Canada existed.  As the best compromise, the nominal year
cantly from this 84-location analysis.  Therefore, as was
for this analysis was chosen to be 1996, (i.e. 1996 meteoro-
found with the earlier dioxin analysis, it is believed that the
logical data were used to drive the model, and 1996
interpolation procedure using these 84 locations is
ambient data were used for model evaluation).  Therefore,
providing estimates of adequate accuracy.
the goal was to utilize emissions inventory data representa-
tive of 1996 to the greatest extent possible.
In these transfer coefficient maps, it can be seen that there
are occasionally small regions around a few of the standard
A mercury emissions inventory for the United States was
source locations that appear to be artifacts of the interpola-
tion procedure.  For example, in the Hg(p) map in Figure
obtained from the U.S. EPA (Ryan 2001).  The inventory
contains annual emissions estimates for most anthropo-
5, there is a small circular region around the standard point
genic sources of mercury.  For coal-fired electricity genera-
in the northern, central portion of the map (in the North-
tion boilers, municipal waste incinerators, and medical
west Territories, about 500 km west of Hudson Bay).  The
waste incinerators, the estimates in this inventory were for
fact that this small region surrounding the standard point
appears to have a slightly lower potential for transport to
1999, while the remainder were reported to be representa-
tive of 1996 emissions.
the lakes in comparison to the surrounding region is
probably indicative of a slight loss of accuracy in the
The inventory was further modified in recognition that one
interpolation procedure due to the fact that the standard
source category (coal combustion in commercial, indus-
points are relatively sparsely distributed in this region and
trial, and institutional boilers and process heaters) ap-
also, possibly because the location is close to the edge of
peared to be under-represented in this inventory, and so
the modeling domain.  More standard points in regions
data from an alternative 1995-1996 U.S. EPA inventory
such as this could be added, and the modeling domain
(Bullock 2000; U.S. EPA 1997) was utilized for this source
expanded; these alterations would make the transfer
type.  Also, because significant reductions in emissions
coefficient map slightly more accurate.  However, it was not
from United States municipal waste incinerators and
deemed necessary to expend the computational resources
medical waste incinerators occurred between 1996 and
to accomplish this, as the contributions of mercury from
1999 (Mobley 2003), data from this alternative inventory of
these areas to the Great Lakes was inconsequential due to
1995-1996 emissions for these categories were used.
the fact that the transfer coefficients were relatively low and
there were no significant sources in these regions.
The coal-fired utility emissions estimates for 1999 were
used however, as they were based on a significant amount
Note that the transfer coefficient maps shown here are
of source testing and were estimated with a much more
slightly different from the transfer coefficient maps shown
sophisticated approach than used in previous inventories.
in earlier analyses (e.g. IJC Priorities Report 1997-99).
Emissions from coal-fired boilers appear to have been fairly
Previously, it was the fraction of the emission being
similar in 1996 and 1999, at least in total (Mobley 2003).
deposited to the entire lake that was being mapped.
Because each of the Great Lakes is a different size, the
The United States inventory contained a total of 17,513
transfer coefficient patterns appeared to be very different,
but this was primarily the lake-size effect.  That is, in the
discrete point sources with specific locations.  As is
previous version of these maps, the size of the lake
common practice in emissions inventories, certain source
categories (e.g. mobile sources, residential fuel consump-
mattered, because all things being equal, more of the
tion) were not estimated at precise locations but were
emissions from any given location would be deposited in a
estimated at the county level.  There were 52,673 of these
larger receptor than a smaller receptor.  In the current
area sources in the inventory, representing approximately
maps, the values have been normalized by the size of the
lake, making the mapped transfer coefficients independent
17 such source types, on average, in each of the 3141
United States counties.  For such area sources, for the
of this factor.  Following the procedure illustrated below,
however, one can convert the values in the current maps to
purposes of the modeling analysis, it was assumed that the
source location was the centroid of each county.
those in the previous form.
For Canada, the latest available emissions inventory was
Mercury Emissions Inventory
obtained (Niemi et al., 2001).  In this 1995 inventory there
were 583 point sources.  Area sources in this inventory
In any analysis such as this, ideally all critical information
were specified on a 50-km grid in the Great Lakes region
would be referenced to the same time period, including the
and a 100-km grid in the remainder of the country.  On the
emissions inventory, meteorological data, and ambient
473 50-km grid-squares and 1140 100-km grid-squares,
monitoring data for model evaluation.  Unfortunately, as
there was an average of approximately 22 area source
discussed below, there was no one year for which compa-
categories per grid square.  Analogous to the United States
rable emissions inventories for the United States and