The
Mole
d Stud
For Acid Deposition
ation
and Fie
-------
EMEFS Sponsors
Atmospheric Environment Service,
Environment Canada (AES)
United States Environmental
Protection Agency (EPA)
- ENSR Consulting & Engineering, Inc.
- Environmental Monitoring
& Services, Inc.
- Desert Research Institute
- Research & Evaluation Associates
Electric Power Research Institute (EPRI)
- ENSR Consulting & Engineering Inc.
- Environmental Monitoring
& Services, Inc.
- Environmental Science
& Engineering, Inc.
- TRC Environmental Consultants
Florida Electric Power
Coordinating Group (FCG)
- Environmental Science
& Engineering, Inc.
Ontario Ministry of the Environment
& Energy (OMEE)
l+l
Environment Environnement
Canada Canada
Atmospheric Service
Environment de I'environnement
Service atmospherique
vvEPA
EPRI
Electric Power
Research Institute
Ontario
Florida Electric Power
Coordinating Group
Ministry of Ministere de
Environment I'Environnement
and Energy et de I'Energie
Issued under the authority of the Minister
of the Environment
Atmospheric Environment Service
© Minister of Supply and Services Canada, 1993
Catalogue No. Ens6-io6/i993E
ISBN 0-662-20832-3
Egalement disponible en franfais sous le titre
«Etude d'evaluation sur le terrain du modele
eulerien pour les depots acides»
This paper contains a minimum of 50%
recycled fibers, including 10% post-
consumer fibers.
-------
950R93003
Creating Confidence
The prevention of acid rain is an
expensive business. Costs to industry
are projected at $20 billion for the
U.S. and $1.7 billion for Canada.1 )ust
one scrubber to clean up the emis-
sions from a smokestack can cost
half a billion dollars. Society must
take into account many factors as it
fights acid rain. Decisions such as
how many control devices are need-
ed, which kind, and where to place
them for maximum effectiveness
should obviously be made with care,
calling on the best science available.
Because the processes involved
in acid rain are so complex, computer
models are invaluable in understand-
ing the cause-and-effect relationship
between the emission of acidifying
substances from particular sources
and the pattern of acid deposition
(source-receptor relationships), and
in predicting which proposed control
strategies can succeed in reducing
acid rain to acceptable levels.
In the past, simple Lagrangian
models were used to determine
source-receptor relationships. But,
following a series of U.S.-Canadian
meetings from 1979-81 concerning the
Memorandum of Intent on Trans-
boundary Pollution, it was decided
that Lagrangian models could not
reliably describe the nonlinearities
inherent in the real atmosphere.
Nonlinearity could mean, for example,
that a 50% reduction in emissions
would result in only a 30% reduction
in acid deposition on the target area.
If nonlinearity were this high, the cost
of controlling acid rain could double.
Participants in these meetings also
felt that the Lagrangian approach
lacked credibility in accounting for the
multiple sources and long distances
involved in the acid rain issue.
Measurements at sites, such as this
one operated by OMEE at Balsam
Lake, Ontario, were used during the
EMEFS evaluation to verify model
predictions for acid rain.
The Eulerian framework, which
can incorporate a more complex
level of science than the Lagrangian
approach, was therefore adopted by
both countries. Two Eulerian models
were developed: in Canada, the
Acid Deposition and Oxidant Model
(ADOM) was supported by the
Ontario Ministry of the Environment
and Energy, the Atmospheric
Environment Service of Environment
Canada, the U.S. Electric Power
Research Institute, and the German
Umweltbundesamt; while in the U.S.,
the Regional Acid Deposition
Model (RADM) was created by the
Environmental Protection Agency.
The model developers were
exploring new ground in terms of
computational techniques and the
scope of process representation. Any
error in the models' representations
of physical and chemical processes
-------
could bias predictions. So, just how
valid were these new model predic-
tions? The most direct answer would
come from a rigorous evaluation of
model performance against real world
observations, made over as wide a
range of atmospheric conditions as
possible. The evaluation would be
carefully designed to stress the model
performance and bring out any short-
comings or artifacts in their design.
The Eulerian Model Evaluation
and Field Study, or EMEFS, was the
appointed challenge. This precedent-
setting collaborative effort was one
of the largest environmental mea-
surement studies in North American
history, providing a comprehensive
air quality and precipitation chemistry
data base for model evaluation.
The EMEFS sponsors spent a
total of $20 million for model devel-
opment, and $35 million on
gathering the data for the evalua-
tion— a lot of money, yet a fraction
of the control costs for acid rain.
A project protocol was devised
specifying tests to address various
model application issues, as well as
the comparisons to be done between
model predictions and observations
in support of each test. Extensive
quality assurance activities were
implemented to determine and limit
the uncertainties in the evaluation
data set, so that these could be sep-
arated from uncertainties in model
predictions. An external review panel
of international experts was assem-
bled to periodically review the model
evaluation program and provide
counsel on its direction and progress.
These measures were taken to devel-
op confidence in the community at
large in the quality of the evaluation,
and eventually, in the fitness of the
Sneak preview -
scenario for the year 2010.
Predicted response to Canadian
and U.S. sulfur dioxide emission
controls. Simulation on the Eulerian
model RADM.
models for use in policy decisions.
As a result of EMEFS, we have
learned much about the process of
evaluating comprehensive air quality
models and about the adequacy of
our current understanding of atmos-
pheric processes.
The EMEFS database itself will
be of great interest to those investi-
gating empirical relations between
emissions and deposition and air
quality, and to those developing
models still more sophisticated than
RADM and ADOM. Its value lies in its
unique combination of geographical
breadth, chemical depth, defined
quality of the data, and two-year
time span. No other existing air qual-
ity data set shares this distinction.
-------
In the Field
Map of Sites
The EMEFS Surface
Measurement Network
Ground-based Network
From June 1988 to May 1990, air and
precipitation samples were gathered
daily in a network of over 100 sites in
eastern Canada and the U.S. Most of
the sites selected were free from the
influence of local sources of pollution.
Within the surface network, a
dense cluster of sites was introduced
across Pennsylvania and into New
York to test the models' ability to
define the strong gradients expected
in pollutant deposition patterns
throughout this region. This cluster
was named the GRAD network.
Another set of sites, called the VAR
network, was selected to probe sub-
grid variability, that is, how well a
measurement at a single location can
represent an area the size of a model
grid cell (80x80 km for RADM and
127x127 km for ADOM).
These networks were used to
gather data for what is called an
'operational evaluation' of the models.
Essentially, model predictions are
compared with field observations that
are averaged over several days to a
year, to answer the question: "Are we
getting good agreement with the
observations, on time scales of impor-
tance to acid rain effects?"
All of the stations in the network
measured the pH and major ions in
precipitation and operated rain
gauges. Almost all of the stations col-
lected 24-hour air samples using filter
packs for paniculate sulfates and
nitrates, as well as gaseous nitric acid
and sulfur dioxide. Most of the sites
provided continuous ozone measure-
ments, and in the U.S. (but not
Canada), local meteorological data. At
several sites, additional pollutants
were monitored, notably ammonia,
PAN, and nitrogen dioxide.
-------
The EPA sampling deck at the
Quabbin Reservoir in Massachusetts.
Routine Surface Measurements
NETWORK
OBSERVABLE EPA
GRAD VAR
Total Stations
Precipitation
Chemistry3
S(IV)
Depth (i h av)
Aerosol Particle3
Mass (< 2 urn)
Mass (<- hr av)
504=
N03~
NH4+
Gaseous
HN03 (24 h av)
502 (24 h av)
NH3 (24 h av)
N02 (24 h av)
N02 (i h av)
NOx (i h av)
03 (i h av)
Meteorology'3 (at 10 m)
3 24-Hour average pH,
14 11
14 11
8
14 11
4
14 11
14 11
11
14 11
14 11
14 11
14 11
14 11
conductivity
OTHER
43
43
i
43
29
43
43
36
43
43
43
43
36
(S04=, N03
b 3-Hour average wind speed, wind direction,
barometric pressure,
EPRI OMEE AES FCG
25 29 11 4
25 19 11 4
25 19 11 4
25
25 11 10 2
25 11 10 2
25 2
25 11 10 2
25 11 10 2
25 2
24 2
1
24 4 9 2
24
- Cr, NH4+, Na+, K+, Ca++, Mg++)
temperature, relative humidity,
insolation (EPA only)
Intensive Measurements
While data from the routine monitor-
ing program can be used to verify a
model's predictions, more specialized
data is required to determine whether
models perform well for the right
reasons, and not merely due to a
compensation of errors. This type of
testing examines the components of a
model individually, and is called
'diagnostic evaluation'.
To provide this added challenge
for ADOM and RADM, in the summer
of 1988 and spring of 1990, Canadian
agencies and the EPA conducted
intensive six-week measurement cam-
paigns, taking measurements from
five instrumented aircraft as well as
making augmented chemistry mea-
surements at six enhanced surface
sites. They were joined in this effort
in 1988 by the Fraunhofer Institute of
Germany and in 1988 and 1990 by the
-------
The skill with which RADM can simulate large-
scale regional distributions and pollutant
gradients aloft was assessed by comparing
the model output with measurements taken
from two planes executing the zipper and cur-
tain patterns. The drawing above shows the
interlocking curtain and zipper, while the
diagram on the right shows an actual flight
superimposed on the RADM grid.
U.S. National Oceanic & Atmospheric
Administration (NOAA) and the
Canadian Institute for Aerospace
Research (IAR). The aircraft flew during
both the day and night in various pat-
terns between the sites. This marked
the first time extensive aircraft data
had been collected for model evalua-
tion. The aircraft data allowed for test-
ing of the models' atmospheric fea-
tures and three-dimensional behaviour.
The aircraft data were subjected
to a rigorous quality assurance and
control program, and there were
opportunities for comparison through
side-by-side Canadian and U.S. flights.
Indeed, throughout the EMEFS, much
attention was given to meeting the
data quality objectives of precision,
accuracy, representativeness and com-
parability, through efforts such as
collocation of samplers and regular
laboratory intercomparisons.
•85
•80
75
•70
Above: A look inside the
Fraunhofer Institute's plane
Right: The EPA flight team.
-------
t •
Daily emissions estimates were provided for most
of the major point sources of sulfur dioxide, shown
on the map above. These estimates were combined
together with other information from remaining
sources to form daily snapshots of sulfur dioxide
emissions, such as the one for September 16, 1988
on the right.
Emissions Inventory
A high quality emissions inventory is
an essential input for proper model
evaluation. Emissions estimates
which closely represent actual emis-
sions were prepared by replacing
"typical" emissions from the National
Acid Precipitation Assessment
Program (NAPAP) 1985 emissions
inventory with hourly data for three
categories: major point sources,
mobile sources and biogenic sources.
Over 200 of the largest stationary
sources in the U.S. were monitored
for sulfur dioxide and nitrogen
oxides, along with the 15 largest
Canadian sources east of
Saskatchewan, thus accounting for a
large portion of sulfur dioxide and
nitrogen oxide emissions in the
region. (Note that in Canada, moni-
toring data for point sources were
confined to the two intensive mea-
surement periods). Plume rise from
point sources was calculated from
source characteristics in the emis-
sions inventory along with hourly
meteorological data. Mobile-source
emissions were adjusted to account
for the daily average temperature and
the daily temperature range. Natural
emissions of organic compounds and
nitrogen oxides were calculated
based on the type of surface land
cover and the hourly meteorological
data. Estimates of the uncertainty of
the emissions data for sulfur dioxide,
nitrogen oxides and volatile organic
compounds were also made.
Data Base
Once thoroughly validated, data from
all the participants were compiled
and archived in the Acid Deposition
Systems (ADS) data base at Battelle
Pacific Northwest Laboratories. This
data set may be obtained from any
of the cosponsoring agencies.
-------
The Verdict, So Far
4 5
OBSERVED
456
OBSERVED
Data from EMEFS set the stage for
the most extensive evaluation ever
attempted on regional air quality
models. This evaluation will continue
through 1993. As a result of the
early evaluation work, a number of
improvements have been made in the
initial versions of ADOM and RADM.
While their predictions for the
wet deposition of sulfate were fairly
good, both RADM and ADOM were
shown to underpredict sulfate
aerosol. Iterative testing on ADOM
indicated that this underprediction
most likely occurred because non-
precipitating clouds were not being
adequately represented in the mod-
els. This led to the discovery that the
role of non-precipitating cumulus
clouds in sulfate production was
quite important, a major finding of
the evaluation. ADOM and RADM
were improved accordingly, and the
Success: The predictive capability of
RADM improved after it was corrected
to include non-precipitating clouds.
The diagonal line would indicate an
exact correspondence between
simulations and observations. The
irregular line is a smoothed running
median of the values.
new models underwent a favourable
peer review in the spring of 1992.
Thanks to these evaluation
efforts and model improvements, a
sufficient understanding of the pre-
dictions of total sulfur loadings has
been achieved, so that RADM and
ADOM are now able to contribute to
policy assessment. For example, the
Canadian control program aims to
reduce the wet deposition of sulfate
to below 20 kg/ha/yr. Nonlinearity
was shown to be small enough by
the models (in which we now have
confidence) that the promised reduc-
tions in the U.S. and Canada of sulfur
dioxide emissions should prove suffi-
cient to meet this goal. Furthermore,
an evaluation of annual average sul-
fur deposition has shown that the
Eulerian models predict sulfur depo-
sition patterns as well or better than
Lagrangian models. The models are
-------
Predictions vs. Observations
The daily variations in the regional
average of sulfate concentrations
(M§/m3) as predicted by ADOM,
compared to EMEFS observations.
Vertical bars show the 95%
confidence interval
beginning to be used for predicting
total nitrogen loading, though there
is a slight bias towards overpredic-
tion. This is most likely because
there is a strong gradient in nitrate
deposition at the surface. The grid
size in both models was too large to
capture this process. Horizontal and
vertical grid resolutions are now
being improved to better describe
this gradient.
Model developers had also antic-
ipated that given the ADOM and
RADM large grid sizes, it would be
difficult to predict for oxidants. This
was confirmed in the evaluation.
Research continues, though. ADOM is
presently being adjusted with a
smaller grid-scale to analyze the cre-
ation and transport of smog through
the Windsor-Quebec corridor.
Furthermore, EMEFS experience will
be heavily relied upon for model
development and field data collection
in the upcoming U.S. Southern
Oxidants Study. Other possible appli-
cations for ADOM and RADM include
addressing visibility, emissions trad-
ing schemes and air toxics.
Thus, having gained experience,
a firmer technical foundation and the
confidence of both the scientific com-
munity and policy-makers, the EMEFS
has placed us in a far better position
to address emerging air quality prob-
lems, and provides a blueprint for
effective international research col-
laboration on these issues.
-------
For further information
Several reports discuss the planning,
execution and model evaluation results
of the EMEFS. Here is a selected list:
"The EMEFS Project Plan"
Available from EPRI
(address below)
"The National Acid Precipitation
Assessment Program Report 5:
Evaluation of Regional Acidic
Deposition Models and Selected
Applications of RADM" (1990)
Can be purchased from:
Superintendent of Documents
Government Printing Office
Washington, DC 20402-9325
(202) 783-3238
"The EMEFS Model Evaluation: An
Interim Report" (1991)
Can be purchased from:
National Technical Information Service
U.S. Department of Commerce
5285 Port Royal Road
Springfield, VA 22161
(7<>3) 487-4650
The following reports will be published
in 1994: "Characterization and Usage
of Sub-grid Scale Variability in Model
Evaluation", written by and available
from EPA, and "The Eulerian Model
Evaluation and Field Study Quality
Assurance Synthesis of Network Data,
Volume i", written by all the sponsors
and available from EPRI.
Contacts;
Jason Ching
Environmental Protection Agency
Atmospheric Sciences Modelling Division
Air Resources Laboratory
Research Triangle Park, NC 27711
Robin Dennis
Environmental Protection Agency
Atmospheric Sciences Modelling Division
D. Alan Hansen
Electric Power Research Institute
3412 Hillview Ave.
P.O. Box 10412
Palo Alto, CA 94303
John Jansen
Florida Electric Power Coordinating Group
P.O. Box 2625
800 Slades Creek Parkway
Burmingham, AL 35202
Norman Kaplan
Environmental Protection Agency
Pollution Control Division
Maris Lusis
Ontario Ministry of the
Environment & Energy
Air Resources Branch
125 Resources Road, West Wing
Etobicoke, ON MgP 3V6
P.K. Misra
Ontario Ministry of the
Environment & Energy
Air Resources Branch
Keith Puckett
Atmospheric Environment Service
Air Quality Research Branch
Environment Canada
4905 Dufferin St.
Downsview, ON M3H
Robert Vet
Atmospheric Environment Service
Air Quality Research Branch
Environment Canada
Notes
1From "Regulatory Impact Analysis of
the Final Acid Rain Implementation
Regulations", EPA, 1992 and "The
1990 Canadian Long-range Transport
of Air Pollutants and Acid Deposition
Report - Part 7, Socio-economic
Studies", The Federal/Provincial
Research and Monitoring Coordinating
Committee (RMCC), 1990.
Acknowledgments
Many thanks to Tina Shapiro for
writing the text and overseeing its
production, to Daniel Pokorn for his
able translation, and to Rushton,
Green and Grossutti Inc. for their
design work.
Photo Credits:
Cover (left to right): the EPRI site in
Shawano, Wl taken by Carol
Westmoreland-Pounds of ENSR; the
AES sponsored Twin Otter, operated
and photographed by The Institute
for Aerospace Research (IAR); and
the EPA sponsored site at Quabbin
Reservoir, MA, run and photographed
by ENSR.
Page i: the OMEE site at Balsam
Lake, ON, taken by Celine Audette.
Page 4: the Quabbin Reservoir site,
taken by ENSR.
Page 5: the EPA flight team pho-
tographed by Chet Spicer of Battelle
Pacific Northwest Laboratories; and
the inside of the Fraunhofer
Institute's plane, taken by the
Institute.
------- |