molar mass (g/mol)
boiling point (°C)
density (g/mL)
vapor pressure (Pa)
solubility (mg/m3)
H (Pa m3/mol)
KOW
KAW
KOA
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
57.68 72.15 86.17 102.88 78.62 107.55 136.55 56.05 78.52 84.22 81.26 84.16 100.79 126.24 75.83 78.11 92.14 106.20 106.20 112.20 126.34 155.51 128.17 142.20
-5 36 69 104 43 95 148 -6 53 60 67 81 102 137 65 80 111 136 138 117 171 223 218 241
0.5640 0.6214 0.6548 0.6812 0.6300 0.6774 0.7094 0.5879 0.6613 0.6747 0.7419 0.7739 0.7618 0.7664 0.7883 0.8765 0.8669 0.8670 0.8611 0.7104 0.8659 0.8723 0.8684 0.8375
194 294 41 683 11 682 2893 37 582 3075 653 195 061 26 239 18 643 13 245 7205 3307 768 17 176 7212 2055 659 606 1525 128 14 5 4
58 000 39 000 10 000 2000 32 000 2000 1000 235 000 127 000 93 000 64 000 55 000 12 000 2000 359 000 1 780 000 515 000 152 000 215 000 7000 36 000 3000 31 000 26 000
193 890 78 114 105 964 119 146 92 058 196 740 176 514 46 454 16 254 16 881 16 910 11 025 27 375 54 756 3625 316 368 461 299 24 432 445 801 20 21
1033 4073 18 617 154 331 491 13 383 160 071 345 991 2159 2981 3980 11 470 41 679 700 195 708 1949 2187 29 305 8705 155 172 3388 10 469
81.78 32.95 44.70 50.26 38.83 82.99 74.46 19.59 6.86 7.12 7.13 4.65 11.55 23.10 1.53 0.13 0.16 0.19 0.13 10.31 0.19 0.34 0.01 0.01
13 124 417 3071 13 161 2150 18 145 303 418 856 993 1805 458 1460 4564 10 034 17 335 2844 46 383 459 394 410 174 1 160 886
FIGURE 3. Chemical space plot showing the location of the component groups. Large dots are groups that comprise more than 10% of the liquid state mixture composition, medium dots comprise 1-10%, and small dots comprise less than 1%. Thick lines show the percent distributions for Level I equilibrium partitioning from Gouin et al. (26). with a log KOW range of approximately 2-5, a log KAW range of -2 to 2, and a log KOA range of 1-7. 2714
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 8, 2005
Degradation Half-Lives. First-order degradation half-lives in air, water, soil, sediment, and sewage treatment plants
TABLE 2. Estimated Half-Lives (h) for Component Groups in Relevant Media group no.
STP
air
water
soil
sediment
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
2 1 1 1 1 2 2 1.5 1 2 1 2 2 2 1 1.2 1.2 1.2 1.2 2 2 2 2 2
35 30 30 30 40 40 35 10 10 10 30 30 30 30 10 300 50 50 50 10 50 50 20 20
300 150 150 150 150 300 300 225 150 300 150 300 300 300 150 180 180 180 180 300 300 300 300 300
900 450 450 450 450 900 900 675 450 900 450 900 900 900 450 540 540 540 540 900 900 900 900 900
2700 1350 1350 1350 1350 2700 2700 2025 1350 2700 1350 2700 2700 2700 1350 1620 1620 1620 1620 2700 2700 2700 2700 2700
(STPs) were selected for each group. These values are for the primary degradation or transformation of the groups into the first degradation product. The selection of half-lives was a matter of professional judgment based on a weight of evidence assessment of values from a number of sources, the assumptions made by these sources, and consistency between environmental media and structural classifications of the groups. Table 2 gives the results for gasoline. For degradation in air, half-life values compiled and considered in the assessment, as detailed in the Supporting Information, included the European Union Technical Guidance Document on Risk Assessment (TGD) (9), the QSARbased model AOPWIN (version 1.90) (Syracuse Research Corp., http://esc.syrres.com/interkow/estsoft.htm) from the EPIWIN suite of models, and those from Mackay et al. (20), which assigns semiquantitative degradation half-lives to degradability classes of chemicals that are determined from an extensive literature review. Biodegradation rate constants in surface water, soil, and sediment are not as well characterized as those in air due to the multitude of variable environmental conditions that impact biodegradation mechanisms, for example, the nature of the microbial community, variability in environmental media composition, and temperature. Drawing from empirically based STP half-lives, semiquantitative data on primary biodegradation rates in water, soil, and sediment from Mackay et al. (20), the TGD (9), as well as expert judgments of biodegradability from BIOWIN QSAR-based predictions (Syracuse Research Corp., http://esc.syrres.com/interkow/ estsoft.htm), a set of compartmental half-life scaling factors was selected for all gasoline groups as detailed in the Supporting Information of 1 (STP):150 (water):450 (soil):1350 (sediment). This is recognized as being simplistic, but these approximations can be revisited in light of the predicted critical exposures and risks. While hydrocarbons in water, soil, and sediment are not subjected to hydrolysis, the compounds may be subject to abiotic reactions, especially oxidation by reactive species such as ozone and peroxides and direct and indirect photolytically induced processes. Because biodegradation is expected to be the primary route of degradation in these compartments, abiotic degradation processes were ignored to provide a conservative assessment of exposure.
The degradation half-lives selected for the gasoline groups in all relevant media are consistent between media and chemical structures. The values selected are in general agreement with predicted and semiquantitative half-lives, but it is emphasized that they are subject to considerable uncertainty. Regardless of the mixture in question, we believe that it is desirable to set out the physicochemical properties and degradabilities of the groups in a consistent and transparent manner as has been done here so that the effects on environmental fate and exposure can be readily interpreted.
4. Assessing Environmental Fate Several multimedia models can be used to predict environmental fate on different spatial scales and with different levels of complexity. In this study, a screening level model, EQC (EQuilibrium Criterian) (Version 2.02) (Canadian Environmental Modelling Centre, http://www.trentu.ca/cemc), was selected. EQC is a fugacity-based, steady-state model that uses an “evaluative” environment, which has realistic properties but is not representative of a particular geographic region. EQC provides insight into a chemical’s relative fate, facilitating comparison between chemicals and identification of chemicals of particular concern. It does not predict concentrations in a geographically explicit or “real” environment for comparison with monitoring data. If this is desirable, a model such as SimpleBox (27), which is part of European Union System for the Evaluation of Substances (EUSES) (28), can be used to predict concentrations at the local, regional, and continental scales. The standard EQC environment has four environmental compartments: air (including aerosol), water (with suspended particulate matter and fish), soil (with air, water, and solid phases), and sediment (with water and solid phases). Output from the EQC model includes environmental concentrations and rates of inter-compartmental transfer and advection and reaction loss. It also provides estimates of persistence and bioaccumulation. All equations in EQC are linear; thus all concentrations are directly proportional to emission rates to a specific medium such as air. Sorption is exclusively linear, and all reactions are first order. Concentrations are, however, dependent on mode-of-entry to air, water, or soil. Emissions to sediment are not considered. This linearity permits the use of a unit emission-model response matrix approach. Rather than specify emissions, it is preferable to run the model three times for each component group with an arbitrary unit emission of 100 kg/h, to air, water, and soil. The results can then be compiled into matrices and scaled to the actual emission rates. Example matrices are given in the Supporting Information. The concentrations and fluxes resulting from each mode-of-entry can be scaled and added to yield predicted values for the total emission. Thus, if unit emissions of a component group at a rate of 100 kg/h give concentrations in a specific compartment such as air (shown as subscript A) as a result of emissions to air (CA1), to water (CA2), or to soil (CA3), then the total concentration in air (CAT) attributable to actual emissions to air (E1), water (E2), and soil (E3) will be:
CAT ) CA1E1/100 + CA2E2/100 + CA3E3/100
(1)
This approach quantitatively describes the contribution of each emission source to the total concentration in a given compartment. It is assumed that all concentrations are below saturation levels, that is, that all component fugacities are less than their vapor pressures. This approach also applies to chemical masses, fluxes, and exposure quantities that depend on concentration. The compartmental concentrations of each group can also be added to give the total VOL. 39, NO. 8, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2715
concentration of gasoline hydrocarbons, although the proportions in the compartments will differ from that of the gasoline emission as a result of differences in the environmental partitioning and degradation behavior of the component groups. To illustrate this approach, the matrices are scaled to the following hypothetical emissions scenario. Consider a total gasoline emission of 15 kg/h with a composition consistent with that of Western European gasoline (see Figure 2). For comparison purposes, the emission was divided evenly between air, water, and soil; thus each has an emission of 5 kg/h. Emissions to air were assumed to be as fugitive vapor from liquid gasoline storage or at a refueling station; thus the calculated equilibrium vapor composition was used. Emissions to soil and to water were considered to be as liquid gasoline discharges; thus the liquid composition was used. It should be noted that the composition of gasoline emitted to air depends on the nature of the emission process. When only a small fraction of the liquid gasoline evaporates (e.g., escaped vapor when decanting), the vapor will consist mainly of the more volatile groups with fast evaporation rates. Emissions are then most accurately represented by the equilibrium vapor rather than the liquid composition. Alternatively, if emissions to air are from liquid gasoline under conditions when almost all of the liquid volatilizes (e.g., a spill to an impervious surface), then the liquid composition may be more appropriate. The results are shown in Figure 4 as a four-by-four set of charts displaying the relative gasoline composition by mass in air, water, soil, and sediment as a result of the three individual modes-of-entry and of the total simultaneous emission, which is calculated as the sum of the individual concentrations. For each mode-of-entry, the percentage of the total gasoline inventory by mass present in each compartment is shown as well as compartmental concentrations. Comparison down the columns shows the partitioning of total gasoline and of the individual groups for the particular mode-of-entry. The contribution of each modeof-entry to the total compartmental composition can be estimated by comparison along rows for a given compartment. The difference between these compositions and that of the original gasoline mixture can be seen by comparison with Figure 2. When the emission is to air only, most of the gasoline inventory remains in air (approximately 100%). This is composed mainly of groups 1 and 5, alkanes with a lower number of carbon atoms that have high KAW values and also make up almost 70% of the gasoline vapor composition. Less than 1% partitions into the other three compartments. In water, groups 16, 17, and 19 dominate the inventory, monoaromatics with low KAW’s that make up a slightly larger proportion of the vapor emission than other groups with similar KAW values. These same groups dominate the sediment inventory, along with group 21, which has a comparatively high emission rate and a high KOW. In soil, groups 17, 19, and 21 make up the greatest proportion; these are mono-aromatic groups with high KOA values. When the emission is to water only, 85% remains in water, mostly groups 5, 6, 10, 17, 19, and 21. Groups 17, 19, and 21 as previously mentioned have low KAW values, and groups 5, 6, and 10 compose approximately 30% of the liquid-phase gasoline emission by mass. 8% of the inventory is in sediment, mainly groups 4, 6, and 7, which are high carbon number alkanes, and group 21, high carbon mono-aromatics. These groups have comparatively high KOW’s and emissions. Less than 1% is in the soil, mostly groups 17, 19, and 21, which dominate the soil composition for all modes-of-entry. These groups have high KOA values and make up about 35% of the liquid emission. Groups 5, 6, 17, 19, and 21 make up 2716
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 8, 2005
approximately 75% of the emission and dominate the composition of the 7% in air. For the mode-of-entry to soil, 91% of the inventory remains in soil. This is dominated by groups 17, 19, and 21, which as previously discussed have high KOA values, are less volatile, and compose a large fraction of liquid gasoline emissions. Less than 1% of the inventory is in water and sediment; these compartments are also composed primarily of groups 17 and 19, and 21. 9% of the inventory is in air composed primarily of groups 5, 6, 17, and 19, with a high fraction of the emission composition and low KOA values. For the total emission with each mode-of-entry occurring simultaneously, 52% of the gasoline inventory is in soil, 31% in water, 14% in air, and 3% in sediment. This is not an indication of the predominant exposure pathways to gasoline; it is an indication of the partitioning and fate of gasoline if emissions are occurring simultaneously and in equal amounts. This scenario is unlikely because the majority of gasoline emissions are introduced via the atmospheric modeof-entry. However, these calculations serve an illustrative purpose. For a given compartment, comparison of the contributing compartmental concentrations from the three modes-of-entry shows that the gasoline inventories in air, water, and soil are primarily a result of their respective emissions. The inventory in sediment originated from the emission to water. It follows that for this example the total compartmental compositions tend to be consistent with those of the emission which is the primary contributor to the overall inventory.
5. Assessing Exposure Ultimately, the primary interest when conducting exposure assessments is the impact that the use of the chemical or mixture may have on human and environmental health. Having determined the relative fate of the component groups in the environment, the concentrations in exposure media such as inhaled air, ingested water, and a variety of foods can be determined. Depending on the food organism, this may require the quantitative assessment of chemical transport from environmental media into and through food webs. These predicted exposure concentrations can then be used to calculate doses. Because exposure and dose are concentration dependent, it is possible to again apply the unit emission-model response matrix approach in the manner described for predicting environmental concentrations. It then becomes apparent which groups contribute most to the exposure. The combined exposure to the total gasoline mixture can also be calculated as the sum of the group doses. This approach has been illustrated by MacLeod et al. (29), who showed how food web models can be used to estimate body burdens of hydrocarbons. Some property data used here differ from that study, a longer atmospheric half-life of benzene being assumed here as well as differences in the nature of the emissions to the atmosphere.
Discussion The focus of this study has been on assessing exposure, but it is important to emphasize that this is only a preamble to an assessment of risk, which necessarily includes toxicological and possibly pharmacokinetic considerations. There are three general approaches for assessing risk, all three requiring actual emissions data. First is an examination of concentrations external to the organism relative to target or objective concentrations or those which are judged to be of toxicological significance. This is the PEC/PNEC (predicted environmental concentration to predicted no effect concentration ratio) approach. Second is the comparison of predicted intakes or doses (µg/day) with toxicologically significant doses. Third is the calculation of an internal dose or critical
FIGURE 4. Gasoline composition in environmental compartments as a result of emissions to air, water, and soil individually at an arbitrary rate of 5 kg/h and simultaneously. body residue (CBR) or tissue concentration. The third is believed to be the most meaningful approach, especially for hydrocarbons that display narcotic or baseline toxicity. Total internal concentrations in the lipid of biota can be calculated and compared to corresponding effects based on concentrations derived from narcosis theory (30). This is valid only if all groups in the mixture have the same mode-of-toxic-action. Because benzene is known to be a carcinogen (18), it is treated as a unique group so that in addition to its narcotic effect this toxicity endpoint can also be addressed. The state of the science in physiologically based pharmacokinetic (PBPK) modeling of mixtures has been described by Dennison et al. (31) for inhaled exposure to five gasoline hydrocarbons and
a sixth “lumped component” showing the necessity of including competitive metabolic inhibition at high exposures. This logical five-stage exposure assessment strategy illustrates that the component groups within a mixture can demonstrate widely different environmental fates that are dependent on three sets of essential data: physicochemical properties including KOW, KAW, and KOA, degradation halflives, and mode-of-entry into the environment. The division of the mixture into component groups that can then be individually assessed results in a more comprehensive and meaningful exposure and risk assessment than is possible if the mixture is treated as a single substance. We also suggest that the unit emission-model response matrix approach is VOL. 39, NO. 8, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2717
invaluable as a method of presenting and processing the model results, especially for situations in which compositions and emissions are variable.
(13)
Acknowledgments We thank CONCAWE and ExxonMobil Biomedical Sciences for financial as well as technical support for this work and the consortium of companies that support the Canadian Environmental Modelling Centre.
Supporting Information Available Average relative mol, mass, and volume compositions of liquid and vapor state gasoline; the subset of representative chemicals in gasoline component groups for which property data were collected; details of temperature corrections used for property data and selecting degradation half-lives in relevant media with supporting tables; examples of unit emission-model response matrices. This material is available free of charge via the Internet at http://pubs.acs.org.
Literature Cited (1) Environment Canada. Assessment Report-acetaldehyde; http:// www.ec.gc.ca/substances/ese/eng/psap/final/acetaldehyde.cfm, 2000. (2) European Chemicals Bureau (ECB). Risk Assessment Reports: benzene, cumene, cyclohexane, ethylbenzene, naphthalene, npentane, toluene; http://ecb.jrc.it/esis/, 1999-2002. (3) Mackay, D.; Leinonen, P. J. Mathematical model of the behaviour of oil spills on water with natural and chemical dispersion; Report EPS-3-EC-7719; Environment Canada, Ottawa, ON, 1977. (4) Yang, W. C.; Wang, H. Modelling of oil evaporation in aqueous environment. Water Res. 1979, 11, 879-887. (5) Mackay, D.; Paterson, S. The physical properties of fresh and weathered crude oils; Publication EE-13: A report to the Arctic marine oil spill program; Environment Canada, Ottawa, ON, 1980. (6) Reijnhart, R.; Rose, R. Evaporation of crude oil at sea. Water Res. 1982, 16, 1319-1325. (7) Eastcott, L.; Shiu, W. Y.; Mackay, D. Modelling Petroleum in Soils. In Petroleum Contaminated Soils V. 1; Kostecki, P. T., Calabrese, E. J., Eds.; Lewis Publishers: Chelsea, MI, 1989; pp 63-80. (8) Lee, C. L.; Eastcott, L.; Shiu, W. Y.; Mackay, D. Petroleum Contaminated Soil: Chemistry and Modeling. In Principles and Practices for Petroleum Contaminated Soils; Calabrese, E. J., Kostecki, P. T., Eds.; Lewis Publishers: Chelsea, MI, 1993; pp 323-339. (9) EC (European Commission). Technical guidance documents in support of Directive 93/67/EEC on risk assessment of new notified substances and Regulation No. 1488/94 on risk assessment of existing substances; Office for Official Publications of the European Community, Luxembourg, 1996. (10) Gustafson, J. B.; Griffith Tell, J.; Orem, D. Total Petroleum Hydrocarbon Criteria Working Group Series Volume 3: Selection of representative TPH fractions based on fate and transport considerations; http://www.aehs.com/publications/catalog/ contents/Volume3.pdf, 1997. (11) Meijer, S. N.; Ockenden, W. A.; Steinnes, E.; Corrigan, B. P.; Jones, K. C. Spatial and temporal trends of POPs in Norwegian and UK background air: implications for global cycling. Environ. Sci. Technol. 2003, 37, 454-461. (12) Meijer, S. N.; Ockenden, W. A.; Sweetman, A.; Breivik, K.; Gimalt, J. O.; Jones, K. C. Global distribution and budget of PCBs and
2718
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 8, 2005
(14)
(15) (16) (17) (18) (19) (20)
(21)
(22) (23)
(24) (25) (26)
(27)
(28)
(29)
(30)
(31)
HCB in background surface soils: implications for sources and environmental processes. Environ. Sci. Technol. 2003, 37, 667672. Ockenden, W. A.; Breivik, K.; Meijer, S. N.; Steinnes, E.; Sweetman, A. J.; Jones, K. C. The global re-cycling of persistent organic pollutants is strongly retarded by soils. Environ. Pollut. 2003, 121, 75-80. Meijer, S. N.; Steinnes, E.; Ockenden, W. A.; Jones, K. C. Influence of environmental variables on the spatial distribution of PCBs in Norwegian and UK soils: implications for global cycling. Environ. Sci. Technol. 2002, 36, 2146-2153. CONCAWE. A Survey of European Gasoline Exposures for the Period 1999-2001; Report 9/02; Brussels, Belgium, 2002. Mackay, D. Multimedia Environmental Models: A Fugacity Approach, 2nd ed.; Lewis Publishers: Boca Raton, FL, 2001. Peterson, D. R. Calculating the aquatic toxicity of hydrocarbon mixtures. Chemosphere 1994, 29, 2493-2506. U.S. EPA. Integrated risk information system (IRIS); http:// www.epa.gov/iris/; Washington, DC, 2004. Peterson, D.; Parkerton, T. F. 2002, personal communication. Mackay, D.; Shiu, W. Y.; Ma, K. C. Physical-Chemical Properties and Environmental Fate and Degradation Handbook (CD-ROM); CRC Press LLC: Boca Raton, FL, 2000. Verbruggen, E. M. J.; Hermens, J. L. M.; Tolls, J. Physicochemical properties of higher nonaromatic hydrocarbons: a literature study. J. Phys. Chem. Ref. Data 2000, 29, 1435-1446. Ferris, S. W. Handbook of Hydrocarbons; Academic Press Inc.: New York, 1955. Reid, R. C.; Prausnitz, J. M.; Sherwood, T. K. The Properties of Gases and Liquids, 3rd ed.; McGraw-Hill Book Co.: New York, 1977. Stephenson, R. M.; Malanowski, S. Handbook of the Thermodynamics of Organic Compounds; Elsevier: New York, 1987. Bokis, C.; Heidman, J. L. 2003, ExxonMobil personal communication. Gouin, T.; Mackay, D.; Webster, E.; Wania, F. Screening chemicals for persistence in the environment. Environ. Sci. Technol. 2000, 34, 881-884. van de Meent, D. SimpleBox: a generic multi-media fate evaluation model; Report 6727200001; National Institute of Public Health and the Environment (RIVM), Bilthoven, The Netherlands, 1993. EC (European Commission). EUSES, the European Union System for the Evaluation of Substances; National Institute of Public Health and the Environment (RIVM), The Netherlands; Available from the European Chemicals Bureau (EC/JRC), Ispra, Italy, 1996. MacLeod, M.; McKone, T. E.; Foster, K. L.; Maddalena, R. L.; Parkerton, T. F.; Mackay, D. Applications of contaminant fate and bioaccumulation models in assessing ecological risks of chemicals: a case study for gasoline hydrocarbons. Environ. Sci. Technol. 2004, 38, 6225-6233. McGrath, J. A.; Parkerton, T. F.; Di Toro, D. M. Application of the narcosis target lipid model to algal toxicity and deriving predicted no effect concentrations. Environ. Toxicol. Chem. 2004, 23, 2503-2517. Dennison, J. E.; Andersen, M. E.; Clewell, H. J.; Yang, R. S. H. Development of a physiologically based pharmacokinetic model for volatile fractions of gasoline using chemical lumping analysis. Environ. Sci. Technol. 2004, 38, 5674-5681.
Received for review August 13, 2004. Revised manuscript received January 16, 2005. Accepted January 18, 2005. ES048734P
