United States
Environmental Protection
Agency
Atmospheric Sciences
Research Laboratory
Research Triangle Park NC 27711
Research and Development
EPA/600/S8-85/016 Sept. 1985
SEPA Project Summary
User's Guide for the Advanced
Statistical Trajectory Regional
Air Pollution (ASTRAP) Model
Jack D. Shannon
The Advanced Statistical Trajectory
Regional Air Pollution (ASTRAP) model
simulates long-range, long-term trans-
port and deposition of air pollutants,
primarily oxides of sulfur and nitrogen.
The ASTRAP model is designed to com-
bine ease of exercise with an appropri-
ate detail of physical processes for as-
sessment applications related to acid
deposition. The theoretical basis and
computational structure of the ASTRAP
model are described. Major simplifica-
tions and assumptions incorporated in
the model are discussed.
The data requirements for ASTRAP
simulations are monthly to seasonal
time series of transport wind and pre-
cipitation analyses and an emissions in-
ventory. ASTRAP consists of three pro-
grams: HORZ, VERT and CONCDEP.
The source code is in standard FOR-
TRAN, while the JCL is appropriate for
an IBM 3033 mainframe computer.
Horizontal dispersion and wet deposi-
tion statistics are calculated in HORZ.
The process of turbulent vertical dif-
fusion within the mixed layer, leakage
to the free atmosphere, chemical trans-
formation and dry deposition are calc-
ulated in VERT. The CONCDEP pro-
gram combines the statistics produced
by HORZ and VERT with an emissions
inventory to calculate primary and
secondary pollutant surface concen-
trations along with wet and dry dep-
ositions.
This Project Summary was devel-
oped by EPA's Atmospheric Sciences
Research Laboratory, Research Triangle
P-rk, 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 infor-
mation at back).
Introduction
The Advanced Statistical Trajectory
Regional Air Pollution (ASTRAP) model
described here has been developed to
simulate the long-term (monthly to
yearly), regional-scale (resolution about
100 km) deposition of oxides of sulfur
and nitrogen, the major contributors to
acid deposition. The ASTRAP tech-
niques can be extended to other pollu-
tants, provided that linear parameteriza-
tions of chemical transformation and
removal processes are suitable. They
also can be extended to other spatial
scales, if appropriate meteorological
data are available. Unmodified exten-
sion of the modeling techniques used in
ASTRAP to shorter, episodic temporal
scales is not recommended because of
certain statistical features of the model.
The ASTRAP model consists of three
submodels and various preprocessors
and postprocessors. Use of particular
preprocessors and postprocessors de-
pends on application and data availabil-
ity. The submodels are the vertical diffu-
sion program (VERT), the horizontal
dispersion program (HORZ), and the
concentration and deposition program
(CONCDEP). The processes of vertical
diffusion, chemical transformation, and
dry deposition are simulated in VERT
through parameterizations that are in-
dependent of horizontal location or par-
ticular meteorological conditions. In
HORZ, the processes of horizontal ad-
vection, horizontal diffusion, and wet
deposition are simulated through the
-------
use of time series of wind and precipita-
tion analyses. The CONCDEP program
combines the statistics produced by the
first two programs with emission inven-
tories to produce fields of average at-
mospheric concentration and cumula-
tive deposition.
The ASTRAP model can be applied in
emission policy assessments, for which
the model can be exercised to predict
how deposition fields would change in
response to changes in the pollutant
emission field. Assuming linearity be-
tween emissions and deposition, the
model can be used to estimate concen-
tration and deposition at specific recep-
tor locations, and it can estimate the in-
dividual contribution from different
sources or source regions. The follow-
ing major simplifications and assump-
tions have been incorporated into the
ASTRAP model:
1. Long-term horizontal and vertical
dispersion can be simulated inde-
pendently.
2. Long-term horizontal diffusion can
be approximated by the spread of
plume centerlines; small-scale diffu-
sion about individual plumes is
ignored.
3. Chemical transformation can be
parameterized as a linear, first-order
process.
4. Wet removal is a function of the half-
power of the precipitation.
5. Transport is two-dimensional.
6. The dry deposition parameterization
is horizontally uniform.
The data requirements for ASTRAP
simulations are monthly to seasonal
time series of transport wind and pre-
cipitation analyses and an emissions
source inventory. Initial preparation of
wind and precipitation fields for AS-
TRAP simulations has been performed
at the University of Michigan by Perry
Samson. His initial wind fields, pro-
duced every 12 h are for 500-m layers
upto3000-m MSLfora 17 x 19 horizon-
tal grid of National Meteorological Cen-
ter (NMC) spacing. Speed is given in
meters per second. Linear temporal in-
terpolation has created fields at 6-h in-
tervals, and the wind analyses have
been combined into three-month-long
files (December-February, March-
May, June-August, and September-
November). The wind components
have been internally converted to com-
ponents along the NMC axes in the
HORZ subprogram.
The precipitation analyses produced
at the University of Michigan are on a 50
x 45 grid with 1/3 NMC spacing. The
original analyses are hourly; there is a
special code for missing data. Precipita-
tion amount is given in millimeters. The
hourly fields have been added to pro-
duce six hourly fields; the missing data
were filled by interpolation and extrapo-
lation. The precipitation data are ar-
ranged on monthly files.
An S02 inventory has been created in
which the seasonal emissions in kilo-
tonnes are given by effective stack layer
and NMC position of the lower left cor-
ner of the grid cell. No separate SO|
emissions inventory is used, so primary
sulfate emission factors are applied in
CONCDEP. The primary sulfate emis-
sion factor for sources in the lowest
layer (0-100 m) is assumed to be 0.05
(i.e., one unit of SO2 equivalent emis-
sion is treated as 0.95 units of S02 and
1.5 (0.05) = 0.075 units of primary sul-
fate; the 1.5 factor arises because the
ratio of the molecular weight of sulfate
to that of S02 is 96/64). The primary sul-
fate emission factor for point sources is
assumed to be 0.03 in Florida and the
northeast and 0.015 elsewhere. The
emission grid has the same spacing as
the precipitation grid, but it is irregular
because the inventory is arranged by
state and province. The emission infor-
mation has been derived from a prelim-
inary version of the National Acid Pre-
cipitation Assessment Program
(NAPAP) inventory for 1980. Wind, pre-
cipitation, and emission data are all
written in binary format for efficiency.
Technical Description
Horizontal dispersion statistics are
calculated in HORZ for a virtual source
grid covering the contiguous United
States and Canada for a particular mete-
orological period and are subsequently
interpolated when concentrations and
depositions are calculated. For a one-
to-three-month sequence of meteoro-
logical analyses, simulated tracers of
unit mass are released from the sources
at 6-h intervals. Calculations in HORZ
are independent of the height of re-
lease. The tracers are tracked for 28 time
steps (seven days) or until they leave
the wind grid. At each successive tracer
position, the precipitation field is
checked to see whether there should be
wet removal.
The statistics calculated in HORZ are
ensemble statistics; each ensemble rep-
resents all trajectory positions of a par-
ticular plume age for each source for the
length of the meteorological analysis.
The statistics generated for a puff de-
scribing the density of the ensemble of
equal age trajectory end points are the
coordinates p.xand jxy of the mean posi-
tion, the standard deviations ax and cry,
the correlation term pxy, and n, the num-
ber of equivalent tracer masses con-
tributing to the ensemble statistic. The
statistics are collected for a puff associ-
ated with airborne or dry tracers and for
another puff associated with wet depo-
sition tracers.
The wet deposition is parameterized
as a function of the half-power of the
precipitation. It contains certain con-
straints such that any precipitation
amount larger than 1 cm/6h has the
same effect as that of 1 cm/6h, while
precipitation amounts less than a mini-
mum threshold value of 1 mm/6h have
no removal effect at all.
As previously mentioned, there are
separate sets of statistics for wet and
dry tracer ensembles. The same trajec-
tory end points are used in calculation
of each corresponding pair of wet and
dry ensembles, but the weights are dif-
ferent (most of the wet ensembles have
zero weight) and, thus, the wet and dry
ensemble puffs differ. The puffs for dry
deposition and surface concentration
tend to be more regular in the trend of
their overlapping positions than do the
wet deposition puffs. This is because
wet deposition is a highly irregular pro-
cess and thus exhibits more statistical
variation for a single season.
The process of turbulent vertical dif-
fusion in the mixed layer, leakage to the
free atmosphere, chemical transforma-
tion, and dry deposition are calculated
in the VERT program. The diurnal varia-
tion of the planetary boundary layer is
parameterized in VERT through a sea-
sonally and diurnally varying stability
profile .(vertical eddy diffusjvity speci-
fied for each layer of the model).
The variations in dry deposition
amounts associated with typical diurnal
and seasonal patterns of both atmos-
pheric stability and surface resistances
are parameterized through average di-
urnal values of dry deposition velocities
for each season. A diurnal variation in
the linear first-order transformation rate
of the primary pollutant (S02 or NO/
N02) to a secondary pollutant (SOJ or
N03) directly or indirectly due to photo-
chemical activity patterns is also com-
pensated for in VERT. The seasonal pat-
terns are intended to include all
chemical transformation pathways ex-
cept those associated with precipitating
clouds. The rates of S02 transformation,
for example, are greater than those nor-
-------
mally found in clear-air experiments,
because the rates include the effects of
chemical processing in nonprecipitating
clouds.
Leakage from the mixed layer into the
free atmosphere is now parameterized
in ASTRAP, but the rates have been set
at relatively low values until more is
learned about the long-term regional
significance of the layer.
The calculated and stored normalized
statistics from the VERT program for a
seasonal simulation include the one-
dimensional surface concentration, the
pollutant mass remaining aloft, and the
pollutant mass deposited by dry pro-
cesses during the time increment. Sepa-
rate sets of statistics are maintained for
S02, primary and secondary SOJ, as
well as N0/N02, primary and secondary
NO3. The eddy diffusivity, chemical
transformation, and dry deposition
parameterizations for a particular sea-
son are applied everywhere, regardless
of latitude or surface vegetation. This is
an obvious limitation of the ASTRAP al-
gorithms, which include some oversim-
plifications to achieve computational
simplicity and efficiency.
The CONCDEP program combines the
statistics produced by HORZ and VERT
programs with an emissions inventory
to calculate primary and secondary pol-
lutant surface atmospheric concentra-
tions and wet and dry depositions. The
concentration and deposition fields are
calculated by overlaying puffs and
adding their densities over the receptor
grid. For this, the puffs must be
weighted by both the emissions rate per
6 h for the source and the number of
normalized tracer masses contributing
to the puff.
Emissions from a hypothetical source
in Oklahoma were combined with sum-
mer vertical dispersion statistics and
with trajectory statistics for June
through August, 1980, to perform a sea-
sonal simulation in a test of the model.
Arrays of the concentration and deposi-
tion fields and a table of total deposition
for each receptor (state or province)
were generated.
Computer Aspects
As currently structured, ASTRAP con-
sists of three programs: HORZ, VERT,
and CONCDEP. The job control lan-
guage is appropriate for the IBM 3033
mainframe computer at the Argonne
National Laboratory; it would require
some modification for use on other sys-
tems. While programming is in stand-
ard FORTRAN, input and output
algorithms may require coding modifi-
cation for other systems. On an IBM
3033, VERT requires 210k (bytes) stor-
age, 5-10 min of CPU time, and one
output file for a seasonal simulation.
HORZ requires 500k storage, 10 min
CPU time, two input tape drives, and an
output file for a seasonal simulation.
CONCDEP requires 500k storage, 10-15
min CPU time, an emission input file, the
output files from the other two sub-
programs, a file used to identify each
receptor cell as a state or province, and
an output file for a seasonal simulation.
The output file from CONCDEP is nor-
mally stored on disk, where postproces-
sors can later be used to display the
results graphically.
Jack D. Shannon is with Environmental Research Division of Argonne National
Laboratory. Argonne. IL 60439.
Terry L. Clark and Jason K. S. Ching are the EPA Project Officers (see below).
The complete report, entitled "User's Guide for the Advanced Statistical
Trajectory Regional Air Pollution (ASTRAP) Model," (Order No. PB 85-236
784/AS; Cost: $11.50, 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 Officers can be contacted at:
Atmospheric Sciences Research Laboratory
U.S. Environmental Protection Agency
Research Triangle Park. NC 27711
-------
United States
Environmental Protection
Agency
Center for Environmental Research
Information
Cincinnati OH 45268
Official Business
Penalty for Private Use $300
EPA/600/S8-85/016
0000329 PS
U S ENVIR PROTECTION AGENCY
REGION 5 LIBRARY
230 S DfARBORN STREET
CHICAGO IL 60«04
------- |