| 4.4 | DIOXIN: APPLICATION OF THE COMPREHENSIVE MODELING TECHNIQUE | |
The polychlorinated dibenzo-p-dioxins and dibenzofurans pollutant group (PCDD/Fs) contains several potent carcinogens and powerful endocrine disruptors. (10) The comprehensive modeling technique is being applied to investigate the transport and deposition of dioxin from sources in the United States and Canada to the Great Lakes. In the discussion that follows, preliminary results are presented, including emissions inventories and the results of the fate and transport modeling.(11)
| 4.4.1 | Case-Study Examples of the Calculations | |
As an introduction, results for two illustrative case-studies will be presented. In these case studies, emissions of dioxin from two different metropolitan regions, Hamilton, ON and St. Louis, MO-IL, are estimated, and their impact on atmospheric deposition to each of the Great Lakes is considered.
Estimated emissions from each of the metropolitan areas are presented in Table 5. The total emissions from the Hamilton area are estimated to be on the order of 30 grams TEQ/year. (12) For the St. Louis area, the emissions are estimated to be on the order of 133 grams TEQ/year.(13)
| Table 5. Estimated Dioxin Air Emissions for the Hamilton and St. Louis Metropolitan Areas | ||||
| Emissions (grams TEQ/year) (the low-range and high-range estimates are given in parentheses below the mid-range estimate) | ||||
| Metropolitan Area and Inventory Year | Waste Incineration | Metallurgical Processes | Fuel Combustion | Total |
| Hamilton (ON) 1995 | 4.2 (1.3 - 13) | 25 (8 - 80) | 0.50 (0.16 - 1.6) | 30 (10 - 95) |
| St. Louis (MO,IL) 1996 | 33 (10 - 100) | 99 (33 - 310) | 1.5 (0.5 - 4.6) | 133 (42 - 420) |
For both areas, metallurgical processes are the largest emissions category. Significant emissions also are contributed by waste incineration processes. Emissions from fuel combustion is relatively insignificant for both areas.(14) Rough estimates of the uncertainty in the emissions estimates are presented in Table 5. It can be seen that there is approximately a factor of 3 uncertainty on either side of the mid-range emissions estimates. The emissions inventories used in this analysis are discussed in more detail in section 4.4.2.
The estimated efficiencies of transfer of emissions from each of the case-study areas to each of the Great Lakes through the atmospheric deposition pathway are presented in Table 6.(15) For example, it is estimated that approximately 3.6% of the dioxin emitted into the air in the Hamilton area will be deposited in Lake Ontario over the course of a year. Due to the proximity of Hamilton to Lake Ontario, this transfer coefficient is higher relative to that for the other Great Lakes. For emissions from the St. Louis area, the largest transfer coefficient is 1.5%, to Lake Michigan, the closest Great Lake. Generally, the larger the lake, the higher the transfer coefficient will be, as there is a greater surface area for atmospheric deposition. However, the distance of the source from the lake and its orientation relative to the direction of prevailing winds play a major role as well.
| Table 6. Model-Estimated Dioxin Transfer Coefficients from Hamilton and St. Louis to the Great Lakes | |||||
| Transfer Coefficients (dimensionless) (fraction of emissions predicted by the modeling to be deposited in a given lake; average value for all sources in the Metropolitan region, on a TEQ basis) | |||||
| Metropolitan Area | Lake Ontario | Lake Erie | Lake Michigan | Lake Huron | Lake Superior |
| Hamilton (ON) | 3.6% | 0.92% | 0.40% | 1.4% | 0.40% |
| St. Louis (MO,IL) | 0.16% | 0.32% | 1.5% | 0.74% | 1.0% |
The final step in this analysis is the use of the estimated transfer coefficients to assess the atmospheric deposition impacts of the emissions on the Great Lakes. This is done by multiplying the emissions by the model-estimated transfer coefficients.(16) Results of this multiplication are presented in Table 7 and are schematically depicted in Figure 3. Thus, for example, of the 130 grams TEQ/year emitted from the St. Louis area, 1.5% - or about 2 grams TEQ/year - are deposited in Lake Michigan. Lesser amounts are deposited in the other Great Lakes from this source area. Analogously, the Hamilton area contributes approximately 1.1 grams TEQ/year to Lake Ontario, and lesser amounts to the other Great Lakes. In this example, due to the importance of metallurgical processes in the estimated emissions, this source category contributes the largest amount of dioxin from these source areas to the Great Lakes.
| Table 7. Atmospheric Deposition of Dioxin to the Great Lakes Arising from Air Emissions from the Hamilton and St. Louis Metropolitan Regions | ||||||
| Atmospheric Deposition to the Great Lakes (grams TEQ/year) | ||||||
| Metropolitan Area | Source Category | Lake Ontario | Lake Erie | Lake Michigan | Lake Huron | Lake Superior |
| Hamilton (ON) | incineration | 0.15 | 0.039 | 0.017 | 0.061 | 0.017 |
| metallurgical | 0.91 | 0.24 | 0.10 | 0.36 | 0.10 | |
| fuel combustion | 0.02 | 0.0047 | 0.0021 | 0.0073 | 0.0021 | |
| total | 1.1 | 0.28 | 0.12 | 0.43 | 0.12 | |
| St. Louis (MO,IL) | incineration | 0.05 | 0.11 | 0.48 | 0.25 | 0.34 |
| metallurgical | 0.16 | 0.32 | 1.5 | 0.73 | 0.98 | |
| fuel combustion | 0.0023 | 0.0047 | 0.022 | 0.011 | 0.015 | |
| total | 0.21 | 0.43 | 2.0 | 0.99 | 1.3 | |
Dioxin emitted to the atmosphere from any source is subject to dispersion (dilution) in the atmosphere, transformation (e.g., chemical reactions), and deposition to the earth's surface. These phenomena all tend to reduce the concentrations at greater and greater distances from the source. However, for dioxin, these influences are not so significant that regional and long-range transport can be ignored. This example shows that emissions from outside the Great Lakes basin can be as or more significant than emissions from within the basin.
In the following sections, this type of analysis is extended to other source regions in the United States and Canada.
| 4.4.2 | Emissions Inventories for Dioxin | |
The starting point of the analysis is an emissions inventory. For the U.S., a 1996 dioxin emissions inventory prepared by the Center for the Biology of Natural Systems (CBNS) at Queens College N.Y. has been utilized. This 1996 inventory was used in an analysis of transport and deposition of PCDD/F to dairy farms in Vermont and Wisconsin (Commoner et al., 1998). Built from a previously-prepared 1993 inventory(17), an attempt was made in this inventory to update the emissions from 1993 to 1996, to include several new source categories, and estimate emissions with a higher degree of spatial resolution. Consistency with a contemporaneous U.S. EPA inventory (U.S. EPA, 1998), where appropriate, was sought.
For Canada, a dioxin emissions inventory for 1995 prepared by Environment Canada and the Canadian Federal-Provincial Task Force on Dioxins and Furans has been used (Environment Canada et al., 1999). Additional assistance in gridding the inventory for model application was provided by Environment Canada.
Overall summaries of the inventories are used to generate the binational emission map for dioxin (Figures 4 and 5). In Figure 4, the total emissions per county in the U.S. and in 50-100 km grid cells in Canada are mapped. Because the counties and grids are different sizes, the "areal density of emissions" has been mapped, i.e., the emissions for each locale (county or grid) are divided by the area of that locale.
The resulting units are grams TEQ per square kilometer per year, analogous to an emissions flux. In Figure 5, the total emissions for each state and province have been mapped as a percentage of the total emissions for the U.S. and Canada.
Maps of emissions from a few selected individual source categories include municipal waste incineration (Figure 6), medical waste incineration (Figure 7) and cement kilns burning hazardous waste and other alternative fuels (Figure 8). In these maps, which represent point sources, the emissions amount itself is mapped; the amount is not divided by the area and is represented by a circle of a particular size and color.
There are significant uncertainties in the estimation of dioxin emissions in the U.S. and Canada, because the number of actual sources sampled is very limited, and many important source categories appear poorly characterized. While the information available appears adequate to generate a first estimate of source/receptor linkages, further resources dedicated to improvement of emission inventories is would be crucial to improve the accuracy of this modeling methodology.
The dioxin inventories are currently being assessed for their degree of completeness in considering all significant emission sources. Preliminary results include the following:
Other shortcomings may exist in the PCDD/F inventories, and an attempt is being made to further correct or at least qualitatively discuss these limitations. In addition, in some cases, it is known that there have been significant changes in estimated emissions from particular sources and source categories since the inventories used in this analysis were compiled. For example, it is estimated that emissions from the municipal waste incinerator in Levis, Quebec (adjacent to Montreal) dropped from 61.8 grams (2 ounces) TEQ/year to less than one gram TEQ per year in Fall 1998, as a result of renovations made at the facility (Environment Canada, 1999). Thus, it is important to recognize that the inventories being used are intended to be representative of the modeling period chosen and do not necessarily reflect the current situation.
| 4.4.3 | Generalized Fate and Transport Modeling of Dioxin | |
The HYSPLIT model was used to simulate the fate and transport of emitted dioxin from sources in the United States and Canada. Examples of the general results of this modeling have been summarized in transfer coefficient maps for overall dioxin TEQ for lakes Superior and Ontario (Figures 9 and 10). These transfer coefficient maps are independent of the actual emissions. They only describe the relative efficiency of transport and deposition between each location on the map and the indicated receptor. A useful way to think about these maps is to consider that they represent the relative depositional impact on the given receptor from each point on the map if the emissions were uniformly equal everywhere in the modeling domain.
In Figure 9, for example, between 0.013 - 0.021 (1.3 - 2.1%) of any dioxin emissions that would be emitted within the red area would be deposited in Lake Superior. At the low end of the range (the yellow area), the fraction of dioxin emissions deposited is less than 0.003 (< 0.3%). Again, this map is not related to the actual emissions, but solely depicts the model-estimated ability of the atmosphere to carry emissions from various source locations and deposit them in the indicated lake.
The maps generally show the influence of the prevailing winds from the west and southwest, as the efficiency of transport from those directions is generally greater than for other directions. It is also evident that while the transfer coefficients diminish with distance from a given lake, they do not drop so precipitously that contributions from regional and long-range sources can be considered insignificant. Note that the transfer coefficient map will be different for each congener or mix of congeners considered, and for each receptor considered (a "congener" is a particular dioxin species; there are a total of 210 different dioxin congeners - see footnote 16). These maps were prepared assuming a mix of congeners typical of the average mix of congener emissions in the U.S./Canadian emissions inventory. In practice, these maps are only created as examples of the calculation, and the myriad variations are handled computationally.
| 4.4.4 | Evaluation of the Dioxin Modeling Results | |
In order to evaluate the overall validity of the modeling results, the modeling predictions at specific locales can be compared against ambient measurements at those locations. Ambient dioxin measurements in the U.S. and Canada have been reviewed by Cohen (1998). For dioxin, in 1996, 30-day ambient rural air measurements at two sites each in Vermont and Wisconsin (Commoner et al., 1998), and one site in Connecticut are available, as are 48-hour samples at several rural sites in Canada (Dann, 1998). Other rural short-term (24-48 hour) measurements are potentially available as well in Mississippi (Fiedler et al., 1997), Ohio (Wagrowski and Hites, 1997), and Vermont (Agency of Natural Resources, 1996). There are no direct measurements of dry deposition of PCDD/F for the time periods or regions of interest. Thus, dry deposition estimates for PCDD/F are typically inferred from ambient concentration measurements. The only wet deposition measurement program in North America appears to be that sponsored by the Ontario Ministry of Environment. Wet deposition data for PCDD/F for 1996 from a semi-rural site in Dorset, Ontario are being sought.
The dearth of ambient monitoring data for PCDD/F does reduce the potential for model evaluation, a significant limitation of this or any other PCDD/F modeling effort. Additional resources should therefore be devoted to measurements of dioxin ambient air concentrations and deposition. Other ambient air and precipitation data for dioxin will be incorporated into the analysis as they become available.
In recent studies (Cohen et al., 1995; Commoner et al., 1998), predicted dioxin concentrations using this methodology were found to be encouragingly close to ambient measurements; most predictions were matched by the measurements within the estimated uncertainties in each. In the present study, the relative agreement has been found to be comparable to that in the previous investigation. Model evaluation results are presented in a freestanding report on this project.
When monitoring and modeling are undertaken as complementary activities, the value of each is immeasurably increased. Ambient measurements are essential for model evaluation and hence model viability. At the same time, monitoring everywhere, all the time is not feasible. Modeling can help to fill in the spatial, temporal, and phenomenological gaps in the monitoring program, interpret available ambient measurements, refine existing monitoring programs, design new monitoring programs, and theoretically predict the impacts of different proposed courses of future action. Without both monitoring and modeling activities being undertaken in a complementary manner, the utility of each is decreased.
| 4.4.5 | Source-Receptor Relationships for the Atmospheric Deposition of Dioxin to the Great Lakes Arising from Emissions in the U.S. and Canada | |
The emissions inventory data have been combined with the generalized fate and transport results to estimate source-receptor relationships between the sources in the inventory and each of the Great Lakes. Maps showing the geographical distribution of contributions of atmospheric deposition to Lake Superior and Lake Ontario are presented in Figures 11-14. Mid-range (geometric mean) estimates of emissions were used to generate these maps. In Figures 11 and 13, the areal density of deposition contribution is mapped (µgrams TEQ/km2-year), and in Figures 12 and 14, the total contributions for each state and province are mapped as a percent of the total estimated deposition.
While maps for deposition to all the individual lake basins are available, these two were chosen for this summary chapter to contrast a relatively undeveloped shoreline (Lake Superior) with an extensively developed one (Lake Ontario). As might be expected, influential sources of persistent toxic substances to Lake Superior are typically at a greater distance than the more localized sources affecting Lake Ontario. Indeed, in the case of Lake Superior, sources as far south as St. Louis are having a significant effect on deposition to the lake, while the large urban centers immediately adjacent to Lake Ontario are responsible for a significantly greater deposition to that lake.
Overall graphical summaries of the relative contributions from different distances and source categories are presented in Figures 15, 16, and 17. Figure 15 notes that the most significant sources of deposition to Lake Superior, accounting for near 40 percent, are located between 400 and 700 kilometers (250 and 435 miles) distant, whereas, in the case of Lake Ontario, approximately 35 percent of the deposition is from sources from 0 to 20 kilometers (0 to 12.5 miles) from the shoreline.
Figure 16 shows that the percent of deposition for each of the lakes arising from emissions in each of the states and provinces in the Great Lakes basin (Minnesota, Wisconsin, Illinois, Indiana, Michigan, Ohio, Pennsylvania, New York, and Ontario) and the percent arising from the rest of the United States and Canada. These data are also presented in Table 8. It can be seen that on the order of 75% of the dioxin deposition to the Great Lakes arises from the Great Lakes states and provinces. However, for Lakes Superior and Huron, only about 25% of the deposition arises from states and province actually adjoining the Lake.
| Table 8. Contribution of 1995/1996 Atmospheric Deposition of Dioxin to the Great Lakes from Great Lakes States and Provinces and from the Rest of the U.S. and Canada* | ||||||
| Percent of Atmospheric Deposition of Dioxin to Lake Contributed by State or Province | ||||||
| State or Province | Lake Superior | Lake Huron | Lake Michigan | Lake Erie | Lake Ontario | Average for Great Lakes |
| Illinois | 16 | 11 | 36 | 6 | 4 | 15 |
| Indiana | 13 | 12 | 27 | 8 | 4 | 13 |
| Michigan | 10 | 17 | 7 | 11 | 4 | 10 |
| Minnesota | 9 | 3 | 3 | 1 | 1 | 4 |
| New York | 4 | 8 | 2 | 8 | 35 | 11 |
| Ohio | 7 | 17 | 4 | 37 | 15 | 16 |
| Pennsylvania | 2 | 3 | 1 | 5 | 8 | 4 |
| Wisconsin | 2 | 1 | 2 | 0.3 | 0.3 | 1 |
| Ontario | 6 | 4 | 1 | 2 | 7 | 4 |
| total for Great Lakes States and Provinces | 69 | 75 | 83 | 78 | 79 | 77 |
| total for all other U.S. States | 29 | 23 | 16 | 21 | 20 | 22 |
| total for all other Canadian Provinces | 2 | 2 | 1 | 1 | 2 | 1 |
| total for all other U.S. and Canadian States and Provinces | 31 | 25 | 17 | 22 | 21 | 23 |
| total | 100 | 100 | 100 | 100 | 100 | 100 |
* Note: Sums may not appear exact because of rounding.
Figure 17 presents preliminary estimates of the contribution of specific source categories to the annual dioxin deposition flux (grams TEQ deposited per square kilometer of lake) to specific lake basins.
In order to summarize these results succinctly, the various source categories in the inventories were aggregated into three general categories: incineration, metallurgical processes, and fuel combustion. The estimated impacts (grams TEQ deposited to a given lake per year from each category) were divided by the area of the lake to get a flux (grams TEQ per km2 per year)(18) This flux amount was then divided by the population of the source country to obtain a per-capita value for the contribution. On average, using 1995-1996 emissions inventory data, various types of incinerators were the major source category of dioxin deposition to the entire Great Lakes basin. Even on a per-capita basis, the U.S. contribution appears to be relatively large compared to the Canadian contribution(19), except for Lake Ontario, where the two are comparable.
It should be noted that this analysis has included only sources in the United States and Canada; sources in Mexico have not been considered. Unfortunately, a dioxin inventory for Mexico is not currently available. Given the distances involved, it is expected that the contribution of dioxin from Mexico to the Great Lakes will be relatively insignificant. However, this hypothesis will have to be verified through an extension of this comprehensive modeling analysis when an inventory for Mexico becomes available. Further extension of the analysis to yet additional countries might also be useful, but is expected to add relatively little to the analysis of the dioxin input to the Great Lakes; in any event, this hypothesis can be tested in future work.
10. PCDD/F's will be referred to as "dioxin" in this chapter.
11. Work on this pollutant, as well as atrazine and cadmium, is ongoing, and a more complete report on all three pollutants anticipated toward the end of 1999.
12. TEQ (Toxic Equivalents) is a weighted average amount of dioxins/furans which takes into account the relative toxicity of the various forms of these compounds in a given mixture.
13. The Environment Canada inventory used in this work is for 1995, so the estimates for Hamilton, Ontario are most representative of 1995 emissions. The U.S. inventory used in this work is for 1996, and so the St. Louis estimates are most representative of emissions for 1996.
14. The relative importance of different source categories varies throughout different regions in the United States and Canada. In the inventories used in this analysis, waste incineration is estimated to be have higher overall emissions than metallurgical processes. Thus, this particular example is not representative of the overall emissions pattern of either nation.
15. These estimates are based on the results of atmospheric modeling, discussed below.
16. There are many different dioxin molecules (congeners), each with a different toxicity and atmospheric behavior. Since there are different congener emissions profiles from each source, the actual calculation is done source by source, congener by congener. The overall, average results, on a TEQ basis, are summarized here.
17. An earlier version of the U.S. inventory, for the year 1993, was developed at CBNS during 1994 -- 1995 in conjunction with an analysis of PCDD/F transport and deposition to the Great Lakes (Cohen et al., 1995; 1997a). This 1993 inventory did not include a number of the source categories that were later included in the 1996 inventory. Area sources also were aggregated at the state/province and metropolitan area level. The earlier inventory was designed to be somewhat consistent -- where appropriate -- with the U.S. EPA Dioxin Reassessment Inventory (U.S. EPA 1994).
18. This normalization by lake area was done so that the deposition contributions to the lakes could be compared on an equal basis. That is, all things being equal, there will be more atmospheric deposition (e.g., grams per year) to a large lake than a small lake (since the surface area for deposition is larger), but, a large and small lake will have the same atmospheric deposition flux (e.g., grams per year per square kilometer of lake).
19. Informal burning of waste (e.g., backyard trash burning) has not yet been included in the Canadian inventory, and its significance is not well understood presently. This omission may partially account for the U.S./Canadian difference.