United States ,
Environmental Protection
Agency
Environmental Sciences Research
Laboratory
Research Triangle Park NC 27711
Research and Development
EPA-600/S7-83-058 Jan. 1984
&ERA Project Summary
Sulfur Deposition Modeling in
Support of the U.S./Canadian
Memorandum of Intent on
Acid Rain
T. L Clark and D. H. Coventry
At the request of the U.S./Canadian
Work Group 2 of the Acid Rain Memo-
randum of Intent, the Eastern North
American Model of Air Pollution
(ENAMAP-1) was applied to simulate
the monthly wet and dry depositions
and monthly averaged ambient concen-
trations of SOzand SO< for January and
July 1978 across eastern North
America. Using these model results,
unit emissions (1.0 Tg S yr'1) transfer
matrices, which describe source/
receptor relationships, were generated
and a model performance study was
undertaken. In addition, a model sensi-
tivity study was conducted to examine
the consequence of model input param-
eter uncertainties.
This Project Summary was developed
by EPA's Environmental Sciences Re-
search Laboratory, Research Triangle
Park, NC, to announce key findings of
the research project that is fully docu-
mented in a separate report of the same
title (see Project Report ordering in-
formation at back).
Introduction
In the mid 1970's, SRI International
developed a Lagrangian-puff air pollution
model, European Regional Model of Air
Pollution (EURMAP)forthe Federal Envi-
ronment Office of the Federal Republic of
Germany (Johnson, et al., 1978). This
regional model was capable of calculating
monthly S02 and SO« concentrations and
dry and wet deposition patterns and
international exchanges of sulfur across
13 countries of western and central
Europe.
In the late 1970's, SRI International,
sponsored by the U.S. Environmental
Protection Agency (EPA), adapted and
applied EURMAP to eastern North
America. The adapted version of this
model. Eastern North American Model of
Air Pollution (ENAMAP), was capable of
calculating monthly SOzandSO^concen-
trations and dry and wet deposition
patterns and interregional exchanges of
sulfur across a user-defined number of
regions (Bhumralkar et al., 1980). Thus, it
was possible to assess the impact of
sulfur emissions from individual sites and
provinces on the sulfur concentrations
and depositions across the same regions.
In 1981, the Atmospheric Sciences and
Analysis Work Group (Work Group 2) of
the U.S./Canadian Memorandum of
Intent on Transboundary Air Pollution
included the ENAMAP-1 model as one of
eight Lagrangian long-range sulfur
pollution models to be applied. Work
Group 2 requested ESRL to apply the
ENAMAP-1 model using January and
July 1978 input data to generate transfer
matrices, assess model performance, and
analyze model sensitivity in input param-
eters. This report summarizes the ESRL
work.
Model Description
Parameterizations
In the design and development of any
air quality simulation model, there are
usually two conflicting goals: maximum
realism and accuracy on one hand, and
minimum computational requirements on
the other. Greater realism and accuracy
-------
usually require more detailed information
and sophisticated formulations of physical
processes, which in turn require more
computer time and memory capacity. It
was clear from the outset that the com-
puter requirements could be severe for
two reasons:
1. The model must treat a very large
geographical area (2870 km north to
south and 3220 km east to west) and
yet preserve acceptable spatial reso-
lution (70 by 70 km).
2. The model must compute monthly
and annual mean concentration and
deposition fields while preserving the
original temporal resolution (12 h) of
standard meteorological data; thus,
the model must make repetitive
calculations for long sequences of
input data.
Accordingly, as a first step, it was
desirable to design a very simple model
having minimum computer requirements
(i.e., a practical and economical model
that would offer acceptable realism in
simulating the most important processes
involved in the transboundary sulfur
pollution problem). More sophisticated
parameterizations can replace the sim-
plistic approaches described here as
knowledge of the appropriate physical
processes accumulates. Later, the sim-
plistic parameterizations of vertical mixing,
dry and wet depositions, and transforma-
tions of S02 will be replaced by more
sophisticated parameterizations by the
end of 1982. Also by the end of 1982, NO,
chemistry will be added to the model.
Results
Model Applications
To accomplish its goals. Work Group 2
requested that each of the eight long-
range transport models generate the
following model output:
1. annual 1978 unit transfer matrices
of wet sulfur deposition, dry sulfur
deposition, ambient S02 concentra-
tions, and ambient SOi concentra-
tions normalized by a 1.0 Tg S yr~1
emission rate from each of the 40
source/receptor regions and 9 sensi-
tive receptors defined by Work Group
2, and
2. January. July, and annual 1978 wet
sulfur depositions and average ambi-
ent SOUconcentrations at monitoring
sites selected by Work Group 2.
The unit transfer matrices, although
strongly influenced by the meteorological
scenarios used in the simulation period,
were to be used by another work group to
assess the merits of several emission
scenarios. From these transfer matrices,
the effects of emissions from individual
regions on the sulfur depositions and
ambient concentrations across sensitive
receptor areas (Boundary Waters, Algoma,
Muskoka, Quebec, southern Nova Scotia
in Canada and northern New Hampshire,
the Adirondacks, central Pennsylvania,
and Great Smoky Mountains in the United
States) and 40 source/receptor regions
of Canada and the United States could be
determined.
The January and July 1978 meteoro-
logical data used to generate the transfer
matrices were analyzed and gridded by
preprocessors as described in the pre-
vious section. The emissions from each
grid cell within a given source/receptor
region were multiplied by a constant
factor so that the total annual sulfur
emissions from that region equaled 1.0
Tg.
Since the meteorological data and
analyses for the entire year of 1978 were
unobtainable in the rather brief time
frame imposed upon the work group, it
was agreed that ENAMAP-1 would be
applied using only Januaryand July 1978
input data. Annual estimates of the
transfer matrices would be based only on
matrices for those two months. In addi-
tion, since the ENAMAP-1 domain did not
include western North America, only 35
instead of 40 source/receptor regions
were considered.
The ENAMAP-1 January and July 1978
unit transfer matrices are presented in
two appendices in the final report. The
transfer matrices indicated that the
sources within any 1 of the 35 regions
contributed significantly to the sulfur
depositions and concentrations within
that region. In January, an average of
68% of SO2 wet deposition in a particular
region resulted from SO2 emissions from
that region. Similarly, in July, an average
of 64% of the S02 wet deposition in a par-
ticular region resulted from S02 emis-
sions from that region. The ENAMAP-1
results indicated that much of the sulfur
wet deposition consisted of S02 wet
deposition (in some cases, wet deposition
of SO2 was an order of magnitude greater
than that of S0<) and that "local" sources
are significant in sulfur wet deposition.
Model Evaluation
An ideal assessment of the perform-
ance of any regional sulfur model would
require an extensive data base of dry
sulfur depositions, wet sulfur depositions,
and ambient SC>2and SC>4concentrations
measured during all seasons of the year
across all portionsof the modeling domain
that are removed from the effects of local
sources. Unfortunately, such a data base
does not exist. However, daily average
concentrations of SC>2 and 364 from 1
Canadian and 53 U.S. sites were available
from the Electric Power Research Insti-
tute's Sulfur Regional Experiment (EPRI-
SURE). Monthly sulfur wet deposition
data for these periods were available
from several Multistate Atmosphere
Power Production Pollution Study
(MAP3S) sites in the northeastern U.S.
and from about a dozen Canadian Network
for Sampling Precipitation (CANSAP)
sites. Together, these data formed the
best available regional sulfur modeling
evaluation data base for North America
during this period.
Work Group 2 screened the data on
ambient SO* and sulfur wet deposition
for January and July 1978. For the
MAP3S network, precipitation samples
with a catch of less than 50% of a nearby
rain gauge measurement were ignored.
Otherwise, the precipitation sample
amount was adjusted to the rain gauge
measurement. Valid monthly samples
required a minimum 90% capture. For
CANSAP, 7 of the 16 operational sites
were ignored because of operational or
siting problems. Valid monthly samples
required a minimum operational time of
20 and a minimum collection efficiency of
25%. Collection efficiency was defined as
the ratio (%) of the precipitation recorded
by the sampler to that measured by a
colocated rain gauge. Since even rain
gauges do not collect all of the precipita-
tion (precipitation collection efficiency is
a function of gauge type, site exposure,
and wind velocity), the rain gauge precipi-
tation data from both networks were
adjusted.
The S02 emission data used in the
evaluation study were obtained from
several sources, since a complete 1978
emission inventory was not available.
Because much of the S02 was emitted
from electric power plants, it was imper-
ative to accurately define the 1978 power
plant emissions. With this in mind.
Environment Canada prepared a 1978
S02 emission inventory for large point
sources. Emissionsfrom all other sources
were assumed to be the same as in 1976.
The 1978 S02 emissions from the U.S.
power plants were estimated from fuel-
use records, while the 1978 emissions
from all other sources were assumed to
-------
be the same as in 1980, the year for
which emissions data were available.
The emission data set was more precise
than the precipitation data set. Due to the
lack of extensive precipitation measure-
ments tn Canada in the winter, the 3-h
precipitation amounts used in the
ENAMAP-1 applications were extrap-
olated for much of the Canadian portion
of the grid domain. The data from the
Canadian sites showed less agreement
with the monthly ENAMAP-1 precipita-
tion amounts across the grid cells en-
compassing each site.
Because the measured data set was
inadequate for concrete conclusions.
Work Group 2 computed the values of
selected statistical parameters to assess
the performance of the models for
January, July, and annual 1978. Models
applied to only the two months were not
evaluated for the annual period. This
evaluation exercise was expected to
identify models that consistently per-
formed unacceptably.
Table 1 presents the values of the
statistical parameters calculated for the
ENAMAP-1 results. The residuals were
determined by subtracting the calculated
value from the observed value. The mean
residual is a good measure of a model's
ability to correctly calculate the higher
and lower observed values. In general, a
model is said to overpredict the observa-
tions when the mean residual is less than
zero. For a "good" model, the correlation
between the residual and the calculation
is low.
Based on the correlation of the residual
and calculation, the model performed
best for July sulfur wet deposition and
January ambient SOl concentrations.
Furthermore, the values of the mean
residuals were comparable.
Model Sensitivity Study
Every numerical simulation model pro-
duces results that do not concur exactly
with observations. These discrepancies
are due partially to uncertainties in the
values of model input parameters (e.g.,
dry deposition rates, scavenging coeffi-
cients, mixing heights, transformation
rates, etc.). Atmospheric scientists cannot
reach a consensus on a single appropriate
value for each of these model input
parameters, but a consensus can be
reached on a general range of "accep-
table" values.
Since there are many "acceptable"
values, the modeler is faced with the
problem of selecting one value to use in
model applications. The magnitude of this
Table 1. Values of Statistical Parameters Calculated for the ENAMAP-1 Results
Parameter
Total Sulfur Wet Deposition (kg ha V
January 1978 July 1978
Sample size
Mean of observation
Mean of calculation
Mean residual
Standard deviation of residual
Correlation of residual and calculation
5
0.84
1.00
-0.16
0.31
-0.62
11
1.20
0.87
0.33
0.46
0.18
Average Ambient SO4 Concentrations (jig m )
January 1978 July 1978
Sample size
Mean of observation
Mean of calculation
Mean residual
Standard deviation of residual
Correlation of residual and calculation
29
6.8
6.3
0.5
1.8
0.14
47
11.6
11.8
-0.2
4.9
-0.36
problem is proportional to the model
sensitivity, or the degree of observed
change in the model calculation from a
unit change in the model input value. If
the model proves sensitive to a certain
model input parameter, the modeler must
carefully select the value of that param-
eter. A model sensitivity study identifies
those model input parameters whose
values must be chosen carefully.
In a sensitivity study, the model is
applied many times; the value of one of
the model input parameters is changed
each time, and the resultant changes in
model output are examined. Typically,
this involves the consumption of consid-
erable computer time. To minimize this
expense, this sensitivity study used an
abbreviated version of ENAMAP-1 and
only 3 values of each of the 15 model
input parameters (Table 2). Two of the
three values represented the low and high
values of the "acceptable" range of each
parameter, while the third value (base
case) was the value used in past
ENAMAP-1 applications.
The abbreviated version of ENAMAP-1
considered (1) only one emission source,
which emitted puffs of S02 and S0~4 at
12-h increments, (2) a continuous 1.0
mm h~1 precipitation rate, and (3) a
uniform transport wind from the south-
west. All other parameterizations were
preserved, with the exception of the SO2
and SOi dry depositions, which reflected
the parameterization used in a version of
ENAMAP-1 currently under development.
This parameterization is based on atmos-
pheric stability and land-use characteris-
tics. The region between southern Ohio
and eastern Quebec was selected. The
Adirondacks sensitive area is located in
the middle of this region, 700 km from the
source.
Conclusions
Any conclusions drawn from only two
1-month periods would not likely apply to
much longer periods. However, for this
evaluation data set, the model performed
rather well. The ENAMAP-1 January
mean sulfur wet deposition was slightly
greater than the monthly mean of the
measurements at the five sites. The
ENAMAP-1 July mean was slightly less
than the mean measurement. The mean
ambient SOi concentrations calculated
by the model (6.3 and 11.8 //g rrT3 for
January and July, respectively) compared
very favorably with the mean measure-
ments (6.8 and 11.6/ug m"1 for January
and July, respectively). The mean resid-
uals for both January and July were less
than 1.0 fjtg irr3. Except for the case of
sulfur wet deposition in January, the
absolute values of the correlation be-
tween the residuals and the model calcu-
lations were less than 0.40, which
indicated that the model performed well
for the two 1 -month periods.
The model sensitivity study assessed
the changes in the model output at 100-
km increments downwind of a single
source due to changes in the values of
one of the model input parameters. Some
significant conclusions of this study were:
1. The 3-h time step used in previous
ENAMAP-1 applications led to a
saw-toothed distribution in the model
output for moderate and high wind
speeds (> 20 km h"1).
2. A 2-h time step led to a saw-toothed
distribution in the model output for
-------
Table 2. Model Input Parameter Values Considered in the Sensitivity Study
Parameter Case 1 Base Case
Computational time step (h)
Wind speed (km h" )
Precipitation rate (mm /i"V
Puff expansion rate (km2 /»"V
Transformation rate (% /T1/
Mixing height (km)
Initial puff size (km)*
SOz wet deposition rate (% mm'1)
SOl wet deposition rate (% mm'*)
SOz dry deposition factor ft
SO 4 deposition factor ft
SOz emission rate (kton />"V
SO* emission rate (kton h'1)
SOz loss rate from mixed layer top (kton /?~V
SO* loss rate from mixed layer top (kton h'1)
1.0
8.1
0.05
29.0
0.50
1.00
60.0
10.0
1.0
0.67
0.167
0.29
0.018
0.005
0.005
3.0
24.2
0.10
36.0
1.00
1.45
70.0
28.0
7.0
1.00
1.000
0.36
0.000
0.000
0.000
Case 2
2.0
40.3
o.ts
42.0
1.50
1.90
80.0
46.0
13.0
2.22
1.830
0.43
0.036
0.010
0:010
'The initial area of each puff is defined as the area of a square of sides EMISCELL
ttThe land-use, stability-dependent dry deposition rates are multiplied by these factors.
3.
only the high wind speed (> 35 km 9.
A 1-h time step did not remove the
saw-toothed distribution in the model
output for the high wind speed, but it
did reduce the amplitude of the
fluctuations.
Within 700 km of the source, S02
wet deposition was sensitive to the
S02 wet deposition rate; at a dista nee
of 100 km from the source, S02 wet
deposition increased about 200 to
about 800 mg m~2 resulting from an
increase in the SO2 wet deposition
rate of 0.10 to 0.46% mm"1
6.
7.
8.
Beyond 200 km from the source, SO*
wet deposition was sensitive to the
SO2 wet deposition rate; at a distance
of 1 500 km from the source, SO* wet
deposition decreased about 25 to
about 10 mg m~2 for the same
increase in the SO2 wet deposition
rate.
Beyond 100 and 300 km, the ambient
concentrations of SOz and SO*, re-
spectively, were sensitive to the SOz
deposition rate.
The SOz and SO* wet deposition
rates affected wet deposition and
concentrations more than the
changes in the SO2 dry deposition
rate (from 67 to 220% of the base
case value).
Changing the SO*dry deposition rate
(from 1 6.7 to 1 83.0% of the base
case value) affected the SO* concen-
trations and wet depositions in the
same way as changing the SOz wet
deposition rate.
The consideration of a small SO*
emission rate, 18th'1 or 5% of the
base case SOz emission rate, resulted
in significant increases in SO* con-
centrations and wet deposition; at
200 km from the source, deposition
and concentration increased 35 and
48%, respectively.
The model sensitivity study also asses-
sed changes in model output at 100-km
increments downwind of a single source
due to changes in all the model input
parameters except those relating to
meteorological and emission scenarios.
This assessment showed that the SO*
wet depositions and SOz concentrations
calculated using the base case values
were very similar to those calculated
using the high case values. Furthermore,
the SO* concentrations beyond 500 km of
the source calculated for the base case
were greater than those calculated for
the other two cases.
References
Bhumralkar, C. M., R. L Mancuso, D. E.
Wolff, R. A. Thillier, K. C. Nitz, and W. B.
Johnson (1980). ENAMAP-1 Long-Term
Air Pollution Model: Adaptation and
Application to Eastern North America.
U.S. Environmental Protection Agency,
EPA-600/4-80-039, 93 pp.
Johnson, W. B., D. E. Wolff, and R. L
Mancuso (1978). Long-Term Regional
Patterns and Transfrontier Exchanges of
Airborne Sulfur Pollution in Europe.
Atmos. Environ., 12:51-527.
The EPA authors. T. L. Clark (also the EPA Project Officer, see below) and D. H.
Coventry are with the Environmental Sciences Research Laboratory, Research
Triangle Park, NC 27711.
The complete report, entitled "Sulfur Deposition Modeling in Support of the U. S. /
Canadian Memorandum of Intent on Acid Rain." (Order No. PB 84-122 837;
Cost: $14.50, subject to change) will be available only from:
National Technical Information Service
5285 Port Royal Road
Springfield, VA 22161
Telephone: 703-487-4650
The EPA Project Officer can be contacted at:
Environmental Sciences Research Laboratory
U.S. Environmental Protection Agency
Research Triangle Park. NC 27711
------- |