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

-------