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

-------