EPA-450/3-75-069-a
July 1975
PHOTOCHEMICAL
OXIDANT MODELING
Volume I - Techniques Applicable
to Highway System Evaluation
U.S. ENVIRONMENTAL PROTECTION AGENCY
Office of Air and Waste Management
Office of Air Quality Planning and Standards
Research Triangle Park, North Carolina 27711
-------
EPA-450/3-75-069-a
PHOTOCHEMICAL
OXIDANT MODELING
Volume I -
Techniques Applicable
to Highway System Evaluation
by
F. A. Record, R. M. Patterson, D. A. Bryant, and A. H. Castaline
GCA Corporation
Bedford, Massachusetts 01730
Contract No. 68-02-1376, Task 14
EPA Project Officer: Thomas McCurdy
Prepared for
ENVIRONMENTAL PROTECTION AGENCY
Office of Air and Waste Management
Office of Air Quality Planning and Standards
Research Triangle Park, N. C. 27711
July 1975
-------
This report is issued by the Environmental Protection Agency to report
technical data of interest to a limited number of readers. Copies are
available free of charge to Federal employees, current contractors and
grantees, and nonprofit organizations - as supplies permit - from the
Air Pollution Technical Information Center, Environmental Protection
Agency, Research Triangle Park, North Carolina 27711; or, for a fee,
from the National Technical Information Service, 5285 Port Royal Road,
Springfield, Virginia 22161.
This report was furnished to the Environmental Protection Agency by GCA
Corporation, Bedford, Massachusetts 01730, in fulfillment of Contract
No. 68-02-1376, Task 14. The contents of this report are reproduced
herein as received from GCA Corporation. The opinions, findings, and
conclusions expressed are those of the author and not necessarily those
of the Environmental Protection Agency. Mention of company or product
names is not to be considered as an endorsement by the Environmental
Protection Agency.
Publication No. EPA-450/3-75-069-a
11
-------
ABSTRACT
This report presents a comparative discussion and evaluation of the
modeling techniques available to assess the photochemical oxidant air
quality impact of highway system modification or development. Five
attributes of eight modeling techniques are discussed. The eight
methods considered, in order of increasing complexity, are: (1) VMT
Change, (2) Linear Rollback, (3) Nonlinear Rollback: Appendix J, (4)
Gifford-Hanna Photochemical Model, (5) Statistical Relationships:
Repro-Modeling, (6) Reactive Environmental Simulation Model, (7)
Diffusion Kinetics Model, and (8) Urban Air Shed Photochemical Simula-
tion Model. For each of these the following five attributes are
discussed: (1) applicability and reliability, (2) data and manpower
requirements, (3) use of the technique, (4) limitations, and (5)
special features. The information presented in this report gives
guidance towards choosing a model best suited to a given need, based
on compatibility with these attributes.
iii
-------
CONTENTS
Page
List of Figures vi
List of Tables vii
Sections
I Introduction and Summary 1
II VMT Change 6
III Linear Rollback 13
IV Nonlinear. Rollback: Appendix J 32
V Gifford-Hanna Photochemical Model 45
VI Statistical Relationships: Repro-Modeling 55
VII Reactive Environmental Simulation Model 64
VIII Diffusion Kinetics Model 77
IX Urban Air Shed Photochemical Simulation Model 92
-------
LIST OF FIGURES . . ,
No. Page
1 Maximum 1-Hour Average Oxidant Concentrations as a 34
Function of 6- to 9-a.m. Averages of total Hydrocarbon
Concentrations at CAMP Stations, June Through September,
1966 Through 1968 and in Los Angeles, May Through
October 1967
2 Maximum Daily 1-Hour Average Oxidants as a Function of 35
6- to 9-a.m. Averages of Nonmethane Hydrocarbons at
CAMP Stations, June Through September, 1966 Through
1968, Los Angeles, May Through October 1967
3 Comparison of Curves for Total Hydrocarbons and Non- 36
methane Hydrocarbons With Appendix J
4 Comparison of Appendix J and Linear Rollback 37
5 Upper Limit of Maximum Daily Oxidant at Three Los Angeles 39
County Stations, May Through October 1967
6 Comparison of Total Hydrocarbon Versus Oxidant Curve of 41
AP-84 With Curve for Denver
7 Comparison of Nonmethane Hydrocarbons Versus Oxidant 42
Curve of AP-84 With Curve for Los Angeles
8 Variation of Diffusitivity With Height for Different 88
Stability Conditions
vi
-------
LIST OF TABLES
No. Page
1 Summary Matrix of Photochemical Oxidant Modeling 4
Techniques and Attributes
2 Data Requirements and Procedure Outline 18
3 Comparison of Observed Hourly Regional Average Concentra- 47
tions of RH, N02, and 03 With Concentrations Predicted by
the Gifford-Hanna ATDL Model for Los Angeles, September
30, 1969
4 Comparison of Observed Hourly Average Concentrations of 48
03, HC, and CO With Concentrations Predicted by the
Gifford-Hanna ATDL Model at Five Locations in Denver,
Colorado, August 13, 1973
5 Impact of Federal Emission Control Standards Mobile 59
Sources Oxidant (in pphm)
6 Effects of Single Day Traffic Reduction Policy on 60
Air Quality (Oxidant, pphm)
7 Input Variable Constraints 62
8 Chemical Dynamics Module for REM 67
9 Estimated Maximum Time and Costs for Obtaining Traffic and 72
Area Source Inputs for REM
10 Correlation Coefficients and Regression Equation for DIFKIN 78
in Los Angeles
11 Percent Change in Predicted Concentrations of 03, NO, N02, 80
and HC for a 50 Percent Increase in Each of Various Model
Parameters
12 Vehicular Emissions Data Required for DIFKIN 83
13 Stationary Source Emissions Data Required for DIFKIN 84
14 Photochemical Reactions and Rate Constants Used in DIFKIN 85
vii
-------
SECTION I
INTRODUCTION AND SUMMARY
This report presents a comparative discussion and evaluation of the avail-
able photochemical oxidant modeling techniques applicable to highway sys-
stm evaluation. The remainder of this report treats each of the eight
techniques in separate sections, starting with the simplest and proceeding
to the most complex.
The information presented in this report provides guidance towards choosing
a model suitable to a given need. While not laying down strict guidelines,
the report facilitates the choice of an appropriate model by discussing
the following attributes of each modeling technique:
Applicability and reliability
Data and manpower requirements
Application of the technique
Limitations
Special features
Once criteria are set for each of these attributes the choice of a model
may become obvious. Alternatively, the discussion of these attributes
presented in this report may help clarify the criteria to be met.
The following modeling techniques for assessing photochemical oxidant
impact are evaluated in this report:
1. VMT Change
2. Linear Rollback
-------
3. Nonlinear Rollback: Appendix J
4. Gifford-Hanna Photochemical Model
5. Statistical Relationships: Repro-modeling
6. Reactive Environmental Simulation Model
7. Diffusion Kinetics Model
8. Urban Air Shed Photochemical Simulation Model
The first three techniques are simple relationships between a surrogate
for oxidant precursor concentrations and expected maximum oxidant con-
centrations. The third technique involves the Appendix J type curves
and is the first to acknowledge the nonlinear relationship between pre-
cursor and oxidant concentrations. None of these techniques requires the
use of a computer.
The Gifford-Hanna model is a transition from these simple models into the
more complex ones. It may or may not require a computer depending on the
characteristics of its use. It is the first technique in the list to
include specific chemical reaction mechanisms.
Repro-modeling is a hybrid technique which develops statistical relation-
ships between input parameters and calculated concentrations from another
model. To date, a repro-model has been built only for Los Angeles based
on the input and output of the Urban Air Shed Photochemical Simulation
Model.
The last three models are all large computer codes., They treat atmo-
spheric transport and diffusion processes, and atmospheric chemistry, with
varying detail and computational complexity. Their predictive capabilities
also vary, although this is verified only for Los Angeles. Techniques (6)
and (7) are trajectory models which compute pollutant concentrations in
a column of air being advected across a region. Thus, at any time, con-
centrations are estimated at only one (horizontal) spatial point and
conversely. The last technique is a grid type model which calculates
-------
concentrations at a number of points at any given time. These three
techniques represent the state-of-the-art of photochemical oxidant
modeling.
Table 1 presents a summary matrix of oxidant modeling techniques and
the attributes listed above. The information presented in this table is
discussed in the remainder of the report.
-------
w
H
§
01
rH
3
H
1
:
t
]
"
B
4
"\
t
J
:
k
!
>
s
i
4
;
i
s
DIFKIN
'
OB
4)
1
U
M
i-l
U
M
Gifford-Hanna
Nonl Inear
rollback
*«
4 4
S£
_] O
u
f
£
u
41 4 1 1*4
T* e « *-4
O W O *4 u 4
D.«£* & ° 0 ! £ 8 JS "°X d
4 « o ^ TS jo £ *> « u i? 8
O4 W 4 u r* w > O » -o 41 O i'U
O £ 4 41 4 *« W W W B r"« 4
B ^ »W " W J *O «) O t* *J - 1 -*.. .-^
41 4* K t* - d d 4. O 3 3 d ° 3
X ** O O> U ifi > K * P 4 *J -* W f-*«
M O fal
g J, ,
SS 8 2 a «
X* b4J4» oo* «
x a. d u v4** « <>
o 3 c PQ -H*-«4 d oj
« WO BO*O 6 vi
U O U C W B *»
O «- *M. O *J 1 H
41 at 4>d vv? 0 *'
»> TJ4JO wow&cr « »-.
o w °9v4 wcvc B o
>* -* O B *J o « 6 -^ w)
B w
5 i o
w a a. i z i T3
« -^ 4 d <-<4t« X
X o 4 w j: *<
X h O 41 U U O
*-< Od w - «o -< 4 d *»
84 «M 4) O WVjCWWQ tf
i »SS J§S-5gf -5!,
w u B 4 -^ w S d at
4>« 3g*i^-« *^ * ' n. S. b ^ £'
V 4» 41 W O 4) HS|)dOd3^> C>
£^ -a a. u 5 « « 6 « u -^ o- i tn
^ '^ h> * tft-
O
3 2 . 3
gl ? » i . : . slzi
**** O d W OlttW J-' i£> % TT
U4J M)v4d5 5 C dtOTt
04 « XOA 3 C C Q Q. C B «'
41 O d .-» ^4B BO40I
4J-HXW 4Wr-tjC C 6- 1 r-. fl T3 al
4 <^t U 4> 4IVU T^ J= O d 4> .-5 W
v a. * v Xi?T3vi w^4 a > en ** *^>.
a.a.f-1^4 I-IBQ.C * *^ B 4i i 3 O
«4O« B3B» Ba> -OeMJO^-i
OB CM O »J CO iff
41 » W M B
^-« w>ow w£h
8X "*-i T3 4) 4 O l-i *+ 4
r-> dWUB «J«O3 B U
u 4 *o a> O c o *o c>
3 5 £ o o -* j 6 6. x O f CNI
W U »4 CM >
3 - a 1
.s2. .t2ai., I
C ? ** "'l^
U Ck. « 4 OWO.(O<-4«BU96 1
tn u >j M
.5,1 |
M W J> 4 O. w B
S^3^^8«-8 d
SoSS'£3'c»'aj?Ol '
uQ.B4Voa.Qwo. o r
« W U |4 N
e g 1
V 4 1 *< «
w w .a 4 a w >
n=ZiJ*jH44 > T
-^T33gd4l»a O d
SJ^Sjr^^*^^ ^ 8 ^i
W M ' O > A
« 3
. It 4)
fr I
2 & i 2 §
« 5 1 S o
U JO C S w
B "
^ «4 A. W
£ <-> « e
K « , a s
< M ' A X O
e
I
s.
s
II
« o
»M
4 0
1!
§s
^.
jg
«*4
w>
B 4
3Jf
ll
X «
> *
Sfr
ES
|4
C S
i ;
a «
« JH
«i
i.- -a i
i i
-------
PQ
H
sa
CO
w
H
O
2
M
s!
1
X
o
3
re
o
o
H
O
W
P*
PM
O
X
M
Prf
H
(U
(3
C
O
,0
cd
H
J
:
4
0
C
i
4
!
i
a
a
>
D
>
j
*»
j
^
<
;
rH
01
M
£
DIFKIN
S
OS
(0
rH
01
o
i
r-l
(0
U
H
*-»
VI
rt
«
4J
W
Gifford-Hanna
11
J*H
°
LlMtr
rollback
1
U
1
M u
»< rH 1 4»
o< a >
u -o o .0 o
TH-CI-H >, o o rt
"O 4) ri *-* CQ H O *H O W W
. OJ V- O! 0) AJ M -rl « ri
r-H *J -H X .CQo.CCAtHClii-iJJta
tH'vcniu exrxCuwuOtfl*
C"ti
*j iJ x o W U *H{fl4J^ft.P. 01 0) ^ *J C 0 W (H
B > qj Vj to at QJ u
tH3-HX 4=O jr«-H
rHuuat a. d. c u w u
voCr-i (A*H
J-aUD- 4-lOC-rl4JUM
D.Q CQQtQCOOOD.
4J t-i X o a; w M 3 x: *j x
o o w u >H a) u 01 a> u
r-( W -H .CO.CC0T4
rHcaco O'd.tJtau
QJ T3 C (AtOWOVi-4
5 -H at uonuor-t
<-i(X cogflJOOCX
4JCQX OlJJrlJ=^X
O > W V-i ffl *J (X O. 0)
»: H
S
01 0
O r-t >.
C i-4 U
CO O < \ IW -H .
W ^ M O rH 0)
^ O Ol 4) 01 O C
4J 13 > -O C & O
HO 'H 0> t3 O -H
rH E « O. 01 i-t M 4J
H C O U W 3 BJ
^> V) Otr-tncQOri
«J C -H O/
r-( eg QJ ei o) e o) -H
at X> c T3 i-i rt > «
OS r-l
*- 4J C X
>i o o n no
i-H C ^ C -H B O
rH t-< O 4J (0 rH
co ri « -H a -H o
u a) w 01 w re c u
r4 4JU}rHdO)COO
c 3 a) a -< v-i x ot
4= D. O B O O T3 *-
u gat-HCu-.o»C4i
0) OCW-HQBflJfl
CO
o> .
rt 01 00 8}
AJ O 0) > 00 0) TJ 8
t c *j c -r-i c vj w e «
CO 01 0* i-( 0 W-H3rHO-H
o o x AJ -o c -r< ec-aS-HB
narHwa o XJ^o^coo
B o, -H o) o) c o atcjriuoioi
Hffi^^TlOrH C tfi a < l-( fi
nJ M
00 41
t 1-1 C rt
« 01 -H 3
O O X C C *O
O) -H 4J 01 01 01
4J r-l -rl O. Ol U
H CXrH X r> O
B CL 1-1 01 U rl
H CO X) C (0 O.
>J M
5S2
10 01 ^ 3
-a o >> C c TJ
W tH *J O> O» O)
4J rH -rl CX Ol U
-r-i a -H x >- o
s a. -rt 01 u M
3 « J3 C tl> D.
-J W
a
1
u 01
ce M
*J 3
H U
5 S
J PL.
-------
SECTION II
VMT CHANGE
APPLICABILITY AND RELIABILITY
Changes in concentration of photochemical oxidants and nitrogen dioxide
have been defined by many simple functional relationships. The simplest
relates changes in the pollutant levels with vehicle-miles of travel
(VMT). The function assumes that changes in these variables, VMT, and
pollutant concentrations, are linearly related. That is, for example,
if VMT were to increase by 10 percent, concentrations would be assumed
to increase by 10 percent.
By its very nature, the use of this approximation is considered to be
limited. It is intended to be used as a substitute for extensive tech-
nical evaluations. It would be useful as an indicator of trends in air
quality in areas not now experiencing violations of air quality stan-
dards for oxidants (the Priority III areas). Its purpose as a screen-
ing procedure allows for a rough estimation of the impact of a facility
on the regional ambient air quality. If the results of the analysis
indicate the possibility of a violation of the NAAQS, a more extensive
analysis must be undertaken. This screening procedure, approximate in
nature, is best restricted to areas where the growth in vehicular emis-
sions is the greatest contributor to the growth in total emissions,
because growth in stationary source emissions are not accounted for but
could cause violations of air quality standards.
-------
DATA AND MANPOWER REQUIREMENTS
Data requirements for the analysis of small changes in VMT due to facil-
ity construction are relatively minimal. Data needs are limited to air
quality data and existing and projected vehicle-miles of travel (VMT).
The necessary air quality data would ordinarily be the second highest
concentration of oxidants, and should be for a year as recent as pos-
sible. These air quality data would typically be available either from
a regional air pollution control agency, state air agency, or EPA.
Vehicle miles of travel data are typically available from the ongoing
transportation planning process. Sources of such data include the high-
way or planning agency. The technique utilized for determining VMT
varies substantially among individual states. The most widely used
and probably the most accurate method involves a highway link system
from which VMT are calculated as a function of individual link length
and average daily traffic (ADT) volume on the link. This technique
appears to be used most widely in larger metropolitan areas where various
regional transportation planning studies and highway inventories have
been undertaken.
The second method involves estimating VMT based on motor fuel consump-
tion. Generally, this is the method used for small urban and rural
areas, and usually the data are presented in terms of statewide or re-
gional VMT. While this particular technique is suitable for estimating
gross VMT on a regional basis, it proves inadequate when knowledge of
local VMT trends is required. Either technique appears satisfactory
for the needs of this analysis, so long as the data are fairly current,
are for the appropriate geographic area, and include all roads.
Manpower requirements are minimal since data are routinely collected
by the planning agencies and can be recovered with minimum effort. To
-------
identify, locate, and collect current VMT data, to adjust it to the year
of interest, and to compare the project and no-project cases would be
expected to require no more than 1 or 2 man-weeks over a calendar period
of 2 to 4 weeks.
APPLICATION OF THE TECHNIQUE
This technique is based upon a simple linear relationship. The rela-
tionship can be expressed as a simple ratio:
VMT VMT + VMT
c c + c
e p e
where VMT = existing VMT
VMT = projected increase to VMT due to facility
c = existing pollutant concentration
c = projected increase in concentration due to added VMT.
Because the purpose is to estimate the impact of a facility or of traf-
fic growth on air quality to predict future concentrations, Equation
(1) can be rewritten as follows:
c (VMT + VMT
c = c + c = e V E !=JL_ (2)
f p e VMT ^ '
where c = estimated projected pollutant concentration.
This relationship is utilized through the following steps:
1. Define the "baseline" year for the analysis. This must
be the most recent year for which air quality data are
available. Because of the time to collect, reduce, cor-
rect, and summarize air quality data, the baseline year
being used may be 2 or 3 years prior to the current date.
For example, transportation control plans now (in 1975)
being prepared typically use 1972 or 1973 as the base-
line year.
8
-------
2. Define the geographic area. There is little scientific
basis for selecting a geographic area for analysis when
using such a crude technique. Too small an area (e.g.,
a minor civil division) would over-state the effect of
traffic growth. The pragmatic approach is to use the
geographic area also used as the base for the best VMT
data avilable. Highway agencies typically report VMT
for urbanized area, consisting of a central city and its
surrounding densely settled districts, or for standard
metropolitan statistical areas. Either of these may be
adequate for defining the area of analysis, but neither
will probably be representative of the AQCR in general.
The use of a state-wide area should definitely be avoided,
as the air quality will be insensitive to activity changes
over such a large area.
3. Obtain the "present" VMT for the geographic area in ques-
tion. Generally VMT data are available for every major
urban area for some year within the past 2 or 3 years.
4. Adjust the VMT to be representative of the baseline year.
This may require obtaining another year's set of VMT
data or consulting with the highway agency to determine
a basis for interpolation or extrapolation. Generally a
linear fit will be perfectly satisfactory for this method.
5. Obtain the VMT for the project being evaluated. If the
method is being used for areawide evaluation, obtain the
VMT for the future year in question. If the project be-
ing evaluated will not be open until some future year,
the VMT data may be for a future year. In such a case it
will be necessary to perform one other step:
a. Estimate the VMT from roads other than the project
road for the future year, using extrapolation or
interpolation. This provides a basis for determining
the contribution to future VMT from the project alone,
but is less accurate than (b) below because it relies
upon the VMT air quality relationship remaining valid
over time, which is doubtful, or
b. Determine what VMT the project would involve if open
in the baseline year. This requires estimating normal
traffic growth, then subtracting the normal growth
from the future year VMT estimate. This provides an
estimate of the change in present air quality from
the project under current conditions. In this case,
VMT f = VMT (1 + G) (3)
-------
where VMT = future year VMT from proposed
P facility
VMT = estimated present equivalent of
PP VMT from proposed facility
G = normal growth in VMT in inter-
vening years.
Typically, growth is a few percent per year. (the
national average is about 5 percent per year.) The
local growth rate should be obtained from the highway
agency.
The project VMT to be used in Equation (3) is then:
VMT
VMT = P-L-
pp 1 + G
6. Use Equation (3) to estimate the effect of tne project.
Because this is such a crude procedure, this technique should only be
used for determining whether more sophisticated tools are needed. If
the method predicts that air quality standards may be violated,
resources should be allocated for better air quality data, analysis,
and projections, perhaps using one of the methods discussed in later
sections of this manual.
LIMITATIONS
This technique is applicable as only a rough screening procedure under
limited conditions. Its validity is restricted to areas where vehicu-
lar emissions are the major source of growth in ambient concentrations
of oxidants. This procedure yields an estimation of the impact on air
quality due to changes in VMT. If results indicate the possibility of
a violation in the NAAQS, extended analyses of the types to be dis-
cussed in later sections of this work would be required.
10
-------
A major limitation of this model is that it neglects the pollution con-
trol effects of the Federal Motor Vehicle Emissions Control Program and
thus tends to over-predict future contributions of vehicular sources to
total emissions and ambient concentrations. This may, however, tend to
compensate for the lack of consideration of nonvehicular sources.
Two other major factors of pollutant concentrations are ignored, vehic-
ular speeds and stationary sources. Vehicular emission rates vary
with speed. Hydrocarbon emissions decrease with speed while NCL emis-
sions have been shown to increase at higher speeds. Uncertainty occurs
when speeds are not explicitly treated. Generally speaking, however,
sources which would generate significant additional VMT necessitating
an air quality review would typically be high capacity limited access
highway facilities. Speeds would tend to be greater than existing
average network speeds. Measured concentrations should theoretically
be a function of speed (the next section deals with this concept) thus
introducing the possibility for error in the prediction of pollutant
concentrations. This error would be an over-estimate for hydrocarbon
emissions and an under-estimate for nitrogen dioxide emissions.
Stationary sources may contribute greatly to regional pollutant concen-
trations. The contributions would vary depending upon local charac-
teristics such as density and amount of industrialization. Without a
proper stationary source emission inventory, these sources cannot be
treated explicitly. Since the procedure is intended to be "quick and
dirty," exogenous consideration is all that is required. As with the
treatment of speeds, uncertainties introduce errors that would appear
as over-estimates.
SPECIAL FEATURES
This technique, although based on gross assumptions, is a reasonable
screening procedure for areas that are not now in violation of the am-
bient standards or that do not have expertise or funds for air quality
11
-------
forecasting. The data are easily obtainable with minimum manpower re-
quirements. Thus the procedure is applicable to many smaller urban
areas where sophisticated data collection procedures have yet to be
implemented. Use of the technique lends itself to restricted budgets,
since the simple linear expression does not involve computer modeling
or difficult computational processing.
12
-------
SECTION III
LINEAR ROLLBACK
APPLICABILITY AND RELIABILITY
This technique employs the same relationships as the previous one, except
that oxidant precursor emissions are substituted for VMT. This technique
is somewhat more reliable than the technique discussed in Section II due
primarily to three basic refinements: (1) the inclusion of speed data,
(2) accounting for the Federal Motor Vehicle Emission Control Program,
and (3) accounting for stationary source emissions. Like the first tech-
nique, this technique is a reasonable screening procedure under limited
conditions. Its use would be applicable to areas where sophisticated
techniques cannot be used, preferably areas not yet in violation of am-
bient air quality standards.
This procedure can be used to produce results exhibiting the emissions
contribution of a proposed facility, or it can be used for assessing
general growth. The simple relationship of the previous technique is ex-
tended by treating an additional function of vehicular emission rate,
that being speed, and also accounting for the Federal Motor Vehicle
Emissions Control Program (FMVECP). The speed data are required in order
to compute actual emissions instead of working strictly with VMT as with
the previous technique. The inclusions of speed permits the use of emis-
sion factors adjusted for variations in speed from such sources as AP-42.
A greater degree of accuracy can be expected since speed is no longer
implicitly fixed in the relationship but rather behaves as a controlling
parameter.
13
-------
Additional accuracy is obtained by the explicit treatment of stationary
source emissions. These sources may contribute significantly to the
total emissions, but are unaffected by changes in the transportation net-
work. Therefore, by accounting for their contributions, results are
obtained which allow for more accurately predicting the role of transpor-
tation projects (or general growth) in total emissions increases. This
therefore allows assessment of the contribution of transportation to
areawide pollutant burden, and provides an indicator of the future air
quality.
DATA AND MANPOWER REQUIREMENTS
The problem of estimating emissions from motor vehicles is basically one
of specifying a number of parameters, and then multiplying these param-
eters together with suitable emission factors to compute roadway emission
strengths. Differences among methods of estimating emissions arise in the
nature and comprehension of the input parameters. The basic algorithm
is essentially the same for all the methods. The adjustment factors vary
depending on specific requirements and data availability.
The basic data requirements for a highway-related emission model consist
of the following: emission factors, base and projected vehicle miles
of travel (VMT), and speed. As for the previous technique, these data
would usually be available from the local, regional or state transporta-
tion planning agencies. In the case that the data are unavailable
from these sources, except for VMT, the data could reasonably be devel-
*
oped from current literature at the expense of additional manpower.
National average statistics could be substituted or acceptable data
could be developed from such sources as the Highway Capacity Manual.
Some models, such as the Special Area Analysis (SAPOLLUT), include
default values as substitutes for unavailable data.
14
-------
Additional variables could be introduced so that the approach would be
responsive to local conditions. Typical parameters would include the
mix of vehicles, type and location of highway, and degree of congestion.
The data could be aggregated according to total vehicle-miles of travel
by class of vehicle on various roadways by roadway functional classifi-
cation. EPA vehicular classification standards, 40 CFR part 85, used to
determine total emissions, are composed of the following categories:
1. Light-duty gasoline vehicles (automobiles).
2. Light-duty gasoline trucks.
3. Light-duty diesel vehicles.
4. Heavy-duty gasoline vehicles.
5. Heavy-duty diesel vehicles.
6. Motorcycles.
The above listing attempts to identify categories of vehicles that gen-
erally tend to exhibit common characteristics with regard to pollutants
emitted per mile of travel. The distribution of vehicle types can be
established through vehicle classification inventories supplemented by
vehicle registration data obtained from state agencies or private con-
cerns such as the R. L. Polk Company.
Since the volume of nitrogen dioxide and hydrocarbons produced per
vehicle-mile is related to vehicular speed and since average speed varies
with the roadway classification, it may be important that the distribu-
tion of total VMT by roadway classification be identified. (Grouping
roads by speed or functional classification reduces the computational
effort compared to calculating emissions separately for each road.)
15
-------
Typically roadway classifications are categorized as: (1). freeways/
expressways; (2) arterials; (3) collector/distributors; and (4) local
streets. These general classifications are associated with different
characteristics of speed and operating efficiency.
The criteria for determining the functional'classification of a street
system has been well established and documented by the Institute of
Traffic Engineers, the Transportation (Highway) Research Board, and
others.
Speed data could be averaged throughout the day or disaggregated into
either an hourly distribution or periods of the day. The value of either
of the disaggregated approaches is the introduction of the effects of
2
congestion on speeds and therefore emissions. The SAPOLLUT Program for
air quality analyses, for instance, has been developed with endogenous
default values for hourly distributions of ADT for utilization if local
data are unavailable. Based on the Highway Capacity Manual and the
findings of other studies, volume capacity ratios are computed to
generalize a relationship to speed. The explicit treatment of speed and
other parameters by SAPOLLUT extends the validity of its results.
Manpower requirements are considerably greater than for the VMT analyses
of Section II. Significant additional data collection, reduction and
processing is required to prepare input for the modeling process. Much
of the data are part of the normal 3-C planning process efforts; however,
some data, not part of the routine output, can be obtained through de-
velopment of appropriate summary programs or clerical efforts. The 3-C
process can be expected to provide access to most of the typical regional
transportation and land use statistics. In most cases where specific
elements of data are not readily available or recoverable, default values
such as those included in the SAPOLLUT program are acceptable for use in
the computation of emissions or emission changes since the intention of
this technique, as with the previous technique, is that of a reasonable
screening procedure.
16
-------
An additional requirement would be to have or to prepare a stationary
source emission inventory. These sources may significantly contribute
to total regional emissions. Stationary sources are unaffected by
changes in the transportation network. Thus, concerned with the impact
of additional highway network capacity upon air quality, contributions of
the stationary sources should be factored out of the total existing
measurements. This inventory should be available for most urban areas.
In the case where it is not available and budgets for local environmental
analyses are constrained, another technique should be chosen. But if
the time and money could be obtained or are available, the completion of
a stationary source inventory would be recommended.
Data requirements and the procedure outline are tabulated in Table 1.
The collection and adjustment for the VMT data is discussed in Section II.
Speed information is usually available from the planning model used by
the local planning agency, but it is only approximately equivalent to
true speeds. The stationary source inventory explicitly accounts for
nonvehicular emissions. The specific data required by a particular emis-
sions model, as previously discussed, explicitly accounts for local con-
ditions which affect local vehicular emissions characteristics (vehicle
age distribution, etc.).
Basic design variables for the highway of concern will provide for the
classification of the roadway type, and the projected VMT, which can be
adjusted either for the base year or for a desired analysis year as dis-
cussed with regards to the previous technique. Emission adjustment
factors will account for variations in speed, vehicle mix, and the vehicle
age distribution.
The total effort for the data collection and processing stages and the
required computations for analysis and evaluation is estimated to require
at least 1 man month. Typically the total requirement would range from
the minimum of 1 month to 3 or 4 man months. In some cases these esti-
mates could conceivably be considerably low. In all cases the use of
17
-------
Table 2. DATA REQUIREMENTS AND PROCEDURE OUTLINE
Data
Task
EXISTING - DATA COLLECTION
VMT
Air Quality Data
Speed
Stationary Source
Inventory
Model-specific
Collect and adjust to base year
Second highest reading for base year
VMT-weighted daily average speed by
roadway classification, or nearest
approximation.
Determine contribution to total
emissions (reactive HC only)
Special data requirements of selected
emissions model
FUTURE - DATA COLLECTION
Highway facility
VMT
Emission adjustment
factors
Obtain design variables for analysis
Projection for analysis year or
base year for facility of concern.
Obtain information needed for deter-
mining emission factors, such as
speed, vehicle mix and age distributions.
ANALYSIS COMPUTATIONS
Final preparation
Model run
Prepare data for selected model's input
requirements.
Computer computations, Analysis and
Evaluations.
18
-------
computer facilities could, to varying degrees, be substituted for
manual computations and processing to reduce the total calendar
months required for the project.
APPLICATION OF THE TECHNIQUE
Emission Computations
There are numerous methods for estimating emissions from motor
,1
18
vehicles. The basic algorithm is essentially the same for the
various methods. By the method given in EPA publication AP-42
emission strengths for exhaust hydrocarbons and oxides of nitrogen are
calculated by:
n + 1
e = C. d. m. S. (4)
np / j ip ipn in ip v '
i = n - 12
where e = emissions (grams per vehicle mile) for calendar year n
P and pollutant p,
c. = emissions (grams per vehicle mile) for pollutant p for
X^ itn model year at low mileage at an average speed of
19.6 miles per hour.
d = emission deterioration factor for the i model year,
calendar year n, and pollutant p for vehicles with
emission controls,
m = weighted annual travel of the i model year during
calendar year n,
and S. = weighted speed adjustment factor for exhaust emissions
of pollutant p for the itn model year.
The weighted speed correction factor, S. , is in turn computed by the
relation:
19
-------
V
3
with S. = weighted speed adjustment factor for exhaust emissions of
mp pollutant p for the ith model year during calendar year m,
f. = the fraction of the total annual vehicle miles traveled at
m speed j during calendar year m,
v. = the average speed correction factor for average speed j and
pollutant p,
and k = total number of different average speeds.
Curves of v. versus speed are shown on Figure 3.1.1-1 of AP-42. Values
J Sr
of the other parameters will be discussed later with other methods.
Evaporative and crankcase hydrocarbon emissions are calculated by:
n + 1
fn=
i = n - 12
where f = combined evaporative and crankcase hydrocarbon emissions
for calendar year n,
h. = combined evaporative and crankcase emission rate (grams per
vehicle mile) for the ±^ model year,
and m. was defined previously.
To calculate actual emissions, e and f must be multiplied by the num-
np n v j
ber of vehicle miles traveled (VMT) for the region of interest during a
given time period.
As mentioned earlier, the real differences among vehicle emission models
do not arise from the computational methods, but rather from the extent
and definition of the input parameters. It can be stated that, in general,
newer methods will provide better results because of improved input data,
20
-------
and perhaps because experience provides for improved specification of
input parameters.
2345
Additional methods ' ' ' have been given for computing motor vehicle
emissions; almost all use the same basic technique described for AP-42
and all require the same basic data.
A recent technique distinguishes between "hot starts" and "cold starts"
in calculating emissions. Cold start data are obtained from the
Federal Test data (AP-42) by assuming a fraction of the test results
of total emissions to have occurred during the first 2 minutes or cold
start portion of the cycle. This fraction varies with year and is
based on General Motors tests results. The grams per mile emissions
during the "hot" portion of the test are calculated from the remaining
portion of the test sample. The hot emissions are subtracted from the
cold emissions to determine the excess emissions attributable exclusively
to cold start. The result is a new set of emissions data (grams per
mile) for use with VMT data, and a set of "cold start" emissions to be
applied where these conditions obtain.
The method has been carried further to define a range of "fractions
of a cold start" which increase with an increase in soak time. Starts
after a 12-hour soak are considered to be fully cold.
Q
EPA has recently published a supplement revising the internal combus-
tion engine sources portion of AP-42. The new methodology for estimating
emissions is quite different from that previously employed and described
above due mainly to three factors :
The new method does not include deterioration factors
"Cold start" and "hot start" emissions are included
Emissions projections are given in an appendix and not listed
in the main section dealing with internal combustion engine
sources.
21
-------
The emission factors which are presented in this supplement are based
on measurements taken during EPA's annual surveillance programs and
cover calendar years 1971 and 1972. Deterioration is accounted for
implicity in the emission factors for the year of measurement. For
example, emission factors are presented as X grams per kilometer of
exhaust hydrocarbons for a 1969 model year vehicle in calendar year
1972.
Other changes in the method of estimating motor vehicle emissions can
best be described by examining the new algorithms used to compute
emissions. For light-duty vehicles (automobiles) and light-duty trucks
this is:
n
npstw . _" .. ipn in ips ipt iptw ^ '
where e ^ = Composite emission factor in g/km (g/mi) for
npstw - , 1n oo
r calendar year n, pollutant p, average speed s,
ambient temperature t, and percent cold opera-
tion w.
c. = The FTP (1975 Federal Test Procedure) mean
P emission factor for the ith model year light
duty vehicles during calendar year n and for
pollutant p,
t~fo
m. = The fraction of annual travel by the i model
n year light-duty vehicles during calendar year
n,
v. = The speed correction factor for the i model
P year light-duty vehicles for pollutant p and
average speed s,
z. = The temperature correction factor for the i
model year light-duty vehicles for pollutant p
and ambient temperature t,
r = The hot/cold vehicle operation correction
factor for the itn model year light-duty
vehicles for pollutant p, ambient temperature
t, and percent cold operation w.
22
-------
The discussion of these variables applies to automobiles and light
trucks except where noted.
FTP Emission Factor (c. ) - These data are divided by geographic area
into : low altitude (non-California), high altitude, and California
only. The tabulated values are applicable to calendar years 1971 and
1972 (only 1972 for trucks). California emission factors are presented
separately since California vehicles have been, in the case of several
model years, subject to emission standards which differ from those
standards applicable to vehicles under the Federal emission control
program. For those model year California vehicles which did not have
separate emission standards, the national emission factors are
assumed to apply in California as well. Emissions at high altitude are
differentiated from those at low altitude to account for the effect
that altitude has on air-fuel ratios and concomitant emissions. The
tabulated values are applicable to calendar years 1971 and 1972 (only
1972 for trucks) for each model year.
Fraction of Annual Travel by Model Year (m.) - No significant change
has occurred from previous emission estimation methods.
Speed Correction Factors (v. ) - Speed correction factors enable the
.. r , , _____ ips
"adjustment" of FTP emission factors to account for differences in
average route speed. Since the implicit average route speed of the
FTP is 19.6 miles per hour (31.6 kph), estimates of emissions at higher
or lower average speeds require this correction.
It is important to note the difference between "average route speed"
and "steady speed." Average route speed is trip-related. It is
based on a composite of the driving modes (idle, cruise, acceleration,
deceleration) encountered in, for example a typical home-to-work trip.
Steady speed is highway facility-oriented. For instance a group of
vehicles traveling over an uncongested freeway link (volume/capacity
23
-------
of, say, 0.1) might be traveling at a steady speed of about 55 mph
(89 kph). Note, however, that steady speeds, even at the link level,
are unlikely to occur where resistance to flow occurs (unsynchronized
traffic signaling, congested flow, etc.).
Previously, the limited data available for correcting for average speed
were presented graphically. Recent research has resulted in revised
speed relationships by model year. To facilitate the presentation,
the data are given as equations of the form:
v± s - exp (A + BS + CS2) (8)
where S is the speed in miles per hour.
The values of the coefficients A, B, and C apply only for the range of
the data, from 24 to 72 kilometers per hour (15 to 45 miles per hour).
Since there is a need, in some situations, to estimate emissions at
very low average speeds, correction factors have been developed for
this purpose for 8 and 16 kph (5 and 10 mph).
Temperature Correction Factor (z. )
t lpt
The 1975 FTP requires that emissions measurements to be made within
the limits of a relatively narrow temperature band (68 to 86° F).
Such a band facilitates uniform testing in laboratories without
requiring extreme ranges of temperature control. Present emission
factors for motor vehicles are based on data from the standard Federal
test (assumed to be at 75° F). The correction factors are expressed
in equational form and can be applied between 20°F and 80°F. For
temperatures outside this range, the appropriate endpoint correction
factor is applied.
24
-------
Hot/Cold Vehicle Operation Correction Factor (r. )
The 1975 FTP measures emissions over three types of driving: a cold
transient phase (representative of vehicle start-up after a long
engine off period), a hot transient phase (representative of vehicle
start-up after a short engine-off period), and a stabilized phase
(representative of warmed-up vehicle operation). The weighting factors
used in the 1975 FTP are 20 percent, 27 percent, and 53 percent of
total miles (time) in each of the three phases respectively. Thus,
when the 1975 FTP emission factors are applied to a given region for
the purpose of assessing air quality, this can be viewed as if 20
percent of the light-duty vehicles in the area of interest are operating
in a cold condition, 27 percent are operating a hot start-up condition,
and 53 percent are operating in a hot stabilized condition. For non-
catalyst vehicles (all pre 1975 model year vehicles), emissions in the
two hot phases are essentially equivalent on a grams/kilometer (grams/
mile) basis. Therefore, the 1975 FTP emission factor represents 20
percent cold operation and 80 percent hot operation.
There are many situations where these particular weighting factors may
be inappropriate. For example, light-duty vehicle operation in the
center city may have a much higher percentage of cold operation during
the afternoon peak when work-to-home trips are at a maximum and
vehicles have been soaking for 8 hours. The hot/cold vehicle operation
factor allows the cold operation phase to range from 0 percent to 100
percent of total light-duty vehicle operations. This correction factor
is a function of the percent of cold operation, w, and the ambient
temperature, t. The correction factor is:
_ w + (IQO-w)f(t)
iptw ~ 20 + 80f(t) ^ '
where f(t) is a function of temperature presented in AP-42, Supplement
No. 5.
25
-------
The new methodology also allows for calculating evaporative emissions
of hydrocarbons and idle emissions of hydrocarbons, nitrogen oxides,
and carbon monoxide.
Emissions from light-duty, diesel-powered vehicles are calculated as
before, except that the emission factors are given for pre-1973 model
years. Projection to future years is given as an appendix.
Emissions from heavy-duty gasoline vehicles are calculated by:
n
e Y] c. m. v. (10)
nps *-* ipn in ips
i=n-12
where the factors were defined previously. For heavy-duty diesel-
powered vehicles the model year distribution is omitted:
n
e - E c. v. <1;L>
nps *' ipn ips.
i-n-12
Values for c. are based on tests of vehicles on-the-road over the
ipn
San Antonio Road Route (SARR). The SARR, located in San Antonio,
Texas, is 7.24 miles long and includes freeway, arterial and local/
collector highway segments. Since the SARR is an actual road route,
the average speed varies depending on traffic conditions at the time
of the test. However, the average speed tends to be around 29 kph
(18 mph) with about 20 percent of the time spend at idle. The test
procedure emission factor is composed entirely of warmed-up vehicle
operation. Based on a preliminary analysis of vehicle operation data,
HDV operation is primarily (about 95 percent) warmed-up.
Of course, it is quite necessary to estimate motor vehicle emissions
beyond calendar year 1972, and this capability is provided in an
26
-------
appendix to the revised AP-42. This is done purposely to separate the
analytical results of EPA's surveillance testing program from what are
acknowledged to be "best guesses" of future emission factors. There
are several reasons for this separation. First, current legislation
allows for limited time extensions for achieving the statutory motor
vehicle emission standards. Secondly, Congressional action changing
the time table for achieving these standards and/or changing their
levels is likely in the future. Thirdly, new data on catalyst-equipped
(1975 automobiles) are becoming available. The methods presented in
the appendix for estimating emissions are similar to those described
above.
There is a final source of data which might prove useful, especially
for estimating emissions from higher speed, highway traffic. This is
9
the Modal Analysis Model developed for EPA. This model predicts emis-
sions from a single vehicle or an ensemble of vehicles for a user-
specified age distribution over any desired driving sequence within the
range of the applicability of the model. The key point for highway
traffic is that the model can predict emissions which reflect a
constant cruise speed, not an "average route speed" which might include
stops and starts. With proper resources, the model could even be used
to estimate average emissions for a given urban area based on the
typical driving cycle for that area instead of on the Federal Test
Procedure.
Estimating Pollutant Concentrations by Linear Rollback
The linear rollback technique provides a means for converting total emis-
sions to pollutant concentrations. The technique assumes that the con-
centration of a pollutant is equal to the background concentration of that
pollutant plus some linear function of the emission rate of the pollutant,
or:
27
-------
c - c. + k e (12)
t b
where ct is the total concentration of the pollutant
cb is the background concentration of the pollutant resulting
from natural sources
e is the total emissions of the pollutant for a particular
averaging period, typically 24 hours
k is a constant
Generally ct is a. very small fraction of c^ and is often neglected. The
total emissions, e, can be calculated or derived from inventories of
emissions. Therefore, in an area where the concentration of a pollutant
is monitored, the only unknown is the value of k. Because all other
elements of the equation are known, it is a simple matter to calculate k.
The equation in its most common form is expressed as e/c = k. The value
of k is often referred to as the e/c ratio.
In determining the reduction in emissions required in a particular area
in order to meet National Ambient Air Quality Standards (NAAQS), the
second highest validated concentration is usually used with the con-
temporary emissions to determine the corresponding k value. The al-
lowable emissions are then determined as follows:
e - c k (13)
a s
where ea is the estimated allowable or "safe" emission rate
cs is the National Air Quality Standard for the pollutant (for
ozone, cs = 0.08 ppm)
k is a constant
The emissions reduction required, then, is equal to the difference be-
tween the baseline emissions and the calculated allowable emissions or,
expressed as a percent reduction:
28
-------
, . . . , (baseline emissions)-(allowable emissions) 1nn
reduction required = , . v . x 100
baseline emissions
(14)
The chief advantages of the rollback model are that it is relatively
simple and straightforward to use, and its theoretical basis is readily
understood. While the technique does have various limitations, these are
judged to be not so severe as to preclude its use in air quality analyses.
The most important limitations of the model include its insensitivity
both to spatial resolution with regard to the source-receptor relation-
ship, and to local variations in meteorology and topography that occur
within a region.
LIMITATIONS
This technique is applicable as only a screening procedure under limited
conditions. The purpose of these models is as an analysis technique to
determine the impact upon regional emissions due to an addition to the
highway network or to estimating regional growth effects. Results could
be interpreted to determine the possibility of violations to the NAAQS,
through the linear rollback approach (as presented). The performance of
air quality modeling efforts, to be discussed in the next sections, would
be indicated if the occurrence of a violation in standards is forecasted
with the method described here.
This technique introduces parameters which improve its reliability as
compared with that of Section II. However, the forecasted pollutant con-
centration remains a linear function of the emission rate. This limita-
tion is judged to be not so severe as to preclude utilization for the
limited condition previously mentioned.
29
-------
SPECIAL FEATURES
These emission computation techniques, relative to Technique 1, are sig-
nificantly more reliable. The models are constructed to consider
variables of speed and stationary source emission endogenously. The
FMVECP is explicitly accounted for through the use of available emission
factors. Data needed are available from typical regional planning data
collection efforts. These emissions models are easily utilized and may
be suitable for smaller areas or those not yet exhibiting violations of
oxidants standards. Also, the data requirements of these models are
consistent with smaller urban areas where data collection efforts may not
be as sophisticated as those for large metropolitan regions.
REFERENCES
1. Compilation of Air Pollutant Emission Factors. EPA Report
AP-42 (Second Edition). April 1973.
2. Special Area Analysis. (SAPOLLUT Model), U.S. DOT, August 1973.
3. Kircher, David S. and Donald P. Armstrong. An Interim Report
on Motor Vehicle Emission Estimation. EPA-450/2-73-003,
October 1973.
4. Wolsko, T. D., M. T. Matthies, and R. E. Wendell. Transporta-
tion Air Pollution Emissions Handbook. Argonne National
Laboratory Report ANL/ES-15. July 1972.
5. Sauter, G. C. and W. R. Ott. A Computer Program for Projections
of Vehicular Pollutant Emissions in Urban Areas, JAPCA,
24:54. No. 1. January 1974.
6. Wendell, R. E., J. E. Norco, and K. G. Groke. Emission Pre-
diction and Control Strategy: Evaluation of Pollution from
Transportation Systems, JAPCA, 23:91. No. 2. February 1973.
7. Cirillo, R. R., J. E. Norco, and T. D. Wolsko. The Effect of
Cold Start on Motor Vehicle Emissions and Resultant Air
Quality. Paper 74-127, 67th Annual Meeting, Air Pollution
Control Association. Denver. 1974.
30
-------
8. Supplement No. 5 for Compilation of Air Pollutant Emission Factors,
EPA Report AP-42 (Second Edition). Unedited copy. April 1975.
9. Automobile Exhaust Emission Modal Analysis Model, EPA Report
EPA-460/3-74-005. January 1974.
31
-------
SECTION IV
NONLINEAR ROLLBACK: APPENDIX J
Linear and nonlinear rollback are based on the assumptions that changes
in concentrations are either directly or indirectly proportional to
changes in total emissions and that such relationships can be defined.
The assumption that changes in ambient concentration levels are
proportional to changes in total emissions, as described in Section
III for linear rollback, is reasonable if the existing spatial and
temporal distributions of emissions remain constant.
The effects of meteorology on ambient concentrations are included im-
plicitly by the dependence on ambient concentrations. For long time
periods or for worst case values, this lack of explicit dependence
on meteorology is not critical.
APPLICABILITY AND RELIABILITY
For highway evaluation on both the project and system level, the re-
quirement for unchanging emissions distributions in time and space is
not met, since both project and system level modifications will likely
lead to VMT changes. An additional problem with the application of linear
rollback to photochemical oxidants is that oxidant concentrations depend,
in a nonlinear fashion, on ambient concentrations of hydrocarbons and
oxides of nitrogen. Nonlinear or modified rollback attempts to account
for this nonlinearity through an empirically derived relationship between
32
-------
maximum daily 1-hour average oxidant concentrations and 6 to 9 a.m. aver-
age concentrations of hydrocarbons. Hydrocarbon concentrations are then
assumed to depend linearly on emissions, so that a relationship between
changes in hydrocarbon emissions and oxidant concentrations is achieved.
One form of this relationship is the Appendix J curve of the EPA regula-
tions on preparation of implementation plans (40CFR51).
2
Figure 1 shows observed relationships between maximum daily 1-hour
oxidant concentrations and 6 to 9 a.m. average concentrations of total
hydrocarbons. The approximate upper-limit observed oxidant curve forms
an "envelope" around the maximum oxidant concentrations.
Figure 2 shows similar relationships of oxidant with nonmethane
2
hydrocarbons; that is, those considered to be reactive in photochemical
oxidant formation. Again, the approximate upper-limit observed oxidant
curve forms an envelope around the maximum oxidant concentrations.
Both of the envelope curves from Figures 1 and 2 are plotted on
Figure 3, converted to yield the percentage reduction in hydrocar-
bon emissions (total and nonmethane) required to meet the oxidant
standard. The Appendix J curve is also plotted on Figure 3, and it
falls between the total hydrocarbon (lower) and nonmethane hydrocarbon
(upper) curves.
Both the linear and nonlinear rollback models have been used in Trans-
portation Control Plan development and evaluation. Figure 4 is a com-
parison of the nonlinear rollback of Appendix J with the results of assum-
ing linear rollback, or a linear relationship between hydrocarbons and
oxidants. The method has been to determine what hydrocarbon emission
levels are consistent with meeting or maintaining the NAAQS for oxidant,
although NO concentrations are also important. An additional problem
X
with the use of proportional models for oxidant is that a period of time
is required for the reactions which form oxidant to proceed. The air at
33
-------
0.30
0.25
B DENVER
CINCINNATI
A LOS ANGELES
O PHILADELPHIA
A WASHINGTON
0.20
E
a
O
X
o
0.15
0.10
O BBOACOB ABO A«OB BOA
BAABO OOABBA«OBA A »B A
O AOAOB BOCOAAO4OAB AOAO ABA
AA BAA ABAAOA0 B4MABB A«A
ABAAOA ABA 4BA««A« A A9« A A
AABABAA ABA«B* AOA B ««AB
A A
0.05
2 3
TOTAL HYDROCARBONS. ffm C
Figure 1. Maximum daily 1-hour average oxidant concentrations as a
function of 6- to 9-a.m. averages of total hydrocarbon con-
centrations at CAMP stations, June through September, 1966
through 1968 and in Los Angeles, May through October 1967^
34
-------
0.30
0.25
0.20
0.15
o
x
o
0.10
0.05
APPROXIMATE UPPER-LIMIT
OBSERVED OXIDANT
PHILADELPHIA-
PHILADELPHIA
WASHINGTON A
T
LOS ANGELES
LOS ANGELES
WASHINGTON A
DENVER-
LOS ANGELES
A A PHILADELPHIA
LOS ANGELES
A
WASHINGTON AAA A * A A AAA
* A A A A
A A VA*A A
WASHINGTON *^ A MA A A A A A
/ -AAA A A A 4f A
A AAA
A A
0.5 1.0 1.5
NOKMETHANE HC, ppm C
2.0
2.5
Figure 2. Maximum daily 1-hour average oxidants as a function of
6- to 9-a.m. averages of nonmethane hydrocarbons at
CAMP stations, June through September, 1966 through
1968, Los Angeles, May through October 19672
35
-------
a. a: a:
a. uu. uiu.
< o o oo
fi
CU
a,
CO
C
o
a
o
t-i
cu
rt
cu
I
o
a
-a
CO
(3
O
cd
o
o
cd
4-1
O
M
O
CD
a).
a
o
co
H
SJ
n)
I"
o
ro
a)
)-j
oc
rl
!N3Da3d'lNVaiXO nVOIH3HOOiOHd UOd SOWN 3HJ.
3A3IHDV 01 03din03d SNOISSIN3 NOaWVOOdOAH Nl NOI10n03M
36
-------
O
cd
O
t-i
o)
0)
a
n)
>->
X
0)
a,
(3
O
CO
H
S
I
O
a)
M
H
fn
JLN30y3d 'XNVOIXO -!VDIW3H0010Hd HOJ SOWN
3HJ. 3A3IHDV 01 a3ym03M SNOISSIW3 NOSdVOOMCUH Nl NOI10n03d
37
-------
a monitoring location in the morning 6 to 9 a.m. period is different from
the air at that location during the afternoon oxidant peak. The two con-
centrations cannot strictly be related. This effect is shown in Figure
3
IV-5. The predominant summer wind in Los Angeles is southwesterly; that
is, from downtown Los Angeles towards the USC Medical Center (3 miles
away) and Pasadena (13 miles away). This figure suggests that higher
concentrations of oxidant are experienced downwind after the photochem-
ical reactions have had time to proceed. It also suggests that this
effect becomes smaller as the mixing of pollutants improves, as evi-
denced by the closing of the curves at lower hydrocarbon concentrations.
DATA AND MANPOWER REQUIREMENTS
Nonlinear rollback models require only slightly more effort to apply than
the previous technique in Section III. This would arise if location-
specific curves are prepared as described later in this section. This
should require no more than about 1 manweek. To use these curves to assess
the oxidant impact from highways, the only data requirements are the se-
cond highest hourly average oxidant concentration and the percentage
change in hydrocarbon emissions brought about by the proposed highway or
system change.
APPLICATION OF THE TECHNIQUE
Methods for determining hydrocarbon emissions have been described in
Section III. To apply a proportional model, only the peak (or second
highest) 1-hour average oxidant concentration needs to be known. Using
a linear model, the expected peak oxidant levels would be determined by
a simple ratioing technique. The present peak oxidant value is multi-
plied by the ratio of projected to present hydrocarbon emissions to
estimate the expected peak oxidant concentration. In doing this, one
must decide upon the size of the area to be used for the analysit. No
i.eal guidance is available to assist in this determination, and this
38
-------
I
o.«o
Zfl.35
O
X
O
00.30
tt
Ul
>
a 0.25
I
50.20
I
2
§0.15
u.
00.10
a o.os
Ul
a.
a.
3
11234
6-9 a.m. AVERAGE TOTAL HYDROCARBON
CONCENTRATION, ppm C
Figure 5. Upper limit of maximum daily oxidant at three Los Angeles
County stations, May through October 1967-*
39
-------
is a weakness of this technique. This method assumes that hydrocar-
bon concentrations are linearly related to emissions and that oxidant
concentrations are linearly related to hydrocarbon concentrations, and
hence to hydrocarbon emissions. Alternatively, under the nonlinear
assumption, one could use the upper-limit curve depicted in Figure 1
or the Appendix J relationship to determine the impact of an increase
or decrease in hydrocarbon emissions on oxidant levels.
Curves other than those shown so far have been put forth for specific
cities. These city-specific relationships are an improvement for the
individual locations to which they are applied; the curves shown thus
far are composites made up of data from a number of urban areas. They
are also based on data taken before the EPA reference method (chemilumi-
nescence) for oxidant was defined. Preparation of city-specific curves
based on more recent data would thus likely improve the predictive capa-
bility of this method due to the methodological and locational compati-
bility of measurements.
Figure 6 shows the difference between the upper-limit curve of
3
Figure 1 and a curve based solely on data for Denver. In all cases,
peak oxidant levels in Denver are greater, for a given hydrocarbon con-
centration, than would be estimated from the composite curve. These
curves are based on concentrations of total hydrocarbons.
Figure 7 shows similar relationships for Los Angeles based on non-
methane hydrocarbons and the upper-limit curve of Figure 2. Again,
Los Angeles experiences higher peak oxidant levels for a given level of
hydrocarbons up to oxidant concentrations of about 0.22 ppm. Not every
city would, of course, experience concentrations higher than those
demonstrated by the composite curves.
40
-------
0)
0
-------
!N30M3d '1NVQIXO 1VOW3HOO.LOHd 80J SOWN 3Hi
3A3IHOV 01 CJ3ain03H SNOaaVDOyQAH 3NVH13W-NON Nl NOI10riQ3M
M
o
9
O
o-
oo
O
01
3
O
C
rt
X
o
CO
3
n
CQ
d
o
nj
U
o
O
cfl
4-1
CU
O
R
o
CO
a a
O rH
w cu
H 60
(X
a w
o o
CJ fJ
r-
o)
§,
H
42
-------
In developing a city-specific curve it is important to have sufficient
data to define the upper-limit curve clearly, since the data points de-
fining this curve generally occurs infrequently.
LIMITATIONS
The assumptions that must be made in using a proportional model for pre-
dicting expected maximum oxidant concentrations were discussed at the
beginning of this section. These assumptions become even more tenuous
when trying to predict concentrations of a secondary pollutant non-
linearly related to its precursors, than when trying to predict con-
centrations of a relatively stable pollutant like carbon monoxide. It
should be remembered that this technique is useful only for providing
an indication of what concentrations may result from a change in emis-
sions of hydrocarbons. It does not consider the dependence of oxidants
on the hydrocarbon-nitrogen dioxide ratio.
The Appendix J relationship is independent of city size, although this
problem can be accommodated by the development of a location-specific
curve. The fact that the Appendix J type relationship depends heavily
on monitor location also indicates the advantage of using location-
specific relationships.
SPECIAL FEATURES
This technique is relatively inexpensive and easy to apply, requiring
perhaps 1 man-week more than the previous technique when a location-
specific upper-limit curve is prepared for the 6 to 9 a.m. average
hydrocarbons, maximum daily 1-hour average oxidant concentration re-
lationships. This is also the first technique to be considered thus
far which attempts to describe the nonlinear relationship between
oxidant and oxidant precursor concentrations, based on empirical data.
It thus also comes closer to modeling adequately the expected peak oxi-
dant concentrations.
43
-------
REFERENCES
1. Air Quality and Automobile Emission Control, Volume 3. The
Relationship of Emissions to Ambient Air Quality. U.S. Senate,
Committee on Public Works, Serial No. 93-24. September 1974.
2. Air Quality Criteria for Nitrogen Oxides. U.S. Environmental
Protection Agency. Report AP-84. January 1971.
3. Air Quality Criteria for Hydrocarbons. U.S. Environmental
Protection Agency. Report AP-64. March 1970.
44
-------
SECTION V
GIFFORD-HANNA PHOTOCHEMICAL MODEL
For the past several years F. A. Gifford and S. R. Hanna of the Atmos-
pheric Turbulence and Diffusion Laboratory (ATDL) at Oak Ridge, Tennessee
have been active in the development and validation of simple methods for
estimating pollutant concentrations in urban areas. Much of their effort
had been restricted to the analysis of chemically inert pollutants, but
they have recently generalized their model to treat pollutants undergoing
photochemical reactions.
APPLICABILITY AND RELIABILITY
The Gifford-Hanna model is basically a "box model" in which average pollu-
tant concentrations within the box are taken to be proportional to the
ratio of the average area source strength to the wind speed. The pro-
portionality constant is set equal to the average width of the region
divided by the average depth of the pollutant cloud over the area. This
depth is allowed to vary as a function of atmospheric stability, and it
may be calculated by integrating the Gaussian plume model over the ex-
tent of the area source.
For purposes of chemical kinetics calculations the pollutant concentra-
tions are assumed to be uniform within the volume defined by the area of
the region and the depth of the pollutant cloud. By nondimensionalizing
the concentrations which appear in the equations governing chemical reac-
tions, it is possible to gain some insight into the importance of a par-
ticular reaction upon the concentration of a particular species. For
45
-------
instance, if the steady state nondimensionalized concentration of a par-
ticular substance is close to unity, it is possible to neglect chemical
transformation for the associated meteorological conditions and emission
rates.
The Gifford-Hanna model is intended to be applicable to any location, al-
though published reports of its reliability describe experience with the
model only in Los Angeles and Denver. Table 3 displays results in
Los Angeles. The ozone concentrations were not given in reference 1;
observed values of ozone (and the other pollutants) were obtained from
reference 2, while calculated concentrations were derived from the
relationship:
N02"|/rNo] pphm (15)
= 2
Calculated concentrations of NCL and NO were used to determine 0- concen-
trations, with a correlation coefficient of 0.84 between calculated and ob-
served ozone levels. Ozone concentrations were also calculated using ob-
served N02 and NO concentrations, and the correlation coefficient is 0.97.
Both of these are significant at better than the 2 percent level.
While each of these correlation coefficients is significant, the ques-
tion arises as to the utility of predicting areal average concentrations
for a large region. Also, although the calculated and observed values
correlate well, the "calibration factor," or the ratio of observed to
calculated concentrations becomes progressively greater as the morning
progresses and higher ozone concentrations are observed. The observed
concentration is almost ten times that calculated at noon.
Similar results were obtained in an application of the ATDL model to
3
Denver, Colorado, as presented in Table 4. In this case, concentra-
tions were compared hourly at five locations. Again, the correlation be-
tween observed and calculated ozone concentrations is good, but the ATDL
model underpredicts severely during peak hours.
46
-------
Table 3. COMPARISON OF OBSERVED HOURLY REGIONAL AVERAGE CONCENTRATIONS
OF RH, NO-, NO, AND 0- WITH CONCENTRATIONS PREDICTED BY THE
GIFFORD-HANNA ATDL MODEL FOR LOS ANGELES, SEPTEMBER 30, 1969
Hour
0600
0700
0800
0900
1000
1100
1200
RH (ppm)
CALC
7
7
6
4
3
2
2
OBS
7
7
4
4
4
4
5
r = 0.55
N02 (pphm)
CALC
6
8
7
6
4
3
3
OBS
6
10
17
18
14
11
15
1
r = 0.05
NO (pphm)
CALC
33
23
14
9
7
5
4
OBS
33
30
22
11
4
3
4
r = 0.97
03 (pphm)
CALC3
0.36
0.70
1.00
1.33
1.14
1.20
1.50
OBS
1
2
4
6
9
12
14
CALCb
0.36
0.67
1.55
3.27
7.00
7.33
7.50
r = 0.84:*
r = 0.97b
lfoJ = 2 [N0_l / ["NO"] using calculated concentrations.
'ToJ = 2[N02l/fNol using observed concentrations.
47
-------
Table 4. COMPARISON OF OBSERVED HOURLY AVERAGE CONCENTRATIONS OF
03, HC, AND CO WITH CONCENTRATIONS PREDICTED BY THE
GIFFORD-HANNA ATDL MODEL AT FIVE LOCATIONS IN DENVER,
COLORADO, AUGUST 13, 1973
Station
A
C
D
E
F
Hour
0600
0700
0800
0900
1000
1100
1200
0600
0700
0800
0900
1000
1100
1200
0600
0700
0800
0900
1000
1100
1200
0600
0700
0800
0900
1000
1100
1200
0600
0700
0800
0900
1000
1100
1200
03 (yg/m3)
Calc
26.7
28.3
29.8
31.0
31.9
32.8
33.8
26.7
28.3
29.8
31.0
31.9
32.8
33.8
26.7
28.3
29.8
31.0
31.9
32.8
33.8
26.7
28.3
29.8
31.0
31.9
32.8
33.8
26.7
28.3
29.8
31.0
31.9
32.8
33.8
Obs
29
59
84
-
167
274
382
39
59
98
118
225
323
421
29
59
176
118
_
235
304
65
78
98
114
147
235
49
65
127
255
314
216
r = 0.85
HC (yg/m3)
Calc
2394
21615
6358
3988
5562
3589
4439
1862
16812
4945
3102
4326
2791
3452
1197
10807
3179
1994
2781
1794
2219
1596
9607
12716
10636
7416
5981
7891
Obs
2394
1862
1729
1995
2128
2128
2394
1862
2261
3059
1995
1862
1795
1729
1197
1197
1130
1130
1130
1197
1463
1596
1463
1729
1729
1862
1729
1463
r = 0.14
CO (mg/m3)
Calc
2.3
21.0
6.23
3.95
5.57
3.63
4.54
2.3
21.0
6.23
3.95
5.57
3.63
4.54
4.0
36.5
10.8
6.87
9.68
6.31
7.89
3.4
20.7
27.6
23.4
16.5
13.4
i;,9
Obs
2.3
1.1
1.1
3.4
3.4
3.4
3.4
2.3
3.4
2.3
2.3
2.3
3.4
4.6
4.0
5.7
5.7
5.7
6.9
6.9
5.7
3.4
3.4
4.6
5.7
5.7
5.7
4.6
r = 0.22
48
-------
Based on these results, it is evident that the Gifford-Hanna model might
predict trends in ozone concentrations reasonably well, but it seriously
underpredicts amounts of ozone, especially at higher concentrations.
DATA AND MANPOWER REQUIREMENTS
The data requirements for using the Gifford-Hanna model are basically the
reactive hydrocarbon (assumed to be propylene) emission density, the
early morning (6 a.m.) concentrations of NO and NO., and hourly wind
speed values. If the emission densities of NO and NO are known they
14
may be used, although in application ' it has been assumed that Q =
0.3 Q and Q = 0.2 Q^, and hence QNQ = 1.5 Q. It is also assumed
that at 6 a.m. QMA /QM. = N09 / NO , so that all that is required is
N0£ NO L ^J L J
the reactive hydrocarbon emission density and the 6 a.m. NO concentra-
tion to determine the remaining parameters.
The emissions data have been gridded in applications of the model, and
8 miles squared ' and 8 km squared grids have been used. The choice
of grid size seems to be up to the user; no guidance or specification
has been given. For areal average concentrations, average emissions
for a whole region could be used.
The additional manpower necessary to run this model, compared to the
models previously discussed in this report, is about 1 man-week to col-
lect the wind data and make the concentration calculations.
APPLICATION OF THE TECHNIQUE
The simple Gifford-Hanna ATDL model for nonreactive pollutants calculates
surface concentrations C as being proportional to the local source
strength divided by the wind speed:
Q/U.
(16)
49
-------
The photochemical model is an extension of this simple relationship to
include chemically reactive pollutants.
The reaction scheme in this model was originally proposed by Friedlander
and Seinfeld and consists of the following four equations:
3t
[NOZ] = [NOZ][RH] (a [NO] -x[No2]) d7)
= -a[N<>2] [NO] [RH] (is)
°
, 1-2-1
where a = -r- ppm sec
1 -1
V - r sec
2.4 x 10
n 1 -1 -1
= T PPm sec
3 x 10*
i 1 -2 -1
A = r ppm sec
3 x 10
3 = 0.02 ppin
The concentrations in equations (17) through (20) are indicated in brack-
ets. No functional dependence with solar radiation intensity is specified for
50
-------
these reaction rates. Equations (17), (18), and (19) are nondimen-
sionalized for performing the actual calculations; nondimensionalized con-
r -]* *
centrations C and times t may be found from the following
trans format ion:
Q Ax
t* = fi (22)
Ax
where Ax = A Z
u = wind speed (m/sec)
3 -2 -1
Q = emission density (cm m s )
The parameter A = £x/Z is a function of stability and equals 50, 200,
or 600 m for unstable, neutral, and stable conditions.
Two approaches are possible for the prediction of ozone levels. The first
option is to assume that the process has reached a steady state so that
the time derivative on the left side of the equations can be set to zero.
The equations may then be solved algebraically for the nondimensionalized
concentrations. These can then be redimensionalized and compared with
measured values. Since the steady state approximation may not strictly
be applied due to the time variation of emission density and solar radia-
tion intensity, it is open to some question which hour should be chosen
for the comparison of measured and "steady state" values. In his Los
Angeles validation study, Hanna picked the noon hour for a comparison.
The second way in which the model may be applied is to solve the chemical
kinetics equation numerically by use of a Runge-Kutta technique to obtain
values for nondimensionalized concentrations as a function of time. If
an hour such as 6:00 a.m. is chosen as the time when all nondimensionalized
concentrations are equal to 1.0, then the time variation of the wind speed
-------
and nondimensionalized concentrations may be used to project actual con-
centrations from their 6:00 a.m. values.
Dimensionalized concentrations are computed from their nondimensionalized
form by assuming that source strengths and stability are constant and
that equation (2) holds at 6:00 a.m.:
u.
where u = wind speed (m/sec)
u, = wind speed at 6:00 a.m. (m/sec)
= nondimensionalized concentration
= concentration at 6:00 a.m.
1°
Ozone concentrations are found using equation (20).
LIMITATIONS
A major limitation of this technique is that although it predicts trends
in ozone concentrations well during the day, it predicts actual concen-
trations poorly, especially as they build up during the day. A second
limitation is that it assumes that ozone concentrations depend only on
the ratio of N0« to NO concentrations, when they have been shown to be
a function of hydrocarbon concentrations also. This dependence is treated
indirectly in the ATDL model by the dependence of NO and NO concentra-
tions on reactive hydrocarbon concentrations and emissions in equations
(17), (18), and (19). However, the assumption is still made that pre-
dicted ozone concentrations are directly proportional to the accurately
predicted concentration ratios of N02 to NO. The last column of Table 3
shows that ozone concentrations are not well predicted by this ratio,
even when using measured values of N0_ and NO.
52
-------
The grid element scale is somewhat limiting to spatial location of
peak ozone values. This can be accommodated to some degree by making
the grid elements smaller. However, this presents two further problems.
The first is that the smaller grid size allows even less time for the
development of the photochemical reactions. Including pollutant trans-
port from a number of upwind grids will help this situation somewhat,
but a computer must be used to handle the data efficiently, and this
technique features the lack of computer requirements.
SPECIAL FEATURES
The Gifford-Hanna model marks a transition point. It is the last of the
models considered here which does not require a computer, while it is
the first to be considered which attempts to model photochemical reaction
mechanisms. It also includes explicit consideration of the meteorological
parameters of wind speed and stability. Although the model fails to
predict concentrations of ozone adequately, it could be useful for
examining trends in concentrations. An additional application of the
model has been in determining "critical" wind speeds below which photo-
chemical smog will be a problem in a region.
REFERENCES
1. Hanna, S. R. A Simple Dispersion Model for the Analysis of Chem-
ically Reactive Pollutants. Atmospheric Environment. Pergamon
Press. 7:803-817. 1973.
2. Sklarew, Ralph C., Allan J. Fabrick, and Judith E. Prager. Mathe-
matical Modeling of Photochemical Smog Using the PICK Method.
JAPCA. 22(11):865. November 1972.
3. Photochemical Oxidant Modeling: Detailed Technical Report. GCA/
Technology Division, GCA-TR-75-ll-G(2), Draft Final Report -
Volume II. April 1975.
4. Hanna, S. R. Application of a Simple Model of Photochemical Smog.
3rd Clean Air Congress of the International Union of Air Pollution
Prevention Associations, Dusseldorf, Germany. October 8-12, 1973.
53
-------
5. Friedlander, S. K. and J. H. Seinfeld. A Dynamic Model of Photo-
chemical Smog. Environmental Science and Technology. 3:1175-1181.
1969.
6. Schuck, et al. Relationship of Hydrocarbons to Oxidants in Ambient
Atmospheres. JAPCA, 20(5):297. May 1970.
54
-------
SECTION VI
STATISTICAL RELATIONSHIPS: REPRO-MODELING
Another technique for relating highway system changes to resulting oxi-
dant levels is the derivation of statistical relationships between the
important independent variables (emissions of hydrocarbons and NO ,
X
meteorology, temporal, and spatial effects) and the dependent variables
(peak hourly-averaged oxidant concentrations) in an "input-output" model.
Development of such models is underway within EPA, although proven tech-
niques are not yet available.
One statistical technique which has been successfully applied to oxidant
concentration prediction is repro-modeling, and the remainder of this
section will deal with this approach.
A repro-model is a "model of a model" which seeks to provide an approxi-
mation to a more complex model. The development of a repro-model fo-
cuses on relationships between (some) input and output variables of the
complex model, and the repro-model itself is a "fit" to these relation-
ships. Strong points of repro-models are computation times that may be
orders of magnitude less than those of the complex models on which they
are based, much greater ease of application because of reduced input
data requirements and overall complexity, and the opportunity to make
many more applications of the model, for example, to strategies analysis
for policy planning. The major drawbacks to repro-models are first,
that since they are "models of models" they are no better than the
basis models, and second, that a repro-model may produce only a portion
of the picture available from the basis model. This latter point is
55
-------
due to the fact that only selected input and output variables are
chosen on which to build the repro-model, such as the relationship of
peak ozone levels to emissions, with other variables held constant.
APPLICABILITY AND RELIABILITY
Repro-modeling has been applied to diverse topics including freeway
on-ramp control, radar cross section determination, and photochemical
diffusion processes. A repro-model has been built for the Systems
Applications, Inc. photochemical advection/diffusion model. This repro-
model is based on 100 runs of the SAI model for the Los Angeles basin.
One of these runs used input variables obtained for September 30, 1969,
a day on which high oxidant levels were observed and which was included
in a previous calibration of the SAI model. Several inputs were then
varied in each of the remaining 99 runs. More specifically, five aggre-
gate input variables were used. These were (1) mobile source NO ,
X
(2) mobile source hydrocarbons, (3) fixed source hydrocarbons, (4) NO
X
initial values, and (5) hydrocarbons initial values. Each of these
independent input variables was expressed as a percentage of the
values of the base case test day. Twelve outputs were examined:
(1) the peak 1-hour average oxidant concentration in the Los Angeles
basin, (2 to 9) the peak 1-hour average concentration at each of eight
specific locations in the basin, and (10 to 12) 10-hour average con-
centrations of NO at three specific locations. The overall repro-
model was made up of 12 repro-models, each of which related the five
independent variables to one of the dependent variables. Ninety of the
100 runs of the SAI model were used to build the repro-models; the
other 10 were used for independent testing of the resulting model. De-
tails of the development and specific functional forms of this repro-
model are discussed in reference 1. A more general description of
repro-modeling is presented in reference 2.
56
-------
It should be noted that the repro-model described in reference 1 of
this section has the following limitations to its applicability:
(1) the original (SAI) model on which it is based was calibrated for
Los Angeles only, (2) the meteorological inputs remained constant,
(3) other inputs to the SAI model remained fixed, such as the distri-
bution of traffic, and (4) the repro-model is applicable only so far
as the SAI model is applicable. This single instance to date of
repro-modeling of a complex photochemical diffusion model has developed
relationships between emissions of NO and hydrocarbons and resulting
X
oxidant and NO,, concentrations. It is strictly applicable only to in-
vestigating the effects of changing the magnitude of emissions under
the conditions that were obtained in the Los Angeles basin on September
30, 1969. It is also useful, however, for determining the direction
and relative magnitude of concentration changes for varying emissions.
Other repro-models could be developed from different inputs and outputs
of the SAI model, to investigate the effects of different emissions
distributions, for example. None of these repro-models would be more
reliable than the SAI model on which they are based.
DATA AND MANPOWER REQUIREMENTS
The extensive data and manpower requirements occur during the develop-
ment of the repro-model. They would be greater than those listed in
Section IX of this report which describes the application of the SAI
areawide model.
A main reason for repro-models is the great reduction in manpower and
data needed to run them compared to the requirements of the complex,
basic model. The repro-model of the SAI model applied to the Los
Angeles basin requires only five input parameters:
Percentage of test day mobile source NO emissions
A
Percentage of test day mobile source hydrocarbon emissions
Percentage of test day fixed source hydrocarbon emissions
57
-------
Percentage of test day initial conditions for N(\
X
Percentage of test day initial conditions for hydrocarbons
For this case the test day used for constructing the repro-model was
September 30, 1969. The repro-model may be used to investigate the
effects of alternate emission reduction strategies on oxidant and NO
concentrations relative to the test day. Absolute values of emissions
need not be known, except if required to determine them as a percentage
of the test day emissions on which the repro-model is based. Thus,
the manpower and data requirements, and the computer costs, are rela-
tively small. The development of the repro-model described in this
section required about 2 man-weeks to decide upon the input variable
changes for the SAI model runs and about 1 man-week for building the
3
repro-model. It is emphasized, however, that these manpower figures
exclude the effort in calibrating the SAI model and running the 100
cases. (The SAI model is discussed in Section IX.)
APPLICATION OF THE TECHNIQUE
As discussed previously, the largest effort involved in repro-modeling
occurs during the actual model development. This step will not be de-
scribed here; discussions on building repro-models are available else-
1 2
where. ' Once a repro-model is built, it is easy to apply.
The repro-model constructed for Los Angeles has specific application
to transportation control policy evaluation. Demonstrations of its use
have been to assess the impact on oxidant levels of motor vehicle emis-
sion controls, and the effects of a 1-day VMT reduction. These two ex-
amples of use were accomplished by changing the emissions inputs rela-
tive to the 1969 base case values. The model input changes and the
model outputs for these two examples are shown in Tables 5 and 6.
58
-------
co
Q
Z
H
CO
* X
rJ S
0 S
P4 CX
H P.
§ C
U -H
x^
2
O H
M JZj
CO <
co o
M M
*H ^
w o
l-j |
P§ cd
Wrrt
WJ
Q W
W O
Pn <*!
CD
Pn O
O CO
H W
0 nJ
< M
P. pQ
§ O
H a
,
in
0)
XJ
cd
H
0
d
o
H
4J
cd
0
o
rH
H
i-l
O
°1
CN
rH
rH
CN
rH
CN
rH
rH
1^
O
CM
r,
0
CN
O
CN
»i
in
rH
rH
CN
*
O
1
^H
O 0)
6 o
i-i
X 3
0 0
2 M
M
cd
0)
CM rH rH rH
vO CM CD r^>.
rH rH rH
CT\ CT\ ON CTN
rH rH rH rH
rH rH rH rH O O O
-* CT\
rH rH r-( iH rH r-l rH
rH
Q)
$>
Ol
rH
CTt
vD
rH
4-1
cd
rrj
rH
0)
f!
rH
01
^
cd
M
4-1
MH
O
cn
Ol
rH
H
e
01
rH
a
H
A
0)
3
ent. All initial conditions set at values typical for
a
rl
01
ex
o
o
rH
4-1
cd
-a
rH
Q)
A
cn
C
o
f>
rl
cd
u
o
rl
0
tr*]
0)
o
M
3
o
tn
T3
0)
H
fn
ft
.
£*>
a
H
rH
O
a.
j
r i
o
r-l
4-1
C
O
O
C
O
H
tn
tn
H
0
01
01
O
3
O
tn
0)
rH
H
f*l
O
a
cu
J"l
4_)
basin employed in the SAI model. See Section IX.
cn
0)
rH
0)
60
d
CO
J_l
0)
*-M
01
o
.
rH
0)
U
C
0)
J-l
Ol
4-<
Ol
pcj
..
01
u
(^
3
o
CO
59
-------
Table 6. EFFECTS OF SINGLE DAY TRAFFIC REDUCTION POLICY ON
AIR QUALITY (OXIDANT, pphm)
Zone
Peak
10,24
10,21
15,20
20,20
7,17
12,17
25,13
12, 9
3,
Baseline
45
45
25
30
16
3
14
16
2
20% reduction
in mobile .
Source emissions
37
37
20
25
13
3
11
14
1
50% of
£
baseline
7
6
3
4
3
2
2
3
1
20% reduction
of mobile
source emissions
from 50% based
7
4
3
3
3
2
1
3
2
Policy: 100,100,100,100,100 (NOx and HC mobile source emissions,
HC fixed source emissions, and NOx and HC initial
condition, respectively.)
bPolicy: 80,80,100,100,100
cPolicy: 50, 50, 50, 69, 50. The 69% value of NOx initial
condition variable results from the uncontrolled
NOx fixed sources.
dPolicy: 40,40,50,69,50
Source; Reference 1.
60
-------
The user of the repro-model is not completely free to vary the Input
values without bound. Valid input changes are constrained by the input
and output values used to build the model. Table 7 shows the con-
straints for the Los Angeles model.
LIMITATIONS
The basic limitation to the repro-modeling approach is that the repro-
model is a functional representation of the inputs and outputs of ano-
ther, more complex basis model The repro-model will reproduce the re-
sults of the basis model, but it will not produce results which are
closer to reality. Also, the repro-model incorporates only a portion
of the input and output variables of the basis model, restricting its
applicability to specific conditions. For the Los Angeles case, the
only changeable variables are magnitudes of NO and hydrocarbon emis-
X
sions. Spatial variation cannot be modeled, nor can changes in meteor-
ological parameters. A final limitation to the use of repro-models,
as opposed to the repro-modeling technique, is that at present there
exists only the model for Los Angeles. Application to other loca-
tions depends on the application and calibration of one of the photo-
chemical diffusion models discussed in following sections to these
other locations.
SPECIAL FEATURES
The fact that a complex photochemical diffusion model must be calibrated
for a location before a repro-model can be built raises the question of
why develop the repro-model at all when the basis model is available.
The answer to this question derives from a major feature of repro-
modeling the low cost and ease of operation compared to the basis
model. For locations lacking the resources to continually apply the
complex basis model, a repro-model could be developed, based on care-
ful considerations of inputs and outputs in light of potential decision-
making applications, which could then be used repeatedly to assess air
61
-------
Table 7. INPUT VARIABLE CONSTRAINTS
x + x > 30
Xl + x2 < 240
XL - x2 < 40
-x± + x2 < 20
x2 > 0
x3 > 0
XB < 100
x. - 0.558x. > 29.2
4 1
x, - 0.682X.. < 46.8
-x5 + 0.756x2 + 0.144x3 < 5
x5 - 0.924x2 - 0.176x3 < 5
x5 > 0
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
(11)
(12)
x1 = 7, of test day's mobile source NOx
emissions
~ = 7> of test day's mobile source hydro-
carbon emissions
x., = 7o of test day's fixed source hydro-
carbon emissions
x, = 7o of test day's initial and boundary
conditions for NOx
x_ = 7o of test day's initial and boundary
conditions for hydrocarbons
Source: Reference 1.
62
-------
quality impacts. Even if adequate resources are available to apply a
complex model repeatedly, in most instances it would be more cost
effective to use repro-models for repeated applications.
REFERENCES
1. Horowitz, Alan, Meisel, William S., and David C« Collins. The
Application of Repro-Modeling to the Analysis of a Photochemical
Pollution Model. Technology Service Corporation Report,
prepared for Office of Research and Development, USEPA,
Washington, DoC. December 1973
2. Meisel, William S., and David C. Collins. Repro-Modeling:
An Approach to Efficient Model Utilization and Interpretation.
IEEE Transactions on Systems, Man, and Cybernetics. Vol.
SMC-3. No. 4. July 1973.
3. Teener, M. D., Technology Service Corporation, Santa Monica,
California. Personal Communications.
63
-------
SECTION VII
REACTIVE ENVIRONMENTAL SIMULATION MODEL
INTRODUCTION
The Reactive Environmental Simulation Model (REM) is the first of three
large computer models described in this report. Each of these models
is based on the general form of the mass conserving advection/diffusion
equation:
f^f^ /Af /^/"i f^r^
W\J , \J\J f \-J\J m \J\J »
1 , x , * i
h u ^ 1- V
/.,
lK
,..
(24)
The mean concentration of pollutant species i is given by C. ; u, v, and w
are the horizontal and vertical wind components; K , K , and K are the
v x y z
horizontal and vertical eddy dif fusivities ; and R. and S. are the average
reaction and emission rates for the i species.
The REM and DIFKIN (described in the following section) models are tra-
jectory models based on a Lagrangian or moving coordinate system. A
vertical air column moves along a trajectory determined by the horizontal
wind components. Pollutant concentrations are assumed to be homogeneous
horizontally in the air column, and vertical advection and horizontal
diffusion are ignored, so that equation 24 becomes:
64
-------
KZ
The REM model makes a further simplifying assumption regarding vertical
diffusion. It considers the effect of vertical diffusion to be solely
the production of a homogeneous concentration of pollutants in the ver-
tical from the ground to the mixing height. Pollutant concentrations
within the column of air thus depend upon the emission of pollutants
into the column, productive and destructive photochemical reactions,
and the mixing height. Hence, equation (25) becomes:
si + Ri + Hi (26)
where H. is the rate of change of the concentration of pollutant
i due to variation in the mixing height.
Changes in mixing height occur from changes of the base elevation due
to changing topography, while changes in the elevation of the top of the
mixing height are computed as functions of the air temperature at the
ground. The effects of changes in volume of the air column due to mixing
height fluctuations (the base area is constant) on pollutant concentra-
tions within the column are the following. If the columnar volume in-
creases the concentrations decrease to conserve mass. If, on the other
hand, the volume decreases, the concentration is assumed to remain con-
stant while part of the mass (proportional to the volume decrease) is
assumed to be lost through the top of the column.
Under the assumptions expressed in equation (26) the REM model treats
the column of air as a moving smog chamber which has a variable volume
and the capacity for fresh contaminant input during irradiation.
Consistent with this treatment, emphasis is placed on adequately des-
cribing the reaction mechanisms occurring during photochemical smog for-
mation as derived from irradiation chamber data. REM uses a more detailed
65
-------
chemical kinetics scheme than either the DIFKIN or Areawide models to
be discussed in subsequent sections. Table 8 presents the chemical
reactions and their respective rates included in REM. These can be
compared with the reaction schemes presented in later sections of this
report
Because of the important dependence of concentrations on the mixing
height of the air column relative to the ground elevation, REM computes
mixing heights along the trajectory. This is accomplished using the
standard technique of intercepting the morning or afternoon sounding
data with the dry adiabatic lapse rate from the surface temperature.
The trajectory of the air column is calculated by inverse distance
squared weighting of wind velocity data from reporting meteorological
stations. The individual components of wind velocity from each station
are divided by the square of the station distance from the air column;
these are summed, and multiplied by the sum of the squares of the
distances.
This introduction has outlined the major points of the REM model. More
details regarding the development of the model and the specific cal-
culational procedures may be found in references 1 and 2.
APPLICABILITY AND RELIABILITY
REM was designed around a photochemical reaction scheme built to repli-
cate smog chamber data as closely as practicable. The smog chamber
simulation was then extended to the atmosphere under the assumptions on
advection and diffusion previously discussed: a Lagrangian, or moving,
coordinate system is attached to a column of air with a fixed base area
and a height which varies with topography and the atmospheric iitixing
height, and in which pollutant concentrations are homogeneously mixed.
66
-------
Table 8. CHEMICAL DYNAMICS MODULE FOR REM
Reaction
Rate (pphrn" min~ )
1 N02+HV
2 0+02+M
3 03+NO
4 N02+03
5 C2H303 + 02
6 N03 + NO
7 C3H6 + OH
8 NO + H02
9 H + 02 + M
10 C2H30 + 02
11 N02 + OH
12 OH + 03
13 OH + CO
14 CH3 + 02 + M
15 CH302 + NO
16 CH30 + 02 + NO
17 C2H303 + NO
18 C2H302 + NO
19 C2H402 + NO
20 CH30 + 02
21 C3H6 + 0
22 C3H6 + 03
23 C3H6 +0+02
24 C2H402 + 02
25 C3H6 + H02
26 C3H6 + ME02
27 C2H30 + M
28 CH30 + N02
29 C2H303 + N02
30 DUMHC + 0
31 HCHO + HV
32 N02 + N03 + H20
- NO + 0
- 03 + M
- N02 + 02
= N03 + 02
= C2H302 + 03
«= 2N02
= CH3CHO + CH3
= N02 + OH
- H02 + M
= C2H303
» HN03
- H02 + 02
- H + C02
- CH302 + M
= CH30 + N02
= CH302 + N02
= C2H302 + N02
= C2H30 + N02
- CH3CHO + N02
= HCHO + H02
- CH3 + C2H30
- HCHO + C2H402
- HCHO + C2H402
«* C2H303 + OH
« CH30 + CH3CHO
= CH3 + MEO + C2H30
= CH3 + CO + M
= CH30N02
= C2H303M02
= CH3 + C2H30
= H + H + CO
- HN03 + "
4 x 10"1
1.4 x 10e
-,-1
4 x 10
1 x 10
,* J
2 x 1
-------
The trajectory of the air column is determined by inverse distance squared
weighting of wind speed and direction data from reporting stations. So
far as the assumptions made in building REM are plausible the model
should be applicable. The suppression of a more thorough description of
atmospheric transport and diffusion of the pollutant species, at least
for Los Angeles, is based on field study da
as a function of altitude over Los Angeles.
for Los Angeles, is based on field study data of pollutant concentrations
In theory, the model could be applied to any location having sufficient
data to make the trajectory and mixing height calculations. The tra-
jectory could be computed using data from one meteorological station,
although the accuracy improves substantially with an increasing number
of stations. The availability of temperature sounding data to use in
mixing height calculations is somewhat limited, although such data are
available for a number of urban areas. These data are routinely taken
by the National Weather Service, and historical records are maintained
at the National Climatic Center in Asheville, N. C. The model does,
however, allow for direct input of estimated mixing heights should lack
of data preclude their calculation.
Application of REM to highway system evaluation would involve making a
number of model runs for different trajectories to develop sufficient
detail. This involves trajectories for different end points, for spa-
tial detail, and different ending times, for temporal detail to ensure
that highest values are determined. The end point and time of the tra-
jectory are basic inputs to the model, so that it is relatively simple
to specify exactly the locations to be modeled.
REM has been tested only in Los Angeles. The results of the validation
runs are summarized as follows:
CO, less than a factor of 2 in more than 80 percent of
the comparisons; 40 percent agreement within one part
per million.
68
-------
0_, within a factor of 2, 75 percent of the comparisons.
NO, within a factor of 2, 75 percent of the comparisons;
90 percent agreement within 0.02 ppm.
NO-, within a factor of 2, 60 percent of the comparisons.
These comparisons are between instantaneous predicted values on the half
hour and the average observed concentration for the hour. The model
has not yet been tested in other cities.
DATA AND MANPOWER REQUIREMENTS
The data inputs which are required to implement REM can be divided into
those which pertain to the particular location being modeled, such as
gridded emission densities, and those which pertain to the operation of
the model, such as the number of lines to be used in printer plot out-
put. The latter types of inputs are not discussed here, since they are
of most concern to the modeler who will operate REM. They are discussed
in detail in reference 2. Details on format of the location-specific
inputs are also excluded in this discussion for the same reasons.
REM uses a 25 x 26 grid element network to locate the moving air column,
the fixed meteorological and air monitoring stations, and to allocate
emissions. Each grid element is 2 miles on a side, although 2 kilometer
squared grid elements may be used. Area source emissions for each grid
element are required as direct inputs to the model for the following
pollutants:
Reactive hydrocarbons (propylene)
"Less reactive" hydrocarbons
Nitric oxide
Mobile source emissions are computed in the model from the following
inputs:
69
-------
Emission factors for
- Reactive hydrocarbons (propylene)
- Less reactive hydrocarbons
- Nitric oxide
- Carbon monoxide
Freeway traffic mileage for each grid element
(daily totals)
Street traffic mileage for each grid element
(daily totals)
Built into REM is the temporal distribution of daily traffic for freeways
and streets. The fraction of the daily traffic counts corresponding to
the time of day at which the air column passes through a particular grid
element is multiplied by the total daily traffic to find the VMT for
that hour for that grid.
Meteorological data required as inputs to the model include, for each
station as available for up to 32 stations are:
Wind speed
Wind direction
Temperature
Dew point
The model computes relative humidity internally from temperature and
dew point data. Relative humidity is used in calculating the photo-
dissociation rate constant for N0«. Also included in this calculation
are cloud type and solar angle. Cloud type must be input to the model
based on the following types:
Cirrus
Cirrostratus
70
-------
Altocumulus
Altostratus
Stratocumulus
Stratus
Fog
The solar angle is required for each hour of the day.
The final meteorological inputs required to run REM relate to mixing
height. These data may be input either as mixing heights by hour of
day directly, or as the previously discussed morning and afternoon
temperature soundings. In the latter case the program computes mixing
heights from the soundings and surface temperature data.
As mentioned earlier, REM places most emphasis on the chemical reaction
mechanisms. These must be supplied as inputs consisting of the chemical
reactions and their rate constants.
The reactions and rate constants were given in Table 8. They were
derived from reaction chamber data and can be used for locations other
than Los Angeles, with the exception of the NO. photodissociation rate,
which should be adjusted to other locations. The final inputs required
are the initial pollutant concentrations in the air column at the be-
ginning time and location of the trajectory.
Because of the relative simplicity in setting up the input data for REM
(DIFKIN and the SAI Areawide models are discussed in the next two sec-
tions) it requires the least effort of these three models. The largest
effort, which is common to all three models, is establishing the grid
network and allocating the VMT and area source data to each grid element.
Table 9 indicates the estimated manpower needed for this effort,
based on GCA experience in Denver, Colorado. These estimates are given
71
-------
as guides with the understanding that there can be variability from
location to location, depending on the availability of recent traffic
surveys and emissions inventories. For locations where the APRAC model
has been applied, the gridding procedure and VMT allocation should be
readily available. Preparation of the remaining inputs on meteorological
variables and initial concentrations should take approximately 80 manhours
at most, and the bulk of this time would be spent collecting the data
from different sources.
Table 9. ESTIMATED MAXIMUM TIME AND COSTS FOR OBTAINING TRAFFIC
AND AREA SOURCE INPUTS FOR REM
Work required
Establish grid, code
traffic links to grid
Determine diurnal
variation in VMT
Allocate area source
emissions to grid
Total
Estimated
man-hours
400b
80
80C
560
o
Estimated cost
Personnel
$2,800
560
560
$3,920
Computer
$800
$800
Total
$3,600
560
560
$4,720
Assumes an average rate of $7.00 per man-hour and $800 per hour of
computer time.
Based on experience in Denver, Colorado, for DIFKIN model.
ft
Assumes emissions are allocated homogeneously to grid elements from
county-wide totals.
APPLICATION OF THE TECHNIQUE
Details on implementing REM for photochemical oxidant prediction are
provided in reference 2. This reference may be consulted for full in-
formation on required data inputs and formats, while means for obtaining
these inputs are discussed here.
72
-------
The main effort in getting REM up and running is overlaying the grid
network and allocating the area source emissions and mobile source VMT
to each grid element. Obtaining these data was discussed in Sections II
and III of this report. Area source emissions can be allocated to each
grid according to area from county-wide emission inventory data from
NEDS. The fraction of county-wide emissions (or some other geographical
detail) could also be weighted into each grid element by population den-
sity. Point source emissions, which are treated as area source emis-
sions because of the homogeneous concentration profile in the vertical
direction which is assumed in the model, should be allocated directly
to the grid element which the point source occupies.
VMT data, as discussed in Section II, are generally available from the
local highway department or planning agency for roadway links. These
data must be allocated to the individual grid elements. This almost
necessitates the use of a computer because of the volume of data to be
handled. Programs have been written to grid the VMT from link data, but
none is generally available. Agencies which have used the APRAC model
for carbon monoxide will usually have such programs available, and, if
APRAC has been run recently, they may have the required VMT data in
gridded form. The local highway department or planning agency should
also be contacted for temporal distributions of street and freeway
traffic.
Meteorological data include station locations, including elevation, wind
speed and direction, temperature, dew point, and temperature soundings.
Wind data may be available from local air pollution agencies, highway
agencies, airports, universities, and power plants. Temperature and
dew point data are usually not so readily available. The National
Weather Service or local airports taking surface observations should be
able to supply these data. Temperature soundings, used to compute
mixing heights, are least available. The National Weather Service should
be contacted to determine if these data are taken for a particular
73
-------
location. REM has the option of using mixing height inputs directly.
These are also available from the National Weather Service as inter-
polated values for many locations which do not have sounding data.
1 2
Solar zenith angle is given in the model ' for Los Angeles. For other
4
locations, Tables 169 and 170 of the Smithsonian Meteorological Tables
can be used to compute the angle for each hour of the day for any day
of the year.
LIMITATIONS
The general application of REM in its current form is limited by its
specificity to Los Angeles and by the fact that validation has been
attempted only for Los Angeles. Neither of these problems is excessively
severe, however. The Los Angeles specific feature which is "built-in"
to the model applies to meteorological stations. The effect is to
disallow the use of pertain station and trajectory location combinations
because of intervening mountains. This is easily changed by removing
7
SUBROUTINE BARIER from the computer program. Validation for cities
other than Los Angeles could be accomplished relatively easily, compared
to the DIFKIN model and the SAI Areawide model.
A limitation of REM for highway system review is its basic structure
around a Lagrangian, or moving coordinate system. Since oxidant con-
centrations are computed only at a specific point at a desired time, or
at a specific time at a desired point, the model must be run repeatedly
to gather enough space and time points to assess fully VMT temporal
and spatial changes.
Another limitation to REM, which can, of course, be eliminated by
modifying the computer code, is its treatment of mobile source emissions.
The model calls for inputs of emission factors for reactive hydrocarbons,
"less reactive" hydrocarbons, nitric oxide, and carbon monoxide. The
74
-------
emission factors are presumably for a mix of light-duty vehicles. There
is no mention of heavy-duty vehicles either for separate factors or
for inclusion in a mix with light-duty vehicles. There is no speed
adjustment for emission factors, although VMT for street and freeway
traffic for each grid element are required inputs to the model.
SPECIAL FEATURES
The REM model has spatial and temporal detail far in excess of any of
the models considered thus far. As a trajectory model, this detail makes
it suitable for the evaluation of highway project impacts on oxidant
concentrations. It can also be used for system evaluation, although
this is somewhat clumsy, since it requires numerous model runs to develop
enough information for full assessment.
While REM is weak in its treatment of atmospheric transport and diffu-
sion processes, it uses a detailed chemical module for explicit calcula-
tion of photochemical processes. It considers almost twice as many
chemical reactions as any of the other models described in this report.
However, this does not necessarily imply that its treatment depicts
reality any better than DIFKIN or the SAI model.
A final feature of REM is that it is easier to apply, requiring fewer
data inputs and having a simpler code, than DIFKIN or the SAI model.
REFERENCES
1. Controlled Evaluation of the Reactive Environmental Simulation
Model (REM). Volume I: Final Report. Pacific Environmental
Services, Inc. EPA Report EPA R4-73-013a. February 1973.
2. Controlled Evaluation of the Reactive Environmental Simulation
Model (REM). Volume II: User's Guide. Pacific Environmental
Services, Inc. EPA Report EPA R4-73-013b. July 1973.
75
-------
3. "1969 Atmospheric Reaction Studies in the Los Angeles
Basin, Volume I," Scott Research Laboratories, Final Report,
Contract No. CPA 70-6.
4. Smithsonian Meteorological Tables, Publication 4014, Smith-
sonian Institution, Washington, D. C. 1951,.
76
-------
SECTION VIII
DIFFUSION KINETICS MODEL
The next level of model complexity is the Diffusion Kinetics (DIFKIN)
model. As in the case of REM, DIFKIN calcuates the trajectory of an air
parcel across an emission grid network and determines time dependent con-
centrations of teactive pollutants. Unlike REM, in DIFKIN the equation
for turbulent diffusion in the vertical direction is solved to obtain con-
centrations and fluxes for as many as ten mesh points from the ground
surface to the top of the mixing layer.
APPLICABILITY AND RELIABILITY
As a trajectory model, DIFKIN is most applicable for project level analysis.
It is ideally suited to relating emissions along an air parcel trajectory
to concentrations observed at later times further along the trajectory.
However, it can still be used for highway system analysis, as was the case
with REM, by repeated applications of the model to develop a sufficient
number of trajectories for adequate spatial and temporal coverage of a
region.
Although DIFKIN considers vertical diffusion of pollutants up to the mixing
height, horizontal diffusion is ignored. This implies that the model is
more applicable to regions with only small emission gradients between
adjacent grid elements. The effect of ignoring horizontal diffusion has
been investigated,* and under "worse case" emission gradient conditions the
errors are about 40 percent. For normal operating conditions the effect is
negligible.
77
-------
Again, as with REM, DIFKIN calculates a trajectory along which the air
parcel travels by interpolating between wind data taken at monitoring
sites in the region. The greater the density of these meteorological
monitoring sites the better DIFKIN will be able to replicate a "true"
trajectory. Thus, in practice, DIFKIN becomes more applicable opera-
tionally as the available meteorological monitoring sites become more
dense in a region.
Tests of the validity of DIFKIN have been performed in Los Angeles,^ San
Francisco, and Denver.3 For the Los Angeles case, validation was per-
formed for the six day period in the fall of 1969 for which REM and the
SAI model were also validated. Three.of these days involved "hands-on"
tests to vary different model parameters and calibrate the model. On the
other three days the model was run "hands-off." Table 10 shows the
correlation coefficients between calculated and observed ozone concentra-
tions for the "hands-on," the "hands-off," and the composite model runs,
and the regression equation for the composite.
Table 10. CORRELATION COEFFICIENTS AND REGRESSION
EQUATION FOR DIFKIN IN LOS ANGELES1
r
n
Hands-on
0.94
= 75
Hands-off
0.88
= 75
Composite
0.92
151
Observed ozone (pphm) = 0.840 calculated ozone (pphm)
+ 2.307 (pphm)
Standard error = 2.109 pphm
78
-------
From the regression equation, it is seen that DIFKIN tends to underestimate
low concentrations and to overpredict at high concentrations, with the
crossover occurring at about 15 pphm of ozone.
During the validation, the emission fluxes of NO were reduced by 75 per-
cent from the emission inventory estimates in order to gain better agree-
ment with observed ozone concentrations. It was also found* that pre-
dictions of NO and N02 had to be sacrificed to improve the ozone
correlations.
Perhaps the most significant result, which was also experienced with
3
the Denver case, was the very high sensitivity of the model predictions
to the initial pollutant concentrations used to "seed" the calculations.
This was found to be very important for trajectories beginning about 6:00
a.m. and running for a time less than 8 hours. A preliminary sensitivity
analysis of DIFKIN has been performed by the California Division of High-
ways. This analysis determined the percent change in predicted concen-
trations of 03, NO, N0~, and HC for a 50 percent increase in each of
various model input parameters dealing with initial pollutant concentra-
tions, chemical reaction rates, meteorological inputs, and growth and
adjustment factors. The results, which are given in Table 11, show
the highest sensitivity to be that of ozone to the initial concentration
of hydrocarbons.
o
In an application for San Francisco, the model worked well, with a cor-
relation coefficient between calculated ozone and observed oxidant con-
centrations of 0.88 for 22 data pairs. These comparisons are for tra-
jectory end points, and they range from noon to 5:00 p.m. The root-mean-
square difference between calculated and observed or interpolated
concentrations was 2.65 pphm. Although NO emissions were not reduced by
75 percent in these model runs, minor adjustments were made in 9 of the
22 trajectories. Each of the trajectories began near the coast or over
the ocean. As a result, it was assumed that pollutants were uniformly
mixed in the vertical direction, and that the initial concentrations were
79
-------
Table 11. PERCENT CHANGE IN PREDICTED CONCENTRATIONS OF
03, NO, N02, AND HC FOR A 50 PERCENT INCREASE
IN EACH OF VARIOUS MODEL PARAMETERS4
50 percent increase in
Initial concentration of :
03
NO
N02
HC
Reaction rates:
Ki
K2
KA
K5
6
7
KS
K9
KIO
KH
K12
K13
Kl5
K16
Basic inputs:
Mixing height
Vertical mesh interval
Growth and adjustment factors:
Stationary sources (all
pollutants)
Automobile growth (all
pollutants)
NOX: Stationary
Power plant
Oil refinery
Surface street
Freeway
HC: Stationary
Oil refinery
Surface street
Freeway
Percent change in prediction of
03
0
-47
-13
187
*
-24
3
147
0
-3
-24
-53
3
0
-16
12
0
0
13
13
81
37
-42
-21
-32
-24
-13
-29
-18
21
5
16
3
NO
0
86
14
-66
*
-3
-3
-59
0
0
24
83
-3
0
24
0
0
0
-14
-24
-39
-21
97
55
45
38
14
45
28
-17
-7
-10
-3
N02
0
-1
0
4
*
3
0
3
0
-1
-3
-13
0
0
1
0
0
0
0
-7
-47
-3
13
8
4
4
1
4
3
2
0
1
0
HC
0
5
1
18
*
1
-1
-14
-1
0
3
8
-4
0
7
0
0
0
-1
-1
-48
3
16
11
1
1
0
1
1
10
1
7
1
Base case concentration (pphm)a 3.8 2.9 15.7 ' 13.4
* Function of sun angle.
* Concentration at 1700 (maximum predicted 0- concentration) after 10-
hour simulation.
80
-------
1 pphm for all pollutants except CO and ozone. CO was assumed to be
present at 1 ppm, while ozone was based on the relationship
[NO ]
- (27)
3
A preliminary application of DIFKIN to Denver yielded a correlation of
0.33 between calculated and observed concentrations for five data pairs.
The DIFKIN calculated values were lower than observed ozone levels in all
cases. The Denver experience also showed the predicted values to be
very sensitive to the initial hydrocarbon concentrations. Additionally,
it was found that the rise in calculated ozone concentrations with
time was relatively insensitive to the trajectory which was followed,
and hence to the emissions being input into the air parcel.
The only change made in DIFKIN for application to Denver (aside from re-
moving mountain barriers) was to increase the photodissociation rate for
the production of ozone and NO from N02 by 15 percent to account for the
increased solar intensity due to the altitude increase. The emission
fluxes of NO were used directly and not reduced by 75 percent.
DATA AND MANPOWER REQUIREMENTS
A rather extensive input data set is required by DIFKIN. Much of this in-
volves job control parameters, while other data apply to emissions, mete-
orology, and the photochemical reaction scheme and rates. A user's guide
to DIFKIN is available^ which describes the required inputs in detail. An
overview of the emissions, meteorological, and chemical reaction data in-
puts is presented here.
The emissions inputs constitute the largest volume of data necessary to
operate the DIFKIN code. Most of these data are contained within the
81
-------
program itself, so that the code must actually be modified when these
emissions inputs are changed. The emissions are categorized by mobile and
stationary source. The mobile source emission parameters, their units,
and the required resolution are displayed in Table 12,,
It is evident from Table 12 that the mobile source emissions are
treated in much more detail than was the case with REM. Diurnal variations
in traffic volumes, speed differentials for fast and slow directions of
travel, and corrections for speed, cold starts, and hot starts are
included.
The emissions data needed for stationary sources are given in Table 13.
These include emissions from power plants and oil refineries as separate
inputs. All other stationary source emissions fall into the category of
"distributed emissions." Again displaying the potential to accommodate
more detailed emissions inputs than REM, DIFKIN requires the hourly frac-
tion of the total daily emissions for each of the three stationary source
types.
The basic meteorological inputs to the model are hourly wind speed and
direction measurements from as many stations as possible so that an air
parcel trajectory can be determined. In addition, DIFKIN employs up to
ten vertical mesh points at which concentrations are calculated, and at
which vertical diffusivity coefficients must be assigned. The mixing
height is the final meteorological input to the model. It is included for
completeness, although it is not used and hence may be omitted. The re-
quirement for vertical diffusivity coefficients, the treatment of ver-
tical diffusion, and the neglect of the mixing height show a fundamental
difference in the approaches of REM and DIFKIN. REM assumes a well-defined
volume of air, bounded by the ground at the bottom and the mixing height
at the top, in which pollutants are well mixed. DIFKIN, on ti a other
hand, uses up to eight vertical increments through which pollutants are
diffused and concentrations are calculated.
82
-------
Table 12. VEHICULAR EMISSIONS DATA REQUIRED FOR DIFKItT
Parameter
Freeway VMT
Surface street VMT
Fraction of total daily
freeway VMT
Fraction of total daily
surface street VMT
Off peak average speed on
freeways
Peak traffic average speeds
on freeways (slow
direction)
Peak traffic average speeds
on freeways (fast
direction)
Peak traffic volume ratio
on freeways (slow/fast)
Fraction of vehicle starts
which are cold starts
Correction factor to
account for emissions
following cold starts
Exponent for emissions
versus speed
correlation
Hot-running emission factors
Cold-start cycle emission
factors
Linear coefficient for
emissions versus speed
correlation
Units
kmi
kmi
-
-
mph
mph
mph
-
_
-
-
grams
mile
grams
mile
_
Resolution
Each grid square
Each grid square
Each hour
Each hour
All hours except
6:00-10:00 a.m.
Each grid square,
each hour between
6:00 a.m. and
10:00 a.m.
Each grid square,
each hour between
6:00 a.m. and
10:00 a.m.
Each grid square,
each hour between
6:00 a.m. and
10:00 a.m.
In 8 time periods over
24 hours
Each pollutant, 1.0
before 6:00 a.m.
and after 9:20 a.m.,
function in 19
linear segments
between 6:00 and
9:20 a.m., suf-
face streets only
Each pollutant, free-
ways only
Each pollutant
Each pollutant
Each pollutant, free-
ways only
83
-------
Table 13. STATIONARY SOURCE EMISSIONS DATA REQUIRED FOR DIFKIN"
Parameter
Power plant NOX emissions
Oil refinery NOX emissions
Oil refinery reactive HC
emissions
Distributed source NOX
emissions
Distributed source reactive
HC emissions
Fraction of total daily
power plant emissions
Fraction of total daily
oil refinery emissions
Fraction of total daily
distributed source
emissions
Units
kg/hr
kg/hr
kg/hr
kg/hr
kg/hr
-
Resolution
Daily average each grid
square
Daily average each grid
square
Daily average each grid
square
Each grid square
6:00 a.m. -6: 00 p.m. average
Each grid square
6:00 a.m. -6:00 p.m. average
Each hour
Each hour
Each hour
DIFKIN employs half as many chemical reactions as REM. The reactions, the
reaction rates, and the stoichiometric coefficients, are all input to the
model and are shown in Table 14. These have been standard in use
except for the N0£ photodissociation rate constant for the first reaction
displayed in Table 14.
The increased data requirements of DIFKIN imply a greater effort to apply
the model compared to REM. Most of this would be expended in actually
understanding the code and getting the model operational, and much of this
effort would be spent in revising the emission data which are maintained
internally in the code. Collection of the emissions data and the gridding
procedure are relatively similar to those of REM, and thus Table 9,
which is actually based on experience with DIFKIN, is a reasonable guide
for both models.
84
-------
Table 14. PHOTOCHEMICAL REACTIONS AND RATE
CONSTANTS USED IN DIFKIN
Reaction
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
HV + N02 = NO + 03
NO + 03 = N02 = 02
0 + HC = 8R02
OH + HC = 8R02
R02 + NO = N02 + 0.125*OH
R02 + N02 = PAN
OH + NO = HN02
OH + N02 = NH03
03 + HC = R02
HV + HONO = OH + NO
N02 + 03 = N03 + 02
N03 + N02 = N205
N205 = N03 + N02
N205 + H20 = 2HN03
NO + N02 + H20 = 2HN02
N02 + PARTICLES =» PRODUCTS
Rate,
pphm~l min-1
1 x 10-*
2.67 x 10-1
1 x 10~6
1 x 102 a
1 x 103
2 x 10°
1.5 x 101
3 x 101
4 x lO-5
3.7 x ID'7
5 x 10-5
4.5 x 101
1.4 x 101
6.05 x 101
1 x 1(T5
1 x 10-3
Variable dependent upon intial ratio of HC to NOX.
y O fi
Experiences with DIFKIN ' ' have had total costs of implementation of ap-
proximately $25,000 to $35,000. This breaks down to roughly 10 to 15 per-
cent for computer time and 85 to 90 percent for labor. Once the initial
experience in applying DIFKIN has been gained by an organization, further
applications to other locations can be made for an estimated $10,000 to
$15,000.
85
-------
APPLICATION OF THE TECHNIQUE
As in the previous section, an overview will be given here on applying
DIFKIN. References 1 and 5 provide the detailed information needed to run
the model. This discussion draws on information presented in Reference 3.
As described previously, certain programming changes are required to run
DIFKIN for locations other than Los Angeles. These changes are needed
largely due to the fact that much of the data which might more conveniently
be read in externally are actually wired into the code itself. The fol-
lowing DIFKIN program modifications are required to incorporate a new data
base:
The appropriate wind station names and coordinates, compass
point azimuths, and distance criteria must be inserted in
"SETIN", a DIFKIN subroutine called upon to initialize the
aforementioned variables.
Subroutine BARIER, a routine which checks whether or not
two given points are on the same side of each of a set
of finite length, straight line barriers, in most cases
must be altered to remove the effect of Los Angeles
topographical features such as the San Gabriel mountains,
Santa Monica mountains, and Palos Verdes hills. Inter-
polation of wind speed and direction measurements for
trajectory calculations is not permitted between those
stations located on opposite sides of a barrier.
BLOCK DATA programs ONE through FIVE must be changed so
that new mobile and stationary source emission parameters
can be entered.
Emissions inputs are needed for stationary and mobile sources for CO, NO,
and reactive hydrocarbons. Emissions of N02 are converted to NO emissions
in the model by multiplying by the ratio of molecular \?eights, or 30/46.
It is also assumed in the model that mobile source hydrocarbon emissions
are composed of a reactive fraction which is about 70 percent of the total
by weight. Different emission factors may be input for hot-start and cold-
start emissions if these are available. If they are not both available,
86
-------
9 10
emissions derived from AP-42 may be used in both cases. A study of
cold-start effects on photochemical smog has shown negligible impact.
Stationary source hydrocarbon emissions are to be input as reactive hydro-
carbons. Studies in Los Angeles?,8 classify 70 percent of stationary
source hydrocarbon emissions (excluding refineries) as reactive, while 50
percent of refinery hydrocarbon emissions are designated as reactive.
In addition to the hourly wind speed and direction data, vertical dif-
fusivity coefficients must be specified at each of the vertical mesh points.
These may also be updated periodically during the simulation. Figure 8
shows the variation of diffusivity with height for different stability con-
ditions. Stability estimates can be made using the methodology outlined
in Reference 11.
The N02 photodissociation rate constant is another parameter which can be
revised periodically. This is important since it is a function of the
solar zenith angle. A separate program to compute this rate constant as a
function of time of day is provided in Reference 5. The computed values
are then input directly into DIFKIN.
123
In all three applications of the DIFKIN model ' ' essentially the same
reactions, rate constants, and stoichiometric coefficients were used, with
the exception of the N02 photodissociation rate constant. No real guidance
is offered in the user's manual as to when or in what manner different
values might be used. In addition, although the model is admittedly quite
sensitive to the initial concentrations inputs, no guidelines are given
for choosing reasonable "seed" values, either in terms of actual concentra-
tions or in terms of concentrations compatible with accurate predictions.
LIMITATIONS
Like REM, DIFKIN is a trajectory model, and hence it has the same limita-
tions for highway system evaluation as were discussed for REM. The model
87
-------
100
UJ
_«
CO
(O
a:
UJ
ui
_j
CD
_J
<
a:
»-
3
UI
a
CD
(S
0
E
X
o
50
2XI04 4X I04 6xl04 BxlO4
VERTICAL DIFFUSIVITY, cmz/t
10'
2xl05
Figure 8. Variation of diffusivity with height for different
stability conditions
88
-------
must be run repeatedly to gather enough space and time points to assess
VMT and emissions temporal and spatial changes fully. The SAI model dis-
cussed in Section IX has relatively long running times (about 1 hour of
computer time for a 10 hour simulation). DIFKIN requires about 2 minutes
of computer time for each trajectory. Thus, DIFKIN is not necessarily any
more economical to run for system evaluation, especially considering that
the individual trajectory calculations must be compiled into an overall
mapping of concentrations. This operation is intrinsic to the SAI model.
The inclusion of emissions data in a "BLOCK DATA" statement within the
program itself is somewhat clumsy operationally. While this is not a
serious limitation to the model, it would be simpler if these data could
be read in as inputs. Another problem, not especially serious, is the
specificity of the model in its present form to Los Angeles topography.
Mountain barriers are included in the program which exclude the use of
certain combinations of meteorological station geographical locations in
computing trajectories.
A more serious problem is the sensitivity of the model to initial pollutant
concentrations, and the fact that this is not even acknowledged in the
user's manual to DIFKIN. More guidance on the magnitude of this problem
and how it might best be handled is needed.
1 12
Neglect of horizontal diffusion has been investigated ' and introduces
only small errors in predicted values. Under most conditions the errors
are smaller than 10 percent and can generally be ignored.12 Under extreme
emission density gradients the errors may approach 40 percent.* DIFKIN also
ignores vertical advection, and this can cause errors of a factor of two
12
under common urban conditions. Finally, neglect of vertical wind
shear, inherent to trajectory models, may introduce errors of overpre-
12
diction of 50 percent or greater.
89
-------
SPECIAL FEATURES
Many of DIFKIN's limitations are also its strengths compared to previous
models which have been discussed. It offers the most comprehensive treat-
ment of atmospheric transport, diffusion, arid photochemical reaction
processes of any of the models described thus far. While the chemical
reaction scheme is not so elaborate as that of REM, DIFKIN predicts ozone
concentrations more nearly accurately. Perhaps this reflects- the current
state of understanding of atmospheric transport and diffusion compared to
photochemistry occurring in the atmosphere.
As a trajectory model, it is well suited to determining the effects of
specific sources (project level analysis) on resulting downwind concentra-
tions. This could be a valuable tool for investigating the existence of
"hot spots."
The best feature of DIFKIN might be the fact that it has been applied in a
number of locations^»^>3 an(j undergone other investigation.3»4,12 Thus,
many of the strengths and weaknesses of the model have been uncovered, and
there exists a body of knowledge available to guide its future application.
REFERENCES
1. Eschenroeder, A. Q., J. R. Martinez, and R. A. Nordsiek. Evaluation
of a Diffusion Model for Photochemical Smog Simulation. Final
Report. Prepared for U.S. Environmental Protection Agency, Contract
No. 68-02-0336. General Research Corp., Santa Barbara, California.
1972. 212 pp.
2. Ludwig, F. L. and J. H. S. Kealoha. Present arid Prospective San
Francisco Bay Area Air Quality. SRI__Proj.e£t,32Z4i. ^Stanford
Research Instiute. December 1974.
3. Photochemical Oxidant Modeling: Detailed Technical Report. Final
Final Report - Volume II. Report No. GCA-TR-11-G(2). GCA/
Technology Division. July 1975.
90
-------
4. Ranzieri, A. J., P. D. Allen, and W. B. Crews. A DIFKIN Sensitivity
Analysis for Transportation Air Quality Studies. Draft Report.
California Division of Highways. 1974.
5. Martinez, J. R., R. A. Nordseick, and M. A. Hirschberg. User's
Guide to Diffusion/Kinetics (DIFKIN) Code. Final Report. Prepared
for U.S. Environmental Protection Agency, Contract No. 68-02-0336.
General Research Corp., Santa Barbara, California. 1973. 240 pp.
6. Ludwig, F. L. Personal communication.
7. Roberts, P. J. W., et al. Extensions and Modifications of a Con-
taminant Emissions Model and Inventory for Los Angeles. Report
No. R73-15, Prepared for U.S. Environmental Protection Agency,
Contract No. 68-02-0339. Systems Applications, Inc., Beverly Hills,
California. 1973. 52 pp.
8. Roberts, P. J. W., P. M. Roth, and C. L. Nelson. Contaminant
Emissions in the Los Angeles Basin - Their Sources, Rates and Dis-
tribution. Appendix A of Development of a Simulation Model for
Estimating Ground Level Concentrations of Photochemical Pollutants.
Report No. 71SAI-6, Prepared for U.S. Environmental Protection Agency,
Contract No. CPA 70-148. Systems Applications, Inc., Beverly Hills,
California. 1971. 86 pp.
9. Compilation of Air Pollutant Emission Factors. Second Edition.
U.S. Environmental Protection Agency. Office of Air and Water
Programs, Office of Air Quality Planning and Standards, Research
Triangle Park, North Carolina 27711. April 1973.
10. Martinez, J. R., R. A. Nordsieck, and A. Q. Eschenroeder. Morning
Vehicle-Start Effects on Photochemical Smog. Environ Sci Technol.
7:10, October 1973.
11. Turner, D. Bruce. Workbook of Atmospheric Dispersion Estimates.
U.S. Environmental Protection Agency, Report AP-26. 1970.
12. Systems Applications, Inc. Development of a Second Generation of
Photochemical Air Quality Simulation Models. Progress Report.
Prepared for U.S. Environmental Protection Agency, Contract No.
68-02-1237. 1974.
91
-------
SECTION IX
URBAN AIR SHED PHOTOCHEMICAL SIMULATION MODEL
APPLICABILITY AND RELIABILITY
This model, which was developed by Systems Applications Inc. (SAI), pro-
vides the most detailed treatment of the physical and chemical processes
important in the formation of photochemical smog of all the models studied.
Concentrations of reactive species are determined by means of a finite dif-
ference scheme for an array of grid elements. Meteorological inputs to
the program consist of wind speed and direction for a number of stations
in a study area. These values are then interpolated to give a flow vector
for each ground level grid element. Each grid element is also assigned an
hourly mixing depth which depends upon the terrain elevation and the mea-
sured mixing depth at one or more meteorological stations. The input of
initial concentrations and hourly meteorological variables for specified
measurement locations is handled by means of data preparation programs.
These programs also determine correlations between corresponding'input
parameters at various stations so that measured values may be extended over
the entire grid system.
As part of these data preparation programs, the SAI model also employs a
rather sophisticated emission inventory routine for mobile and stationary
sources. In addition to standard inputs such as vehicle miles traveled
(VMT) for different road types, the emissions submodel requires such
detailed information as airport emissions for different aircraft types,
mole fraction of NO in auto NOX emissions, and correction factors for each
pollutant species to account for the nonuniform distribution of vehicle
starts.
92
-------
The SAI model predicts pollutant concentrations for all grid elements in
a 25 by 25 element array at discrete times. This capability makes it
ideally suited for highway system evaluation among it, REM, and DIFKIN.
Like REM and DIFKIN, the SAI model contains features which are specific to
Los Angeles, some of which make application to other locations slightly
more difficult than for the other two models.
The SAI model has been validated for the same six days in 1969 in Los
Angeles for which tests were made of REM and DIFKIN. One report of this
validation^ compared calculated and observed ozone concentrations for the
SAI model on two different data sets. The first was composed of data from
monitoring locations. These data were used in the evaluation of REM. The
second set was made up of data at trajectory points used in the DIFKIN
evaluation. These did not necessarily coincide with station locations.
Using the first set of data, the SAI model produced a correlation coef-
ficient of 0.60, while REM yielded 0.50. For the second data set, the
correlation coefficient was 0.69 for the SAI model compared to 0.91 for
DIFKIN. Further analysis of the SAI model results showed that it tended
to underpredict the station values while tending to overpredict the
trajectory values, although not strictly so in either case. REM and
DIFKIN exhibited these same tendencies, respectively.
DATA AND MANPOWER REQUIREMENTS
The user's manual^»^ for the SAI model provides exhaustive explanations of
data requirements and operating procedures for the model. Again, this
should be consulted for detailed information. The following description
of the data requirements and outputs is intended to provide an overview
for comparison with the requirements of other models.
The Urban Air Shed Photochemical Simulation Model is made up of four com-
puter programs: (1) the Atmospheric Pollution Simulation Program, (2) the
Emissions Data Preparation Program, (3) the Meteorological Data Preparation
93
-------
Program, and (4) the Data Plotting Program. The heart of the SAI model
is the Atmospheric Pollution Simulation Program (APSP). This program
accepts the input data supplied by the two data preparation programs and
performs the actual pollutant concentration calculations. The Data
Plotting Program is a useful accessory which plots calculated and observed
concentrations as a function of time at specified locations.
The Emissions Data Preparation Program requires the following information
to develop inputs to APSP:
Automotive emissions
- Daily miles driven on all streets (excluding freeways)
in each grid square.
- Daily miles driven on freeways in each grid square.
- Hot and cold start emissions factors.
- Fraction of vehicles cold started.
- Correction factor for the nonuniform distribution of
trip starts.
- Freeway driving speeds in the fast and slow directions.
- Emissions - driving speed correction factors.
- Ratio of freeway vehicle miles driven in the slow direction
to the number driven in the fast direction.
- Temporal distributions of freeway and nonfreeway traffic
activity.
Aircraft emissions (ground operations)
- Number of daily aircraft flights at each airport.
- Temporal distribution of aircraft activity.
- Emissions/aircraft engine.
- Number of aircraft engines/aircraft.
94
-------
Fixed source emissions (ground-based)
- Total distributed fixed source emissions from each
grid square.
As output, this program creates a data file for use by APSP, and a print-
out of the emission fluxes by hour.
The Meteorological Data Preparation Program uses hourly data on the fol-
lowing parameters to prepare inputs to APSP:
Wind speeds
Wind directions
Mixing depths
The data file which is created as output from this program provides data
on wind speeds, wind directions, the time rate of change of the mixing
depths, and the time interval for which these data are applicable.
As mentioned previously, the actual concentration calculations are done by
APSP. Meteorological inputs and ground level emissions inputs are accepted
from the data preparation programs. In addition, the following inputs are
required:
Grid characteristics, including number of vertical
strata, horizontal grid spacings, and shape of the
region.
Starting and stopping time of the simulation.
Time interval between instantaneous concentration
printouts.
Location of monitoring stations.
Rate, location, and temporal variation of major
elevated point source emissions.
Parameters applying to the operation of the finite
differing scheme.
Parameters applying to the chemical reaction scheme.
95
-------
Initial conditions
Concentrations aloft
Boundary concentrations at points of horizontal inflow
As output, APSP provides the following information:
Printed hourly-averaged ground level concentration maps
and a printed summary of the hourly-averaged concentra-
tions predicted at each monitoring station.
Printed instantaneous ground level concentration maps
and a printed summary of the instantaneous vertical
concentration distribution above each monitoring station.
Punched cards for input to the Data Plotting Program con-
taining the hourly-averaged concentrations predicted at
each monitoring station.
The emissions data requirements of the SAI model are similar to those of
DIFKIN, while the meteorological inputs are like those of REM. The efforts
involved in preparing the data to run the model are thus of the order of
the efforts described for these other two models. The major extra effort
to employing the SAI model involves actually getting the program opera-
tional and validating it for a location other than Los Angeles. To date,
only SAI has attempted to use the model outside of Los .Angeles, and such
attempts have not yet come to fruition. Based on experience applying the
DIFKIN model in Denver,^ and the much longer running time of the SAI model,
it is estimated that as much as an additional $10,000 above the DIFKIN
efforts might be required for initial implementation of the model.
APPLICATION OF THE TECHNIQUE
Ample information on the methodology and operation of the various programs
in the SAI model is given in References 2, 3, 5. These must be consulted
for any detailed information. General information on emissions data is
given in previous sections of this report. Methods for developing the
meteorological data are described in Section VII.
96
-------
Data on the frequency of operation of different aircraft categories are
required for the emissions program of this model. These may be taken
from AP-42.6
This model calculates vertical diffusion coefficients internally as a
function of wind speed and height. The horizontal diffusion coefficient
is fixed. On the other hand, the N(>2 photodissociation rate change during
the day is handled by inputting fractional values of the nominal, input
rate constant for each hour.
As with REM and DIFKIN, mountain barriers must be removed from the program,
or changed, to allow full use of the wind data. In addition, the SAI model
divides the Los Angeles region into different subregions to be used in
interpolating between air quality monitoring stations and specifying initial
concentrations for each grid cell, in performing a similar function for
wind speed and direction, in specifying mixing depths based on inversion
height and general features of the underlying terrain, and in estimating
concentrations at the boundary of the region. These subregions are built
into the program and must be removed or modified to fit other locations.
LIMITATIONS
Perhaps the greatest limitations to the use of the SAI model are its com-
plexity and expense. These limitations detract from its applicability to
general use both in getting the model operational in a new location and in
making successive runs once the model is actually functioning. A simula-
tion for one "day" (10 hours of daylight) requires over an hour of
computation time on an IBM 370/155 computer. This makes the model much
more expensive to implement initially than REM or DIFKIN. The cost of
getting similar amounts of information does not differ so much among the
models, however, because multiple trajectories must be run for DIFKIN and
REM.
97
-------
The initial application of the model to any given location is made dif-
ficult by the fact that some aspects of the program are quite specific
to Los Angeles. In particular, for the correlation of wind speeds,
wind directions, and mixing depths, the Los Angeles basin is subdivided
within the program itself into a number of topographically similar
regions. Another drawback to the application of the SAI model from the
viewpoint of a potential user is the inability to specify different
values of horizontal and vertical diffusion coefficients without making
modifications to the program itself. This situation seems curious in
light of the significant role played by these variables in the deter-
mination of ozone of NOX concentrations and the extensive amounts of
computer time required for the simulation of these diffusion effects.
Although there are plans to implement the model in a number of cities,
under the sponsorship of the U.S. DOT, the model has been run for Los
Angeles only thus far, and only SAI has had experience with its use. This
has limited the development of experience and knowledge of its capabilities.
Two recent studies^*? have chosen to use DIFKIN over the SAI model, partly
due to its expense. One of these studies^ concluded that in spite of the
fact that the SAI model is an excellent research tool, the long running
times and extensive data requirements make it difficult to use the model
for analysis of different transportation strategies without a considerable
investment in time and manpower.
One other limitation to the SAI model involves the finite differencing
scheme used to solve the atmospheric diffusion equation for this Eulerian,
fixed-grid type model. Intrinsic to these models is the production of
numerical or "pseudo" diffusion of material between grid cells due to the
finite difference approximation to the solution of the diffusion equation.
The effect in the SAI model is to introduce a harmonic error which is am-
plified with time, and which approaches 50 percent after a 9-hour
simulation.^
98
-------
SPECIAL FEATURES
The most important advantage of the SAI model over the other two complex
computer models investigated is its ability to predict concentrations for
all grid squares at a particular time. This feature makes it especially
well suited to highway system evaluation.
The SAI model treats the atmospheric transport and diffusion process in
more detail than any other model. It also uses the most complete emis-
sions data set. The capacity to accommodate different topological fea-
tures and boundary concentrations, while increasing the difficulty of im-
plementation, attempts to replicate real-world processes more closely than
the other models.
REFERENCES
1. Systems Applications, Inc. Development of a Second Generation of
Photochemical Air Quality Simulation Models. Progress Report Pre-
pared for U.S. Environmental Protection Agency, Contract No. 68-02-
1237. 1974.
2. Reynolds, S. D. Urban Air Shed Photochemical Simulation Model
Study: Volume II - User's Guide and Description of Computer
Programs. Systems Applications, Inc. U.S. Environmental Protection
Agency, Publication No. EPA-R4-73-030f. July 1973.
3. Whitney, D. Urban Air Shed Photochemical Simulation Model Study:
Volume II - User's Guide and Description of Computer Programs:
Appendix - Data Preparation Programs. Systems Applications, Inc.
U.S. Environmental Protection Agency, Publication No. EPA-R4-73-030g.
July 1973.
4. Photochemical Oxidant Modeling: Detailed Technical Report. Draft
Final Report, Volume II. GCA/Technology Division. GCA-TR-75-ll-G(2).
April 1975.
5. Reynolds, S. D., Mei-Kao Lui, T. A. Hecht, P. M. Roth, and J. H. Sein-
field. Urban Air Shed Photochemical Simulation Model Study.
Volume I - Development and Evaluation (With Appendices). Systems
Applications, Inc., Beverly Hills, California. Prepared for Office
of Research and Development, U.S. Environmental Protection Agency,
Washington, D.C. EPA-R4-73-020a-e. July 1973.
99
-------
6. Compilation of Air Pollutant Emission Factors. Second Edition.
U.S. Environmental Protection Agency, Office of Air and Water
Programs, Office of Air Quality Planning and Standards, Research
Triangle Park, North Carolina 27711. April 1973.
7. Ludwig, F. L. and J. H. S. Kealoha. Present and Prospective San
Francisco Bay Area Air Quality. SRI Project 3274. Stanford
Research Institute. December 1974.
100
-------
' TECHNICAL REPORT DATA
(Please read iHHtnictums on I/if reverse bcjore. completing)
1 REPORT MO 1?
EPA-450/3-75-069-a |
4: TITLE AND SUBTITLE
Photochemical Oxidant Modeling: Volume I - Techniques
Applicable to Highway System Evaluation.
7. AUTHOR(S)
F. A. Record, R. M. Patterson, D. . .. Bryant, and
A. H. Castaline
9. PERFORMING ORGANIZATION NAME AND ADDRESS
GCA Corporation
GCA/ Techno logy Division
Bedford, Massachusetts 01730
12. SPONSORING AGENCY NAME AND ADDRESS
Environmental Protection Agency
Office of Air and Waste Management
Office of Air Quality Planning and Standards
Research Triangle Park, North Carolina 27711
3 PEC'P'ENT'S ACCESC'OI*MQ
5 REPORT DATE
April 1975
6. PERFORMING ORGANIZATION CODE
8. PERFORMING ORGANIZATION REPORT NO
GCA-TR-75-ll-G(la)
10. PROGRAM ELEMENT NO.
11. CONTRACT/GRANT NO.
68-02-1376
13. TYPE OF REPORT AND PERIOD CO VERFD
Final
1«. SPONSORING AGENCY CODE
15. SUPPLEMENTARY NOTES
16. ABSTRACT
This report presents a comparative discussion and evaluation of the modeling
techniques available to assess the photochemical oxidant air quality impact of
highway system modification or development. Five attributes of eight modeling
techniques are discussed. The eight methods considered, in order of increasing
complexity, are: (1) VMT Change, (2) Linear Rollback, (3) Nonlinear Rollback:
Appendix J, (4) Gifford-Hanna Photochemical Model, (5) Statistical Relationships:
Diffusion Kinetics Model, and (8) Urban Air Shed Photochemical Simulation Model.
For each of these the following five attributes are discussed: (1) applicability
and reliability, (2) data and manpower requirements, (3) use of the technique,
(4) limitations, and (5) special features. The information presented in this
report gives guidance towards choosing a model best suited to a given need,
based on compatibility with these attributes.
17. KEY WORDS AND DOCUMENT ANALYSIS
.1. DESCRIPTORS
Highways
Automotive Emissions
Oxidant
Oxidant Precursors
Air Pollution Forecasting
13. DISTRIBUTION STATEMENT
Unlimited
b. IDENTIFIERS/OPEN ENDED TERMS
Photochemical Oxidant
Models
Appendix J Relationships
Linear & Non-linear
Rollback
Proportional Models
19. SECURITY CLASS (Tins Report)
Unclassified
20 SECURITY CLASS (This pa^ej
Unclassified
c. COSATI 1 icId/Group
21 NO OF PAGES
105
22 PRICE
FPA Form 2220-1 (9-73)
101
-------
3D
ct>
"0
C|
w
r1
ii
n
5
I-H
O
z
2
O
M
^
>
4x
v^i
O
OJ
I
-^1
V^l
6
ON
vo
I
p
NO 33
CD
W CD
a. ^
Q.
t s
-^
CO O
en Q.
3 °
2 2:
CD ,-.
&H
^
en
3"
D
B. S
o
O CD
CD CD
CQ 3
CD »
CD
^ ^
O m'
(Q -
m
D
O
-Q
-D
O
H
O
-<
m
3J
O
TI
T1
O
CD
00
m
Crt
en
<
31
°
02 a n ^~
5" re ?L H
"I?- ^
-*l?5
o m
5 O
3 §
o
"
m m
fli
oo H -n
m m m
O m
-I en
11
>0
O
m
2
O
m
2
O
------- |