A ''.
United States
Environmental Protection
Agency
Atmospheric Sciences •*«
Research Laboratory "^
Research Triangle Park, NC 27711 '
Research and Development
EPA/600/S3-88/017 July 1988
&EPA Project Summary
Simulated Effects of
Hydrocarbon Emissions
Controls on Seasonal Ozone
Levels in the Northeastern
United States: A Preliminary
Study
Robert G. Lamb
The Regional Oxidant Model
(ROM) was developed to assist the
formulation of emissions control
strategies for attaining the primary
ozone standard. The ROM was
designed to simulate hourly con-
centrations under meteorological
conditions associated with the
highest ozone levels during a year.
Typically, these conditions span a
period of one to two weeks in the
summer months. Recently, efforts to
promulgate a secondary ozone
standard, which would protect crops
and forests, have created the need
for a seasonal ozone model; specif-
ically, a model capable of estimating
the effects of emissions changes on
the frequency of multi-hour aver-
aged ozone concentrations during
the ozone "season," the seven-
month period from 1 April to 31
October of each year. Using a
method developed here, the pre-
dicted ozone changes that result
from emissions changes are extrap-
olated from three, 2-week periods to
the entire ozone season in 1980. The
implementation of the extrapolation
method is preliminary, but the details
of the technical foundation of the
method are described fully.
This Project Summary was
developed by EPA's Atmospheric
Sciences Research Laboratory, Re-
search Triangle Park, NC, to announce
key findings of the research project
that is fully documented in a separate
report of the same title (see Project
Report ordering information at back).
Introduction
The Regional Oxidant Model (ROM)
was developed to assist the formulation
of emissions control strategies for
attaining the primary ozone standard
Since the primary standard is expressed
in terms of a maximum one-hour
averaged ozone value that is not to be
exceeded more than once per year, the
ROM was designed to simulate hourly
concentrations under meteorological
conditions associated with the highest
ozone levels during a year Typically,
these conditions span a period of one to
two weeks in the summer months A
model capable of simulating concen-
trations with hourly resolution is often
referred to as an episodic model
Recently, efforts to promulgate a
secondary ozone standard, which would
protect crops and forests, have created
the need for a seasonal ozone model,
specifically, a model capable of
estimating the effects of emissions
-------
changes on the frequency of multi-hour
averaged ozone concentrations during
the seven-month period from 1 April to
31 October of each year.
The most accurate way of
generating seasonal concentration sta-
tistics is simply to run an episodic model
for the entire season. This brute force
approach is presently impractical in the
case of a model like the ROM that
requires 2 hours of CPU time on an IBM
3090 mainframe computer to simulate a
single day. However, this technique will
eventually become feasible as computing
power continues to increase and com-
puting costs continue to decline.
A second method of modeling
seasonal concentrations is to run an
episodic model for selected short period
intervals within the season, and then
"extrapolate" the results to the entire
season. This is the approach that we will
develop in this paper. This method also
requires the adoption of basic assump-
tions, but their impact on the accuracy of
the model overall can be controlled to a
large extent through proper design of the
extrapolation scheme and judicious
selection of the subinterval episodes.
The principal objective of this project
is to describe estimated effects of VOC
emissions changes on seasonal ozone
levels in the Northeastern United States.
The study is based on ROM simulations
of three, 2-week episodes in April, July
and August of 1980, extrapolated to the
ozone season, 1 April - 31 October,
using a scheme developed here Each of
the three episodes was simulated (and
the results extrapolated to the season)
for each of three emissions distributions:
the base, or reference, case; and two
"control strategies." The latter consist of
different percentage changes in VOC
emissions that could be achieved under
control plans contained in the 1982 SIPs.
The model results are regarded only
as interim estimates, partly because they
are based on a very simplified imple-
mentation of the seasonal extrapolation
scheme; and partly because the 1980
emissions inventory, which is used as
the base state in these simulations, is not
regarded to be of high enough quality to
serve as a basis for control strategy
development. Furthermore, the verifi-
cation analyses of the ROM are not yet
complete, and therefore the accuracy of
the model has not yet been established.
Both the ROM verification studies and
the development of a significantly im-
proved emissions inventory (for 1985)
are expected to be completed by late
1988
Extrapolating the Results of
Simulated Emissions Control
Strategy Effects from an
Episode to a Season
Next, we consider the problem of
extrapolating concentration changes, or
delta concentrations as we shall refer to
them, simulated by an episodic model
over a period of 3 to 4 weeks to a season
consisting of 7 months.
First we define basic terms:
Season = 7-month period from the
beginning of April to the end
of October;
Episode = the period, usually 3 to 4
weeks in length and not
necessarily continuous, for
which model simulations of
emissions controls are
performed.
Considering only species A for the
purpose of discussion, let
AA(x)
be some average concentration
difference or some net benefit of controls
predicted by the model for a given point
x in the episode. Our task is to estimate
the value
AA(x,t)
that the model would predict at the same
point averaged over the season. Strictly
speaking we are not concerned at this
point with estimating what the true value
of
AA(x)
actually is. This question is related to the
issues of model accuracy, predictability,
and so on which are not topics of
concern in this analysis. The principal
objective is the development of a method
of extrapolating model results outside the
simulation period irrespective of the
accuracy of the model.
There are three basic problems that
obstruct the extrapolation of the model
results from the episode simulation to the
entire season (1) differences in the
meteorology in the episode and season,
(2) differences in the chemistry of the
delta concentrations in the episode and
season, and (3) differences in the
emissions changes AS within the
season. The last problem is generally not
significant because any changes in
emissions that are required under a
control plan apply equally throughout the
year.
When one considers the problems
created by differences in the meteorol-
ogy and differences in the chemistry, it
appears that the largest extrapolation
errors will occur when the receptor in
question is influenced strongly during the
season by emissions sources other than
those that impacted it during the episode.
This suggests that if the estimation of
seasonal effects is the principal goal of
the model simulations, the episodes
should be selected so that they contain
at least one sample of each of the
meteorological states that cause the
major source areas in the vicinity of a
given receptor to affect the concen-
trations that receptor observes.
A Proposed Extrapolation
Procedure
Consider three inert tracer species,
T-), T2, and T$ emitted by three different
source areas in the model domain. For
any receptor site x in the model domain
and any hour t in the season we define a
binary number B(x,t) = (b-|,b2,b3) whose
elements bn are binary numbers defined
as follows:
bn = 1 if the concentration of Tn at
(x,t) is nonzero;
bn = 0 otherwise.
Thus, B(x,t)
= 010 if TT = T3 = 0, T2*0 at (x,t),
= 000 if TT =T2 = T3 = Oat (x,t);
= 111 if T^Ta, T3*0at (x,t); etc
Keep in mind that the function B is
derived from the output of the model
produced by a simulation of the three
inert tracer sources over the entire
season with no chemical reactions
Consider a receptor, R, at which we
are interested in extrapolating the results
of simulated control strategy effects from
an episode to the full season. The
influence of the three source areas on
concentrations at R over the season can
be summarized in the form of a Karnaugh
table as shown in Table 1 Here we are
Table 1. Representation of the Function B
Evaluated at Receptor R Inte-
grated Over the Season. Numbers
Indicate the Percentage of Hours in
the Season When B(x=R,t) Had the
Value Indicated. A Dash (-) in the
Table Indicates Less than 1
Percent Occurrence of that Value
bi
0
t
00 01
55
24
11
15
2
10
4
-
using the Karnaugh table (which is
borrowed from logic analysis) as a basis
-------
for expressing the potential influence that
each major source and combinations of
major sources can have on a given
receptor. The numbers in Table 1 are the
percentages of time in the season that
the function B, as determined by the
model predictions, had the value
specified by the row and column
coordinates. In order to display B, which
has three elements bt, bg and b$, on a
2-D chart, element b-| is split out to
form the row entry to the table and the
pair b2b3 make up the column entry.
Thus, the number 55 in the upper left
corner of Table 1 indicates that during
55% of the season all three tracers T-| ,T2
and TS were simultaneously absent at R,
i.e., b-|b2b3 = 000. The number 24 just
below the 55 indicates that 24% of the
season, T-| was predicted at R but at
those times neither T2 nor T3 was
present (B = 100). Similarly, T2 and T3
were present at R with T-| absent (011)
15% of the time; T2 was present with T-|
and T3 absent (010) 4% of the time; and
on 2% of the hours in the season all
three tracers were predicted at R
simultaneously. The total of the entries in
the table is 100 percent. We can
construct tables like this for any number
of sources, although for numbers larger
than n = 8 the tables become quite
unwieldy. Note that the number of
elements, or slots, in a table is 2n, where
n is the total number of source areas
represented.
We define the dominant entries of a
table as the first M members of the
rank-ordered set where M is the
number such that the sum of the first M
entries is closer to some limit, say 90%,
than the sum of the first M-1 or m +1
entries. Using the example shown in
Table 1 as an illustration, the rank
ordered entries are 55, 24, 15, 4, 2. The
sum of the first three of these is 94,
which is closer to 90 than the sum of the
first two (55 + 24 = 79) or the first four.
Hence, 55, 24 and 15 are the dominant
entries.
We define the dominant elements of
the table to be the elements associated
with each dominant entry, and we
indicate the dominant elements by
checks in a dominant element table.
Table 2 shows the dominant elements
corresponding to Table 1.
The Karnaugh tables that we have
discussed up to this point were compiled
using the inert tracer concentrations
simulated by the model for the entire
season of interest. Thus, we will refer to
these as seasonal tables and to their
elements as B = (bib2ba). By applying
the same procedures to any subset of
Table 2. Dominant Elements, Indicated by an
X of the Karnaugh Table, Table 1,
for Receptor R
bi
0
1
00 01
X
X
11
X
10
the simulated tracer concentrations over
the season, we can produce episode
tables. We will refer to the elements of
an episode table as B' = (b'ib'2b'3).We
emphasize that the subset of the season
that is used as the episode does not
have to be continuous in time. One can
select one time interval at one point in
the season and intervals of different
lengths at different times and treat the
combined set as a single subset or
episode. We also want to emphasize that
the seasonal and episode tables pertain
to specific receptor sites, and in general
the entries in either table will vary from
receptor to receptor
We now advance the following
hypothesis:
The simulated effects of emissions
controls can be extrapolated from an episode
to a season for any receptor site whose
episode table covers its seasonal table.
We say that Table A covers Table B if all
dominant elements of B are also
dominant elements of A. Thus, in simple
terms this hypothesis states that the
extrapolation process is possible if the
dominant element table of the episode
contains an X in every box that the
seasonal table has one. The episode
table may contain more checks (dom-
inant elements) than the seasonal table,
but this is not necessary to achieve
coverage.
Suppose that we have found an
episode whose dominant elements cover
those of the season shown in Table 2
Since there is a unique value of B' =
(b'-ib^b's) at R for each hour of the
episode, we can identify all those hours
in the episode associated with each of
the three dominant elements (b'-|b' b's)
= (000), (011), and (100). Let Epoq be
the predicted total benefit of emissions
controls at R for all hours associated with
(000) in the episode, and let E0n and
EIQO be the corresponding benefits
predicted for all hours associated with
(011) and (100), respectively. The benefit
might be the average delta concentration,
e.g.,
where the summation is over all K hours
t|< in the episode for which (b'ib'2
b'3) = 011, and AA is the predicted
change in the concentration of a
particular species; or the benefit might be
some dollar value of crop yields. In any
event, the corresponding benefit E at R
for the season is extrapolated to be
i(R)=eoooEooo+eoiiEon + eiooEioo
where e is the percentage occurrence
of (bib2D3) = (000) in the season, as
given in the present example by the
entries m Table 1 (6000= 55). Similar
meanings apply to 6011 and e-ioo
(eioo = -24. eon =.15). This is the
essence of the proposed extrapolation
technique.
In practice we usually do not have
the luxury of being able to change the
modeling episode at will. Often the
episode has already been selected and
modeled, and our task is to extrapolate
the results of that simulation to the
season for any or all receptor sites in the
model domain. In general we will find that
the episode table does not cover the
seasonal table at every receptor At those
locations we can attempt to achieve
coverage by one of two methods (1) we
can ascertain whether those dominant
elements in the seasonal table that are
not covered by elements in the episode
table are irrelevant insofar as the eval-
uation of the control strategy of interest is
concerned, and therefore are "remov-
able" elements; or (2) we can attempt to
"fill in" necessary elements in the
episode table by borrowing from the
episode tables of other receptors.
Interim Implementation of the
Extrapolation Scheme
A rigorous implementation of the
seasonal extrapolation method just
described would require that the ROM
generate the inert tracer concentration
fields, used to define the states B, for the
entire season. This is not possible at the
present time, so we present here an
interim implementation of the scheme
that utilizes the ATAD trajectory model.
We also adopt several strong assump-
tions to simplify the analysis. These
include the neglect of vertical wind shear,
vertical turbulent diffusion, and vertical air
motion including the vertical motion
associated with convective clouds We
also resolve the spatial and temporal
variations in the tracer concentration
fields with limited resolution. We treat
blocks of 5x5 ROM cells as single
"receptors"-this has the effect of
smoothing the tracer concentrations over
-------
roughly 100 km subdomains-and tracer
concentrations are sampled every 6
hours rather than every 30 minutes as
the ROM would provide.
In order to minimize the data
handling requirements of the interim
scheme, we consider only 27 source
sites, selected because they are the
locations where the largest absolute
reductions in VOC would occur under the
control strategy identified as #1 in the
on-going rural ozone studies. Beginning
at hour 0000 Z on 1 April, and at every
subsequent six-hour interval 0006,
0012, etc., until 2400 Z 31 October, a
particle is released from each of the 27
source sites and tracked by the ATAD
model for a period of 5 days, or until it
passes outside the model domain. Every
six hours a count is made of the particles
in each receptor. These consist of 98
rectangular cells arranged in a 12x9
array. When the particle counts are
performed, a separate running tally is
maintained for each of the 27 sources. At
the end of the season, the source with
the highest total count at a given
receptor is the source that affects that
receptor most frequently. The source
with the next highest count is the second
most significant source, and so on. In
this way we identify for each receptor
(I,J) the eight most significant sources.
This information is then used to define
the state variable (B(l,J,tn) =
bsbybgbsb^babi described earlier.
Recall that the bk are binary numbers
with b|< = 1 if at time tn, where tn is any
hour in the season, tracer material is
present at receptor (I,J) from the k-th
significant source. Otherwise, bk = 0.
For example, if at a given time, tracer is
present from the most significant source
(k = 1) simultaneously with material from
the second and eighth most significant
sources, then B(l,J,tn) = 10000011. If
there were no tracer material present at
tn from any of the top eight sources,
B(l,J,tn) = 00000000.
Note the following important aspects
of B: (1) it is a discrete valued function
with 28 = 256 possible values; (2) the
specific sources (of the set of 27)
represented by each bk may differ from
one receptor to another; (3) the most
significant source is represented by the
rightmost bit of B while the 8th most
significant source is represented by the
leftmost bit; and (4) at each receptor
there is a unique value of B associated
with each hour of the year.
As we noted earlier the ozone
season is currently defined as 1 April
through 31 October In the present study
the episode consists of the following
three disconnected periods: 14-29 April,
12-26 July, 20-31 August. Table 3
contains the seasonal/episode state
tables for two receptors, (I,J) = (8,6) and
(8,7), computed by the procedure
described above using meteorological
Table 3. Sample Episode/Seasonal State Table for Two Receptor Sites (I.J) See Text for Explanations of the Table Entries
SEASONAL MID EPISODE KARIAU
SEASON* 11 AFRIL-11
SEASON SUM- 156
TABLE FOR RECEPTOR
OCTOBER 1»IO)
ROW LABEL (1.1,1,1)
REPRE
111.
EENTS SOURCES
5, 1, 7)
/ oooo
0000 / 17:419 19
«001 / 2:11
0010 / 4:13
0011 / 3
0100 / 5:19
0101 / 1:5
0110 / 1
0111 / 1
1000 / 1:20
1001 / 1:
1010 / 1:
1011 /
1100 /
1101 /
1110 /
1111 /
0001 0010 0011
:102 1:31 2:19
2
2
1:5 1:1
1:1 1
( I- 1 J- « )
EPISODE-I 12-24
EPISODE stm- is»
COLUMN LABEL (1,1,1
0100 0101 0110
(:41 1 1
2:7
1:5
2
1
1:1
1 1
1:5
1
1 1
JULI , 20-
,1 IKE PRESENTS
0111 1000
2:25
1:2
1:2
1
1:1 1
1:1
1
1 AUGUST, 1
SOimCES(10.17,
1001 1010
2:10 7
2
1
1:2
1
1
4-29 APRIL, If 10)
12.1O
1011 1100 1101
4:19 1:3
1
1
1
1
1:1 1
1110 1111
1 1
1
1
SEASONAL AND EPISODE KARNAU TABLE FOR RECEPTOR (I- • J- 7 )
SEASON.(1 APRIL-11 OCTOBER 1910) EPISODE-I 12-24 JULY,
SEASON SUM- 156 EPISODE SUN- 159
ROW LABEL (1,1,1,1)
REPRESENTS SOURCES
( 2,17, 5, 9)
/
0000 / 95
0001 /
0010 /
0011 /
0100 /
0101 /
0110 /
0111 /
1000 /
1001 /
1010 /
1011 /
1100 /
1101 /
1110 /
1111 /
0000
:500
4:27
1:4
1:1
1:24
4
1:1
1:4
1:1
1
0001
(:66
1:5
1:4
1:2
1:1
1
COLUMN
0010 0011 0100
1:26 5:20 3:21
1:4 1 2
1
2 1
1
111
1:2
LABEL
0101
7:11
1:2
1
1
(1,1,1
0110
3:10
1:5
1
1:2
1
1
,1 (REPRESENTS SOURCES! 7,10,12,16)
0111 1000 1001 1010 1011 1100
10 2:14 1:3 2:9 2
2:2
1:6 1 2 1 2
1
1
2
1
1 1:2 1
1 1:1
1:1
1101 1110 1111
1
1
1
1:1
1:1
1 1
-------
data for 1980. The column entry of each
table is the rightmost four bits of B(I,J),
i.e., b4D3b2bi, which represent the four
most significant sources at receptor (I,J);
and the row entry of each table is the
leftmost four bits, bsbybebg. The specific
sources represented by each bit are
shown just above the column and row
headings.
Each table has three types of
entries: (1) two numbers separated by a
colon, (2) a single number, or (3) a blank.
A blank signifies that the state B
represented by that particular column/
row entry in the table did not occur in the
season, nor the episode. A single num-
ber signifies that the state occurred in
the season but not the episode. In this
case the integer shown is the number of
occurrences, in units of quarter days (6
hours), of that state in the season. Fin-
ally, an entry of the form n:m indicates
the number of occurrences of the state
found in the season (m) and the episode
(n). The state tables provide an easy
means of assessing how well the chosen
episode represents the season, at least
insofar as the source impacts are
concerned.
Since there is a unique value of B
associated with each hour, regardless of
whether the hour is inside or outside the
episode, for extrapolation purposes we
can equate unmodeled hours in the
season to modeled hours in the episode
using the B values as the basis. Another
way of saying this is that for any receptor
(I,J) we can map the simulated effects of
emissions controls at a given hour tn of
the episode into any hour tm of the
season as long as B(l,J,tn) = B(l,J,tm).
Table 4 shows the mapping for the first
12 days of the season at 12 receptors
(I.J), 1 si s12, J = 2, derived from the
interim implementation of the
extrapolation scheme that we have
described to this point.
The second column from the left
contains 5 digit numbers of the form
DDDHH where ODD is the Julian date in
the season (1980), and HH is the ending
hour (EST) of a 6-hour block. Columns
3-14 contain the corresponding day and
hour in the episode from which model
results can be used to estimate benefits
at time ODDHH at the receptor identified
at the top of the column. The mapping
relationships shown in Table 4 were
generated under the following
constraints:
(1) The 6-hour morning period ending
at 1300 EST can be extrapolated
Table 4. Episode-to-Season Extrapolation Matrix for the 12 Receptors Indicated by the Column Headings for the First 12 Days of the Season.
See Text for Further Explanation
NJWIM MCTOU rot uccms
1
M
n
tcu- |
l_
10
11
12
1)
14
IS
1C
n
11
i»
20
21
22
ZJ
24
JS
2(
27
21
2*
10
11
32
1)
14
IS
1C
17
11
it
40
41
42
41
44
45
4(
47
41
4*
XME-( ODOUR)
O-JULIAX DATI
-(-IOU*. rtXIOD
(CST) UDIin
*201
*207
*211
1219
»}01
*307
*313
1119
>401
9407
»413
Ml*
9S01
9S07
9S13
9519
9(01
9607
9613
9(19
9701
9707
9711
9719
9101
9107
9111
9119
9901
9907
9911
9919
10001
10007
10011
10019
10101
10107
10113
10119
10201
10207
10211
10219
10101
10107
10111
10119
10401
1
1
| ( 1,2)
10*01
10(07
10(11
10(19
10701
10707
1(713
10719
10101
11001
11013
1111*
11201
11301
11411
11419
11501
11(01
11(11
11(19
11701
12001
19(11
1971*
19101
19(07
19111
19119
19*01
19907
1*91)
19919
20001
20007
2001)
2001*
20101
20107
2011)
2011*
20201
20207
2*21)
20219
20)01
20)07
20111
2011*
20401
I 2,2)
11207
11401
11411
1141*
11(01
11(07
11*1)
1*41*
19(01
20107
*
2)31*
2)401
21407
21411
0
21*07
24001
24011
24019
24101
24107
24111
2411*
24101
24107
24111
24119
24401
10(07
10(11
10(19
10701
10787
10713
10919
11107
11201
11111
11319
11401
11407
11411
11419
11(01
11(07
11(11
1*41*
20001
( 1,2)
10(07
11(07
20S1)
21119
23401
23*01
21*13
24219
23(01
0
0
2)71*
23101
23107
23113
2421*
24301
10707
10713
1111*
11407
12007
20013
2021*
20301
10107
20313
2041*
23101
21*07
21111
2421*
24101
24107
2411)
24)19
24401
10(07
10(13
10(1*
10701
10*01
11313
1171*
1*701
1*707
1*713
1*719
19101
( 4,2)
10(01
19(01
20111
24119
24201
24207
11113
11519
11(07
1
0
1*41*
19501
19S07
20111
20119
21401
23701
21111
21119
21901
11(07
11(11
11(19
1*501
19S07
20111
2011*
20201
20207
20211
2021*
2*101
20307
20314
2041*
20 SOI
20107
23513
23S1*
24101
24201
11113
1141*
1*501
19507
20113
2011*
20201
1 S.2)
10(01
11707
1*513
1*519
19707
20107
20213
24119
1*707
0
1141)
11*1*
12001
1*501
1*(13
19(19
19701
1*101
11713
0
12001
1*501
1X13
11(19
20107
20201
20213
20219
20301
20307
2*311
2031*
20407
20501
20713
2341*
23S01
23S07
23S13
23S1*
23(07
24201
24213
2421*
24307
24401
10(13
10(11
10701
I (,2)
12001
19(01
20(13
23111
0
24107
24113
1111*
11701
12001
2341)
0
23S07
23(01
24013
2411*
24401
10707
10713
1071*
10101
20(01
20713
0
24301
24401
10(13
10(1*
10701
10001
10*11
1101*
11107
11101
1*71)
lill*
19907
20107
20211
2021*
20101
20307
20313
20S1)
0
24301
24313
2431*
24401
( 7,2)
0
0
23(11
2141*
19(07
0
1*111
1**19
20701
21407
21711
21519
24101
24107
24111
24119
24401
10(01
10(11
10719
11101
11107
0
24119
20501
20101
24111
24119
24401
10(01
10(11
10(1*
10701
10707
11111
11*1*
11*01
11*07
20111
2041*
20(01
24001
20111
2041*
20(01
24201
20111
2041*
20101
( *,2I
10(07
10(11
10(1*
10101
10*01
11313
«
0
1*(01
19(11
11119
11401
11(01
11(11
11(19
11707
0
11111
11(1*
0
10*07
10*11
0
0
20107
20111
2031*
20401
20407
20413
2041*
20501
20507
20S1)
20519
20*07
23407
2)413
1101*
11101
0
«
1»«1»
1*901
1*907
20113
2031*
20401
I »,2)
10(07
10(13
10(11
10701
11001
11013
2071*
20107
23401
23*13
!*(!>
1*101
1*107
1991)
1*919
2*007
10707
11013
11019
19707
24001
0
10719
10*07
0
11*13
11*1*
11101
11201
11513
11(1*
11701
11707
11711
11*1*
11*01
11*07
11*11
11*1*
12001
1M01
0
2401*
24401
0
10(11
10(19
10701
11*, 21
10(07
10(11
10(1*
10701
10(07
10*11
0
11107
1*507
19513
19519
20(07
21401
21411
21419
21S01
0
19(11
0
19507
11001
0
0
0
24007
0
24219
24)01
24107
24313
24319
24401
10(01
10(13
10(19
10701
10*07
10113
10(1*
11007
11101
0
24019
24101
24301
24313
24319
24401
111,2)
10(07
10(13
10(1*
10701
10707
10113
1101*
0
11101
11111
1111*
11401
11507
11511
11519
0
19907
0
1*419
19701
19707
0
19719
10901
11)01
11411
11419
11507
11(01
11(13
11(1*
11701
11707
1171)
1171*
11101
11(07
11(13
11*1*
11*01
11*07
11*11
24019
11107
0
0
1111*
114*1
(12,2)
10(07
10(11
10(1*
10701
10707
10(11
10(1*
0
11007
0
!*(!»
1*701
1*707
1*713
1*719
19*01
19(07
24011
24119
24201
24207
24211
19(19
19701
19707
19711
19719
1M01
1M07
1*(11
1*11*
1*901
19907
19913
lt*li
20001
20007
20113
2011*
20201
20207
20211
0
20101
0
0
2011*
2*401
-------
only from 6-hour morning periods
in the episode.
(2) The 6-hour afternoon period ending
at 1900 EST can be extrapolated
only from 6-hour afternoon periods
in the episode.
(3) The nighttime period 1900-0700
can be extrapolated only from
nighttime periods in the episode.
These constraints help maintain accurate
diurnal variations in concentration pat-
terns. To help maintain realistic temporal
variations over multiple days, the
episode states B(l,J,tn) are examined
sequentially in forward time order, i.e., tn
increasing in unit step, from the point tno,
say, that was extrapolated into the
immediately preceding quarter-day of
the season. To help maintain realistic
seasonal variations in concentration only
the April episode period is used to fill in
the Spring (1 April - 20 June) and
autumn (20 September - 31 October)
months, and only the July and August
episodes are used to fill in the Summer
(20 June - 20 September).
The "0" entries in Table 4 indicate
that the state 6(l,J,tn) prevailing at the
given receptor and time in the season
did not occur in the episode, under the
constraints described above. There are
innumerable ways to "fill in" these gaps
of missing values. Here we have chosen
simply to march across the gap by
continuing the episode time sequence
that was last in effect. If in this process
we reach the end of the episode period
before we have completely crossed the
gap, "we wrap around" and continue the
interpolation from the beginning of the
same episode period. In cases where the
season begins with missing states, as is
the case at receptor (7,2) as indicated in
Table 4, the first mapping time is taken
from any one of the closest receptors
whose factor is not missing. In this way
we obtain a complete mapping of
episode into the season.
The extrapolation sequence derived
for each receptor and each hour in the
quarter-day intervals in the episode is
mapped one-to-one into the hours in
the season. When the final values are
applied to the concentration differences
predicted by the model for given
emissions changes, the outcome is
hourly values of delta concentrations at
each hour of the season in each ROM
grid cell. These values may then be
processed in any desired manner to
arrive at estimates of the seasonal
impact of given emissions changes.
-------
RG. Lamb is on assignment to the U.S. Environmental Protection Agency from „
the National Oceanic and Atmospheric Administration, U.S. Department of
Commerce.
K.L Schere is the EPA Project Officer {see below).
The complete report, entitled "Simulated Effects of Hydrocarbon Emissions
Controls on Seasonal Ozone Levels in the Northeastern United States: A
Preliminary Study," (Order No. PB 88-195 706/AS; Cost: $14.95, 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:
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/S3-88/017
0000329 PS
U S EHVIR PR0TSCTIQH AGENCY
REGION $ LIBRARY
230 S DEARBORN STREET
CHICAGO It 60604
------- |