| 4.3 | A PROJECT USING THE COMPREHENSIVE MODELING APPROACH | |
In this context, seeking to develop the tools necessary to take a strategic approach to the aforementioned binational agreements, over the past four years, the International Air Quality Advisory Board of the International Joint Commission, co-chaired by Dr. Gary Foley of the U.S. EPA and Dr. Don McKay of Environment Canada), have initiated a collaborative, multi-agency effort to comprehensively determine the sources responsible for atmospheric deposition to the Great Lakes through application of an atmospheric model. The contaminants under current investigation with this methodology include dioxin, atrazine, and cadmium. This report will examine, in some detail, the development of linkages between U.S. and Canadian sources of dioxin and individual lake basins in the Great Lakes region.
As the modeling analysis is not yet complete, the discussion that follows is in the form of a progress report. Preliminary results, where available, will be presented. As the dioxin analysis is furthest along at the present, the development of linkages between U.S. and Canadian sources of dioxin and individual basins in the Great Lakes will be discussed here. A more detailed discussion of this project is available as a free standing report.
| 4.3.1 | Availability of Emissions Inventories | |
As the board noted in its chapter of the IJC 1995/97 Great Lakes Priority Report, emissions inventories are critical to the successful execution of any modeling effort. Geographically and temporally resolved emissions inventories are needed for any compound for which a comprehensive modeling analysis is to be performed. Thus, one of the first steps in this work was determining the availability and adequacy of emissions inventories for each of the compounds of concern in the Great Lakes. At the present time, for all of the Binational Toxics Strategy pollutants, relatively accurate binational inventories only appear to be available or potentially available for dioxin, cadmium, mercury, hexachlorobenzene (HCB), and possibly one or more polycyclic aromatic hydrocarbon (PAH) compounds. For these contaminants, the development, verification and subsequent mapping of a truly binational inventory is a required first step in the modeling process. In addition, a binational emissions inventory is available for atrazine, a widely used herbicide on corn and sorghum.
Given the need for a relatively comprehensive emission inventory, the compounds chosen for further consideration in this phase of the project include dioxin, atrazine, cadmium, and mercury.(3), (4), (5) Results for the first three of these compounds are anticipated by Fall 1999, and results for mercury should be available in the following year.
The inventories selected for use in this phase are summarized in Table 4. The International Air Quality Advisory Board has established a binational body, the Emissions Inventory Working Group, to provide expert overview and comment on the inventories considered for use and to support the development and mapping of binational inventories. Each of the inventories is subject to a preliminary evaluation for completeness and accuracy using a variety of approaches, including the following:
| Table 4. Summary of Inventories Currently Identified for Use in Project | |||||
| Inventory Developer | Pollutant | Spatial Extent | Spatial Resolution | Temporal Resolution | Notes |
| Center for the Biology of Natural Systems, Queens College NY | PCDD/F (dioxins) | United States | Major point sources and county-resolution area sources | Annual emissions for 1996 | |
| Environment Canada | PCDD/F | Canada | Major point sources and both census division and gridded data for area sources (50 km grid around Great Lakes, 100 km grid in remainder of Canada). | Annual emissions for 1995 | |
| U.S. EPA National Toxics Inventory | Cadmium | United States | County resolution for all source categories | Annual emissions, nominally for 1993 | |
| Environment Canada | Cadmium | Canada | Major point sources and both census division and gridded data for area sources (50 km grid around Great Lakes, 100 km grid in remainder of Canada). | Annual emissions for 1995 and for 1990 | Estimates of emissions for 1993 may be developed by Environment Canada |
| Center for the Biology of Natural Systems, Queens College NY | Atrazine | United States and Canada | County resolution for the U.S., and a 127x127 km emissions grid in Canada | Weekly emissions for March-July 1991 | Based partly on emissions factors developed by Ortech Ltd. |
| U.S. EPA National Toxics Inventory | Mercury | United States | County resolution for all source categories | Annual emissions, nominally for 1994-5 | |
| Environment Canada | Mercury | Canada | Point sources for major facilities and both census division and gridded data for area sources (50 km grid around Great Lakes, 100 km grid in remainder of Canada). | Annual emissions for 1995. | |
Corrections or qualitative discussion of any limitations identified through the above process is being included in the analysis.
| 4.3.2 | Atmospheric Fate and Transport Modeling | |
The NOAA HYSPLIT (Hybrid Single Particle Lagrangian Integrated Trajectory) computer model has been chosen to simulate the atmospheric fate and transport of pollutants from sources in the United States and Canada to the Great Lakes. HYSPLIT was originally developed at the U.S. National Oceanic and Atmospheric Administration (NOAA) for medium and long-range transport modeling of accidental releases of radioactive materials. It is currently used operationally for emergency response at NOAA. The development, validation, and operation of HYSPLIT are described elsewhere (Draxler et al.,1998, 1991ab; Draxler, 1987, 1992).
HYSPLIT is a Lagrangian model, in which puffs of pollutant are emitted from user-specified locations, and are then advected, dispersed, and subjected to destruction and deposition phenomena throughout the model domain. It has been used to simulate many different atmospheric processes, including sulfur transport and deposition in the U.S. (Rolph et al.,: 1992, 1993) and dispersion of pollutants from Persian Gulf oil fires (Draxler et al.,1994; McQueen et al., 1994). HYSPLIT has been specially modified for the work discussed here, so that it can better simulate the fate of the pollutants being considered. Simplified schematics of its structure and operation are shown in Figures 1 and 2. Similar to most atmospheric fate and transport models, HYSPLIT uses gridded meteorological data computed by an external model.
For the current simulations, output from NOAA's Nested Grid Model (NGM) is being used. The NGM -- a primitive-equation meteorological simulation model -- takes weather-related observations and uses these data to estimate meteorological conditions existing between the observations (spatially and temporally). The data provided to HYSPLIT from the NGM include wind speed and direction, amount and type of precipitation, temperature, and humidity, as well as other meteorological data.(7)
HYSPLIT was chosen for this analysis for several reasons, including the following: (a) meteorological data necessary to drive the model were readily available for the period 1990-1999; (b) the model is computationally efficient - as it is designed for operational use - allowing year-long simulations to be performed practically; (c) the model and associated methodology have been specifically tailored to provide source/receptor information, a feature not commonly found in most fate and transport models; and (d) the model has been successfully applied in the past to simulate several of the compounds being considered here.
The methodology used in this study has been developed in a series of studies performed previously (Cohen et al., 1997ab, 1995; Commoner et al., 1998). The simulation of atmospheric fate of a selected contaminant includes provision for vapor/particle partitioning, wet and dry deposition, reaction with the hydroxyl radical (OH-), and photolysis. (8)
It must be noted that other models could be used to perform this type of analysis. Indeed, the EPA has used the Lagrangian model RELMAP to simulate dioxin and cadmium fate and transport in the U.S. and is using an Eulerian framework (Models-3) to simulate atrazine transport and deposition to Lake Michigan.
| 4.3.3 | Evaluation of Modeling Results | |
The modeling results will generally be evaluated by comparison with the two types of available primary monitoring data -- measurements of atmospheric ambient concentrations and wet deposition. The availability of ambient atmospheric monitoring data is somewhat limited for many pollutants of concern in the Great Lakes; however, for the pollutants and time periods selected for this study, sufficient data exist for limited model evaluation purposes.
Given the substantial uncertainties in the emissions inventories and the fate and transport modeling - and sometimes in the ambient measurements themselves -- it is not expected that model estimates will precisely match ambient measurements. Reasonable agreement with ambient measurements was found in previous work on dioxin and atrazine (Cohen et al., 1995, 1997ab; Commoner et al., 1998), and it is anticipated that satisfactory results will be obtained in the present analysis. A comparison of the ambient measurements with the model predictions will be presented to provide some context within which to understand the relative degree of matching. If there are significant differences, their potential reasons will be hypothesized and attempts to reconcile the discrepancies will be made.
| 4.3.4 | Uncertainties | |
There are substantial uncertainties involved in this analysis. Emissions inventories and the fate and transport simulation of the compounds studied in this project have not yet been well characterized by the scientific and/or regulatory community. Throughout the work, attention is being paid to this central limitation, and attempts continue to be made to provide qualitative and/or quantitative estimates of the magnitude of the uncertainties involved. Where possible, suggestions will be offered on areas of research and/or data collection that are critically needed to reduce uncertainties. Despite the uncertainties inherent in this work, it is anticipated that the results will be useful in the discussion of strategies to reduce loadings to the lakes from particular sources and source regions.
| 4.3.5 | Summary of Methodology | |
Emissions inventories for the U.S. and Canada are being combined with an atmospheric model (HYSPLIT) to estimate the fate and transport of emitted pollutants, including deposition to each of the Great Lakes. Existing ambient monitoring data are being identified and assembled for use in evaluating the modeling results. For each compound, the modeling predictions of ambient air concentrations and wet deposition will be compared against available actual measurements. The simulation time periods chosen for this project are the following (9): (a) dioxin - 1996, full year; (b) atrazine - 1991, March-July; and (c) cadmium - 1993, full year.
3. HCB has an extremely long atmospheric lifetime (~ years) and is generally believed to be globally distributed. Therefore, an analysis which considered only U.S. and Canadian inventories of this contaminant would be insufficient; rather, a global inventory and modeling effort would likely be necessary. Since the current effort is limited to U.S. and Canada, HCB was eliminated from consideration in the current phase of this project.
4. The U.S. National Toxics Inventory gives data only for a group of PAHs ("16-PAH"). The Canadian inventories for PAHs contain some individual PAHs, such as benzo(a)pyrene and anthracene. The two inventories cannot be combined in a straightforward manner; thus, PAHs have not been included in the current phase of this project.
5. Elemental gaseous mercury is similar to HCB in that it is believed to have an atmospheric lifetime of approximately 2 years. However, reactive gaseous mercury (RGM) (e.g. HgCl2) has a much shorter atmospheric lifetime, as it is deposited very efficiently. As a result, atmospheric deposition of mercury may be dominated by the relatively short-range phenomena associated with RGM. Thus, a substantial part of the mercury deposition in the Great Lakes may come from sources in the U.S. and Canada. Therefore, it may be possible to perform a useful modeling study based primarily on U.S. and Canadian inventories. However, this hypothesis would have to be evaluated in any investigation. Such an analysis would require incorporation of a suitable method for including "global background" contributions.
6. Source categories are particular processes, e.g., municipal waste incineration, coal combustion, cement production, copper smelting, etc.
7. Gridded model outputs have been archived at a spatial resolution of 180 km (112 miles), with determinations at ten (10) vertical levels up to a height of 5,000 meters (16,400 feet) in altitude, with a temporal resolution of two hours.
8. The current work for dioxin represents an update on previous work (Cohen et al., 1997a, 1995; Commoner et al., 1998) in the following ways: (a) use of more complete and up-to-date emissions inventories for the U.S. and Canada; (b) more monitoring data are available, allowing a more rigorous evaluation of the model results; (c) use of an updated version of the HYSPLIT model, reflecting numerous incorporated refinements; (d) substantial additional effort is being directed towards the visual display of model output; and (e) results for the Great Lakes are being estimated for 1996 (as opposed to 1993).
9. The decision regarding which time period to simulate revolved around a consideration of the following: (a) availability of adequate inventory information; (b) availability of ambient monitoring data for model evaluation; and (c) availability of meteorological data necessary to drive the model. No one year or time period perfectly satisfied all criteria for a given pollutant, but it is believed that the time periods chosen are reasonable, given the limitations in the availability of data. For example, at the time this project began, a binational atrazine inventory was only available for 1991. Moreover, the most substantial ambient monitoring data set was available for 1990-1991. Thus, 1991 was the logical (and really only) choice feasible for modeling this pollutant. Similar considerations apply to the other pollutants.