VALIDATION OF AN IMPROVED
PHOTOCHEMICAL AIR QUALITY SIMULATION MODEL
PES Document TP-014
Pacific Environmental Services, INC.
1930 14th Street Santa Monica, California 90404
-------
VALIDATION OF AN IMPROVED
PHOTOCHEMICAL AIR QUALITY SIMULATION MODEL
by
Peter J. Drivas and Lowell G. Wayne
PES Document
-.TP-014
March , 1977
PACIFIC ENVIRONMENTAL SERVICES, INC.
1930 14th Street
Santa Monica, California 90404
(213) 393-9449
-------
TABLE OF CONTENTS
Chapter Page
1. INTRODUCTION 1-1
1.1 Background r 1-1
1.2 Pacific Environmental Services Modeling
Approach 1-2
2. DESCRIPTION OF IMPROVED PHOTOCHEMICAL
MODEL, REM2 2-1
2.1 Principles of Simulation 2-1
2.2 Photochemical Mechanism 2-5
2.3 Model Assumptions 2-8
2.4 Model Operation 2-9
2.5 Model Applications 2-13
3. REM2 VALIDATION RESULTS 3-1
3.1 A Tale of Three Cities 3-1
3.2 03 Validation Results 3-5
3.3 N02 Validation Results 3-7
3.4 NO Validation Results 3-9
3.5 NMHC Validation Results . 3-11
3.6 CO Validation Results 3-13
4. SUMMARY 4-1
. REFERENCES A-l
-------
LIST OF TABLES
Table Page
2-1 REM2 34 - REACTION PHOTOCHEMICAL MECHANISM 2-6
2-2 REM2 HYDROCARBON REACTIVITY CLASSES 2-7
3-1 REM2 VALIDATION RUNS 3-2
3-2 REM2 VALIDATION RESULTS 3-3
4-1 SUMMARY OF VALIDATION RESULTS 4-2
LIST OF FIGURES
Figure Page
2-1 REM2 MODEL DYNAMICS 2-2
2-2 EMISSIONS GRID AND VALIDATION TRAJECTORIES
FOR PHOENIX AREA 2-10
2-3 TYPICAL MODEL OUTPUT FOR PHOENIX AREA 2-11
3-1 03 VALIDATION RESULTS 3-6
3-2 N02 VALIDATION RESULTS 3-8
3-3 NO VALIDATION RESULTS 3-10
3-4 NMHC VALIDATION RESULTS 3-12
3-5 CO VALIDATION RESULTS 3-14
-------
1. INTRODUCTION
1.1 Background
Photochemical air quality simulation modeling is a mathe-
matical attempt to predict very complicated physical and
chemical atmospheric processes, and it might be consider
as a rather technical type of art form. Generally, photo-
chemical models use some form of solution to a conservation-
of-mass equation, with varying degrees of complexity.
Photochemical models have used a fairly simple "box model"
approach (Hanna, 1973), Lagrangian trajectory techniques
(Wayne el; al_., 1973; Eschenroeder ejt al_., 1972), a quasi-
Lagrangian "particle-in-cell" approach (Sklarew et a!.,
1972), and complex Eulerian grid K-theory techniques
(Reynolds et al_., 1973; MacCracken and Sauter, 1975).
The photochemical models which have had the most extensive
validation studies have been the Lagrangian model REM
(developed by Pacific Environmental Services), the
Lagrangian model DIFKIN (developed by General Research
Corporation), and the Eulerian SAI "urban airshed" model
(developed by Systems Applications, Inc.), These valida-
tion studies for the Los Angeles area have been published
by the Environmental Protection Agency (Wayne et^ .al_., 1973;
Eschenroeder et al_., 1972; Reynolds e_t al_., 1973). All
three models showed fairly good agreement, typically within
a factor of two, with measured 03 and N02 concentrations.
However, the validations were limited specifically to the
Los Angeles area, and several model "calibration" runs
were permitted before final validation. It should be noted
that the Pacific Environmental Services REM model was not
designed for, and did not use, any "calibration" runs for
its validation results.
1-1
-------
1.2 Pacific Environmental Services Modeling Approach
The establishment of objective criteria for the evaluation
of photochemical air quality simulation models is a diffi-
cult matter, involving both quantitative comparison and
subjective judgment. The modeling approach that Pacific
Environmental Services (PES) has followed has concentrated
on the following criteria:
0) Economical simulation accuracy
A photochemical model should employ fundamen-
tal and realistic principles consistent with
present knowledge of physical and chemical
atmospheric processes. In addition, the model
should be simple and economical to run on
normally-available computation facilities.
(2) "Hands-off" validation accuracy
A photochemical model should have good agree-
ment in comparisons of model predictions with
observed concentrations at air monitoring
locations. However, the model should be used
in a "hands-off" manner for all validations,
i.e. there should be no internally adjusted
parameters which must be "calibrated" for
optimum validation.
(3) User-oriented adaptability
A photochemical model should be easily adapt-
able to different applications and conditions
of use and different degrees of data avail-
ability. The model should be adaptable for
use in any urban or rural location, without
changing any internal programming.
1-2
-------
Based on these criteria, the PES technical staff originally
developed, with EPA funding, an efficient photochemical
air quality simulation model, REM (Reactive Environmental
Model). REM was a Lagrangian model which was designed
for the prediction of photochemical contaminant levels
specifically in the Los Angeles Basin (Wayne et_ aK, 1971).
REM was tested by comparing its predictions with the actual
measurements observed by the extensive air monitoring net-
work in the Los Angeles Basin. Results of this validation
study have been published by the EPA (Wayne et^ al_., 1973;
Kokin jet jaJL, 1973); they showed that REM yielded good pre-
dictions for typical smog situations in Los Angeles.
The current.photochemical model, REM2, is an improved
version of the original model, and it can easily be used in
any location. The improvements have been in both simulation
accuracy (e,g-> horizontal diffusion) and user-oriented
adaptability (e.g., variable grid size). The improved
photochemical model, REM2, is discussed in detail in
Section 2. In Section 3 of this paper, recent "hands-off"
validation results for REM2 are discussed, including vali-
dation for ozone (P3), nitrogen dioxide (NCk), nitric
oxide (NO), non-methane hydrocarbons (NMHC), and carbon
monoxide (CO). Section 4 presents a summary of the REM2
validation results.
1-3
-------
2. DESCRIPTION OF IMPROVED PHOTOCHEMICAL MODEL. REM2
2.1 Principles of Simulation
REM2 is a regional photochemical air quality model which
simulates a 34-reaction photochemical mechanism in a
Lagrangian (moving-coordinate) frame of reference. The
basis of the model is a moving parcel of air, which is
bounded by a mixing layer [inversion base) above and the
ground below. The basic model dynamics are shown in
Figure 2-1. Pollutant emission sources are input into
the moving air parcel from an Eulerian emissions grid,
. and pollutants can diffuse in and out of the moving air
parcel by horizontal diffusion.
The location of the base of the moving column at successive
moments generates the path or trajectory that the air
parcel traverses across the region. Either forward or
reverse trajectories can be computed by special routines
contained in the REM2 program from wind velocity and
direction information, given in the data base as a function
of time of day and location. The moving parcel of air is
assumed to be well-mixed vertically between ground level
and the inversion base. Both the ground terrain level
and the inversion base height can be entered as functions
qf location and time of day; thus the model can accommodate
varying ground terrain and varying inversion heights.
Because of the Lagrangian formulation which follows an
air parcel in a moving-coordinate frame of reference, the
basic equation is simply that of conservation of mass in
the air parcel for each pollutant of interest:
2-1
-------
MIXING
HEIGHT
GROUND
LEVEL
HORIZONTAL
DIFFUSION
u(x, y, t)
EMISSION
SOURCES
Figure 2-1. REM2 MODEL DYNAMICS
2-2
-------
N - hi
LcfiTJ Total ~ L^fJ re
reaction LdtJ horizontal
+ fe]
volume LdtJ emissions
change input
diffusion
(2-1)
The reaction term is handled in the conventional manner,
(2-2)
reaction
J
where k.. is the reaction rate constant. The horizontal
* J
diffusion term involves the use of the semi-empirical turbulent
diffusion equation or K-theory,
horizontal y Q 2 vt"OJ
diffusion y
where K is the horizontal diffusion coefficient and y is the
direction perpendicular to the trajectory direction.
The REM2 computer program is modular in design, with separate
modules linked to form a complete atmospheric simulation system.
Modules presently in the system determine the necessary
meteorological parameters, the rate of absorption of ultra-
violet light by N0?, emissions due to traffic and area sources,
and solution of the conservation-of-mass equations. The
ultraviolet absorption module calculates a diurnal ultra-
violet irradiance function based on measurement of cloud
cover, latitude, and local calendar time.
The source emissions module calculates the pollutant inputs to
the column of air as it passes over vehicular, stationary, and
area emission sources. The emissions from freeway traffic,
2-3
-------
street traffic, and area sources are represented by a
Eulerian grid system, whose size is adjustable.
Currently, three types of pollutant emissions are con-
sidered: nitric oxide (NO), carbon monoxide (CO), and
non-methane hydrocarbons (NMHC), Separate emission fac-
tors and diurnal distributions for freeway and street
traffic are input into the model. The NMHC emissions
are divided into two reactivity classes.
2-4
-------
2.2 Photochemical Mechanism
The chemical kinetics mechanism is the heart of REM2
and simulates the complex photochemical reactions
occurring in the moving parcel of air. The model contains
a 34-reaction mechanism, shown in Table 2-1, which is based
on stoichiometrically valid elementary reactions (Wayne
e^t al_., 1973). Twenty-four different chemical species are
considered; of these, twelve are free radicals.
Non-methane hydrocarbons are grouped into two reactivity
classes - more reactive hydrocarbons and less reactive
hydrocarbons. Methane is assumed non-reactive and is not
included in the reaction scheme. The types of compounds
assigned to the REM2 reactivity classes are given in
Table 2-2.
The conservation-of-mass equations, which include the
chemical kinetics expressions, are solved by the efficient
Gear numerical integration routine (Gear, 1971); this
routine has found widespread use in photochemical kinetics
simulations.
2-5
-------
Table 2-1. . REM2 34 - REACTION PHOTOCHEMICAL MECHANISM
REACTION
RATE CONSTANT (25°C)
1.
2.
3.
4,
5.
6.
7,
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
N02 4 h. -^NO + 0
0, + 0 + M **00 + M
2 3
NO + 03 -^N02 + 02
N00 + 0~ *-NO., + 0,,
23 32
NO + NO,, *-2N00
3 2
N09 + NO. + H90 t^2HN00
232 3
N00 + OH *-HN00
2 3
NO + H0? *-N09 + OH
c 2
0, + H + M *-H00 + M
2 2
0, + OH *-H09 + 09
3 22
CO + OH i^C02 + H
HCHO + hv >CO + 2H
C H + 0 ^CH, + C9H,0
O O o i. o
*"»*£ "" U * CHo T" CoH^U
36 323
C' H + O »- n/>iir> i f» U n
onr Uo ** nLnU T Orvil.Uo
« m.vr»«w XtLX
C3H6 + OH ^CH3CHO + CH3
C3H6 + H02 i»-CH30 + CH3CHO
Co"c "*" CH00O ^ CH-% + CH«0 + C^H^O
OD Jt J " O CO
Cu 4- n 4- n tfc i ir*i in j. f* u n
o**c *^o nunu * ^o^/i^'o
C9H,0 + M >-CO + CH. + M
*- * .3
32 2 '
c o £- 233
C2H3°3 + 02 -^C2H3°2 + 03
C H,,00 + 00 >-C HO + OH
242. 2 233
CH.09 + NO ^^CH-0 + N00
32 32
CH30 + NO + 02 ^CH302 + N02^
C,H,00 +' NO »*C0H,0 + N00
232 23 2
C0H,0. + NO B-C0H000 + N00
233 232 2
C2H4°2 + N0 >-CH3CHO + N02
CH.,0 + N00 ^CH.ON09
32 32
C9H,0.j + N09 *- C0H-00N00
233 L 2332
NO + Radical *- Products
Radical + Radical * Products
Depends on light intensity
6,7 x 10" ppm" min"
1 -1 -1
4.0 x 10 ppm mm
-3 -1 -1
1.0 x 10 ppm min
,4 -1 -1
2.5 K 10 ppm mm
-2 . -1
1,0 ppm mm
4 -1 -1
1*0 x 10 ppm min
3 -1 -1
1.0 x 10 ppm mm
621
4.8 x 10" ppm" min"
3 -1-1
1.0 x 10 ppm min
2 -1 -1
3.0 x 10 ppm mm
1/133 k]
3.5 x 103 ppm"1 min"1
2 -1 -1
7.0 x 10 ppm min
5.0 x 10" ppm" min"
5 -1 -1
1.5 x 10 ppm min
2-11
1.0 x 10 ppm min
, ~ -1-1
1,0 ppm mm
-3 -2 -1
8,3 x 10 ppm min
1,0 x 10 ppm" min"
-4 -1 -1
9,5 -x 10 ppm min
6.7 x 10"6 ppm"2 min"
4,8 x 10 ppm" min"
-5 -1 -1
9,5 x 10 ppm min
.1.4 x 10" ppm" min"
2 -1 -1
2.0 x 10 ppm mm
-3 -2 -1
4.8 x 10 ppm mm
3 -1 -1
2.0 x 10 ppm min
,2 -1 . -1
2.5 x 10 ppm mm
4 -1 . -1
1.0 x 10 ppm mm
,2 -1 . -1
1.0 x 10 ppm mm
i -i . -i
2.0 x 10 ppm mm
-1 -1
5.0 ppm mm
, « ,»4 -1 -1
1.0 x 10 ppm mm
*Less reactive hydrocarbon
2-6
-------
Table 2-2
REM2 HYDROCARBON REACTIVITY CLASSES
Unreactive Less Reactive More Reactive
methane £ + paraffins olefins
acetylene aldehydes
benzene cycloparaffins
acetone aromatics (other than
... , benzene)
methanol
ketones (other than
acetone)
alcohols (other than
methanol)
2-7
-------
2.3 Model Assumptions
As with all models, the REM2 model includes certain
assumptions. The assumptions regarding atmospheric
motions include:
1. A minimum effective mixing depth exists which
may be assumed operative in instances of surface
inversion.
2. Effects of wind shear are unimportant and may be
neglected.
3. Effects of lag in vertical mixing within the
mixing layer are unimportant on a regional scale
and may be neglected.
Assumptions regarding photochemical contaminants and their
chemical behavior are the following:
1. Only contaminants emitted or produced chemically
within the mixing layer are involved in the
photochemical reactions.
2. Effects of temperature changes on the rate of
photochemical reactions are unimportant and may
be neglected.
3. The non-methane hydrocarbons involved in photo-
chemical reactions can be adequately simulated
in terms of two reactivity classes.
4. Vertical contaminant concentration profiles are
uniform within the mixing layer; i.e., the effects
of variations in the vertical dimension are
negligible.
2-8
-------
2.4 Model Operation
The REM2 program can accommodate up to 2200 emissions
grid squares (e.g., a 50 x 44 grid), and the size of
the grid squares is adjustable. In three different
validation studies, three diferent emissions grid
sizes were used: 2 miles x 2 miles, 1 mile x 1 mile,
and 1 km x 1 km. The emissions grid used for the
Phoenix, Arizona validation study, discussed in detail
in Section 3, is shown in Figure 2-2; this grid was
55 miles by 40 miles and contained 2200 1 mile x 1 mile
emissions grid squares.
The REM2 program can accommodate up to 32 meteorological
stations supplying hourly data on wind speed, wind
direction, temperature, and humidity. Trajectories are
normally automatically computed from the given wind
speed and direction data by means of an inverse-square
distance relationship among the meteorological stations.
If desired, artificial trajectories can be entered as an
alternative. Trajectories can be calculated forward in
time from a specific starting point, or backwards in time
("reverse" trajectory) from a specific receptor point
for validation comparisons. Examples of reverse trajec-
tories calculated for the Phoenix validation study are
shown in Figure 2-2; these trajectories were based on
data from 15 meteorological stations in the Phoenix
area.
The main output of the REM2 model is a record of all
chemical species concentrations as a function of travel
time along a specific trajectory. Graphical output is
provided for the main pollutants 0.,, NOp. NO, NMHC, and
CO. Typical model results are shown in Figure 2-3 for
the Phoenix area; the NO peak in the morning was due to
emissions from a major airport.
2-9
-------
40
30
Y. MILES 20
10
Y i i i | i i i i i i i
117 '
i i i i i
START-TYPICAL DAY
TRAJECTORY
START-SEVERE DAY
TRAJECTORY
I i I I I I I I I I I I I I I I !
O- WIND STATION
i i I i i i i I i i i i I i i i i I i i i i I i i i i I i i i
10
20
30
X. MILES
40
50
Figure 2-2. EMISSIONS GRID AND VALIDATION TRAJECTORIES
FOR PHOENIX AREA
2-10
-------
16
12
PREDICTED
CONCENTRATION
\ I I I I I
NO2
(pphm)
I I I I I
6 78
9 10
TIME OF DAY
11 12 13
Figure 2-3. TYPICAL MODEL OUTPUT FOR PHOENIX AREA
2-11
-------
The REM2 program requires 220 K computer memory for
operation, which is common on normally available
computers. The real time to computer time ratio, on
an IBM 370/158, is about 600:1, i.e., ten simulation
hours require about one minute of computer time.
Thus, the REM2 model is extremely cost-effective in
use.
2-12
-------
2.5 Model Applications
The usefulness of any regional photochemical air quality
model is determined by its ease in adaptability for
various applications. Potential uses for photochemical
models fall generally into three categories:
0) Environmental impact assessments
(2) Control strategy evaluations
(3) Scientific research
The improved PES photochemical model, REM2, is easily
adaptable and extremely cost-effective for each of
these applications.
The Lagrangian approach of REM2 is ideally suited for
determining the-air quality impact of new or proposed
sources on regional photochemical pollution, REM2 has
recently been used in a number of environmental impact
assessments of proposed sources, including both point
sources and a proposed new highway.
An important use of a photochemical model is in evaluating
alternative control strategies for the abatement of air
pollution problems. This application requires an extremely
versatile model, since all emission sources must be
easily adapted for future years and a large number of
alternative strategies must normally be run. Wayne et al.
(1971) present an excellent discussion of using an efficient
model similar to REM2 for determining control strategies
in Los Angeles.
Other uses of REM2 which can be classified under scienti-
fic research include model validations, short-term air
quality forecasting, prediction of pollutant concentra-
tions at locations not covered by air monitoring stations, and
2-13
-------
identification of source areas that contribute to the
air quality of specific receptor points. REM2 is easily
adaptable and very economical for use in these applica-
tions.
2-14
-------
3. REM2 VALIDATION RESULTS
3.1 A Tale of Three Cities
REM2 is normally validated as a first step in its use in
a particular location. The validation procedure involves
running reverse trajectories to specific air monitoring
locations, and comparing the predicted concentrations with
measured pollutant levels at the air monitoring stations.
In three recent modeling applications, REM2 was validated
in three very different locations:
(1) a high-density urban area - Los Angeles, California
(2) a medium-density urban area - Phoenix, Arizona
C3) a low-density rural area - Goleta, California
As was discussed in Section 2-4, a different size emissions
grid was used in each location: a 2 mile x 2 mile grid
size was used in Los Angeles; a 1 mile x 1 mile grid size
was used in Phoenix; and a 1 km x 1 km grid size was used
in Goleta (a small town about 8 miles west of Santa Bar-
bara, California).
Validation runs in each location were made on days with
"severe" meteorology and also "typical" meteorology. The
meteorological inputs to the model were determined from
actual measurements made on specific days in each location.
The number of validation runs and the specific days simu-
lated are shown in Table 3-1. Four validation runs were
made in the Los Angeles area, four runs were made in the
Phoenix area, and two runs were made in the Goleta area.
Predicted pollutant concentrations at specific air moni-
toring locations were compared with measured pollutant
levels at the monitoring sites on the specific days simu-
lated in Table 3-1. The overall results for 03> NCy NO,
NMHC, and CO are shown in Table 3-2. It should be noted that
3-1
-------
Table 3-1
REM2 VALIDATION RUNS
Location
1. Goleta
2. Goleta
3. Phoenix
4. Phoenix
5. Phoenix
6. Phoenix
7. Los Angeles
8. Los Angeles
9. Los Angeles
10. Los Angeles
Meteorology
Typical
Severe
Typical
Typical
Severe
Severe
Typical
Typical
Severe
Severe
Date Simulated
November 19, 1975
September 25, 1975
June 11, 1976
June 11, 1976
May 17, 1976
May 17, 1976
July 24, 1975
July 24, 1975
July 11, 1975
July 11, 1975
3-2
-------
Table 3-2
REM2 VALIDATION RESULTS
Location (meteorology)
1. Goleta (typical)
2. Goleta (severe)
3. Phoenix (typical)
4. Phoenix (typical)
5. Phoenix (severe)
6. Phoenix (severe).
7. Los Angeles (typical)
8. Los Angeles (typical).
9, Los Angeles (severe).
10. Los Angeles (severe).
03(ppm)
lleas.
0.03
0.08
0.06
0.07
0.13
0.15
0.04
0.16
0,07
0,27
Pred.
0.02
0.06
0.10
.0,08
0.11
0,13
0.07
0,16
0,11
0,27
N02(ppm)
Meas.
0.02
0.05
0.02
n.d.
n,d,
n.d.
0.05
0,07
o,n
0,10
Pred.
0.01
0.04
0.04
0.04
0.05
0,07
0,05
0.06
0,14
0.19.
NO(ppm)
fleas.
0.01
0.02
n.d.
n.d.
n.d.
n.d.
0,02
0.01
0,02
0.01
Pred.
0.01
0.01
0.01
0.01
0,01
0.01
0.01
0,01
0,02
0.01
NMHC(ppmC)
Meas.
0.2
0.9
n.d.
0.8
1,0
1.3
n.d.
0,7
n.d.
..1,0 .
Pred.
0.2
1.1
0.8
0.9
1.0
1.1
1,4
1.3
2.0
1.8
CO(ppm)
Meas.
0
1
1
1
0
1
3
2
4
3
Pred. .
0
1
1
1
1
2
2
2
4
5
n.d, -.no data available for comparison
CO
CO
-------
the REM2 model was always used in a "hands-off" fashion;
the model was not changed and there were no internally
adjustable parameters which were "calibrated" for any of
the validation runs in the three different locations.
3-4
-------
3.2 03 Validation Results
The results from Table 3-2 for ozone (0-) are plotted
in Figure 3-1, The solid line in the figure corresponds
to theoretical perfect agreement between measured and pre-
dicted 03 values. As can be seen, actual agreement was
excellent over the entire range of measured 03 values.
The linear correlation coefficient for the ten 03 vali-
dation points was 0.94, which is significant at the
0.001 level. From Table 3-2, the average absolute error
in 03 prediction was 0.02 ppm, with a standard deviation
(or) of only 0.01 ppm. Thus, the REM2 model dynamics and
kinetics assumptions appear ideal for the prediction of
urban and rural ozone levels,
3-5
-------
0.4
0.3
PREDICTED
63, ppm
0.2
0.1
r = 0.94
0.1 0.2 0.3
MEASURED 03, ppm
0.4
Figure 3-1. 03 VALIDATION RESULTS
3-6
-------
3.3 N00 Validation Results
£
The results from Table 3-2 for nitrogen dioxide (N(L) are
plotted in Figure 3-2. The solid line in the figure indi-
cates theoretical perfect agreement between measured and
predicted NCL values. With limited data, agreement was
excellent for low N(L concentrations, however the two
model predictions above a measured N02 level of 0.1 ppm were
somewhat high. It should be noted that three of the vali-
dation runs in Phoenix had no measured NCL data available
for comparison.
The linear correlation coefficient for the seven NCL vali-
dation points was 0.89, which is significant at the 0.01
level. From Table 3-2, the average absolute error in HQy
prediction was 0.02 ppm, with a standard deviation (
-------
0.2
PREDICTED
NO2, ppm
0.1
I
0.1
MEASURED NO2, ppm
0.89
0.2
Figure 3-2. N02 VALIDATION RESULTS
3-8
-------
3.4 NO Validation Results
The results from Table 3-2 for nitric oxide (NO) are
plotted in Figure 3-3, The solid line in the figure
indicates theoretical perfect agreement between
measured and predicted NO values, As can be seen,
agreement was good for low NO values, however the limited
data cannot provide an adequate validation for higher NO
levels. It should be noted that all four of the valida-
tion runs in Phoenix had no measured NO data available
for comparison.
The linear correlation coefficient for the six NO vali-
dation points was 0.45, which is not significant: From
Table 3-2, the average absolute error in NO prediction was
less than 0.01 ppm, with a standard deviation (
-------
0.1
PREDICTED
NO, ppm
0.05
r = 0.45
I
0.05
MEASURED NO. ppm
0.1
Figure 3-3. NO VALIDATION RESULTS
3-10
-------
3,5 NMHC Validation Results
The results from Table 3-2 for non-methane hydrocarbons
CNMHCl are plotted in Figure 3-4, The solid line in the
figure indicates theoretical perfect agreement between
measured and predicted NMHC values. With limited, data,
agreement was reasonable, with some scatter at higher
NMHC levels. It should be noted that two of the valida-
tion runs in Los Angeles and one validation run in Phoenix
had no measured NMHC data available for comparison,
The linear correlation coefficient for the seven NMHC vali-
dation points was 0.67, which is significant at the 0.10
level. From Table 3-2, the average absolute error in NMHC
prediction was 0.3 ppmC, with a standard deviation (
-------
PREDICTED
NMHC, ppmC
T
o
r = 0.67
.MEASURED NMHC, ppmC
Figure 3-4. NMHC VALIDATION RESULTS
3-12
-------
3.6 CO Validation Results
The results from Table 3-2 for carbon monoxide (COl are
plotted in Figure 3-5, The solid line in the figure
indicates theoretical perfect agreement between measured
and predicted CO values. As can be seen, agreement was
very good over the entire range of measured CO values.
The linear correlation coefficient for the ten CO vali-
dation points was 0.84, which is significant at the
0.005 level. From Table 3-2, the average absolute error
in CO prediction was less than 1 ppm, with a standard
deviation (cr) of 1 ppm. These excellent CO validation
results verify the REM2 model dynamics assumptions, since
CO is a fairly inert pollutant. Carbon monoxide is included
in the photochemical mechanism given in Table 2-1, however
the CO reaction rates are quite slow.
3-13
-------
10
PREDICTED
CO, ppm
r = 0.84
10
MEASURED CO. ppm
Figure 3-5. CO VALIDATION RESULTS
3-14
i r
-------
4, SUMMARY
In three recent modeling applications, an improved photochemical
air quality simulation model, REM2, was validated in three very
different locations:
Ola high-density urban area - Los Angeles, California
(2) a medium-density urban area - Phoenix, Arizona
(3) a low-density rural area - Goleta, California.
Four validation runs were made in the Los Angeles area, four
runs were made in the Phoenix area, and two runs were made in
the Goleta area. The validation procedure involved running
reverse trajectories to specific air monitoring locations,
and comparing the predicted concentrations with measured pollu-
tant levels at the air monitoring stations.
The validation results are summarized in Table 4-1. Model
agreement with measured concentrations was excellent for ozone
(CO, nitrogen dioxide (N02), and carbon monoxide (CO), with
respective linear correlation coefficients of 0.94, 0.89, and
0.84. Agreement was reasonable for non-methane hydrocarbons
(NMHC); limited measured data for nitric oxide (NO) prevented
an adequate validation except at very low NO levels.
The REM2 model was always used in a "hands-off" fashion. The
model was not changed and there were no internally adjustable
parameters which were "calibrated" for any of the validation
runs in the three different locations.
These validation results verify the REM2 model dynamics and
kinetics assumptions as appropriate for regional photochemical
air quality simulation modeling. REM2 is economical to use,
and is easily adaptable for use in any location for a variety
of applications, including environmental impact assessments
and regional control strategy evaluations.
4-1
-------
Table 4-1
SUMMARY OF VALIDATION RESULTS
Number of Correlation Ave, Absolute
Pollutant Validation Runs Coefficient Error (ppm) o-(ppm)
03 10 0.94 0.02 0.01
N02 7 0.89 0.02 0,03
NO 6 0,45 <0.01 0.01
NMHC 7 0.67 0,3 0,3
CO 10 0.84 <1 1
4-2
-------
REFERENCES
Eschenroeder, A.Q., J.R, Martinez, and R,A, Nordsieck 0972)
Evaluation of a Diffusion Mode] for Photochemical Smog
Simulation, Final Report, General Research Corporation,
Santa Barbara, California, EPA No, R4-73-012, Volume a,
Gear, C.W, (.1971) The automatic integration of ordinary differen-
tial equations. Comrn. A.C.M.. 14, 176-179,
Hanna, S.R. (1973) A simple dispersion model for the analysis
of chemically reactive pollutants. Atmos. Environ., _7,
803-817.
Kokin, A,, L.G. Wayne, and M. Weisburd (1973) Controlled Evaluation
of the Reactive Environmental Simulation Model (REM), Volume
II: User's Guide. Pacific Environmental Services, Inc.,
Santa Monica, California, EPA No. R4-73-013b, NTIS No. PB
220 457/6,
MacCracken, M.C. and G.D. Sauter, editors (.1975) Development of
an Air Pollution Model for the San Francisco Bay Area,
Final Report to the National Science Foundation, Volume 1.
Lawrence Livermore Laboratory, UCRL-51920 Vol. 1.
Reynolds, S.D., M.K. Liu, T.A. Hecht, P.M. Roth, and J.H. Seinfeld
(1973) Urban Air Shed Photochemical Simulation Model Study,
Volume 1: Development and Evaluation. Systems Applications,
Inc., San Rafael, California,EPA No. R4-73-030a.
Sklarew, R.C., A.J. Fabrick, and J.E. Prager (1972) Mathematical
modeling of photochemical smog using the PICK method.
J. Air Pollut. Control Assoc., 22, 865-869 (1972).
Wayne, L., R. Danchick, M. Weisburd, A. Kokin, and A. Stein (1971)
Modeling photochemical smog on the computer for decision-
making. J. Air Pollut. Control Assoc., 2K 334-340.
Wayne, L.G., A. Kokin, and M.I. Weisburd (1973) Controlled Evalua-
tion of the Reactive Environmental Simulation Model (REM),
Volume I: Final Report. Pacific Environmental Services, Inc.
Santa Monica, California, EPA No. R4-73-013a, NTIS No. PB
220 456/8.
A-l
------- |