United States      Office of Air Quality       EPA-450/4-81 -031 a
           Environmental Protection   Planning and Standards     September 1981
           Agency         Research Triangle Park NC 27711
           _
vvEPA       The Sensitivity Of
          Complex Photochemical
         Model Estimates To Detail
            In Input Information

-------
This report was furnished to the U.S. Environmental  Protection
Agency by Systems Applications, Incorporated in fulfillment of
Contract 68-02-2870.  The contents of this report are reproduced
as received from Systems Applications, Incorporated.   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.

-------
                             EPA-450/4-81-031a
   The Sensitivity Of Complex
Photochemical  Model Estimates
 To Detail In Input  Information
          EPA Project Officer: Edwin L Meyer, Jr.
                 Prepared for

           U.S. Environmental Protection Agency
            Office of Air, Noise and Radiation
         Office of Air Quality Planning and Standards
         Research Triangle Park, North Carolina 27711

                September 1981

-------
                                  PREFACE
     This report and its companion appendixes present the  results  of  a
large number of air quality sensitivity  simulations  as well  as  supporting
analyses.  Given the volume of information contained herein,  the reader
may wish to first gain an overview of the study  by reading the  introduc-
tion and conclusions, chapters I and VII, respectively.
                                  n

-------
                              ACKNOWLEDGMENTS
     Several individuals made significant contributions to the technical
work presented in this report.  Assistance from the technical staff of  the
California Air Resources Board—in particular, Messrs. Andrew Ranzieri,
Paul Allen, and Bill Loscutoff—is gratefully  acknowledged.  Mr. Ron Taira
of the California Department of Transportation was especially helpful in
formulating and supplying inputs for certain motor vehicle emission
inventories.  Contributions from individuals at the Office of Air Quality
Planning and Standards, notably Dr. Edwin Meyer, and Messrs. John
Summerhayes, David Barrett, and James Southerland, were central to the
direction of this study.  Special appreciation is given to several members
of the SAI technical staff, particularly Drs. G. Z. Whitten, S. D.
Reynolds, and M. K. Liu, for their advice and consultation throughout the
course of the study.  Production of the final report was due largely to
the dedicated efforts of Karyl Pullen and Carol Lawson, the technical
editors, as well as several members of the Publications Center.
                                111

-------
                             EXECUTIVE SUMMARY
     Considerable effort has been devoted in recent years to the develop-
ment of urban-scale photochemical simulation models for analyzing the
impacts of alternative emission control strategies on ambient oxidant
concentrations.  Because of the complexity of the physical and chemical
processes governing the formation, transport, and dispersion of secondary
pollutants, complex grid-based atmospheric simulation models are attrac-
tive tools for relating air quality concentrations to pollutant emission
levels.  Although evidence related to the performance of these models  is
limited, it appears that uncertainties in the model predictions fall
within the bounds of current accepted practice; computed ozone levels  are,
on the average, within 25 to 40 percent of the measured values when the
measurements exceed the National Ambient Air Quality Standard of 0.12  ppm
(12 pphm).

     Clearly, reduction in the uncertainties associated with model
performance is desirable; however, a main concern regarding the use of
these urban-scale models for air quality simulation is the cost and
complexity of this approach.  To simulate urban photochemical pollution, a
considerable amount of emission, meteorological, and air quality data  is
required, and the difficulties involved in acquiring such a data base  have
discouraged wider application of these models.  It is quite important,
therefore, to ascertain the extent to which the amount and quality of
input data may be reduced from the level considered desirable in current
practice, without substantially reducing the accuracy of the model
predictions.  To acquire such knowledge, studies of the sensitivity of
urban-scale air quality models to the level of detail of the input data
are required.  Several sensitivity studies have been performed with the
Systems Applications, Incorporated (SAI), Airshed Model and with the
Livermore regional air quality model (LIRAQ).  However, these studies
focus on a limited number of sensitivity cases; they do not provide
sufficient information about the relative importance of the various sets
of input data (meteorology, air quality, emission inventories) on model
performance.

     In this report a general meth^ology is developed to define a
priority ranking of the input data needed for urban airshed simulation,
which is then applied to air quality simulations of the Los Angeles Basin.

-------
METHODOLOGY FOR RANKING DATA NEEDS

     The methodology presented here  is based on the  definition  of  a
sensitivity-uncertainty index, and requires information  on  the  sensitivity
of the model to the level of input data detail  (e.g.,  number  of monitoring
stations and emission inventory), on the uncertainty of  the model  input
variables (e.g., wind direction and  nitrogen oxide concentrations),  and  on
the cost required for acquiring the  input data.  These factors  are
considered in the definition of a sensitivity-uncertainty  index that
provides the basis for ranking data  needs.  A sensitivity  index may be
defined as follows:
                                                                       (1)
where Aj^i is a measure of the deviation  in model  prediction,  for
instance, the absolute deviation  in ozone  levels  for  the  input data j  and
base case simulation i, and Alj is the corresponding  perturbation  in the
input data.

     Next, the sensitivity index  is normalized with respect  to a cost
difference (AK.J) corresponding to the change  in input data AI.., because
the ranking of data needs clearly must include such information.   The
normalized sensitivity index  is then defined  as follows:

                                  AI    AJ
                        Sij - rij ^ - AIT1    -                      (2)
                                    J     J

This index is a measure of the effect of the  cost  (associated  with
improving the level of detail of  input data j) on  model predictions for
episode i.

     The need for increasing the detail of input  data should be weighed,
however, against the accuracy available for corresponding model input
variables.  Accordingly, the following uncertainty index  is  introduced:





where 
-------
 input  variable  P^,  which  has  the  greatest  effect  on  model  performance
 during the  sensitivity  study,  is defined  by i  and  j.   For  instance,  in  the
 case of a perturbation,  Alj  is the number of meteorological  stations and
 P.JJ may refer to  the wind  field or the  mixing  height.
                   r             f            '
     The larger the  uncertainty in an  input variable,  the  less  additional
 Input  data  is needed, since  expenditure of  considerable  resources  may not
.materially  reduce the uncertainties associated with the  input parameter.
 Hence,  we weight  the.sensitivity index  S^,-,  by means  of  the  uncertainty
 Index  U^,  and  a  sensitivity-uncertainty  index is  defined  as follows:
                                                                      (4)
     A  schematic  representation  of  the  sensitivity-uncertainty analysis  is
 shown below.



INPUT DATA
t
3
t
r


INPUT VARIABLES
_^h r
t
Inherent



AIRSHED MODEL
SIMULATION





MODEL PREDICTIONS
1
U
1
Perturbation al, uncertainty Perturbation AJ^
"u
SENSITIVITY  STUDIES OF  THE  URBAN  AIRSHED  MODEL

     The SAI Airshed Model  has  been  applied  to  the  urban  areas  of Los
Angeles, Denver,  Las Vegas,  Sacramento,  and  St.  Louis.   Since the Los
Angeles basin offered the most  complete*  readily available  data base, it
was chosen over .other urban  areas for  the sensitivity studies.   From the
outset of the study, the influence of  different  meteorologies on the
sensitivity  study results was of  concern.   In Los Angeles,  the  close
proximity of coastal and desert environs  leads  to shallow mixing layers
(from 240 to 300  meters) along  the coastal margin and to  deeper mixing
layers (from 900  to 1100 meters)  farther  inland  during  the  midday hours.
To determine whether the selection of  different  meteorological  episodes
might alter  a subsequent ranking  of  data  needs,  two different air pollu-
tion episodes were considered:  26 June  1974 and 4  August 1975.
                                 vi

-------
     The level of airshed model performance for these two days  is compar-
able to the general performance obtained in other air quality studies
conducted with the model.  The average absolute difference between
predicted and observed concentrations for these two simulations  is 31 to
35 percent (7.5 to 8.7 pphm) for ozone levels  above 20  pphm.

     A total of 22 sensitivity studies was carried out  with these simula-
tions of the Los Angeles basin air quality.  The studies  involved pertur-
bations in the meteorological air quality and  emission  data, as  well as
simplification of the model  structure.  The resulting changes in the model
predictions have been analyzed to provide information on model  sensitiv-
ity.

     Assessment of model sensitivity to input  data may  be considered in
two parts:

     >  The sensitivity of the model output to changes  in the
        input variables (e.g., wind fields, diffusivities,
        emission rates and mixing depth), may  be evaluated through
        successive model simulations, each involving a  prescribed
        set of input variables.

     >  The sensitivity of these input variables to perturbations
        in the basic data used to construct them (for instance,
        gridded wind field inputs to the SAI Airshed Model will
        vary, depending on the number of meteorological stations
        used to prepare wind model inputs.)

Thus, interpretation of airshed model sensitivity is seen as the combina-
tion of the sensitivity of the model to its input variables and  of the
sensitivity of the input preparation procedures (or models) to their basic
data requirements.  A complete discussion of the sensitivity results using
this approach is presented in appendix A.

     The sensitivity cases considered in this  study are summarized in
table 1.  These simulations may be grouped into four general categories:

     >  Meteorological data:  Sensitivity cases 1 through 4, and
        sensitivity case 5-2.

     >  Air quality data:  Sensitivity cases 5 through  10.

     >  Emission data:  Sensitivity cases 11 through 19.

     >  Model structure:  Sensitivity cases 20 through  22.
                                 VII

-------
     An important step in quantifying model sensitivity is the definition
of appropriate measures.  The following measures are useful for this
purpose:

     Signed deviation is calculated as follows:


                        N'     N   c   • •    cb  - •
                                  _ilili-I	liii   _                 (5)
where C$ ^ ^ and C^ ^ j are the ozone concentrations for the sensitivity
case and'th'e base case, respectively, at station i for the hour j; N is
the number of stations; and N1 is the number of simulation hours.

     Absolute deviation (in ozone levels) is calculated as follows:

                        N1      N
                    IV  I
                    N'Z-f  N
(6)
     These sensitivity measures are presented for ozone levels > 12 pphm
in table 1.  Other sensitivity measures were also considered: overall
maximum ozone levels, the magnitude and time of occurrence of peak ozone
levels at various stations, and dosage based on ozone levels above 12
pphm.  Although results obtained with these different sensitivity measures
are consistent, in some cases (e.g., for maximum ozone predictions) they
are less pronounced.

     Simulations involving a reduction in the upper air meteorological
data or a change in the magnitude of reactive hydrocarbon initial condi-
tions result in the greatest signed and absolute deviations.  Such devia-
tions range from 15 to 45 percent.  The change in grid size results in an
absolute deviation of about 20 percent.  Other perturbations in the input
data result in absolute deviations of less than 15 percent.

     A reduction in meteorological data results in the lowest temporal
correlations, with values as low as 0.023 and -0.183 for the 26 June 1974
simulations.  Changes in the hydrocarbon initial conditions also result in
a low temporal correlation coefficient for the 4 August 1975 simulation
day.  The same simulations (reduction in meteorological data, change in
hydrocarbon initial conditions) also lead to low spatial correlation
coefficients.  The change in grid size results in a relatively low spatial
correlation (i.e., PS < 0.721), as expected, since reducing the grid


                                                              330R/1
                                v i i i

-------
   TABLE 1.   SENSITIVITY MEASURES FOR OZONE CONCENTRATIONS ABOVE 12 PPHM
Sensitivity                                               Signed      Absolute
   Study      	Description	    Deviation    Deviation

     1        Reduced upper air meteorological data  (J)    -0.177        0.300
     2        Reduced upper air meteorological data  (A)    -0.094        0.164
     3        Reduced surface and upper air meteoro-       -0.172        0.320
              logical data (J)
     4        Reduced surface and upper air meteoro-       -0.146        0.189
              logical data (A)
     5-1      Reduced surface air quality data (J)          0.000        0.028
     5-2      Reduced surface air quality, surface and     -0.208        0.329
              upper air meteorological data (J)
     6        Reduced upper air quality data, more          0.105        0.106
              precursor material aloft (J)
     7        Reduced upper air quality data, less         -0.073        0.073
              precursor material aloft (A)
     8        Reduced surface air quality data (J)         -0.141        0.145
     9        Reduced initial hydrocarbon concentration    -0.344        0.344
              data (J)
     10       Reduced initial hydrocarbon concentration    -0.450        0.450
              data (A)
     11       Reduced hydrocarbon emission speciation       0.046        0.046
              data (J)
     12       Reduced hydrocarbon emission speciation       0.121        0.125
              data (A)
     13       Outdated mobile source emissions (J)          0.027        0.037
     14       Estimated mobile source emissions (J)         0.064        0.067
     15       Estimated mobile source emissions (A)         0.115        0.127
     16       Reduced data of point source emission        -0.034        0.035
              temporal distribution (J)
     17       Reduced data of area source emission          0.054        0.055
              spatial distribution (J)
     18       Reduced data of area source, emission          0.118        0.127
              spatial distribution (A)
     19       Reduced data of area source emission         -0.005        0.007
              temporal distribution (J)

-------
Sensitivity                                               Signed      Absolute
  Study      	Description	       Deviation    Deviation

     20       Larger grid size (J)                          0.148        0.193
     21       Two-grid layer model (instead of 4)          -0.007        0.053
     22       One grid-layer model (instead of 4)          -0.067        0.207
  J and A refer to the 26 June 1974 and 4 August 1975 simulation, respectively.

-------
resolution tends to spread out the predicted ozone  concentration  fields
(i.e., the pollutant levels are more diffuse).

     The large effect on model predictions of  a  reduction  in  upper  air
meteorological data is due in part to the fact that the  wind  field  used
with this urban air quality model  is obtained  by solving a boundary value
problem for a potential function.  Therefore,  the simulation  results of
cases 1 and 2 are more sensitive to meteorological  data  located  at  the
boundaries of the airshed than to  the intensity  of  surface data  acquisi-
tion in midbasin, for example.

     Reductions in surface meteorological data (sensitivity cases 3 and
4), which affect mixing depths, wind direction,  and wind speeds,  lead to
some perturbations in the ozone levels.  However, the  overall  effect is
not as large as that caused by reductions in the upper air meteorological
data.

     The reduction in the number of air  quality  monitoring stations (sen-
sitivity case 5) does not have a major effect  on model predictions.   It
should be noted, however, that the Los Angeles ba--in monitoring  network  is
rather dense, and it is not surprising that the  model  is relatively
insensitive to the intensity of surface  air quality data.   It  appears,
when comparing the sensitivity measures  of cases 3,  5-1, and  5-2, that the
reductions in air quality data and meteorological data are not additive.
This reflects the complexity of chemical and physical  processes  in  the
atmosphere and the importance of including in  a  mathematical model  all
relevant phenomena.

     An increase and reduction in  precursor levels  aloft lead  to  higher
and lower ozone levels, respectively, as expected (sensitivity cases 6 and
7).  Similarly, lower ozone levels result when the  initial  and boundary
conditions are reduced to background levels (sensitivity case  8).

     The deviations in ozone levels due  to the prescription of initial
non-methane hydrocarbon concentrations are of  the order  of 30  to  40 per-
cent.  This large perturbation confirms  the idea that  these data  are an
important input to an urban photochemical model.

     In sensitivity case 11, an average  hydrocarbon  speciation (represen-
tative of motor vehicle emissions) was considered for  all  source  cate-
gories.  The main effect was to increase the chemical  reactivity  of  the
hydrocarbon emission mixture from petroleum refinery sources that,  in
turn, results in higher ozone levels downwind.

     The sensitivity of the model  to mobile source  emissions has  been
investigated in cases 13, 14, and  15.  In case 13,  an  outdated emission

                                 xi

-------
inventory was used; the perturbations  in model predictions  are  relatively
small.  However, the use of a simpler  inventory, based on fuel  sales  and
demographic distributions, results  in  a greater effect on ozone levels
predictions (up to 16 percent in absolute deviation for case  15),  as
demonstrated by the sensitivity measures of cases  14  and 15.

     Changes in the temporal resolution of point source emissions  (case
16) do not lead to any major perturbations in the  simulations.   The  spa-
tial resolution of area source emissions was modified in cases  17  and
18.  Larger perturbations were observed with the 4 August 1975  conditions
(case 18), and also for case 15 (in contrast to case  14) and  case  12  (in
contrast to case 11).  This is due  to  the lower mixing heights  on  4  August
1975 compared with 26 June 1974; consequently, a perturbation in the
emission levels is associated with  a smaller mixing volume  on 4 August
1975, thus inducing larger deviations  in pollutant concentrations.
Changes in the temporal resolution  of  area source  emissions  lead to  a
negligible effect on model predictions, as indicated  by the  sensitivity
measures of case 19.

     The model structure was modified  in the last  three sensitivity
cases.  An enlargement of the grid  size introduced an absolute  deviation
in ozone levels on the order of 20  percent.  The reduction  of the  number
of grid layers from 4 to 2 had relatively little effect on  model perfor-
mance.  This simplification might be used in some  model applications  to
reduce computational costs without  degrading the accuracy of  the simula-
tions.  It is important, however, to have one layer below the inversion
and one layer above.  Simulation with  only one layer  below  the  inversion
lead to notable deviations in ozone levels.

     It is also important to note at this point that  the use  of single-day
simulations, in contrast to multiple-day simulations, distorts  the
importance of initial conditions and underestimates the importance of
emissions.  This distinction is interpreted in light  of recent  multiple-
day airshed model simulations of the South Coast Air  Basin  (Killus et al.
1980; Souten et al. 1980).
APPLICATION OF THE METHODOLOGY TO THE LOS  ANGELES  BASIN    "

     The methodology introduced earlier for ranking  the  input  data  needs
of an urban airshed model was applied to the Los Angeles basin.   From the
results of the sensitivity studies presented previously, and from uncer-
taintly and cost estimates,  sensitivity-uncertainty  indexes were  calcula-
ted.  Sensitivity studies that involved the gridded  structure  of  the  model
are not considered in this section.
                                 XII

-------
     The results of the sensitivity-uncertainty  analyses  are  presented  in
table 2 for the 12 sets of input data, along with the corresponding ozone
level deviations, estimated costs for data  acquisition,  and uncertainties
in the model input variable.

     Only the ranking of meteorological  input data  (upper air and  surface)
is affected by the choice of oxidant episode.  Although  the value  of  the
sensitivity-uncertainty indexes of the other input  data  (air  quality,
chemistry, emissions) vary depending upon the episode, their  relative
ranking is not affected.  This suggests  that the meteorology  of  an urban
area must be taken into account primarily when considering the need for
meteorological input data.

     Considering the 26 June 1974 sensitivity results, it is  seen  that
updating the Los Angeles mobile source inventory ranks the highest,
because the cost involved is relatively  low and  because  the accuracy  of
mobile source emission rates is reasonable.  Upper  air data acquisition
can be important; this depends, however, on the  meteorology of the urban
areas.  For a single-day simulation, the initial conditions for  reactive
hydrocarbons are also ranked high, despite  the relatively high costs  of
such a large monitoring effort.  The latter, however, have been  shown to
be a key input to the airshed model.  Predictions may be  greatly affected
by variations in hydrocarbon initial and boundary conditions.  Reactive
hydrocarbons, an essential precursor of  photochemical smog, are  difficult
to measure accurately in ambient air; they constitute a  small  amount  of
the total mass of hydrocarbons, which is composed primarily of methane.
They are usually estimated via assumed HC/NOX ratios or  as a  specified
fraction of the total hydrocarbon concentrations.   Obviously,  considerable
uncertainty is involved in these empirical formulas.

     Spatial resolution of area sources, upper air  pollutant  concentration
data, detailed hydrocarbon speciation, and a detailed point source
inventory generally are of similar importance to model performance.   The
importance of surface meteorology is of  the same order,  though it  varies,
depending on the episode simulated.

     Surface air quality is ranked relatively low.  This  is because we
assumed a 10-station network in the Los  Angeles  basin for analysis.   The
sensitivity relationship is probably nonlinear and, if only two  or three
stations had been considered, the importance of  surface  air quality would
likely have been higher.  This underscores the recommendation  that the
results of the sensitivity-uncertainty analysis  must be considered in
light of the assumptions made in the sensitivity study.   It would  be mis-
leading to assume that surface air quality data  have little effect on
model predictions in other applications.
                                 xm

-------
                         TABLE  2.   SENSITIVITY-UNCERTAINTY INDEXES
Oione
Input Ditj
Upper air meteorology

Level
Deviation
0.33
0.
17
(J)
(*)
Cost for Data
Acquisition
Model Input
(dollars) Component Affected
SO, 000-100. 000 Wind
field (direction)
emissions source
Mixing depth
Surface meteorology

Surface »ir quality
Upper air quality
1C for HC

HC speciation

Ho&ile source
updating inventory
Ho&ile source
gas iaiev
Point sources
Area sources
vpatisl resolution
Area sources
tajnporal resolution
0.
0.
007
06
0.03
0.11
0.
0.
0.
0.
0.

0.
0.
0.
0.
0.


3?
43
045
14
0«

07
16
01
OSS
14
01

(J)
(A)
(J)
(J)
(0)
(A)
(J)
(A)
(J)

(J)
(*)
(J)
(J)
(»)
(J)

20.000-30,000 Wind
field
SensHtv1ty:uncert*tr,ty
Uncertainty
25'-50'
10-501
SO- 701
10-501
^missions sources
75.000-150.000 NO,,
SO. 000-125. 000 NO,,
7S.OOO-1SO.OOO RMC

20,000-100,000 RHC

5,000-20,000 NO,,

250,000-1,000,000 NO,,

10.000-50.000 NO,,
50,000-100.000 NO,,

60,000-150,000 NO,,

RMC
RHC




RHC emissions.

RHC emissions.

RHC emissions.
RHC emissions.

RHC emissions.

20-601
20-601
20-601

20-601

10-301

10-301

10-301
10-301

10-301

l«de«T » 10*
6
2

0
4
0
1
3
4
0
2
6

0
0
.6
.4

.5
-
.33
.5
.6
.6
.75
.3
.7

.23
.53
0.67
1
4
0

.8
.6
.22

- (21) - 66
- (4.

- (1.
(11)
- (0
- (4.
- (B.
- (11
- (2
- (9.
1) - 6.8

3) - 3.5
- 30
.82) - 2
1) - 11
7) - 21.3
.?) - n.t
.9) - 11.25
0) - 35
- (23) - 80

- (0
- (1
- (2

.8) - 2.E
.8) - 6.4
6) - 1C
- (4.5) - 11
- (H
- (0

.5) - 28
.6) - 1.7

J and A refer to the 26 June  1974  and 4 August 197S simulations,  respectively.
Lo»«r bound, geometric  nein value, upper bound.
                                          XIV

-------
     A detailed specification of transportation patterns  is  a  rather
expensive task, and this results in a very  low ranking of  the  need  for
such input data.  This result may be compared with  the simple  updating  of
the same inventory.  This shows the importance of the assumptions that
have been made and indicates that data  needs will vary according to the
existing data base.  Quantification of  the  temporal distribution of sta-
tionary source emissions is expensive to obtain and appears  to have little
effect on model performance.  This results  in a low sensitivity-uncer-
tainty index.

     The sensitivity-uncertainty analysis that has  been presented should
be seen as a procedure for defining input data needs; the  results obtained
for the Los Angeles basin should be considered as an  illustration of  the
method.  It is recommended that costs and uncertainties be estimated  on a
site-specific basis for the photochemical model application  area of
interest; such cost and uncertainty estimates can then be  used with the
results of the sensitivity studies (AJs) to compute the sensitivity-
uncertainty indexes.

     The results of the sensitivity simulations depend on  the  perturba-
tions introduced in the input data, the area chosen for the  study,  the
atmospheric conditions of the prototype episode, and  the modeling condi-
tions (e.g., type of wind model and the duration of the simulation).  The
effect of atmospheric conditions has been considered  for several sensi-
tivity studies carried out under different  meterological conditions
(simulations of 26 June 1974 and 4 August 1975).

     Limitations of the sensitivity studies result  from the  specific
attributes of the Los Angeles basin and the length  of simulation time.
The influence of the air quality data and emission  inventories on model
predictions is affected by the length of simulation time.  When the
simulation is extended from a single-day to a multiple-day run, the
importance of air quality data decreases, whereas the effect of changes in
emission inventories on model predictions increases.  It might be pos-
sible, however, to use the information obtained for the Los  Angeles area
in another location by evaluating the specific attributes  of both areas.
For instance, a classification of urban areas in terms of  atmospheric
attributes has recently been developed  (Hi 1 Iyer, 1980), and  it should
provide guidance in the generalization of these sensitivity  results.
                               xv

-------
                                   CONTENTS



PREFACE 	        i 1

ACKNOWLEDGMENTS	       iii

EXECUTIVE SUMMARY	        i v

LIST OF ILLUSTRATIONS	        xx

LIST OF TABLES	       xxi

     I  INTRODUCTION	       1-1

    II  CHARACTERISTICS OF THE SOUTH COAST AIR BASIN	      II-l

        A. Study Area Selection	      II-l
        B. Photochemical Oxidant Formation in the
           South Coast Air Basin	      II-2
           1. Meteorology:  The Inversion and the Sea Breeze	      II-3
           2. Emission Distribution and Oxidant Formation	      II-4
           3. Effects of Topography on Pollutant Transport	      11-5
           4. The SCAB as a Case Study	      11-6
        C. Description of Two Los Angeles Oxidant Episodes	      II-7
        D. Available Data Resources	      II-7
           1. Emission Inventory	     11-12
           2. Aerometric Data	     11-12
        E. Previous Sensitivity Studies in the
           South Coast Air Basin	     11-13

   III  STRUCTURE OF THE SENSITIVITY SIMULATIONS	     III-l
        A. Generic Classification of Sensitivity Runs	     III-l
        B. Procedural Issues	     III-l
           1. Modeling Region	     III-l
           2. Multiple Base Cases	     III-5
           3. Simulation Duration	     II1-5
           4. Selection of the Reduced Set of Air
              Quality Monitors	     III-7
                                 xvn

-------
    C.  Attributes of Each Sensitivity Run	    III-9
       1.  Simulations Focusing on a Limited Number of
          Aerometric Monitoring Stations	    III-9
       2.  Simulations Focusing on More Specialized
          Aerometric Monitoring Activities	   111-10
       3.  Simulations Focusing on Details in Emissions
          Inventories	   II1-12
       4.  Simulations Focusing on Model  Grid
          Mesh Configuration	   111-30
    D.  Concluding Remarks	   II1-31

IV  BASE-CASE SIMULATIONS	     IV-1
    A.  Identification of Analysis Procedures	,	     IV-1
    B.  Base-Case Simulation Results	     IV-4
       1.  Accuracy of Computed Peak Concentrations....	     IV-4
       2.  Estimates of Systematic Bias	     IV-7
       3.  Estimates of Gross Error	    IV-11
       4.  Temporal Correlation	    IV-14
       5.  Spatial Alignment	    IV-15
    C.  Simulation Results for Specific Monitoring
       Stations	    IV-15
    D.  Evaluation of Ground-Level Ozone Concentration
       Fields	    IV-30
    E.  Summary	    IV-45

 V  SIMULATION RESULTS	      V-l
    A.  Introduction	      V-l
       1.  Review of Sensitivity Analyses of
          Air Quality Models	      V-l
       2.  Sensitivity of the Airshed Model to Input Data	      V-2
    B.  Measures for Ascertaining Model Sensitivity.....	      V-2
       1.  Signed Deviation	      V-4
       2.  Absolute Deviation	      V-5
       3.  Temporal Correlation	      V-5
       4.  Spatial Correlation	      V-6
       5.  Overall Maximum Ozone Level	      V-6
       6.  Maximum Ozone Statistics	      V-6
       7.  Dosage	      V-7
       8.  Isopleths of Maximum Ozone Deviation	      V-7
       9.  Ozone Profiles at Air Quality Monitoring
          St at i ons	      V-8
      10.  Summary of Sensitivity Measures	      V-8
                            xvm

-------
        C. Summary of Sensitivity Results	      V-9
        U. Sensitivity Results	     V-21

    VI  INTERPRETATION OF RESULTS	     VI-1
        A. Ranking of Data Needs through Sensitivity-
           Uncertainty Analysis	     VI-1
           1. Definition of a Sensitivity-Uncertainty Index	     VI-1
           2. Cost Estimates for Data Acquisition	     VI-6
           3. Uncertainty Estimates for Model Input Variables....     VI-6
           4. Ranking of Data Needs	     VI-9
           5. Conclusions	    VI-18
        B. Generalization of Results	    VI-21
           1. Specific Attributes of the Simulations	    VI-21
           2. Limitations of the Sensitivity Results	    VI-24
           3. Generalization of the Results	    VI-28

   VII  CONCLUSIONS	    VII-1
        A. General Procedure for Sensitivity Analysis	    VII-1
        B. Specific Results of This Study	    VII-3
        C. Future Needs	    VI1-6

REFERENCES    	      R-l
                                   xix

-------
                                ILLUSTRATIONS
III-l  Modeling Region Used in the Airshed Model  Sensitivity
       Study	  111-6

 IV-1  Estimates of Systematic Bias in Computed Ozone
       Concentrations as a Function of Observed
       Concentration Level	  IV-10

 IV-2  Estimates of Error in Computed Ozone Concentrations
       as a Function of Measured Concentration Levels....	  IV-12

 IV-3  Distribution of Residuals (Predictions Minus
       Observations) for the Ozone Simulation Results	  IV-13

 IV-4  Number of Grid Cells for Which Model Predictions Bracket
       Observed Ozone Concentrations for 18 Stations during
       the Period 10:00 a.m. to 5:00 p.m.  PST	,	  IV-16

 1V-5  Topographical Features and the Locations of Air
       Monitoring Stations in the South Coast Air Basin	  IV-19

 IV-6  Calculated and Observed Ozone Concentrations for
       26 June 1974	  IV-20

 IV-7  Calculated and Observed Ozone Concentrations for
       4 August 1975	  IV-25

 IV-8  Ozone Isopleths for 26 June 1974	  IV-31

 IV-9  Ozone Isopleths for 4 August 1975	  IV-38

  V-l  Sensitivity Analysis of the Urban Airshed  Model	    V-3

  V-2  Signed Deviations for Ozone Concentrations Above
       12 pphm	   V-16

  V-3  Absolute Deviations For Ozone Concentrations Above
       12 pphm	   V-l7
                                 xx

-------
VI-1  Comparison of "Perturbation Specific" and Classical
      Sensitivity Indexes for a Hypothetical  Case.,	    VI-4

VI-2  Sensitivity-Uncertainty Indexes—Simulation of
      26 June 1974:  Sensitivity to Ozone Levels
      above 12 pphm	   VI-14

VI-3  Sensitivity-Uncertainty Indexes—Simulation of
      4 August 1975:   Sensitivity to Ozone Levels
      above 12 pphm	   V1-15

VI-4  Sensitivity-Uncertainty Indexes—Simulation of
      26 June 1974:  Sensitivity to Air-Shed-Wide Peak
      Ozone Level	   VI-16

VI-5  Sensitivity-Uncertainty Indexes—Simulation of
      4 August 1975:   Sensitivity to Air-Shed-Wide Peak
      Ozone Level	   VI-17

VI-6  Hypothetical  Case of Needs for Data Acquisition	   VI-19
                              xxi

-------
                                      TABLES
    1  Sensivity Measures for Ozone Concentrations
       above 12 pphm	        i x

    2  Sensitivity-Uncertainty Indexes	      xiv

 II-l  Summary of Peak Hourly Averaged Concentrations Observed
       During the Los Angeles Episode Days Examined in
       this Study	      11-7

 II-2  Summary of Aerometric and Emission Data Bases	      II-8

III-l  Description of the Airshed Model Sensitivity
       Simulations	     111-8

III-2  Air Quality Monitoring Stations in the South Coast
       Air Basin and the Number of Times They Were Selected
       for Inclusion in the Reduced Aerometric Data Set	     111-8

III-3  Hydrocarbon Splits by Source Category	    111-13

III-4  Categories of Emissions Sources and Temporal
       Distributions	    111-25

 IV-1  Summary of Maximum Computed and Observed 03
       Concentrations	      IV-5

 IV-2  Summary of Model Performance in Computing Peak
       Ozone Concentrations	,	      IV-6

 IV-3  Summary of Airshed Model Performance Measures
       for Ozone for the Los Angeles Simulation	      IV-8

  V-l  Summary of Sensitivity Studies	      V-10

  V-2  Sensitivity Measures for Ozone Concentrations Above
       12 pphm (Station Statistics)	      V-ll
                                xx 11

-------
  V-3  Sensitivity Measures for Ozone Concentrations Above
       20 pphm (Station Statistics)	     V-12

  V-4  Sensitivity Measures for Ozone Concentrations Above
       12 pphm (Grid Statistics)	     V-13

  V-5  Sensitivity Measures for Ozone Concentrations Above
       20 pphm (Grid Statistics)	     V-14

  V-6  Overall Maximum Ozone Levels	     V-18

  V-7  Maximum Ozone Statistics	     V-19

  V-8  Dosages	     V-20

  V-9  Summary of Airshed Model Sensitivity Results	     V-22

 VI-1  Cost Estimates for Data Acquisition	     VI-7

 VI-2  Uncertainty Estimates for Model Input Variables	    VI-10

 VI-3  Sensitivity-Uncertainty  Indexes:  Sensitivity to
       Ozone Levels above 12 pphm	    VI-11

 VI-4  Sensitivity-Uncertainty  Indexes:  Sensitivity to
       Air-shed-Wide Peak Ozone Level	    VI-12

 VI-5  Hypothetical Cost Estimates for Data Acquisition	    VI-20

 VI-6  Possible Limitations of the Sensitivity Results	    VI-25

VII-1  Results of the Sensitivity/Uncertainty Analysis
       for the Los Angeles Basin	    VII-4
                                xxm

-------
                             I   INTRODUCTION
     The Office of Air Quality Planning  and  Standards  (OAQPS) of  the  U.S.
Environmental Protection Agency (EPA)  has for  some  time  been examining the
feasibility of using grid-based complex  atmospheric  simulation models that
estimate concentration patterns of reactive  pollutants for  regulatory
applications.  Atmospheric modeling requires relating  changes in  emissions
to changes in air quality and present  grid-based  simulation models  provide
a fairly detailed representation of the  spatial and  temporal features of
atmospheric background species and pollutants.  The  development and
application of such complex models has been  considered because of the
complexity of the physical and chemical  processes governing the formation,
transport, and dispersion of secondary  pollutants.   Caution in recommend-
ing such models, in general, derives from several considerations:

     >  The limited extent of testing  of the predictive  perfor-
        mance of these models.

     >  The imprecision and possible bias associated with their
        predictions; however, this imprecision  and  bias  has not
        been shown to be greater—and  may well  be less—than that
        associated with the predictions  of models presently recom-
        mended by EPA for general use.  Such recommendations have
        been conferred only on models  used to  predict  the concen-
        tration distributions of inert pollutants.

     >  The difficulty and cost of acquiring field  data  (meteoro-
        logical data, emission inventories,  air quality  data) used
        as input information for the model.

     >  The costs associated with the  setup  and execution of the
        models;
  Primary pollutants are directly emitted  into the atmosphere,  whereas
  secondary pollutants are not  directly emitted,  but  are  formed  in the
  atmosphere by the chemical  reaction of primary  pollutants.
                                   1-1

-------
     Although evidence related to model  performance  is  limited,  it does
suggest that uncertainties in model  predictions fall well within the
bounds of current, accepted practice.   For  instance, the  Systems Applica-
tions, Incorporated (SAI)  Airshed Model  shows  relatively  little systematic
bias for ozone, and computed ozone levels  are, on  the  average, within 25
to 40 percent of the measured values when the  measurements exceed the
National Ambient Air Quality Standard  (NAAQS)  of 0.12  ppm.   Computed CO
and N02 concentrations agree with the  observations within 40 to 65 percent
(Reynolds and Roth, 1980).  Of course, reduction in  the uncertainties
associated with model performance is clearly judged  desirable.

     The remaining concerns attendant  with  use of  these complex models are
cost-related; of the various cost elements, by far the  dominant one is
that associated with field observations  (and data  collection)  and the
compilation of an emission inventory.   The  difficulty  and cost of acquir-
ing such data is perceived to be suficiently great to  discourage wide-
spread adoption of grid-based modeling approaches  for  routine  regulatory
purposes.  Therefore, it appeared necessary to systematically  investigate
the sensitivity of complex photochemical models to the  detail  of input
information.  There are several points to be addressed  in such a study.

     First, a general methodology must be developed  to  combine in a
rational manner the various components of  a definition  of model  sensitiv-
ity to input data.  These  components are:

     >  The sensitivity of the model predictions to  the level  of
        input data detail  considered.

     >  The uncertainty inherent in the  model  variables that are
        related to the input data.

     >  The cost of collecting the input data.

Once such a methodology has been developed, it can be  used to  determine
with detail and accuracy which input data must be  obtained so  that the
appropriate emphasis can be established  during the development of the data
base.

     This general methodology must also be  applied to  a specific example
to provide an illustration of its use, as well as  a  demonstration of its
utility.  Information on the relative  importance of  input data obta.ned
  For example, uncertainty in wind direction must  be considered  in
  assessing the importance of the amount  of  meteorological  data  in model
  predictions.
                                   1-2

-------
from a specific case study can be used  to  indicate  those  simplifications
that can be made in a complex model  without  substantially affecting
predictions.  This type of information  is  also  useful  in  assessing which
input parameters are the most important  and  which should  be considered
explicitly in "simple" models.  This report  describes  work carried out to
address these issues.

     An important aspect of this study  is  the selection of the  air quality
simulation to be used as a demonstration of  the general methodology.
Although it would have been desirable to carry  out  the most general type
of sensitivity assessment, it was necessary  to  restrict the inquiry in
several important ways:

     >  The selection of one particular grid-based,  complex air
        quality simulation model—the SAI  Airshed Model.   This
        model was chosen by OAQPS because  it is presently being
        evaluated in several U.S. cities.   Clearly,  however,
        another grid-based simulation model  could have been used.

     >  The selection of the Los Angeles area as a  prototype.
        This choice was based primarily on considerations of cost
        and timeliness.  Evaluation  of  model performance  had been
        completed for a suitable test day  for Los Angeles; this
        work would not have to be repeated,  and sensitivity stu-
        dies could commence without  undue  delay.  It was  acknow-
        ledged, however, that the study of one  to three additional
        cities, located in other geographical areas and having
        different climatologies, would  be  highly desirable.

     >  The selection of particular  days for study.  Because each
        day has its unique meteorological  and air quality charac-
        teristics, the days selected may not typify other days  of
        interest--!.e., "bad" or "worst" days under various mete-
        orological regimes.  However, two  reasonably representa-
        tive bad (i.e., high ozone)  days were chosen for  study—in
        fact, these days were selected  as  prototypical in at least
        two prior research studies as well.

     >  The study of a fifteen-hour  simulation  period. Unbiased
        sensitivity estimates can be derived only from multiple-
        day simulations, as usually  emissions (and  emission
        changes) from prior day(s),  as  well  as  emissions  from the
        present day, influence the simulation results. Thus, one-
        day simulations underestimate sensitivity to changes in
        emissions and overestimate sensitivity  to initial condi-
        tions.   Multiple-day simulations were not carried out
        during the course of this study because of
                                   1-3

-------
        -  The unavailability of a tested base case.

        -  Lack of adequate experience with  such  simulations.

        -  Cost restrictions.

     >  The limited range of applicability of sensitivity esti-
        mates.  Calculated "sensitivities" apply  only for the
        perturbations studied; the effects of other  perturbations
        or variations of differing magnitudes can not be  directly
        extrapolated because of the nonlinear nature of the model.

Despite these limitations, it is worthwhile  to undertake  such  a  sensitiv-
ity analysis for the following reasons:

     >  The present lack of any useful sensitivity results or
        analyses pertaining to data base degradation.

     >  The potential usefulness of sensitivity results in
        defining relative priorities in data needs.

     >  The need for information of this type in  estimating the
        overall costs of photochemical grid  model use.

     >  The inferences that might be drawn about  the utility of
        simple models that do not explicitly use  several  of the
        input parameters required by the airshed  model.

     >  The commonality of meteorological conditions in varied
        geographical areas that lead to the  highest  observed con-
        centrations of ozone.

     >  The ability to bring available information and knowledge
        to bear in modifying the findings, through qualitative
        arguments, so that they apply more generally.

Hence, the sensitivity analysis was planned  and carried out.

     In chapter II of this report, certain important characteristics  of
the Los Angeles Basin as they relate to the  formation of  photochemical
oxidant are discussed and the data resources that were available to
support the present study are summarized. Chapter III introduces the
sensitivity simulations and provides the rationale behind the  selection  of
each and the methodology employed in preparing model  input files.   Next,
the two base case simulations are discussed.  Evaluation  of model
                                   1-4

-------
performance in simulating ozone concentrations  on  the  days  of  26  June  1974
and 4 August 1975 is presented in chapter  IV.

     The results of 22 sensitivity runs are discussed  in  chapter  V.
Certain measures that are helpful in  quantifying model  sensitivity are
introduced and then applied to these  results.   Overall  sensitivit.es are
summarized in tabular and graphical form.   Detailed  results  of the indivi-
dual simulations are compiled in appendix  A of  this  report.

     The general methodology used to  evaluate the  relative  importance of
input data for a complex air quality  model  is developed and  appl /ed  in
chapter VI.  The chapter begins with  the definition  of  a  sensitivity-
uncertainty index.  After estimating  costs  for  acquisition  of  various
types of data, the index and cost estimates are applied,  yielding a
ranking of data needs.  The chapter concludes with a discussion of the
specificity, generalizability, and limitations  of  the  results.

     Chapter VII summarizes the principal  findings and  conclusions.
Appendix B discusses model input requirements,  the level  of  detail in data
bases currently available throughout  the United States, and  results  of
recent photochemical grid model sensitivity studies.
                                   1-5

-------
             II    CHARACTERISTICS  OF  THE  SOUTH  COAST  AIR  BASIN
     This chapter discusses the selection of the South Coast  Air  Basin
(SCAB) as the urban area prototype for the airshed model  sensitivity
study.  Unique features of the SCAB, as they relate to oxidant  formation
processes and the preparation of a photochemical modeling data  base, are
considered.
A.   STUDY AREA SELECTION

     At the inception of this study, two urban areas possessed photo-
chemical modeling data bases of sufficient "richness" to justify their
selection for subsequent sensitivity analyses.  These data bases were  for
the Los Angeles, California and St. Louis, Missouri, urban complexes.

     Determination of the most suitable study area involved a number of
trade-offs and consideration of previous experiences in oxidant episode
simulation.  For Los Angeles, for example, detailed mobile, area, and  sta-
tionary point source emission inventories were readily available in  a
format well-suited to modification.  (Modification to the baseline inven-
tory would later be made to reflect reduced levels of detail  in the  infor-
mation data base used to compile an emission inventory.)  The St. Louis
inventory, in contrast, did not offer an equivalent degree of flexibility
in altering the emission rates in various source categories.   Indeed,  it
was unclear whether any changes to the St. Louis mobile source inventory
could be made (e.g., reflecting outdated VMT estimates) without a costly
and time consuming effort to obtain the relevant emission files.

     The physical setting and meteorology of the South Coast  Air Basin
presented another trade-off situation.  The meteorology of this area is
more complex thfan that of many other urban areas in the United States  for
which photochemical modeling may be contemplated.  One consequence in
selecting Los Angeles, then, is the possibility that the need for detailed
meteorological data may become overstated (relative to other  data needs)
as a result of its complex meteorology.

     Previous experience in simulating oxidant episodes in Los Angeles  was
also a factor in determining which of the two urban areas should be  the
                                   II-l

-------
focal point for the study.  Adequate airshed  model  performance  in  simula-
tions of the South Coast Air Basin  had  been demonstrated  in previous
studies (e.g., Tesche and Burton,  1978; Reynolds  et al. 1979).  At  the
time the decision was made as to which  urban  area was  to  serve  as  the
prototype, model performance evaluation studies  in  St.  Louis  were  still
underway.  Establishment of a suitable  base case  appeared to  be a  less
costly endeavor for Los Angeles than for St.  Louis. Furthermore,  a number
of previous airshed model sensitivity runs had been carried out already
with the Los Angeles data base, and these studies could be  used to
supplement the findings of the present  work.  Based upon  the  above
considerations, the South Coast Air Basin was chosen for  the  sensitivity
analysis.

     The following discussion provides  an overview  of  the more  important
oxidant formation processes in the  Los  Angeles basin,  for the reader who
may be unfamiliar with certain unique features of the  area.
B.   PHOTOCHEMICAL OXIDANT FORMATION  IN THE  SOUTH  COAST  AIR  BASIN

     Important elements in the formation of  oxidant  in  the  South Coast  Air
basin are the unique meteorology,  complex topography, and varied spatial
and temporal distribution of precursor emissions.   In the presence  of  an
intense summer sun, these variables assume values  that  produce  frequent
high concentrations of photochemical  oxidants.   Two  major meteorolog cal
mechanisms—a persistent low-level subsidence  inversion  and  a thermally
driven sea breeze circulation—contribute to the accumulation and trans-
port of both primary pollutants and secondary  photochemical  oxidants.
Topography influences the transport of polluted  air  masses.  Because of
wind channeling by the surrounding mountainous terrain,  certain receptor
areas (e.g., Pasadena, Upland, Fontana) often  experience high oxidant
levels.  Also, a varied spatial distribution of  precursor emissions,
particularly influenced by temporally varying  automotive exhaust, combines
with the two physical mechanisms mentioned above to  form a  dense chemi-
cally reactive pollutant cloud, thereby creating a potential for high  of
oxidant concentrations.

     The following subsections elaborate on  the  roles of meteorology,
emissions, and topography in the oxidant formation process.  The interac-
tion of these physical and chemical mechanisms  leads to the typical daily
oxidant formation pattern in the SCAB during the summer-to-autumn smog
season.
  Figure IV-5 on page IV-19 identifies important landmarks and cities  in
  the South Coast Air Basin.
                                   II-2

-------
1.   Meteorology:  The Inversion and the Sea Breeze

     Formation of oxidant in the basin during the months of May through
October (the smog season) is dominated by the intensity and height  of  a
persistent low-level temperature inversion and by the strength  of the
land-sea breeze.  Southern California weather is  strongly influenced by
the semi-permanent east Pacific high pressure system.   Throughout the
summer, this intense high pressure system generates anticyclonic flow
aloft diverging toward the southeast.  This divergence aloft creates air
mass subsidence, thus creating an intense inversion as the sinking  air
heats through adiabatic compression.  The warm sinking air is extremely
bouyant when contrasted to the layer of cool moist marine air beneath  the
inversion.  A stable situation results; the inversion acts as a virtual
"lid" that both defines and restricts the size of the cooler marine layer.

     The summer inversion over the SCAB suppresses the vertical dispersion
of pollutants, causing emissions to become "trapped" in a shallow layer
the size of which is defined by the inversion height.   The size of  the
marine layer varies with time and location; that  is, its depth  at a par-
ticular location generally increases during the day.  In addition,  as  the
layer moves inland its depth increases.  Although inversion base heights
range from an average of 450 meters in the morning to  750 meters in the
afternoon (AQMD, 1979), the most severe cases of  oxidant formation  occur
when the base height [measured by the 0700 PST Los Angeles International
Airport (LAX) rawinsonde sounding at the coast] lies between 250 and 500
meters and the inversion strength (measured as the top temperature  minus
the base temperature) is greater than 10° C (Zeldin et al., 1976).

     A major component of pollutant dispersion in the SCAB is the ther-
mally forced sea breeze.  The sea breeze is generated from the  temperature
contrast between warm inland areas and the relatively cool coastal  waters
of the Pacific Ocean.  The sea breeze commences midmorning and  moves
inland during the day, persisting into the early  evening.  As evening
temperatures diminish inland, the thermal gradient reverses, and a  weak
offshore flow is established.  Several trajectory studies of this sea
breeze system (Neiburger, 1969; Holmes et al., 1956; Drivas and Shair,
1974; and AngeTl, Dickson, and Hoecker, 1976) have verified this persis-
tent pattern.  In general, the direction and strength of the sea breeze
determines the potential for dilution of polluted air masses.  Under
typical summertime conditions, morning emissions  of precursors  (say from
the 0600 to 0900 PST period) accumulate in the metropolitan and coastal
source areas under generally stagnant winds.  The ensuing sea breeze
gathers strength and advects the pollutant cloud  inland.
                                   II-3

-------
     The daytime pollutant transport pattern usually results  in  peak
oxidant concentrations in the early afternoon at  downwind  receptor  sites
in the eastern half of the basin with secondary maxima in  the San  Fernando
Valley.  As midafternoon temperatures increase, surface heating  weakens or
destroys the inversion, causing trapped oxidants  to mix within a deeper
layer, thereby diluting surface concentrations.  In contrast, on days  when
the sea breeze is weak, transport and dilution of the pollutant  cloud  are
negligible.  Under severe conditions, a weak offshore flow occurs  during
the day, confining the polluted cloud over the downtown metropolitan
area.  These conditions often coincide with an intense low-level subsi-
dence inversion, resulting in several days of high concentrations  of
photochemical smog (Zeldin and Cassmassi,  1979).
2.   Emission Distribution and Oxidant Formation

     The spatial and temporal distributions of precursor emissions  bear
directly on the formation of oxidant in the basin.   The  primary  source of
precursor emissions in the SCAB is automobile traffic.   Currently,  over  9
million people live in the basin,  approximately 7 million of  these  in  Los
Angeles County (AQMD, 1979).  Because of limited mass  transit facilities,
the automobile dominates commuter  transit,  resulting in  high  levels of HC
and NOX emissions during the morning and evening rush  hours.   1974  traffic
contributions to precursor emissions are estimated  at  approximately 44
percent of the hydrocarbon inventory and 51 percent of the total  NOX
emissions.  Estimated 1987 emissions from mobile sources are  24  percent
and 37 percent, respectively (Souten et al. 1980).

     Although the spatial distribution of the traffic  emission inventory
varies with the time of the day, traffic emissions  are concentrated gene-
rally around downtown Los Angeles  and along the major  traffic arteries
leading into the metropolitan area.  Emissions from traffic increase dur-
ing early morning rush hours (0600 to 0900), a period  typically  char-
acterized by light stagnating winds and a low-level inversion.  Simultane-
ously, contributions from several  point sources, particularly power plants
(NOX) and refineries (NOX and HC)  clustered along  the  coast from Lennox  to
Costa Mesa, add to the spatially varied distribution of  precursor emis-
sions.

     Although a significant number of these sources are  elevated, most
precursor emissions fail to penetrate the inversion; rather,  they remain
within the marine layer.  From the combination of  the  traffic, and point
sources, precursor emissions begin to accumulate,  forming a large reactive
pollutant cloud.  With time, the reactive cloud "ages/1  and chemical
transformation of primary pollutants takes  place.   Concurrently,  the
pollutant cloud is advected inland by the sea breeze.  Although  oxidant
                                   II-4

-------
production is evident in the polluted cloud,  high levels are not  usually
observed in the metropolitan area (downtown Los Angeles).   This  is  due  in
part to NOX scavenging by continuous automotive emissions.   In addition,
by midday the sea breeze has advected relatively clean marine air over  the
west-central basin.
3.   Effects of Topography on Pollutant Transport

     Transport of photochemical oxidants in Los Angeles is largely
constrained by the topography of the basin.  Elevated terrain,  including
the San Gabriel, Santa Monica, and San Bernardino mountains,  and the
Puente Hills, influence local wind patterns.  Although the mountains
virtually surround the basin, they do not present an impenetrable barrier
blocking dispersion of regional pollutants.   A major effect  of the
topographic relief is the channeling of wind flow, which forces pollutants
to converge in some areas (e.g., the San Fernando Valley)  and to disperse
in other areas (southeast of Chino).  As a consequence, three frequent
transport patterns are observed commonly.

     The dominant oxidant transport pattern in the SCAB follows a course
from coastal source areas inland to the San Gabriel  Valley.  The route is
directly related to the east-west ridge line transversing the San Gabriel
and Santa Monica mountains.   In effect, the orientation of the  ridge line
and the basic direction of the sea breeze coincide,  thus enhancing
transport of the polluted cloud inland.  The channeling of high con-
centrations of oxidant and precursors in the direction of Pasadena,
Fontana, and San Bernardino  is verified by the occurrence of  persistently
high concentrations of oxidant in the east basin (Davidson et a!., 1979).

     In contrast, Riverside  is an east basin inland  receptor  that is
affected primarily by pollutant transport through the Santa Ana Canyon.
When offshore flow forces the major pollutant cloud  into a more southerly
transport trajectory, Riverside can experience some  of the highest oxidant
concentrations in the SCAB (Zeldin and Cassmassi, 1979).

     The San Fernando Valley represents an alternate receptor for the
polluted cloud.   Areas of the east San Fernando Valley (e.g., Burbank)
frequently expeVience higher values of oxidant and precursor
  Although a general accumulation of photochemical  oxidants and precursors
  is observed along the foothills of the San Gabriel  Mountains, intense
  solar heating of the mountaintops will usually generate an upslope  wind
  system of sufficient strength to penetrate the inversion and to  vent
  surface pollutants aloft.
                                   II-5

-------
concentrations, than receptors in the western portion of the  valley.   High
levels of oxidants can be observed when southerly flow exists over
Southern California (Zeldin and Cassmassi,  1979).
4.   The SCAB as a Case Study

     Oxidant formation in the SCAB represents  a unique situation  in  which
routinely monitored high concentrations  are  directly related  to the  inter-
action of the three principal component  mechanisms:   meteorology,  topo-
graphy, and emissions.  Although other metropolitan  areas  (e.g., the
Northeast megalopolis, Albuquerque, and  Denver) experience periods of high
oxidant concentrations, the severity of  oxidant formation  and the  condi-
tions affecting it are different from those  observed for the  SCAB.  For
example, the latitude in Albuquerque is  similar to that  of Los Angeles
and, as a result, these cities experience approximately the same  solar
input.  However, Albuquerque is not subject  to the effects of a low-level
subsidence inversion or to a similar intensity of precursor emissions.
Consequently, oxidant episodes are less  frequent and severe,  compared with
those in Los Angeles.  Although the emission density and frequency of
stagnant summer periods for the Northeast megalopolis (including Washing-
ton, D.C., Philadelphia, and the tristate region of  New York, New  Jersey,
and Connecticut) parallel those of the SCAB, stagnation  in the Northeast
is not exacerbated by the presence of an intense subsidence inversion.

     The SCAB also illustrates a direct  cause-effect situation.   Unlike
the case for the Northeast, where precursor  emissions in Philadelphia may
affect oxidant concentrations observed in New York,  oxidant. formation in
Los Angeles is largely attributed to local emissions; "upwind" pollutant
concentrations typically reflect background  concentration  levels  attribu-
table to previous emissions from the area.

     In summary, the SCAB represents an  interesting  case study area  for
conducting photochemical modeling and sensitivity analyses and for examin-
ing specific features of the oxidant formation process.   However,  as
discussed later, certain unique features of  the basin make it necessary to
exercise care in generalizing the sensitivity study  results to other urban
areas, where the importance of various physico-chemical  oxidant forming
processes may differ.
                                   II-6

-------
C.   DESCRIPTION OF TWO LOS ANGELES OXIDANT EPISODES

     The airshed model  sensitivity simulations  are  based  on  two ozone
episodes  that occurred in Los Angeles during the summers of 1974  and
1975.  During the 26 to 29 June 1974 episode, a land-sea  breeze regime
prevailed—a common occurrence in the SCAB (Kieth and Selik, 1977).  At
night, pollutants were  advected out over the Santa  Monica Bay; with  the
ensuing morning sea breeze, some of this material was reintroduced into
the air basin.  Such conditions are representative  of a "typical"  smoggy
day in Los Angeles.  On 4 August 1975, during another high ozone episode,
onshore winds prevailed throughout the entire 24-hour period.  Analysis of
the synoptic meteorology and local wind station measurements revealed per-
sistent onshore winds for the two days preceding 4  August, as well.
Selection of this day allows us to examine a set of meteorological condi-
tions that are characteristically different from those occurring between
26 and 29 June 1974.  This, in turn, provides some  insight into how
closely tied results of the sensitivity analyses are to a given set  of
meteorological conditions.  Table II-l summarizes the peak pollutant
concentrations recorded during the two episodes.
       TABLE II-l.  SUMMARY OF PEAK HOURLY AVERAGED CONCENTRATIONS
                    OBSERVED DURING THE LOS ANGELES EPISODE  DAYS
                    EXAMINED IN THIS STUDY
                                 1974                  1975
           Pollutant       26 June   27 June         4 August

           NO?            26 pphm     36 pphm       19 pphm
           0^             34 pphm     49 pphm       32 pphm
           CO             13 ppm      14 ppm         8 ppm
           Sulfate        7.2 yg/m3   No data       28.9 ug/m3
           SOo            10 pphm     17 pphm       12 pphm
D.   AVAILABLE DATA RESOURCES

     Two of the major reasons the Los Angeles basin was chosen  for  this
study are the existence of an extensive meteorological  and  air  quality
monitoring network and the recent compilation of a detailed emission
inventory.  Table 11-2 summarizes the extent  of the data available  for
  The rationale for selecting two episodes instead of one is  discussed  in
  the next chapter.
                                   II-7

-------
                           41
                           U

                           c
                           01
                                         in
                                         1/1
                                         O
                                         OC.
                                                                               O
                                                                              OC
                                                                                         O
                                                                                         oe
 O
QC
                                                                                                                     O
                                                                                                                     C£
CO

«t

«r
Q
00
00
H-1

LU

O


         ID
        Q

         to
         O)
        r—
         (U
         cn
         to
         O
                   «   —• «
                        C O
                    fc.  O
                    O •!- 01
                       4-> C
                    fljj  IQ «^
                    U  IM 4-1
                    u ••- o
                    3  C Ol

                    LO  oi-
                                                                                         Q.



                                                                                         CO
                                         fO
                                         >

                                         •a
                                                O>
                                                Oi
                                                (Q
                                                U
                                                (11
                                         1-  O  3
                                         a> M-  o
                                         > -C  -c
                                         •o TE   i
                                                w
                                         X   •  CM
                                         •—  O)
                                         t. 4->  "O
                                         3  «
                                         O H_  i/l
                                         £ •—  "3
                                             3
                                         OJ  i/l  *C
                                         t.     a>
                                         IO 4_»  4J
                                             a.  i-
                                         «J  OJ  O
                                         4->  (J  Q.
                                         
10
CE

.
f— i
w
O
o

o


O)
4-J
i


UJ

1—
i^n
a.
T
0
I
,
h-

»

O

CM
~
O



X
«*
—1
^.
0
c

T3
C


^_^
»—

Q-
J5

^_4

8
O
ft






^— ,
h-
1/1
ex

8
*~4


^^
«
«

01
c
i
LU
4-1
«O


ai
•o
c
3
0
l/l

U

^J
t/1
3
O
a

Ol
in
0>

cc
4-t
«o

;

•o

Ol
(J
IO
It-
l-
3
CO

Ol
	 1

L. I/I
oi -a
0. C
o. ••-

                                                                                                              o.
                                                                                                              Ol
                                                                                                              o
                                                                            II-8

-------
           t
           •I
                                                                                           CD
                                                                                           r*-
                                                                                           CTl
                                                                                                   «t
                                                                                                   <_>
                                                                                                     e
     isj!
     
TO
X

1.
3
0
_c
0>
i-


TO

IB
O


VI
c
o
4-*
TO
4-1
VI

r—
TO

O>
>
Hi
VI

4-»
TO

VI
cu


4-1
TO
i-
OJ
0.
1
h-
OJ
W
3

OJ *C
U U
«C OJ



CO 4->
0
0
CD
VI
01
3
TO
>
•o
»
TO
t,
01
TO
>i

^
3
O
_c:
cu
s-
TO

TO
4->
10
O




»
^
-J

4-»
•O

1/1
4-»
c
i

j_
=j
*/>
»o
i1

c
o
4->
TO
TO
^






L.
TO
r—
O
CO
















cu
TT
en
t_
QJ
.,_
^;

•o
c
TO


V)
CU

O^
c

!
<9
L_
OJ
*tJ
^

t.
15
O
.C
QJ

«3


*t>
0





1/1
c
o
•r-
4->

in

*o
u
QJ

QJ
W
4->
fO

^

T3
1
X



X
4->

TD

c

X
I/O
z
v>
Ol
3
TO
>
VI
3
O
Ol
c
TO
C
TO
4-J
VI
C
••-
cu
t-
TO

TO

TO
O









t/l
4->
^
O
o.
s_
TO

r^

4^
TO
X
4^
'^.

J3
VI
3>

>,
4->
.^*
r—

XI

VI

>•


X

L-
3
O
JC
•D
O'
•o
1-
0
U
CU
1-






































^1
7
z
VI
Ol
3
TO
>
VI
3
O
0>
c
TO
4->
C
TO
4->
VI
c
.1-
QJ
L.
TO

TO
4->
TO
0








VI
4->
l_
O
o.
V-
TO

f^v

4-»
TO
fc
Ol

o
u

•s
o
CJ
u
OJ
>
o
u

•^1
5
o

CJ


X

w
3
O
.c
T3
O>

t-
0
U
OJ







































                                                         VI

                                                         i
1
VI
TO
0
•a
c
TO
TO
CU
O-
K-
14.
0
TO
QJ
a.
E
o
1_

T3
OJ
>

S-
QJ
T3
VI
Ol
4->
TO
E

4-1
VI
QJ

C
0
VI
V)
.,—
E
LiJ




U
• r—
i4-
<4-
TO
S~
(—
on rates for THC, S02, NOX,
d CO are available. Percent-
es of hot and cold starts and
t soaks are included in the
ventory with spatial and
mporal resolution. Vehicle
eed di itributions and vehicle
pes (4 classes) are included

VI

^—
QJ

g

c

•r~
4->
TO
4-J
0
O.
VI
C
TO
I.


CO
^—
ex.
er
— 1
01

4->











c CD o c o) a. x
TO TO XT If~ 4-* V) 4J




VI
c
O

VI
VI

'i


«— <
(—
o

CO
fy*
•dc

OJ
4-1
•a
c
TO











issions are based on nominal
E
LU

^3
C
TO
X
O
z

l-
0
l*~
VI
cu
4->
TO
E

4->
VI
Ol

c
o
en
*^-
^
LU



X
t_
cu
c
>^-
k)_
CU

VI
cu
TO
L-

•o









V)
C
o
XI
1-
TO
U
O


^^
r-

CU
>
4J
TO
01
l-















1_
O
14-
in
Ol
4-1
TO
_e

4->
V)
01

C
o

VI
VI
•r- VI
§4-)
S_
Q
QJ 1-
T3 ••-
•O TO
•r-
l- «•
O CO



4->
kk-
TO
^
U
V,
•r—
<
VI
1
<*-
o
c
o
VI
TO
C
u

0


*•>
CO
5?

*
X



0
**-

(/I
4-J
C
«o
»—
Q.

Sr-
$
g

co
~
*/l
C
«T3
^_
o.

I-
m

o
o.
ons based on inspection of
ily operating logs
•r-   4->  O
                                                                                                                     O  VI  V>
                                                           II-9

-------
                                 •O en
                                 C r—
                                 01  C
                                 £  O
                                 o  *->
                                                                   0)  C
                                                                   JC  O
                                                                   U  •*-"
                                                                   I/I  t-
                                                                   0)  =
                                                                   K-  CD
                                                                                    01  C
                                                                                    £  O
                                                                                    U  4->
                                                                                    I/I  U
                                                                                    O>  3
                                                                                    t—  CD
             •-
             o —  cr
                «j  c
             ti  «  —
                 C
             O  «0 r-
              t—
                                                                   Q.  IE
                                                                   <  O.
                                                                                                                      CO
                                                                                                                      cc
 C
 o
CM
  I
LU


CQ
         i/l
         «
        CQ
 
r™1
 OJ
 cn
         1/1
         O
IT)
r^-
a\
                                 •a  m
                                 41  .-
                                 cn
                                     03
                                     "-  O
3  T5
o 14-
                                     3
                                  Cg  t/i
                             Q.  1-
                         ^j  a>  o
                         *j  o  ex
                         

                                  c
                                                                            >> c
                                                                            3  Ul
                                                                            O  C
                                          a; -a
                                          T3  C  i/>
                                          3  «  D
                                          —      3
                                          u -a •—
                                          c  ID  «>
                                          —  01  >
                                              «
                                          »O  I-  I/I
                                          4->  OJ  3
                                          <0  >  O
                                          O  T3  QJ
                                                   QJ
                                                   C
                                                   o
                                                                                                                                          •o
                                                                                                                                          8.
 -
3
O
                                                                                                                                                  QJ
                                                                                                                                                  i-
                                                                                                                    OJ
                                                                                                                    3
o-
c

i-
o
4-1
c
i

-O
C

X
Q)
u
IO l/l
14- C
I- O
3 ••-
m »_>
10 2
ir> i/i
t—
on
a.
O
ro
l»4
^.
X
<
_J

4^
•a
1/1
OJ
T3
C
O
I/I
o
•o
«
a:
4->
O-
>—
1/1
a-
0
PO
CVJ
»— (


o
t— i
§


QJ
4->
C
i
«
1—
00
Q.
in
1C


T3
c
•o





f
o
in
<->
0

3
*—
(/)
a.
ID
O
«T
»«4

•o
C
"0
I/I
•"^
in
•c
r^
O
f
u

z

c
10
00
Ol T3
r- 3 C
UJ
X
A3
T3
C
10
V—
00
a.

o
^~
O^
*t
C
nj
XI
i^
3
CO

4-1
10

I/I

Q_
t—
oo
a.
O
n
O
T3
C
iO
tn
•e
rn
0


O
c
T3
U
10
c
1_
QJ
CO
C
10
00
(U
-o
lA
U
01
>
ce
4-J
K3


»o
t.
a.
*/i - — .
»—
4-> OO
u- Q-
fO
LJ 25
V- f^>
•r- O
^. -^1-
!/•>
a.
o
fi
o
»
o
r->
ir>
O
s
_1

4^
10
I/I
QJ
t)
C
O
in
0
•o
iO
ce
cr
£ c
*o
• i/i
4^
Q- -a
c
> 10
>—
00 «
o- . — .
V—
O 00
ro Q.
CXJ
— < in
«—4
• vn
o —
«-H
8»
— o
o
O) —
4->
c •
0 0
£ m
CO
i— O
UJ 	
t—
00
IX

s
*r
vH

T3
C
IO
I/I


I/I
(O

o
_c
u
z
m
|

4-1
(13
.*->
l/l
LO

4-J
•^3

ai
L.
3
+j
t5

a.
1
»—
                                                                                                                               I/I
                                                                                                                           T?  QJ     ^3
                                                                                                                           QJ  3      U



                                                                                                                           L.  >      i-
                                                                                                                           Ol         Ol


                                                                                                                               c-
                                                                                                                           >1  1O      3*1


                                                                                                                           i-  Q)      W.
                                                                                                                           3  >      3
                                                                                                                                                            •si
                                                                                                                                                            C.  ns
                                                                                                                                                            i
                         ai
                     •0  3
                                                                                                                                                             OJ -o

                                                                                                                                                             E —     4J
                                                                                                                                                    •a  X)
                                                                                                                                                     •o   c
                                                                                                                                                    ce   >o
                           <0

                          T3
                                         in
                                         c
                                         o
    c  \~

O  4J  C
«  3  oj
l«-  •-  U


3  o  §
l/l  O.  u
                                           c
                                           o
                                           u  i/i
                                               c
                                           4-)  O
                                           c —
                                           ra *->
                                           4->  13
                                           3  I-
                                                   o
                                                   Q.
                                                                  en
                                                                  o

                                                                 "o
                                                                  u
                                                                  o
                                                                  a;
                                          a>
                                          u
                                                                     -
                                                                    3
                                                                   l/l
                                                                                          a>
                                                                                          >
                                                                                          OJ
                                                                                  tt> T)
                                                                                  Q. C
                                                                                     a.
                                                                                     OJ
                                                                                     01
                                                                                     c
                                                                             11-10

-------
•t
o
I
•I

                                                           03
                                             
\-
i
4.

"O
r^

4->
IO
X
4-*

.0

in
••—
:»















>^
w.
0)

Ol

•o
Ol
•D

O
u
or
i.

m
Ol
3 i.
•— 3
















in
c
0

4->
1C
u
o
L.


+j
O
1
«t

on
••
on
3
Z








in
3
O
o>
c
*o

c

C
O
O.
•^-
10

f~
4->
IO

u
Ol
>
o
o
-o
3
o

o















^^
i-
Ol

Ol

•^
Ol
TT
V-
O
O
01
w

l/l
o>
3 k.
i— 3
10 0
















I/I
C
o

4->
1C
o
o
V-
Ol
JZ
4->
0




CO



1
l/l
1

r—
10
4-*
o
4->

•o
c
10

0
±t
IO
Ol
a.
14-
i+-
0

»
^
10
Ol
o.





E
o
u
14-
•o
Ol
•>

I-
*

in
Ol
4->
IO
e

4-1
1/1
111

C
o
in
in

E
UJ







i
4-1
• C
X QJ
O 0
Z U
01
> a.
OJ
O
on •
a>

O J3
I <0
>— •—

k. «
o >
S- 10

in Ol
Ol J.
4^1 
c
o -a
in 10


^o
c
1C

^—
01 ^-
If
c:
o c
— o
4-) *^-
CO 1/t
4-1 \r\
l_ •*-
0 E
CL 93
l/l
C r~*
^ p—
I- O
^5
on u_
>—
CC. CO
5%

01 aj
JC JT
4-> 4->







•o
c
IO Ol
JE
in 4J
4-1
I- C
1C -r-
4->
VI T3
01
TJ -a
,_ 3
O •—
u u
c
•O "-
c
1C 0)
u
4-> 10
o
c. in
-^
<4- 10
o o
in
in
Ol 4->
cn o

r—
m

*J C
10 O
a.—
in 4^
3
JC —
*-> O
>*— in
X Oi
I.
>,
I- •—
O *O
4-> C
C O
OJ Q.
> E
c a>
•*- 4J


































Ol
•— ID
u 01
"i?
Ol j—
> u
c
-o •—
c
* 01
i^
t/l tl
c
o - — •
•T- 1/1
4-> IIJ
3 l/l
.O I/I
•r- IO
^ r*-
4-J U
1/1
.1- «r
X3 * 	

-O l/l
Ol O>
Ol O.
S. x
I/I 4->






























£
f
CJ
on
































•
X
0 I

4-1
• I/I
l/l Ol
01
in m
IO OJ

10 a>
O •— "-
"- 3 4J
C O •—
IO •>- ' —

S_ i- O
O 10 <0
o. >t-

O -O CD
U- C OJ
t/l i-
c •• o
o o >*-
•— o
l/l "O
in » 01
•.- 0-J4J
E O 10
uj on £



g
•4
O
vn



































s_
o
lt-
4.J

E

in
Ol

c
o
• r-
in
in
•«— i/i
Si L.
O
•o a.
Ol t-
"O •'^
•o 10

w ^>
t5 ro
tn
Ol

*J
•r-
4-1
3






'i
0

l/l
C
o

4-1
3
£

\-
4J
in

T3



O
Z
^
o

Ol
40
^

^J
cu

c
o
in
in

E
UJ










in
CTl
I O

OJ
Q. O>
t/> C
C -r-

IO
C S-
0 0)
a.
-o o
o>
in X
IO r-~
["l -t—
1C
I/I T3
C
o •+-
••- 0

in c
•i- O
Ol 4-i






l/l
4-1
c
1C
a.

m
r- 1
j_
0


S


c

•t
10-
o
on
*
CM

on
1/1
at

4->
•~~
4->
3



































^3
c
^
ct

s_



ai
*j

1/1
o>

c
0
l/l
(/I
••- X
E O
UJ Z
in
Ol

4-J
r""
4-1
O



































U
•r- 1
C t-
o- a.
u
o -D
c
I- 10
o
o
in O
Ol
1C CVJ
E O
••- on

in *
o> x
O "/i
c z 01
0 4->
in in • —
VI Ol 3
•i- l/l U
E -o •-
UJ cr 4->
V
IO
o
CO

Ol
U 4-1
t- o
3 •—
O l-
in 4-1
O> \Si
o
s_
•^- CT1
«t c
II w
o
CO 4-J
ce -^
 on
C "
CJ (X)
O 
1/1 i-

a> i-
— OJ
— -C
W^ 4J
tn *O
              L.

              01
•3
o
                       c
                       o
                                    X
                                    l-
                                    o>
                                    c

3

>^

^j
in
•r-
o


in
Ol
U
3
O
in

 4->
O in






i/i
Ol
u
L.
3
0
in

u
<*-
•f-
u
a.

u

(j
h—
X
CL.
*
10
C
0
•*-
1o
z

II

on
3
z

                                     n-n

-------
photochemical modeling of the June 1974 and August 1975 episodes.   Impor-
tant aspects of the data base are mentioned briefly.
1.   Emission Inventory

     The 1975-1976 emission inventory for the SCAB is a result of the Air
Quality Management Plan (AQMP)  development process initiated  in 1976  by
the Southern California Association of Governments (SCAG)  and the South
Coast Air Quality Management District (SCAQMO).   The development of the
emission inventory was separated organizationally and procedurally into
several components:  stationary sources,  area sources and  off-road
vehicles, and mobile sources (Souten, Anderson,  and Ireson,  1980). The
stationary source inventory development was initiated at the  SCAQMD using
their files on source activity.  Much of  the work performed  at the
district involved manual transcription of emission records and inventory
reformatting necessitated by the model.  District data were  then modified
by a contractor to approximate  average summer day emissions.

     Area source and off-road motor vehicle inventories were developed at
SCAG, primarily using demographic data, whereas  the on-road  mobile source
inventory was generated by starting with  SCAG demographic  and transporta-
tion system data.  These data were used as inputs to the California
Department of Transportation (Caltrans) traffic  modeling system, which in
turn provided the necessary inputs for the Direct Travel Impact Model
(DTIM), a model provided by the California Air Resources Board (CARS)  to
generate actual mobile emissions according to vehicle fleet  mix, hot  and
cold starts, hot soaks, and so on.  The ARB was  also involved in assembly
of the inventory components through a system designed to allow automated
generation of gridded emissions rates as  required by photochemical
simulation models.  Details of  the 1975-1976 emission inventory are
discussed in the SCAB Non-Attainment Plan (NAP)  and its supporting working
papers, prepared jointly by SCAG and SCAQMD.
2.   Aerometric Data

     As table II-2 indicates, meteorological  and air quality data for the
SCAB are available from several sources.   Data from these sources were
acquired in the course of previous applications of the airshed model  to
the basin (Reynolds et al., 1979; Tesche  and  Burton, 1978).
                                                                33QR/2
                                   11-12

-------
     One element of the aerometric data base not listed in table II-2
warrants special mention.   Routine reporting of hydrocarbon concentration
data in the basin includes total hydrocarbons (THC)  and non-methane hydro-
carbons (NMHC), the latter obtained primarily from the difference between
THC and methane determinations.  The data are reported to the nearest  part
per million.  Lack of precision in the reactive hydrocarbon data (due  to
rounding to the nearest ppm) was partially ameliorated by a special
reactive hydrocarbon monitoring program sponsored by the CARB during the
summer months of 1974 and 1975.  From June to September in both 1974 and
1975, a program was conducted to determine the nature and extent of non-
methane hydrocarbons through Cig in the atmosphere of the SCAB.  Sampling
sites selected for the study were Long Beach, downtown Los Angeles, El
Monte, Azusa, and Upland.   Atmospheric samples were collected using auto-
matic sequential samplers at the following times:

     >  0200-0500 hours (pre-dawn).

     >  0600-0900 hours (peak traffic).

     >  During a three hour interval corresponding to peak oxidant
        occurrence at each site.

Total hydrocarbon monitoring was also performed, thereby enabling statis-
tical correlations to be drawn between ambient THC and RHC concentration
levels.
E.   Previous Sensitivity Studies in the South Coast Air Basin

     Several studies of photochemical grid model sensitivity to vanat  on
in input parameters have been reported in the literature in the last
several years.  Most of these studies have involved application of the  SAI
Airshed Model or the LIRAQ model developed by Lawrence Livermore National
Laboratory to various urban areas (i.e., Los Angeles, San Francisco,
Denver, St. Louis, Sacramento).  Reported sensitivity analyses focusing on
the South Coast Air Basin have been performed almost exclusively with the
Airshed Model.  Appendix B of this report discusses the principal  findings
of sensitivity studies conducted for the South Coast Air Basin, as well as
for other urban areas.
                                   11-13

-------
              Ill   STRUCTURE OF THE SENSITIVITY SIMULATIONS
     This chapter introduces each of the sensitivity simulations  and pro-
vides the rationale behind the reduction in the level  of  detail of  infor-
mation available for preparing airshed model  inputs.   Appendix A  compares
in detail the changes in input files for sensitivity runs and base  case
runs.  Chapter V discusses what impacts these changes  have  on predicted
oxidant levels.
A.   GENERIC CLASSIFICATION OF SENSITIVITY RUNS

     The 22 sensitivity runs performed in this study may be  grouped  into
the following general categories:

     >  Meteorology

     >  Air Quality

     >  Emissions

     >  Model Structure, (i.e., grid resolution, number of
        vertical layers).

Table III-l describes each simulation briefly.  In a following section  the
input preparation procedures for each run are summarized.   First,  however,
a few procedural issues that bear on the overall structure of the  sensi-
tivity study need to be identified.
B.   PROCEDURAL ISSUES
1.   Modeling Region

     Two major considerations governed the selection of the modeling
region.  First, substantial computing cost reductions could be gained  by
limiting the size of the region (by eliminating grid cells over the ocean,
mountains, or desert, for example).  The second consideration  was  that  the
                                    III-l

-------
o


                                                                                                           o>
                                                                                                           C  wi CT*
                                                                                                           —  4) C
                                                                                                           X t3 —
elds and
te radio
ten moni
                                                                                                          •O
                                                                                                           C —  *"
                                                                                                           *- UJ  4)
                                                                                                           X     U
                                                                                                             O  3
                                                                                                          *• rn  *-•
                                                                                                          l-x. m  "O
                                                                                                          CTi -*  W
                                                                                                          .-4     4)
                                                                                                             •a  a.
                                                                                                           41  c  6
                                                                                                           c  
laracteri
o




* c
,— - «
-§""

^* *~1 c
8 S on
$_o.
««- x o •**•
*o t-» a.
£ c? g" 8

> m o ~<
QC


r*- ^ —
!/! O O Wl
41 — ^~-
*C 01 3 ^
0 § X ^
"O * U
a •— 4^ *«-
oc uj a. ae

>

^"
u


41
I
UJ

V.


§
o
^
u
Acousti



a, ^
C JC
H-
on
a.

8
8
4i
•o
U.
4)
er

%


L.

"*

Aircraf





o» ^- — ^ c
4-» • T3 t- O *
c *j c on tf) t/i

t— !?( »-«irtonoH-D^ C t— C »— CTI
onp r— •— i Oi/i3 O on*« on 3
a-X LU. -oma.x — a.O a-X
cowico C4-» m--. c
• C • — • Q.U CO -i- _ c t— CT • — • *< (3 — . C • C • — •
- . — . *) t— 3 f CL. o * • — • «on c «*O*—O *^^«t—
O ^— 1/1 O CT> z m o>— a.-*- qj 	 iiconx O •— <^">
*°"H8 SJ 2^ l.0"^^ 5 'C 5 o 8 " §.°"?s

	 10 _j«fe cm _i ^^ • ae u. _j£


"gfc(2 ^QCC 32- * 3 • »2 "» *°S!c^ 3 * S • ^

•oo •oXH-4-*c •o'^*'o u -o -o ' ^c

O *J •« O — * i— i Q_ -a O *J *O >*- O *J CT> "fl ••- C -*-1 ^
oooo oo<^'in *— no oooo 3 oaHTo *" cccc
« r— O •— ^lOc—* •»» 4t ••- «o .— O •*- O « »— *j -^- W'"~ «j — O "-
acLu< — • x cc — • « •— - a.03 *c ce UJ — z us oc uj is. a- <<« QCUJ— ~z

i $

«— 41 •—
1^*^ CjC  C JT

CQ

-------





















•o
0)
3
C


-M
C
O
o


1
»— 4
NH
1 t 1
_J
00
p£





















1
—
Ji
c
iX
V
£

(J

—
OJ
U
fc_
JC











o>
VI


•1
B
j>
**
•5
w
S
f
S

6








Q
U
£


g|
•c •-

£ or
JD u
1/1 i^. t/t
£ 55 §
O. Vt •-
« 4->
•^ ^ «
C 4-J
C
X 41 C
•T- "O •>—
Ecu
o o

•o g
vi to 6
•c u
•— c
41 4, 41
•*- C
£4>J
*
C
B UJ OJ
in o 3
cri rw 

c O c to «—
o o « c a, * o
Vt X «-< •— «0 -— • 3 JC
§1/1 l/l Q h- C?t O
•<- *"" o^ ""* m a! ^ *^
4-- m vi c o
4-" .— •«. o •— h~ ro 4-» «o
vi h- — o ^» t/i 4> O CL i/i
1/1 jC h- Q. -O •-»
O* CL. 3 O t/l •*— • ^

L. ^*) Z O O OJ **) t/l
O — « C 59- 1-1 > V£> O. <--.
c *""* D. — c ro Q.

-i-^S "cm m _j *"* i£ P
C 4->  O •— • CL.
* v. o £ co ~" *£. v> 8 » cS
Qj OJ m O- *— o vt OJ -— O «f
LJ T) C\J t— 4J C "O O «-«
•o C'-'tnco *o .»- 4-1 c o» «-• — -
*£ O^^HO. -0 M~ O 4J ^
3 o o 12 in -^ )B u oEo"c:
TSlOTSV JDU t. "C m^-


o>
OJ

"u "c CX£ *X Q.
f » s" » ir -S
L.
3 0

•^ er
S£
o
u o
**- *J
= i
vt «o
U 4^
^ '* -S
0
0)
Vt
5
v>
§
4^
•o
c
o
u
£
•o
c
I

c
to

*J
c
v
'*-
O)
c

v£>




(4-
o
VI
VI
f
6
I/I VI
c
S c

•*•) 4J
Vt VI
•D
in jo
v>
•o o
Vt 4J
ft; .^
\. -o
3 C
4J O

o>
c
3
o
&








VI
c
I

c

o
c
i

1)
•o

o









1


VI
c

o
£
i
>,
~
L.

R

g
I.
t*_
S




L.
n vi
•D C
§ °
s5
•o
TD C
C 0
•0 O
i


— o
i §s"
.-g.5
vi vi V
JC •«
4-» ja oj
ex u
«Vi *Q
C H-
cn-2 ^
.^ -ti "*
X ^3 "V
E O *
u
- vt
vi X OJ
•o i_ -o
•— tj C
4) ^ O
~- C Vi
W- 3 O
o ••-
T) J3 -0
C t)
'* "c u

rv. * c
2 £ £
«.2-
C 4J UJ

^•oo
§0
^ rn












o
*
•

c
ro
t/l
c
3
o
c
o
CL

u
v<
•g
OJ

L.
o


* " 5
•o c «
3 — >.
u 'oi^
* — =

01
u
3
t/>
§
t.
c
0>
O Vt
II

C 4-)
to i/i

3 CT>
-— c
&I

T3 ••-
c c

3 4->
u c
CX TD
*O CTI
w o

— €
o
5 i
* T3
OJ
C 4->
.2 i
i— I/I
1 1
I/I 4->
«w o
'^ *«
I— <
. — VI
c ••- ^~

-^ OJ C
4-- O
\£> *0 I.





OJ
I/I
01
u
1
2

•C vt
E 3
>
•D
I 3
L.
. U
«0
: 1
. VI

o
4-*
JC
£• -^
JC
u


2


•0
3
cr
u
L.
0. 2
tD -^
5
OJ

i
ecursor
O. v>
c
V) O
VI •»-
OJ 4->
•o
£ g
"»
C l-
o *
— T3
«3 3
*3 J3
E
•*- i-
VI "-
tO
in
r-. c
S 4;
4^ U
Vi
3 ••

« |o






t.
«*-
»
«O

O
I/I
vt
OJ
JC VI
4-> C
x -2
<0
o tl
^ CT>
41 C
VI •«-
2 0
c c
S 1

£ x
0 "3
3
to t-
c *
3
O CO
CO *9
Vt
1


u c
— o
to (J
g f
4> *j
w i
m o
f*- J3
•e




vt
ex c
o
VI 4-1
41 -.-
C
5 S
* X

§10
•o
•*• c
4-» 3

i t.
Vt T)
«r c
fv. IO
u
3 4-T
"3 *-
O






w

4-1


o
Vt
4/1
JZ VI
C 0
6 °
to
O VI
I/I •*•
55
c c
s. i

^ X

3
to u
o cr
CD r">

o
4-J

•o
•5 1
i ?
"1
« o
r- jD
1 VI
o c
L- O


J§ * '-D
* -^ o
U U (J
e? v
"x i fe
JC U -D

•* -o j§
•*• S v
•^^ tfl S3
VI
•g C 0)
u
m oj **-

*3 * °
E J3 0
VI V» 4-»
c «
«• o v-
r^ •—
L. ar
o» 4^ -^.
c c c
3 4> 0
^ W J3
C i.
vC O 'O


1
ttJ

3
Vt
to
i

I
ec

^ 41
41 •*- V
t/> j3 r*.
S 1-5
0 U
4-» in
w ai cr>
at w
o "° "*o
U 0.
JD C C
k. 0 JC
«J -t- O
U 4J VI
O ^ fc-
^ ai x
X L. «
jr a:


* ex c
OJ E OJ
cc OJ E

i"


cc
VI
vi o

*r to
r^- t-
i t/i
e E
c -^ —
O X 4->

u •— o
O i- U
L. 4->
•O m X
W "O
* S §
Z -D J§
c E t.
Vt *}
jC vi
X C 0)

0 C
— o •-
^- 41 «*-
3 Vt O
E -0
•»• J3 O
VI •*•
VI — >
f^ O i-
cr -^
4-» Z Z
VI 4J -^
3 C C
O1 OJ O
3 W JO
< C *-
o 10


1
OJ

3

i

cc
c
•^ 41
oj ••- in
Us
£ 85
0 W
4-i vn
i. at en
C '^- »— '
a> L.
U OJ •
c "° ^o
u ex
C JC OJ
J3 C C
1- C JC
TJ — O

O TJ I.
u — ai
•a a x
X I- »O
JZ Z


cc a; E

o"

CJ
n:
Of
•c
10 "^
^C u



"ex
Vt X
c -
0 E
VI C
•*- ja
e t-

§s
J3 *-
W t/l
m o
0 t
"D U
X
jc x;
>1
4-1 V)
fO
u S
•a- o
r** 4-1
C
3 L.
*^ O
U



JC
u
u *-•
41 >o
4, 4.,
cl
°«

Vi
5 fc
§ §
4J •
i- 41
41 >O
CL, L.
vt c
ex
•c w
§ S
o c
c
"- i-
C 4)
•C 4,
cr c
i- —
o cr
c
— cc
O X
> -C
§


C (J
o £
jO &
t VI
II
c .£
S E
III-3

-------
w
a.
I
l










3
1
^.

i
-C
L.
3
t/i

U

1/1
•o
c
o
cr*
c
o
o
u
•o
s
t/1
JQ
I
i

ource

0)
2
i

r*^

1

1
S

3
u.
a
t«
i
^
.S i
'-*
OJ Qj
t/1
§ S
^ *
1/1 01
i/»
l/i
»
o, *
u OJ
u k,
i *

— §
•- c
JC O
i QJ

CT 3

C —
£.


'E
u
0)
c
«
J3
OJ
o
1
u

g
i/T
o
u
tj
c
t/t
s

c
I/I
I
'i
OJ
souro

J^
JO
U->

2
*•»
3

0)
"0
>
19
X
C
O
u
OJ
o>
£
QJ
2

>
QJ
I
t/l
Of
2
c
Ifl
k.
c

o

T)
c
o
VI

i
OJ
!«
£ 'e
_
3
c
o
'5

c

0.
i
o
a.

2

c
3

•«•*
1
TJ
C
3
•o
k,
>
TJ

!
O)
u
_
t/1

i

L. V*-
QJ O
> k.
§
)urces spatially
jraphic distribute
 O
3 *J
k. C
tn -G
— k.
T3 0
V <->

O*
3 U
*"> O

C
o
iources spatially
jraphic distributi
*o E
QJ o;
fO
T3 JC
QJ 4J
IS 0

k. en
4^ C
£ -o
•o k.
iT> °
f* 

*o
o


c
o
£3
k.
TS
U
£
£
U1



o
k.
c
5
a.


k.

s

01

I
o
y
e




C
s §
s f-)

£S
oc eg

^
> o
£ -r
c
o
4-»
H3
U
^
VI

§
S
i.
g
^
•O

1

C
k.

on

g
	 i "-
0)
u ~a
f l
S|

^ s
> 6

1. Z



41 go
« S
- QJ
^i £
QJ
Sc
1C

£ -a
e o
UJ 1

QJ
U
^
S

41
j3
i
«•
§
^


U
O
a.

c
•Q
u

to

5

ai
O) -O
5 1
1 §
u —

> i

L. 3C
*i


Oi 0=

I a,
t/1 5

c|

1/1 -o
£ i



i

«w
t/1
1/1
•w
cr
v
c
o
ZJ


1

c
k,



5
-J —

11
i i

i/i
•o *-


k. x:
•o *—
Q

O) CD
^) S
E a;
"t/l 4-1
OJ
§1
1/1 i —

ii i



Lrt
OJ

V
i/>
^t
T3
cr
U">

^


O
a.

o
QJ


k.

.£
t/1
S

1
k.
>

X



^
.


1

X
O



QJ
u
k.
3
o
i/>
c
o







OJ

*J
^

"3
J=
*^

l-
^


V
k.


^
0)
C


o

«5
*










•c


o *-•
, — u
QJ -
> k
-o ^
^
0 *-•
4-1 C
c fl)
QJ E
> QJ

si

o *^

?I


* <

3 4-1

k- O


3 •-*
xT iS -
a> o

*0 4-> ^
4J X tO
Ci J3 -— •


OJ
o
k.
3
O
l/t
>o
QJ
k.
ra
^>

"a

a.
o *-*
> k.
OJ 4-1
"O t/i
J_
0 4-»

C Oj
> 
t/t ^
TJ «Q
QJ 3
U, O

k.


3 4->

i- o
*/> CO

3 «-«
* O
"D t/1 »
o* o

•0 4-* <
^ X ^


o>
(-)
k.
3
o
w^
«o
Ot
k.
•w
U-l

•o

CL

> k.

k.
O 4-«
*J C
C OJ

c cy
si
u
3 X
O 4J
i/i .^
TJ (^
t. O
TJ

•85
4-)
3 *J

k. O
t/i en

3 — •
•O un

4-> LJ
QJ X un
O J3 • — -
§
3
*0
flj
k.

•«
k.

1













J
^

O
c
o

3
a
1/1
QJ
k.
•Q



C
O
Nl

*







1
0

















t/1

OJ
u
•^

k.
CTi
*0
»/1

"oi
>
QJ

b
c
U-

_
a>
•ci
i'

k
T)
' —
i'

















>/t

QJ
u
•o

k.
CTi
o
I/I

OJ
>
QJ

b
0
u.


%
i

L.
i?
•~
c
I!
                               III-4

-------
region should not be made too small.   If the modeling region  was  defined
as only a small fraction of the overall  urban area,  thereby excluding
certain monitoring stations, emission sources, and so forth,  it  is  pos-
sible that the importance of a particular model  input could be missed.
Figure III-l defines the modeling region.  It encompasses nearly  all of
the major emission sources in the SCAB and has sufficient coverage  of the
Santa Monica Bay and Pacific Ocean to contain aged pollutants advected
westward with the nocturnal land breeze.
2.   Multiple Base Cases

     The influence of different meteorologies on the sensitivity study
results was of concern from the outset of the study.  A Los  Angeles  appli-
cation would seem to provide a more extreme test of model  sensitivity to
the availability of input data.  Reduction in the number of  wind monitors
in St. Louis, for example, would probably not degrade the resultant  wind
field as much as would a corresponding reduction in Los Angeles  wind
monitoring stations.  Experience in preparing mixing depths  for  St.  Louis
generally indicates no substantial  difference between the estimated  mixing
depths at each of the reporting sites.  In Los Angeles, the  close proxim-
ity of coastal and desert environs  leads to shallow mixing depths (from
250 to 300 meters) along the coastal margin and to deep mixing depths
(from 900 to 1100 meters) farther inland during the midday hours.  To
investigate whether the selection of different meteorological  episodes
might materially alter subsequent ranking of data needs, both  the 26 June
and 4 August episodes were chosen as base cases.
3.   Simulation Duration
     To keep overall computation costs to an acceptable level,  the
simulation duration was limited to less than one day.   The implications  of
this constraint for the meaning and generality of the  sensitivity results
are discussed in chapter VI.  Analysis of the ozone concentration levels
for 26 June and 4 August indicated that nearly all monitoring stations
observed the peak concentrations well  before 1800 PST.   Therefore, 1900
PST was selected as the time to terminate the runs.  The simulation
starting time Was selected after examination of the emission inventory;  at
0400 emissions are relatively low compared to the levels during the
ensuing morning rush-hour.  It was felt that starting  at 0400 would not
bear adversely on the sensitivity simulation involving  temporal redistri-
bution of emissions.
                                   III-5

-------
FIGURE III-l.   MODELING REGION USED IN THE AIRSHED MODEL SENSITIVITY  STUDY.
               (Calculations not performed for the crossed areas.)
                                  III-6

-------

-------
TABLE III-2.   AIR QUALITY MONITORING STATIONS IN THE SOUTH COAST AIR
              BASIN AND THE NUMBER OF TIMES THEY WERE SELECTED FOR
              INCLUSION IN THE REDUCED AEROMETRIC DATA SET
Station
Downtown Los Angeles
Azusa
Burbank
West Los Angeles
Long Beach
Reseda
Pomona
Lennox
Whittier
Newhall
Lancaster
Pasadena
Lynwood
El Toro
San Juan Capistrano
Los Alamitos
Thousand Oaks
Simi Valley
Santa Paula
Port Hueneme
Number of Times
Selected
8
6
6
1
5
6
2
6
2
3
0
1
0
0
0
2
0
0
2
0
Station
Ventura
Ojai
Pt. Mugu
Redlands
Fontana
Upland
Chino
Hemet
Banning
Pern's
Anaheim
Costa Mesa
Laguna Beach
La Habra
Redondo Beach
Camarillo
San Bernardino
Riverside
Prado Park
El Monte
Number of Times
Selected
2
0
0
1
2
3
1
0
0
1
5
1
0
0
0
2
4
1
4
3
                                III-8

-------
4.   Selection of the Reduced Set of Air Quality Monitors

     As discussed in the following section,  certain simulations involve
the use of a limited set of surface aerometric monitors.   The reduced  set
of 10 surface air quality monitors was selected by the following methodo-
logy.  Eight SAI technical staff members with individual backgrounds  in
meteorology, atmospheric chemistry, emission inventory preparation,  aero-
metric monitoring, modeling, and so on were  asked to identify on figure
III-l the locations of 10 sites at which they would locate aerometric
monitoring stations (measuring 0^, NO, N02,  RHC, S02,  sulfate, wind  speed,
wind direction, and temperature).  Objectives of the site  selection  were
the following:

     >  Determination of the peak basin wide 0^ levels.

     >  Determination of RHC and NOX concentrations in major
        source regions.

     >  Determination of characteristic flow patterns in the
        basin.

The closest existing air quality monitoring  stations to each of the  ten
sites selected by the staff members are listed in table III-?.  Of the
stations listed, the following were used in  analyzing model sensitivity to
aerometric data inputs:

     >  Downtown Los Angeles                >  San Bernardino

     >  Azusa                               >  Reseda

     >  Burbank                             >  Lennox

     >  Long Beach                          >  Prado Park

     >  Anaheim                             >  Upland
     Next, each of the airshed model  sensitivity simulations is identified
in greater detail.
                                   III-7

-------
C.   ATTRIBUTES OF EACH SENSITIVITY RUN

1.   jimulations Focusing on a Limited Number of Aerometric Monitoring
     Stations
a.   Run 1--1974 Upper Air Meteorological Data

     This simulation is intended to address the situation in which upper
air meteorological data acquisition in a city is limited to routine,  twice
daily vertical soundings, customarily performed at airports.  Upper air
winds (about 300 meters above mean sea level) were based on the twice
daily soundings (0700 and 1300 PST) at El Monte.  Mixing depths for the
entire region were based on these two soundings and on the hourly averaged
temperatures at all available surface wind and air quality stations.
b.   Run 2--1975 Upper Air Meteorological Data

     This run is the counterpart of run 1; it focuses on the 4 August 1975
meteorological regime.
c.   Run 3--1974 Surface and Upper Air Meteorological Data

     This simulation is intended to address the situation in which both
surface and upper air data acquisition in an urban area is sparse.  Mete-
orological fields were constructed using data available from the 10 sur-
face monitors identified earlier and from the 0700 and 1330 PST El Monte
soundings.
d.   Run 4--1975 Surface and Upper Air Meteorological Data

     This simulation is the counterpart of run 3; it focuses on the
4 August 1975 meteorological regime.
e.   Run 5.1--Ground Level Initial and Boundary Conditions

     For this situation, initial and boundary conditions for 26 June 1974
were prepared from concentration data available from the reduced set of 10
aerometric stations.  Meteorological fields were those of the June base
case.
                                   III-9

-------
f.   Run 5.2--1974 Reduced Meteorological  and  Air  Quality Data

     The purpose of this simulation was  to examine the  impact on model
predictions of limited data from both  meteorological  and  air quality moni-
toring networks.  Meteorological fields  from simulation 3 were  used  in
conjunction with the initial and boundary  conditions  prepared from air
quality data from stations nearest to  the  10 sites identified in run
5.1.  The 26 June episode was used for this simulation.
2.   Simulations Focusing on More Specialized  Aerometnc  Monitor.ng
     Activities

     In establishing the June and August base  cases,  a  detailed  examina-
tion of ambient air quality data was  carried out.   Among  the data resour-
ces utilized in preparing initial and boundary conditions for  the base
cases were the following:

     >  Monitoring data from the ARB  station  located  atop Mt.  Lee.

     >  ARB early morning hydrocarbon speciation measurements  at
        various stations during the summers of 1974 and 1975.

     >  Air quality observations at monitoring stations located
        upwind and outside of the modeling region.

     >  Special studies data collected in the  course  of airborne
        monitoring over the SCAB (for example  see  Husar et al.,
        1977; Blumenthal, White, and  Smith, 1978;  and Edinger,
        1973).

These information sources, together with the  routine  data gathered on the
days of interest, were of great value in preparing  inputs for  the base
cases.  The existence of all of the above information was not  known, how-
ever, at the time the first base case runs were made.  It was  only through
subsequent diagnostic analyses and further investigation that  certa.n of
the additional information resources  became known.  Thus, though some of
the preliminary attempts at establishing June  and  August  base  cases were
set aside based upon additional information, these  simulations are of
value from the viewpoint of model sensitivity. Five  of these  runs are
presented below.
                                   111-10

-------
a.   Run 6--Upper Air Quality Data

     Run 6 was a simulation of 26 June in which the hydrocarbon  and  NOY
                                                  "ft                   "
precursor concentrations in level three grid  cells  were increased  (rela-
tive to the base case) based upon the detailed aircraft studies  reported
by Husar et al. (1977).  In the absence of upper air concentration data on
26 June, pollutant concentrations aloft could only be speculated.
b.   Run 7--1975 Clean Air Boundary Conditions

     Run 7 differed from the 4 August base case in that lower precursor
concentrations aloft were considered and inflow boundary conditions  near
Thousand Oaks and Laguna Beach-Costa Mesa were estimated to be similar to
"clean air" concentrations.  In contrast, the base case run involved the
use of actual hourly concentration measurements at Thousand Oaks,  Santa
Paula, Simi, Costa Mesa, and Laguna Beach.
c.   Run 8--1974 Clean Air Boundary Conditions

     The initial and boundary conditions differ from the base case in that
lower hydrocarbon and nitrogen oxide concentrations were considered
("clean air" conditions).
d.   Run 9--1974 Assumed RHC/NOV Ratio
                            	  A 	

     Prior to the acquisition of the detailed aerometrk hydrocarbon  data
reported by Mayersohn et al. (1975, 1976)  for the summers of 1974 and
1975, the initial conditions for the 26 June simulations were based upon
an assumed hydrocarbon to NOX ratio of 7 (Reynolds et al., 1979).   Run 9
differs from the 26 June base case in that the former simulation  involved
"clean air boundary conditions" and did not take into account the observed
empirical relationship between RHC and THC derived from Mayersohn's data.
e.   Run 10—1975 Assumed RHC/NOX Ratio
     This run is the August counterpart of run 9.
  The vertical structure of the modeling region consisted of four layers,
  whose combined height was a constant 1000 m above local ground  level.
                                   III-ll

-------
3.   Simulations Focusing on Details in Emissions Inventories.

     Altogether, nine sensitivity runs focused on details in  the various
components of an airshed model  emission inventory.
a.   Run 11--1974 Hydrocarbon Emissions Speciation

     The potential need for detailed hydrocarbon speciation of the major
emission sources (as compared to overall  basin  wide hydrocarbon splits)
was addressed in Run 11.   Table III-.3 gives the hydrocarbon splits used  in
both base cases by source category.

     Composite hydrocarbon splits were developed by determining the per-
cent contribution of olefins, paraffins,  aromatics, carbonyls  and  ethylene
to the total reactive hydrocarbon inventory.  The resultant breakdown
between the various species groupings are as follows:

                                     Percent by Weight
                   Species              (as carbon)

                  Olefins                    5
                  Paraffins                 70
                  Aromatics                 17
                  Carbonyls                  5
                  Ethylene                   3
b.   Run 12—1975 Hydrocarbon Emissions Speciation

     The overall hydrocarbon splitting factors in Runs 11 and 12 were
identical.
c.   Run 13--1974 Mobile Source

     The base case motor vehicle emission inventory employed in this study
was generated by the Direct Travel  Impact Model  (DTIM),  a successor  to  the
Caltrans Link Emissions Model (FWY011).   Advantages of the DTIM code are
its ability to allocate emissions due to cold  starts,  hot starts,  and hot
soaks based on the origin, destination,  and type of trip.  The trip
assignment program accounts for intrazonal  and terminal  traffic volumes in
addition to the link volume records treated by its predecessor, FWYOll.
DTIM uses hot and cold start and hot soak files  for emission estimations
based on the correspondence between the  traffic  zones  and grid cells.
                                   111-12

-------
o:
o
CD
                    or
                    fx
cs
1
LU
V
CO
o cs
1 I
1U LJ
r- CD
to r~
ta
i
UJ
m
CO
n
CS
i
LJ
CM
Is*
C
1
U-J

CJ
CB
l
UJ
E.
CB
rj
Q
i
LU
l£>
LD
G
i
UJ

p-
G
1
LU1
&
r-
G
l
LJ
C,
r-
IN.'
C.
1
LL
r
«r
1_



•"^
  ro
                               OD
o.
oo
                                          LU UJ
                                          cr> «-
                                          r- r-
                                                                            KI
                                                                            K
                                                                    cr

                                                                 s rn
                                                                                                                          E s — c:
                                                                                          P-     r-    «• CO — IT
                                                                                          tr     r-    r- r-  a:
                       LJ
                       in
                       hj
                       o
                                               Z  •=! —•

                                               —  O -I
                                               U.  O D
                                        I z  1  i:
                                                           a
                                                           CJ

                                                           a:
                                                              z
                                                           7- CC
                                                           ox r>
                                                           O QJ

                                                           o: -J
                                                           ^- o i
                                                           cr a:
                                     o 2  _/
                                     ^ o  — "z
                                     c: '^  o c,
                                            co •-
                                                                                        tn
                                                                                        _j
                                                                                        cr
                                                                                        i
                                                                          j  •— c: t_i o:
                                              D-  LT,  (_•  —  LJJ  C"'

                                              i-  £  LJ  c:  y  d

                                              cr  x  —  _j  —  o
                                                     _j  LJ  _j  ~
                                mi—  ^-—-at—
                                                      •— o a. o  a.  a.  o
                                                                                                              oca  ^  —  — o u_ •-
                    O t-
                    
-------
                             r~    r\j
                             cs    &
                                          SI
                                           I
                                          bJ
                                          in
                                                    rg     rj
               ID
               OC
               CT
               LJ
               a a
                 a:
                 O
 a>
 3
 C
 C
 O
o


CB
e
e
L
C
C
e <
&
£
j
2
3
r s
e
r-

LC C
UJ UJ
fB CM
— r-
&. ts in — CB c
s s E:
i i
UJ UJ Ul
w r- CR
a _ _. ,£ 5, -
UJ CC
u r.
a: o
ID UJ
O t-
in cc
u











yr
t^

a
UJ
i

CET
X
(J
o



CC











UJ

—
cr


CL
CL

LLJ
^
LO
LU
a



f~~
ro






CD
bJ
_J
LJ

X
LU

tv

*— a
0 —
X X
in
CC 0
o z
u- r
O -



CD cn
rn m


o


LT

LO

LO
cr
o

a

LJ
LT
cr
or
O
LO
t—
U-'
CL-



C1
fT


C


f_
cr
o


LJ
LJ
a
CC





o
z

t—
CC
0


UJ

CC
CC




-
1
—



^-



z
0

t—
cr

cn

tn
a
to
h-
CT

_J
LJ
^3
U.
CC
X
LU
•



U*l



o
z
-J
LJ


{£
Q

UJ
Z
UJ
>.
x

UJ
o
or
O
r
CC
LU




UD






O


tO
CC
UJ
Ql
O
a


LJ
LD
0
— 1
X
2
O
^y




r*



t~
a

o

LU
(_J
Ct
U-
C£
^>

O
LJ
i —
CC
UJ
a:

a
LJ




CD



LU
\—
CC
UJ
x

a
z
a:

cn
UJ
_j

a
CD

CD
o
X
UJ



cn






Z
o

t—
a
CL
LU
z
\- LJ
LD ™

a
X -I
^ cr
LU —

LJ O
o z
a: -



C —




UJ UJ

0 0
Ct Ci

(/I '-'"'

«*• fv

>- >-
^- «-
__' 	 1

h* *-~


U '-'

LJ UJ
c L:



r^j n~i
iT> LO



Ct
LJ


O
CD

^_
Z
er
Q_


0
LU

tn
— o
in H-
LJ Ct
X 2
o .
a •



V LT>
LO U~)


CC
LJ


O
CD

>—
Z
CC
1
Cl-
3
cr;
CL


^_
o
0
CT.
UJ




LT





CC
LU
_J

C
CC

t~
z
-J
ex

ct

2
c^

a
o




r--






r




cr
o
Cl
LL-
L_
O


^
—
•-
c,
-



CC'






^



o
cz
o
tr
.^
LI


z

—
LJ
E
5



c-




I
LJ


O
a

r-
|
L-
0

_J
r~

3;
"
=
C
-



C;












X
U-1

Q

^
C_
1
LL.
O
I
u_



_
-D




z:
C'


0

c:.
z:
GJ

_j
L-.'
—j


'-
t-
t1
5



r '
Ll



>—

cr


— ~ cr

- — C i.
•- _i LV
c tn c 2
o a_ . ' 2 — CL
L «— « i_ — cr/

.* •_•• IL- cr
U- —! — LJ •— Ll
cr c. cr
_ L_ 3- Cl _J l-
~ .— cr c. — LJ
Zl C UJ OO C1
i- ce — 2 — d:
~j o") cr z • •



p" T- LP 'JT r*~ c^1



^
_J T

C '-


CZ Jj

^ ^
^— L_
<-- <.-
£ H


T- --
C *f
LJ _'

LJ , 	
C O
Z Z
• •



CT1 "H



                                                111-14

-------
(NJ
Si
1
UJ

— '
— EC
(NJ

UJ
O)

2 si — s
K)

UJ
CD
tn
C3 CJ)
en TNJ
Q GJ

UJ UJ
CO LO
fTl tO
(T) cn
s
i
UJ
rn
in
en c
E3

LJ
S3
in
s og
CM — fO
C2 S> D
1 1 >
UJ LJ LJ
h- i i r"
cn TT ro
m jj un
C2 G- £1. C- Q Cj E1 C


^ Cfi 1- f J C C C.
r.g r- r- cr LT r~ r r-

f-j
c


c,
r-

t\J rsi
C S
i i
1 — L_
cr •"
- r-

CO
C3

LJ
C
^
2 r- ts
E

LJ
f\J
—
—
                  cr
                  a
eo EJ s  GJ  eo s
 11        ii

CO rn cc  Psj  LC cr,
tC LT> CD  CC  LD LT>
t» CO Q  CS  S)
 I      til
UJ LU LU  LU  LU
                  O
                  a:
                  a
                                                                     S) (S CB  S-
J UJ
r-

ijj

OD
L-1
CD
fSJ
UJ
CD

LJ U
LO U
LO r
J u.
3 a>
- s
L-1 _-• u

cr. ca u
L. UJ LLJ UJ
T INJ r"' r-vj
n in LO in
u-

tn
UJ
rr,
LI
U-i

U".
L_ LJ '_
CC c- S.
LC cr, ^
                                      CB PsJ (S3  OB  —IS
                                                          CD Si K»  Ci CD a

 3
 C
 O
m
 i
it S)
2 Ci

J LU

•' CT'
- r-j

CO
i
LJ
ID
—
Ki C£i Si S
(S S C2 SI

LJ LU UJ UJ
CO r-si r" LTI
c^j u-j cn cn
csJ n-) — CM e

C3
1
UJ
JD

s cn s> —

i i
aJ UJ UJ UJ
u, r- ro cn

j .-NJ CD r\*  CO
CM
IS
1
LU

-*
— i —• S C: (O
Ou S S G (S1
1 1
LU LJ l_ LJ LJ
cc) s o TJ ;-
r- CM x -j 
bJ O
o
a_ ui
L- Ll,
LJ UJ
cc it:
SE

1B BOILER
D

UJ
O
O
C£
H-
LJ
0-
CO

O
UJ
tn
o
u
,_

z
cr
ttL
UJ
0
tn
r-

t—
CT
Q-
O

C3
ft:
d
O
CD

U
2
CC
5

Z
_J
a
U-
INOUS
CD

o
z
UJ
CO
Z
o

z
_J
a:
<_j
aj
r
x:
o
r\>
CD

MB ENGINES
LJ
f-
O
O
U-
LU
CL
CO

in >-
ILJ Cc:
^ to
(3 o
:c o
UJ 2!
ID
O C
UJ
i- s:
z
I.L- —
uj a:
£t (X
•V LT-
00 CD

. INDUSTRY
CL
UJ
D'
a
0
Lu1
Ln
CU

PROCESSES
y
0
c
L_
LU
Q-
CO

C
LJ
L j
C.
c.
u:
X
e.
L_
L.
ft

CL
cc
LU
c
u
(-
CD
U-
o
1 	
±'
u

UJ
LJ
Q_
—

0
U.
LJ
a.
:

a.
a


CJL
C
LT
U.
L-
C.
C-

a
tn
cr
o

X
U-
;

CL.
IL. CL. a_ cr
cr c ~ > o_
> ^ i- ^_ cr
L- LJ U_' ^
^ '-• r " C
Cu t. L- ~
LJ 1_ LJ CT CC
C C LJ
C 1. ^ U.

•~ _J —) L~ L_
L^ U- L_
t~ t—
— : _J LJ — ' LJ
u. i_ _ i. a.
cr< a. : f ~*

CL
c
t"
LL.
C
o

cr
-
LJ
(L
CT

C_
C
L^1
U
C.
LJ

d
	 t
1 	
LJ
C*
CD

C
LJ
CL
C
t-


i JXdOW
t—
UJ
CL
«-,

a
=:
-
T
UJ
CU


0-
CE
LJ
LC
Z
CL
T*
LU
CL


GuoiNii rjun '
rr
~
C1
t—
uJ
CL
c:

^
L,
L- _
c n
L- ~
O C
^ -
cr a
D a
E£

                                                       111-15

-------


I
t~
UJ


*-*
cs

LU

in
q-
— m

i i
LJ LJ
r\j —
m m
f\j CT>
« —.
S SJ
1 1
LU UJ

.-. «r
— (NJ
m
^

UJ
tr
«»-
f\l C
i -
IS

UJ

'?•
a 'r
__
si

UJ
r\j
r-
m (S
 ea   UJ
                                         —I  1.
                                         O  Q
                                         I/I  a


                                         0  O
er  —
z  oc
LU  Q-

o  o
UJ  UJ

(r  cr  or ex d
                                          -z.
                                          a.
                                                                         0.
                                                                         uj
                                                                         UJ
                                                                         CO
                                       O
                                       ^ (/I
                                                                                      a.
                                                                                      a
                                                            •— U- UJ     Q-
                                                    O —  en  i-
                                                    o o  a  L^
                                                       1C
                                                                 c ^
                                                                 IT, <.-
                                                                 C. =)
o a
or cc

a: a

o o
z z



a a
C^ Q;

(x a

a o
"Z. Z



c cr
_j f
— a
0 3

i z:
r z:

LJ CJ

X
C"
l_ LJ






. a.
K i~
> c
UJ — 3
tv: •-

> »- a

a: o —

. > a
tt
	 1 ^ XZ
	 1 UJ H
— _j a
u. o »--
Q: i

2! LU O

-i : '
-~ c_
CL LjJ
a:
CL —•
LU t

^ lit
f"~ J


1


LU



x

>
C7
_J ~

>-
L.1 **'

O
>— ^
c: c
^ L
IT -_
_ _

L_ -
Q '-

r r

•_> '-_•
* *
cr t'
LJ
f- C1
CI Q_
T CL,

*^- r-
t— O
U LJ

„ —
cr

	 i_

c. t~
LJ '/"
o: ^

a CL
. *-
                                                                        —  if  IX f-
                                                               111-16

-------
Running emissions are allocated based on link volumes of traffic.   The
DTIM code provides estimates of up to 6 classes  of  RHC,  CO,  NOV, SOV,
                                                              A     A
particulates, and fuel consumption.  DTIM estimates emissions  for  eight
vehicle types instead of the four vehicle types  treated  by FWY011.  The
base case motor vehicle inventory is based upon  an  updated (as of  1976)
origin-destination survey.

     For the 1974 mobile source sensitivity simulation,  an outdated motor
vehicle emission assessment model (FWY011) was utilized.  Features  of the
FWY011 Link Emissions, Model include the following:

     >  Emissions are calculated by link.

     >  Emission estimates are provided for five pollutants--RHC,
        THC, CO, NOX, SOX.

     >  Four vehicle types are treated.

     >  Cold starts, hot starts and hot soaks are assigned by  set
        percentage of link volumes.

     >  Intrazonal and terminal volumes are not  taken into
        account.

The latter two features represent deficiencies in the FWY011 approach when
it is compared with DTIM.  Emissions due to cold starts  should be  assigned
to the grid of origin.  Emissions due to hot soaks  should be assigned to
zone or grid of destination.  Inputs to the FWY011  code  included 1967
origin-destination data projected to 1974 assumed population growth condi-
tions.

     In essence, the major differences between the  two motor vehicle
inventories is that the sensitivity run inventory is based on  outdated
origin-destination data and does not adequately treat trip end emissions.
d.    Run 14—1974 Gas Sales

     The objective of this sensitivity run was to examine the effect  of
using data related to vehicle fuel  mileage to estimate daily emissions of
major pollutants from on-road motor vehicles.  Although these estimates
should be less precise than those derived from sophisticated motor  vehicle
emission models, such as the DTIM,  emission estimates derived from  fuel
mileage are usually much easier to  calculate.
                                    111-17

-------
     The estimation of motor vehicle emissions for Run 14 was based on
three major parameters:

     >  Vehicle fuel mileage

     >  Vehicle emission factors

     >  Area fuel sales.
1)   Vehicle Fuel Mil

     Fuel mileage in miles per gallon (mi/gal)  was derived from the EMFAC5
computer program.  EMFAC5 has been developed by the ARB to estimate motor
vehicle emission factors and fuel mileage,   The EMFAC5 program is similar
to the EPA's MOBILE .1  program, but it includes, among other things, the
motor vehicle emission standards applicable to  vehicles sold in Cali-
fornia.  EMFAC5 was used to estimate average fuel  mileage values for both
gasoline and diesel vehicles.  A mix of the following vehicle types for
the year 1974 was used in the calculations:

                                                   Percent of
                       	Vehicle Type	Total Vehicles

         jjasoline
                       Light duty automobile          80.4
                       Light duty truck               12.1
                       Medium duty truck               1.4
                       Heavy duty gasoline             2.5
                       Motorcyle                       1.1
         j)iese1
                       Heavy duty diesel               2.5

The vehicle mix percentages listed above were developed for the SCAB by
Caltrans.  Based on this vehicle mix and other  assumptions described below
for EMFAC5, average gasoline and diesel vehicles had the following fuel
mileages:

     >  Gasoline:  12.73 mi/gal

     >  Diesel:     5.5 mi/gal.
                                   111-18

-------
2)   Vehicle Emission Factors

     Emission factors in grams per mile (g/mi)  were derived  from  EMFAC5
for hydrocarbons (HC), carbon monoxide (CO),  and  nitrogen  oxides  (NOX).
In addition to the mix of vehicles identified above, the following  EMFAC5
assumptions were made to calculate emission factors:

     >  Average vehicle speed: 19.6 miles per hour.

     >  Percentage of cold start, hot start,  and  hot stabilized
        operation: 21%, 27%, and 52%, respectively.

     >  Anbient temperature: 75° F.

The assumptions for vehicle speed and percentage  of vehicle  operation  in
different modes were based upon the Federal Test  Procedure for motor vehi-
cles.  If better data in these areas had been available, it  might have
been possible to calculate improved motor vehicle emission estimates.

     Emission factors were calculated for cold start, hot  start,  and hot
stabilized operation as well as for crankcase,  diurnal,  and  hot soak emis-
sions.  EMFAC5 also provided a non-methane hydrocarbon (NMHC)  to  total
hydrocarbon (THC) ratio of 0.9496 for gasoline vehicle exhaust emis-
sions.  Using all of these assumptions, the following emission factors
were calculated from EMFAC5:
     Gasoline vehicles
       Exhaust + crankcase, g/mi

     Diesel vehicles       «
       Exhaust, g/mi

     All  vehicles
       Di urnal, g/day
       Hot soak, g/soak
                                     THC
4.41
9.42
6.63
                                               Emissions
        NMHC
 CO
NO,
6.52    6.19    69.92     3.69
34.08    21.29
  These are emission factors, per vehicle, for the two types of evapora-
  tive HC emissions.  Diurnal emissions cover evaporative HC during
  expansion of vapor in the fuel tank with changes in daily ambient
  temperature.  Hot soak emissions include vapor loss from the  carburetion
  system after completion of the vehicle trip.  These HC emissions must be
  added to the exhaust emissions to get total motor  vehicle emissions.
                                   111-19

-------
3)   Area Fuel Sales

     Estimates of the sale of motor vehicle fuel for a specific portion of
a state can be difficult to calculate.  Recognizing the need for such
estimates, Caltrans has compiled gasoline sales data by county for various
years.  The county gasoline sales data have been developed primarily from
vehicle registration data.

     Initially, we attempted to construct fuel sales data for the SCAB.
Because the SCAB contains portions of counties, and because information
pertaining to the SCAB portion of county vehicle miles traveled (or vehi-
cle registrations) was not readily available, fuel sales were estimated
instead for the Los Angeles Regional Transportation Study (LARTS) area.
Moreover, fuel sales were calculated for a modified LARTS area by deleting
Ventura County and estimating sales for the following counties:

     >  Los Angeles

     >  Orange

     >  San Bernardino--LARTS portion only

     >  Riverside--LARTS portion only.

Caltrans also provided 1974 data indicating that 72 percent and 77 percent
of county vehicle miles traveled occurred in the LARTS portion of San
Bernardino and Riverside counties, respectively.

     No information was available to differentiate average weekday sales
from weekend sales, so average daily gasoline sales were calculated by
dividing the annual data by 365.  State wide information was available,
however, from the California State Board of Equalization (CSBE) to season-
ally adjust the sales data to June 1974.  Therefore, on the basis of these
data, the estimate of gallons of gasoline sold on an average June 1974 day
in the modified LARTS area was 12.31 million.

     County wide sales data for diesel fuel were not available for this
study.  However, the CSBE estimates total state wide diesel and compressed
natural gas (CNG) sales on a monthly basis.  No precise information was
available to subtract CNG sales from these figures, but the CSBE estimated
CNG sales to be less than 5 percent of the total figures (West, 1979).
Thus, no adjustment was made to these sales data to account for CNG sales.
                                 111-20

-------
     To estimate 1974 diesel fuel sales in the modified LARTS area, the
following equation was used:

             1974 daily gallons of diesel  in modified LARTS =
          1974 annual gallons of  /   1974 daily gallons of
             diesel  in state    / \gasoline in modified LARTS*/    n n
                19/4 annual gallons of gasoline in state^ " '


This number was then seasonally adjusted to June 1974 using the CSBE
diesel sales data described above.  On the basis of these data, the esti-
mated gallons of diesel sold on an average June 1974 day in the modified
LARTS area were 1.178 million.
4)   Diurnal and Hot Soak Emissions

     Two additional numbers, vehicle registrations and trips, were needed
to calculate HC emissions from vehicles during diurnal and hot soak condi-
tions.  Total automobile and truck registrations were calculated from  1974
county registration data compiled by Caltrans and the California Depart-
ment of Motor Vehicles, for use with the diurnal emission factor.  Cal-
trans also provided the percent of vehicles registered in the LARTS por-
tion of San Bernardino and Riverside counties.  Using these data,
6,149,152 vehicles were registered in the modified LARTS area in 1974.

     Caltrans has estimated that 24,211,000 vehicle trips occurred daily
in the LARTS area in 1974.  Adjusting this number on the basis of vehicle
registrations to account for trips occurring  in Ventura County, 23,218,000
daily trips were estimated to occur in the modified LARTS area in 1974.
5)   Summary of Emissions Based on Fuel Sales

     On the basis of the data previously discussed, the following motor
vehicle emission estimates were calculated for June 1974 in the modified
LARTS area.
  Not seasonally adjusted.
                                  111-21

-------
                            Emissions,  tons/day

                             THC        1,390
                             NMHC       1,330
                             CO        12,310
                             NOX          790

The grid used for this air quality modeling study is a subset of the modi-
fied LARTS area.  A comparison of the motor vehicle emission estimates
derived from fuel mileage for the grid area to those derived from DTIM is
given below.

                                       Emissions, tons/day
                                      RHC        CO      N0y

            Fuel mileage technique    1,260    11,890    760
            DTIM                        850     8,250    720
e.   Run 15—1975 Gas Sales

     The motor vehicle inventories for runs 14 and 15 are identical.
f.   Run 16—Point Source

     The baseline emission inventory supplied by the ARB contained power
plant emissions for a "typical summer week day".  However, in preparing
inputs for the 26 June and 4 August base case simulations, we deleted the
nominal emission rates and, in their place, substituted the actual hourly
emission rates estimated by the Southern California Edison Company and the
Los Angeles Department of Water and Power.  Power plant emissions for this
sensitivity run were based on nominal annual emission rates compiled by
the District (SCAPCD, 1976) and were temporally distributed according to a
diurnal profile developed by General Research Corporation (GRC, 1974).
g.   Run 17—1974 Area Source

     The objective of this simulation was to examine the need for spa-
tially resolved area and point source emission data.  A total of 323 point
sources are included in the baseline inventory.  Sixty-nine of these are
thermal electric power plants.  These sources are identified by UTM
  Reactive hydrocarbons.
                                     111-22

-------
coordinates and various stack parameters.  The  rest  of  the  emission
sources in the inventory, whether they are line sources,  small  point
sources, or diffuse area sources, are allocated according to  a  25  km^  grid
mesh.

     In this simulation, emissions from the following sources were
unchanged from the base case:

     >  All 323 elevated point sources

     >  All on-road motor vehicles

     >  All natural or geo-genic sources.

The emissions from the remaining sources were aggregated  on an  hourly
basis and allocated to the grid based upon the  1975  demographic distribu-
tion.
h.   Run 18—1975 Area Source

     The area source emissions files  for runs  17  and  18  were  identical
i.   Run 19--Tempora1 Resolution

     In this study, the temporal distribution  of  categories  of  area
sources was varied to examine the change  in overall emission  patterns  and
resultant air quality modeling concentrations.  For each  of  the  source
categories constituting the area source emission  file,  a  temporal  distri-
bution of diurnal emissions was estimated from data sources  described
below.
1)   Temporal Distribution of Existing Emissions  Inventory

     The area source portion of the Los Angeles Air Quality Maintenance
Plan (AQMP) -emission inventory provided to SAI for use  in this  study  had
been disaggregated into categories of emission sources  (CES) with  assigned
CES numbers.  A system of codes was established to readily identify the
temporal distributions used to resolve daily emissions  into specific  hours
  The spatial resolution of area sources was unchanged from the  base  case
  for this run.
                                 111-23

-------
of the day.   Four coding series were used to  identify  the  applicable
temporal distributions of emissions in the AQMP inventory.

     >  0,1,2,...22,23 refers to the individual clock  hours of  the
        day;  for example, 13 means 1300 to 1400.

     >  Numbers in the 40 and 50 series are derived  from the  equa-
        tion:

            40 + n = 0700 + lOOn from 0700 to  the  final time.        (3-2)

        For example, 41 means 0700 to 0800 and 50  means 0700  to
        1700.

     >  Numbers in the 70 and 80 series are derived  from the  equa-
        tion:

              70 + n = 0.5(100n) equally distributed  around 1200.     (3-3)

        For example, 78 means 0800 to 1600.

     >  98 refers to a distribution over each  of the 24 hours of
        the day.

In addition,  each of the hours that is assigned emissions  by  an individual
code receives an equal distribution of the total emissions.

     The AQMP temporal distributions represented by  the codes were  deter-
mined from survey responses of sources in the  Los  Angeles  area  and  from
assumptions regarding typical source operations.   As seen  in  table  III-4,
some source categories within the AQMP inventory had emissions  distributed
in only one manner (e.g., CES number 38), though other categories had
emissions distributed by more than one distribution  (e.g., CES  numbers 121
and 19).  Furthermore, many categories had as  many as  12 applicable tempo-
ral distributions, presumably as a result of  responses to  the source
survey.  However, the AQMP inventory provided  to SAI did not  differentiate
the distributions that had been assumed for a  source category from  those
that had resulted from the source survey.
2)   Temporal Distributions Developed by SAI

     For each source category, an assumed temporal distribution  represen-
tative of a summer season was chosen.  The principle behind  each choice
was to select one specific distribution for a source category  that was
  For purposes of this discussion, "0700" and  "0700 to 0800" mean  one  hour
  of emissions beginning at 0700 and ending at 0800.

                                  111-24

-------
        TABLE III-4.  CATEGORIES OF EMISSION SOURCES  AND  TEMPORAL  DISTRIBUTIONS
                                               CES
                                              Number

   Emission	    1

      Transportation	  130

      • Motor Vehicle	    2

      •  • Catalyst gasoline exhaust	   14
      •  • Non-Catalyst gasoline exhaust	   29
      •  • Gasoline evap. loss carb		   23
      •  • Gasoline evap. loss fuel tank	-  122
v     •  • Gasoline crankcase	   21
      •  • Diesel exhaust	   34
*     •  • Diesel evaporative	   37
      •  • Motorcycle exhaust	  123

      • Off Road Motor Vehicle	   38

      •  • Industrial	   58
      •  • Construction	   59
      •  • Recreational	   60
      •  • Farm	   61

      • Shipping	    3

      •  • Purging	   30
      •  • Off loading--	-	-	--   33
      •  • Ballasting		-   28
      •  • Transit	--   39
      •  • • Boilers non-tankers	  119
;     ... Boilers tankers	  120
      •  • • Pleasure craft	  121

      • Railroad	    4
.?
      • Aircraft	    8

      •  • Jet exhaust	   20
      •  • Jet fuel evaporation	   63
      •  • Piston exhaust	   19
      •  • Piston fuel evaporation	  129
      •  • Rocket	   50

      Stationary	   65

      • Petroleum	    6

      •  • Production	   13
      •  • • Ext. combustion boilers	   78
   AQMP
Distribution
  Code*

   98(56)

     §

    **
  Assumed
Distribution
   Code-

    112
   47(100)
     49
   78(100)


   78(100)
     50


     78
   98(100)

   98(100)
   98(100)
84(67),78(29)

   98(100)
     98

     98
     98
     82

     98
    **
                    **
   51(85)
     51
   98(82)
     98
                                            111-25

-------
                          TABLE III-4 (continued)
                                      CES
                                     Number
   AQMP
Distribution
  Code*
  Assumed
Distribution
  Codet
• • Int. combustion engines	-  83             98(100)            98
• • Industrial processes	-  87             98(96)             98
• . Seeps	-	 118             98(100)            98
• • Crude oil evap. fixed roof	  88             98(99)             98
• • Crude oil evap. floating roof	  89             98(100)            98
• Refining	  12
• • Ext. combustion boilers	  77
• • • Boilers residual  oil	-  73             98(100)            98
• • • Boilers distillate oil	  74             98(94)             98
• • - Boilers natural  gas-	-  75             98(83)             98
• • • Boilers process  gas	  76             98(100)            98
• • Internal combustion engines	  84             98(100)            98
• • Industrial processes		  90             98(75)             98
• • Storage evap.		  91             98(94)             98
• • • Crude oil evap.  fixed roof	  93             98(100)            98
• • . Crude oil evap.  floating roof--  95             98(100)            98
• . . Gasoline evap. fixed  roof	  92             98(94)             98
• • • Gasoline evap. floating roof---  94             98(100)            98
• Marketing	—  10             98(99)             98
• • Storage evap.	-	— 100             98(64)             98
• • • Crude oil evap.  fixed roof	  97             98(100)            98
• • • Crude oil evap.  floating roof--  99             98(100)            98
• • • Gasoline evap. fixed  roof	  96             98(83)             98
• • • Gasoline evap. floating roof---  98             98(100)            98
• • Loading and Unloading	— 103          98(49) ,78(16)        112
• • • Gasoline evap.	 101          98(59) ,78(26)        112
• • • Crude oil		-	102             98(62)            112
• • Underground storage @ stations—  40             52(99)             52
• • Vehicle refueling  @ stations	  45

Commercial & Institutional-	   7             98(100)           112

• Internal combustion  engines	  82             98(100)            98
• Ext. combusion boilers &  space heat 124             98(100)            98
• • Residual oil		125             98(82)             98
• • Distillate oil		126             98(70)             98
• . Natural gas--	- —	 127             98(78)             98
• • Process gas--	 128             98(100)            98
• Printing	-	 113     98(23) ,78(23) ,56(23)      112
• • Flexigraphic	-	- 112     78(26) ,98(18) ,56(13)      112
• • Gravure	 114     98(42) ,78(17),54(17)      112
• Surface coating  air  dried achit.—  16
• • Oil  base including  solvent	- 110             78(100)            50
• • Water base	-	Ill             78(100)            50
• Dry Cleaning			  22
• • Petroleum base perchlorethylene--  46  78(22),49(12),47(11),44(10)   50
• - Synthetic		  43  78(23) ,47(15) ,46(11) ,44(11)   50
• Degreasing		-	  11        78(44) ,56(22)          101
• . Halogenated			  42        78(26) ,56(22)          101
• • Non-Halogenated	  47        78(23),56(12)          101
                                    111-26

-------
                            TABLE  III-4  (continued)
                                         CES
                                        Number
   AQMP
Distribution
  Code*
  Assumed
Distribution
   Code-
.  Industrial			    5          98(35) ,78(20)        112

.  .  Internal  combusion  engines	   81             98(71)            112
.  .  External  combustion boilers  & heaters  49
.  .  .  Residual  oil				-   69             98(64)            112
.  .  .  Distillate oil	   70             98(60)            112
.  .  .  Natural  gas	   71             98(66)            112
.  .  .  Process  gas	   72
.  .  Chemical				   15      98(36) ,56(12),78(10)      98
.  .  Metallurgical	   35
.  .  .  Primary  Metals	   85             98(53)            112
.  .  .  Secondary Metals	   86             98(62)            112
.  .  Mineral	   31          78(34),98(23)        112
.  .  Wood Processing--			--   25
.  .  Elec.  generation  boiler	   18
.  .  .  Residual  oil	   56
.  .  .  Distillate oil	   67
.  .  .  Natural  gas	   55
.  .  .  Process  gas	   68
.  .  .  Coal	   57
.  .  Elec.  generation  Inter. Comb.	   79             98(56)             98
.  .  Surface coating	   44          98(60) ,44(40)        101
.  .  .  Heat treated	   48             78(27)            101
.  .  .  Air dried		—   41
.  .  .  .  Paint	-		   80   44(18),78(1 5) ,43(14),42(12) 101
.  .  .  .  Varnish and Shellac		104   78(25),98(18),46(11),44(11) 101
.  .  .  .  Lacquer	  105          78(28),44(14)        101
.  .  .  .  Enamel			106      78(28),44(12),42(10)     101
.  .  .  .  Primer	  107          78(35) ,56(16)        101
.  .  .  .  Solvent	  108          78(43) ,98(15)        101
.  .  .  .  Adhesives	  109          78(26),56(14)        101
.  .  Incineration	   51   43(18),42(18) ,41 (14),78(11)  52
.  .  Land fills	117             98(100)            98

.  Agricultural			    9

.  .  Agricultural  control  burn	   17
.  .  Vegetative  forest  & citrus	115             98(100)            98
.  .  Animal  wastes			116             98(100)            98
.  .  Pesticides--				   24             52(100)            50
.  .  Food processing			   32          98(38),78(19)        112
.  .  Orchard heating	   36
.  .  Waste burning or wildfires	   27
.  .  Wine processing	   66

.  Domestic	   54

.  .  Solvent use	   26             52(100)           134
•  .  Utility equipment  2 stroke	   53
                                      111-27

-------
                                TABLE 111-4 (concluded)
                                          CES
                                         Number

         Utility equipment 4 stroke	   52
         Fuel  combustion		-   62
         Structural  fires	   64
   AQMP
Distribution
   Code*
  98(100)
  Assumed
Distribution
    Codet
6,7,8...21,22
*  Temporal  distribution codes provided with the emissions  inventory.  Numbers
   in parentheses represent the percent of the total  number of grid cells that
   have emissions (for that source category) temporally distributed by that code;
   only those codes and their corresponding percentages most responsible for
   distributing the category's emissions are listed.   See text for explanation
   of codes.

+  Temporal  distribution codes assumed by SAI.  See text for explanation of codes,

§  No emissions associated with that CES number were provided in this  portion  of
   the emissions  inventory.
**
   The temporal distribution of emissions provided with the emission inventory
   was used for the assumed distribution.   For CES #2,  each of the 24 hours
   received almost equal  weighting (i.e.,  about 4 percent).  For CES #20,  each of
   the 24 hours received  some weighting;  from 0700 through 2200 inclusive, each
   hour received almost equal weighting (i.e., about 6  percent).
                                          111-28

-------
typical of the normal operating schedules of that source type.   In  other
words, an individual  who was familiar with typical  source operations,  but
was lacking site-specific operating data, would be expected to  chose
temporal distributions similar to those chosen by SAI  for this  study.   For
some source categories, no change in source operation  was assumed (e.g.,
CES number 59).  However, the majority of sources were assumed  to have a
single temporal distribution different from the distributions used  in  the
AQMP inventory (e.g., CES numbers 121 and 19).

     Three new temporal distributions were developed to account for the
complex nature of many source operations.  These distributions  were used
to accommodate the differing operating schedules of individual  sources
within a category.  The three new distributions and their codes,  which
were developed by SAI, refer to the following hours and weighting of those
hours.

     >  101 refers to a distribution of hours from 0700 to 2200;
        each hour from 0700 to 1700 receives twice the weighting
        of the hours from 1700 to 2200.  Thus,

        -  0700 to 1700:  each hour receives 8 percent of the
           emissions.

        -  1700 to 2200:  each hour receives 4 percent of the
           emissions.

           All other hours receive no emissions.

     >  112 refers to a distribution over each of the  24 hours  of
        the day; each hour from 0800 to 1600 receives  twice the
        weighting of the hours from 1600 to 0800.  Thus,

        -  0800 to 1600  :  each hour receives 6.25 percent of the
           emissions.

        -  1600 to 0800:  each hour receives 3.125 percent of the
           emissions.

     >  134 refers to a distribution of hours from 0700 to 2100;
        each hour from 0700 to 1800 receives twice the weighting
        of the hours from 1800 to 2100.  Thus,

        -  0700 to 1800:  each hour receives 8 percent of the
           emissions.
                                   111-29

-------
        -  1800 to 2100:  each hour receives 4 percent of the
           emissions.

        -  All other hours receive no emissions.

     The distributions chosen for each source category were based  on four
primary sources of information:

     >  The temporal distributions used in the AQMP  emission
        inventory provided to SAI were partially  developed from
        responses to a survey of sources in the Los  Angeles
        area.  From this inventory, as indicated  in  table III-4,
        SAI calculated the percent of the total number of grid
        cells that had emissions.  These were calculated for each
        source category that was temporally distributed by a par-
        ticular code.  This information was useful  in estimating
        the portion of a category's sources that  operated on a
        particular schedule.

     >  Temporal distributions that were developed  by the ARB for
        a study in the Sacramento, California area  provided addi-
        tional information on common operating schedules for many
        categories of sources (Reynolds et al., 1979).

     >  The Regional Air Pollution Study (RAPS) (Littman, 1978)
        provided temporal  distributions for selected categories of
        area sources.

     >  Area source distributions used during the development of
        the Sacramento AQMP were also useful for  comparison purpo-
        ses.  These temporal distributions were derived from sur-
        vey responses and  estimates of typical source operation
        (Skelton et al., 1977).

In addition, SAI has been  developing a comprehensive emission inventory
for a modeling study in central  California.  Temporal distributions deve-
loped in conjunction with  that study also assisted  this effort.
4.   Simulations Focusing on Model  Grid Mesh Configuration

     Three simulations were carried out to explore the impact of model
configuration on predicted oxidant  levels; these are discussed briefly.
                                    111-30

-------
a.   Run 20—10 km

     For this simulation, all gridded model  inputs were averaged to yield
10 km spatial resolution instead of the 5 km resolution of the base case.
b.   Run 21—2-Layer Model

     In this run, the vertical resolution of the model  consisted of two
layers separated by a temporally and spatially varying  inversion base.
c.   Run 22—1-Layer Model

     In this run, the vertical resolution of the model consisted of one
single layer below the inversion base.
D.   CONCLUDING REMARKS

     This chapter identifies the 22 model sensitivity simulations investi-
gated in the present study.  Most of these simulations correspond to those
originally identified in our preliminary project work plan.  Some do not,
however.  In certain instances (e.g., simulations involving 2.5 km grid
resolution or motor vehicle emissions based on inadequate transportation
models), simulations were not carried out either as a result of a direc-
tive from the EPA Project Officer or because compilation of the requisite
input files was too expensive.

     Some simulations discussed in this report were not originally contem-
plated.  They arose from efforts to establish suitable June and August
base cases.  Previous experience in developing acceptable base case
results in several urban areas—Los Angeles, Denver, St. Louis, Sacra-
mento, and Las Vegas—indicates that an iterative process of input
preparation, model simulation, and diagnostic analysis is generally neces-
sary.  When simulation results fall short of the model performance goals
established in the evaluation effort, additional analyses are performed to
ascertain the sources of inadequate performance.  Often, information is
                                   111-31

-------
discovered that was not initially considered in preparing model  inputs.
Thus, the base case simulation is generally preceded  by several  model
runs.  These intermediate simulations may be regarded then,  iin some sense,
as based on a lesser level  of detail  in input information.  Consequently,
a few of the runs performed in the model  evaluation effort have been
included in the sensitivity analysis  presented in later chapters.
                                   Ill  32

-------
                        IV   BASE CASE SIMULATIONS
A.   IDENTIFICATION OF ANALYSIS PROCEDURES

     Several model performance measures and graphical procedures have been
developed to display and interpret the airshed model simulation results
(Hayes, 1978; Mil Iyer et al.  1979).  In this section analytical procedutes
that are most informative in  evaluating model  performance in simulating
the base case results are identified.

     Because of the vast amount of output information available from a
grid model simulation, the performance of the model may be evaluated from
a variety of perspectives.  Hayes  (1978) reported on a detailed examina-
tion of candidate model performance measures for air quality dispersion
models.  Five attributes of desirable model performance were identified.

     >  Accuracy of the calculated peak concentration

     >  Absence of systematic bias

     >  Lack of gross error

     >  Temporal correlation

     >  Spatial alignment.

     The accuracy of the calculated peak concentration can be evaluated in
different ways.  The observed peak concentration can be compared with the
highest calculated value at a specified monitoring station, or the calcu-
lated peak value can be taken as the highest value calculated at any one
of the monitoring sites, or even as the highest value in any ground-level
grid cell.  The basis for comparing observed and computed peak concentra-
tions must therefore be specified.  Here the focus is on the correspon-
dence between the peak computed concentrations and observed concentrations
at each monitoring station.

     Absence of systematic bias means that a model does not consistently
underestimate or overestimate pollutant concentrations.  The presence of
systematic bias can be inferred qualitatively by plotting pairs of
                                    IV-1

-------
calculated and observed concentrations on a scattergram plot.   If  the
locus of the prediction-observation pairs falls  along a 45°  line  (the  so-
called perfect correlation line),  the absence of systematic  bias  in  model
calculations is indicated.  If the locus of the  points falls above or
below the line, a systematic bias  toward underestimation or  over-estimation
is suggested.  Obviously, it is quite possible for  a model to  exhibit  a
bias toward overestimation in a particular concentration range and toward
underestimation in a different range.  The estimates of systematic bias
are calculated in the following manner.

CALCULATION OF SYSTEMATIC BIAS:
                     Mean deviation = 77
                                     i
         Mean normalized deviation = TT  >    —=-*7	^^-    ,     (4-lb)
where C- and Cm are the computed and measured  concentrations,  respec-
tively, and N is the total number of comparisons.

     Continuing with the scattergram concept,  the  absence  of gross  error
can be determined by the "dispersion" of the prediction-observation pairs
about the perfect correlation line.   If the mean absolute  deviation of  the
pairs about the perfect correlation  line is small,  the model  is  said to
exhibit "skill" as a predictor.   If  the mean absolute  deviation  is  large,
the model suffers from the presence  of large gross  errors. Both the mean
signed deviation (a measure of systematic bias) and the mean absolute
deviation (a measure of gross error) can be determined as  a function of
concentration level.  These measures can be presented  either as  quantities
normalized by the observed concentration level or  as nonnormalized
values.  The mean absolute deviation and the mean  normalized absolute
deviation are given by

CALCULATION OF GROSS ERROR:


                                              N
                Mean absolute deviation
= 1  V  1C      C
  N  Z-r  1Lc,i    Cm,i
                                              N   1C     -  C   I
     Mean normalized absolute deviation = 77  /    —^	^^—    ,   (4-2b)
                                   IV-2

-------
     Temporal  correlation refers to the  "timing"  or  "phasing" of the
observed and computed ozone levels  at  a  specified station.  The temporal
correlation for a given station can be determined by using the pairs of
hourly observed and calculated  concentrations  to  define daily mean
values.  A correlation coefficient  can then  be calculated according to
routine statistics.  Lack of temporal  correlation can  be ascribed to one
or more causes, including inadequate characterization  of emissions, wind,
or mixing depth inputs.

     To calculate an average temporal  correlation coefficient, p  , we
perform the following change of variable (Hoel, 1962):


                               1    l  +  ri
                          *  =    *n   ~      '                     (4"3)
where r^ is the computed correlation  coefficient  for  Station j on the
basis or hourly pairs of predictions  and  observations.   Next, the mean
value of the <|>.'s is estimated  from
              J
                                   M
                             - ff E
where M is the number of monitoring stations.   Since  the  values of  <(>,• will
be approximately normally distributed,  it  can  be  shown  that
                                                                    (4-5)
where p is the average value of the temporal  correlation  coefficient
Thus, p can be determined from the following  equation.
TEMPORAL CORRELATION COEFFICIENT:
                          P =         -       .                       (4-6)
     The spatial  alignment of observed and  calculated  concentration fields
is another useful  measure of model  performance.   For a given hour, imagine
two concentration  isopleths, one constructed  from observed  pollutant
                                   IV-3

-------
concentrations and the other from the corresponding model  calculations.
If one isopleth were placed over the other,  the degree of  spatial
misalignment would be easy to discern (at least qualitatively).   Spatial
alignment can be quantified by considering a sequence of  "time  slices."
For instance, for a particular hou"% mean values of calculated  and
observed concentrations can be computed from the ensemble  of monitoring
stations.  Spatial correlation coefficients  can then be computed for each
hour according to routine statistics.  Common sources of  spatial
misalignment include discrepancies between the modeled and observed  wind
velocities, inaccuracies in the emission inventory, and the treatment of
photochemistry.  Estimation of the average spatial  correlation
coefficient follows the procedure describe*  above.
B.   BASE CASE SIMULATION RESULTS
1.    Accuracy of Computed Peak Concentrations

     The simplest comparison that, can be made between computed and
observed ozone concentrations involves the maximum hourly average  values
at each monitoring station.  In table IV-'J, the peak one-hour-average
ozone predictions are compared with the h'ghest observed  concentrations
for the 4 August 197!) and 26 June 1974 Los Angeles simulations.  The times
at which the peaks occur do not necessarily coincide.  Included  in the
table are the percentage differences between computed and observed values
for each of 17 monitoring stations.  The percentage difference varies from
-53 percent to +500 percent.  As noted in the table, the  absolute  concen-
tration level of the prediction-observation pair should be considered when
examining percentage differences.  Percentage differences are useful
measurements at high concentration levels; absolute differences  are more
relevant at low concentration levels.

     A more useful comparison of the peak concentration predictions can  be
made by considering only those values above a particular  level.  For exam-
ple, in table IV-2 the average percentage differences between peak compu-
ted and observed concentrations are presented for:

     >  All stations reporting peak ozone concentrations  greater
        than or equal to 2 pphm.

     >  All stations reporting peak ozone concentrations  greater
        than or equal to 12 pphm (the National  Ambient Air Quality
        Standard).
                                   IV-4

-------
OJ
O *""^
c 4->
QJ C
U OJ
QJ U
•4- J-
•4- QJ
•«- ex
0 •-'
QJ
u
m c '— -
£ ¥ 1
CT> L. f
<^> ^ £! a"
2; vt- Q.
O 4-> V>- *•— ••
IH 30




^H *^ ix^ CTI in in CD in ro CM in
rofM CMCM«-«- E1
o *- -c
Z OJ CX
o «" °-
° o


roino>r^«-CMa>«-inro^-
«— 1 CM t— t CM CM «— 1 CM


ro
o
f^\ CU ^""**
£ SI
C£ Q. CL
LU E CX
CO O • 	
CO CJ


cncr,oinro«vvorof-com
«-l CM •-< CM «— 1



o
O
Z OJ <-*
to: o 4J
c c
O OJ OJ
UJ i. O
(— OJ J-
1D U- OJ
£ £,5
o o





fx. i^> ^o o O co itf o "^ CM ro
CM ro •— < «• ro r— t o CM »-H co
i in i


0
=3 0
Z «*• c -—*
•— i r»« OJ E
x cr* y f^
< «— I OJ O.
£ vv ex
OJ ««-'—'
LU C '•-
O 3 O



ro co ro co CM ^^ ^f in in ro in





rs
Oi VO "D
S CM QJ •*-»
! -^
»_> OJ ^^
CO i^ CX
X3 *— •
o

- QJ .^
*- 4-> E
uu at.
~J ED.
CO 0 «—
2 *->


t— i ro CM «-iro t-i CM f— i



t—




en
c c
f- O
t~ f
O 4->
4-> 10
•r* +*
c ns>
o
aE



QJ

QJ
CT>
1^
^t _f
u
C l/> 10 4_ TO
r .X O OJ QJ C T3
O C i gi^ TO T? >C ••" OJ O
4-» TO (O •DCO4-»TDp
§t/1jD4->CT>QJQC+JTOi
31-VICl^fcC'<-l/IC
ONI3QJOQJOQJjrTO>-,
O i
ro ri CM c-i i i CM ro t-i i
i i i i i i i i i






•— I VD VD CO 1 1 ^- CT> in 1
1— 1 II 1






ft r~i r^ ro i i VD CO r^ i
?o ro CM CM i i i— t CM CM i





; 3 in •— < o f)^ r*> CM cri CM i
CM CM CM CM »-H «— 1 CM 1









'™- VD C5 1 ^^ C5 '"•^ CO rO ^^
vroro i«-fMrocMCMro
i i i i i i i i






*r cr* co icTicM^-covDO
i— t i t— i







•o-mr^ icr>orocr t.
l» •»- 1C
•o E m c
H3 |^L C^ 'O "C^ ^K ft)
C TO t- f— C QJ QJ
TO C O O O «T TO CD _J
4-» «O C TJ t— •»- t—
C r— -t- /O COET3C •

U.:3OCl-UJ_JCOQ£l/OZ-





















..
CO

C
QJ

L.
QJ
O.

OJ
JZ •
»- »«
8
• .— t
TO
4-> K
TO
T3 *"™^
T3
O- OJ
C >
••- u
in QJ
1/1 i/i
• •*— r^ *-^.
E 0 -0
«M^ Qj
I/I >
QJ 1 U
4-> QJ

O TD JD
••- OJ O
•O 4J • 	
c u
T3
JC QJ
I/I L.

-------
          TABLE IV-2.  SUMMARY OF MODEL PERFORMANCE IN COMPUTING
                       PEAK OZONE CONCENTRATIONS

(Average absolute percentage difference between computed and observed values*)

                                    	QKidant^ Episode

       Basis for Average Value      	

       All stations  > 2 pphm              31
26 June 1974
4 August 1975

      26
       All  stations reporting
       peak concentrations
       > 12 pphm
      29
      25
       All  stations reporting
       peak concentrations
       > 20 pphm
      27
      24
  *  The  values  reported  in  this  table are calculated as follows:
                                                     1   N
            Average  Absolute Percentage Difference = —
                                                     N i=
                          c,i
                                     IV-6

-------
     >  All stations reporting ozone concentrations  greater than
        or equal  to 20 pphm.

     For both simulations, the accuracy of peak  predicted  ozone levels
increases slightly with increasing concentration  level.  For ozone
concentrations at or above the federal  standard,  peak  predicted concentra-
tions are accurate to within  25 to 30 percent  of  the observed  levels.

     An evaluation of the four other generic model  performance measures
discussed in the  preceding section—systematic bias, gross error, temporal
correlation, and  spatial alignment—is presented  in  table  IV-3.   In the
following sections, we provide preliminary interpretation  of these
results.
2.   Estimates of Systematic Bias

     Measures of potential systematic bias can be calculated  as either
nonnormalized or normalized quantities.   The  latter  are  normalized by the
observed concentration level.   In table  IV-3,  measures of  bias are
presented for each simulation  for conditions  when the observed ozone
concentrations equal or exceed the 10 and 20  pphm levels.   The nonnormal-
ized bias is estimated by calculating the average (signed)  difference
between pairs of computed and  observed concentrations  (computed minus
observed).  The normalized bias is estimated  from equation  (4-lb).
An appraisal of systematic bias as a function of measured  concentration
level can be made from figure  IV-1, in which  the results of the 26 June
and 4 August simulations are presented.   In this figure  we  plot the mean
normalized deviations.  The following conclusions can be drawn from the
results.

     >  Computed ozone concentrations for 26  June exhibit  tenden-
        cies toward overestimation at low concentration  levels
        (between 2 and 7 pphm) and underestimation at high
        concentration levels (_>_ 20 pphm).

     >  Computed ozone levels  for 4 August generally exhibit  a
        tendency toward underestimation  throughout the entire
        range of observed concentrations.

     >  Computed ozone concentrations exhibit greater mean
        deviation for 4 August than for  26 June.
                                   IV-7

-------







oo
LU
OL
—~t
OO
«t z
LU O
s: —i
i
r^
LU —
c:
•— ' K
<: o
u_
u_
O LU
z
>- o
f*ff Ih^J
< o
s: or
ID O
oo LL.
•
ro
i
^— *
LU
_J
m
 Z
QJ QJ
^r ^-i A
E
SZ.
CL
CL
s I?
CO <^>
B »
in
i
E
JC
CL
CL
•* JS
VO Ch
(M •— •
• *
r^
i


E
JC.
CL
CL
ro CM
0 0
«3- If)
• •
«— «
1



E
SI
CL
CL
O LO
ro ^j-
ro co
* •
tr>
i




i
> to
OJ 0) E
CT> "O »•
ro 0
U • C
QJ T3 «=
> 4-> O
ro 00 Z
A A






























ized-average
E
JC
CL
CL
CO
ro
•c-
Lf)

E
JC
CL
Q.
0
CO
CT>
LO



E
JC
CL
CL
CO
O
r— 1
f-.




E
jC
Q.
CL
i— <
LO
ro
cc





•
>
QJ
T3
-o
^->
OO
A


^~
O
^-
•



CM
CM
ro
•





*r
^3-
ro
•






CM
If)
ro
•





^lean absolute
deviation
> Normal ized-
average


ro
vo
CvJ
•



CM
«a-
CM
•





VO
r— 1
CM







CM
CTi
«— t






>
QJ
T3
•o
•M
OO
A
E
CL
CL
OO
VD
CNJ
•
VD

E
CL
CL
VO
CO
in
i^.



E
CL
CL
f-»
VD
CO
in




E
jC
CL
CL
VO
0
^.
co





ro
O
C
C
o
A






























ized-average
E
JC
CL
CL
r— 1
f*.
CM
•
^J-

E
.c
CL
CL
ro
ro
in
Lf)



E
f"
J^.
CL
CL
VO
^a-
CM
<3-




E
jC
CL
CL
CM
"3"
r— 4
Lf)





>
QJ
•o
T3
4-)
OO
A
 OJ
 (J   QJ
 C  4J
 tO   3
 E  -C
 v.  •<-
 o   s_
VI-  4->
 4-  -t->
 QJ  <:
D_
O  —
    J->
QJ  ro
O  E
C  O)
QJ  4-J  01
I/I  I/)  fC
    O
    i-
    QJ
x^  I/I
O  O
<0  S_
                                              IV-8

-------



«*
r^.
en
t— i
-4->
in
3
cr>
3

 <^
1
»««4
i . i
u_j
1
CC
— i en
co in
^H CO
• •



E
x:
QL
DL

O
(\J

/\
no
O
E
XT
Q.
n.
o
^-i
/\
ro
O
E
d
CL
O
C\J
/\

CO
0


0
4->

CO ^-1
O en
tn r^
• •


o
4-J
^H ^-1
kC r-.
ur co
1


o
4->
13- r-l
ID in
* *
i


vo
CO
vo
•








O
^M
r^
•




ro
cv
«=r





 -f-
10 O

"a! 
ro
4->


JZ
(J
ro
LLJ

A

C
O
•r—
••->
ro
+J
CO
1
£3
«t

A





QJ
en
ITS
S-
QJ
>
ro









,^ 	
ro
4->
ro
a.
CO



c
o
•r-
+_)
ro

'aJ
s^
o
o


t/1
-p
C
QJ

O

l£
14-
ai
o
o



s_
3
O
x:

j^
o
IO
LU

A




i_
3
O
JT
1
^

 1-
 O
QJ
Q.
    C
    O
 ro  ro
 ^  r—
 O  CU
 D. S_
 E  w
 QJ  O
h-  O
    tr
•—  OJ
   cr
 ro  •—
 Q.  •—
tO  T3
                                            IV-9

-------
g
g
 z.o
 '••
 0.0
    0     5     10    15    20     25    90    35    40    45    5C
•2.0
     "  DA
           Tft fflNTS 213
     -  B
                           B_D     m
                                 1  i
                                               1  J   1   i   1
                                                                  1.0
                                                                  1.0
                                                                  0.0
                                                                -1.0
  •U0     S    10    IS    20    25    30    35    40    45    50
                        CiNCENTRflTliN CPPHH1
                        (a) 26  June 1974
    2.0
  '•"
 £  0.0
     0     5     10    15     20    25    30    35    40    45    50
   -2.0
     \- *3
       DflTfl P0INTS  213
           1   1   1  1  1  I   1   1   1   I   1   I   1   I   1   1   1
                                                                2.0
                                                                  1.0
                                                                0.0
                                                                -1.0
                                                                r2.0
     0     5     10    15     20    25    30    35    40    45    SO
                         CINCENTRflTUN (PPHM)
                        (b)   4  August 1975
 FIGURE IV-1.   ESTIMATES OF  SYSTEMATIC BIAS  IN COMPUTED  OZONE CONCENTRA-
                TIONS AS A  FUNCTION OF OBSERVED CONCENTRATION LEVEL
                                IV-10

-------
3.   Estimates of Gross Error

     The second generic model  performance  measure  summarized  in table IV-3
is the mean absolute deviation.   This  indication of gross error is
estimated by averaging the absolute (unsigned)  difference between the
pairs of computed and observed concentrations.   In table IV-3 these
measures are presented as both normalized  and  nonnormalized quantities.
figure IV-2 presents the mean  normalized absolute  deviation as a function
of measured ozone concentration level  for  the  two  simulation days.  For
the August base case the mean  deviation generally  diminishes with increas-
ing concentration level.  Typically,  at  low concentrations (5 to 10 pphm)
the discrepancies are about 50 to 60  percent.   However, near the peak
concentration levels (about 25 pphm or higher)  the discrepancies are
reduced to roughly 35 to 40 percent.

     In contrast, the 26 June  results  suggest  a trend  toward greater gross
error at higher concentration  levels.  Above 25 pphm,  the errors increase
from about 25 percent to 60 percent at 33  pphm.  This  increase in gross
error is primarily the result  of large point sources of NOX located
directly upwind of three of the ozone  monitors  that recorded  high ozone
concentration levels.  Titration of ambient ozone  in the model by direct
emission of large quantities of NOY leads  to lower predicted concentra-
                                  A
tions than if the plumes from  these^sources were not immediately diluted
into the grid volume upon emission.

     The two preceding performance measures—estimates of systematic bias
and of gross error—are useful in examining the extent of model bias and
the accuracies that exist for  various  observed  concentrations.  However,
the simulation results can also be viewed  from  an  overall perspective by
considering differences between predictions and observations  without
regard to concentration level.  This  perspective is achieved by plotting
the distribution of residuals—that is, the computed minus observed
concentrations.  These distributions  are given  in  figure IV-3.

     Two attributes of overall model  performance can be estimated from the
distribution of residuals:

     >  Accuracy is the degree of conformity of a  particular model
        prediction to an observed value  (a surrogate for the true
  Work is presently underway,  under sponsorship  by  the  Electric Power
  Research Institute,  to eliminate  this  problem  through  incorporation of a
  sub-grid-scale reactive plume model  in the airshed  model.
                                   IV-11

-------
                 10    15    20    25    90    35    40    45    50
gl.S
  1.0
1
£0.5
£
                       i   '   I
        DATA MINTS 213
      "  B  B  B  „

i   I   i   I   i   I   i
                   D  •  B  °
                   i   i  i   i
                                               1   i   1
        10    15    20    25    SO    95
                 CiNCENTRflTIiN  IPPHH)
                                                    40     45
                                                                 2.0
                                                         1.5
                                                         1.0
                                                         0.5
                                                      ro°
                                                                  .0
                          (a)  26 June  1974
    2.0
  5 1*5
    1-
                  10    15    20    25    SO     35     40    45    50
  o
  r>
    0.5
    0.0
iI  •   1  «

OflTfl MINTS 213
                         1   I   |
          B  B
             I  j.   I
                D  "        D        B
                      B  B     o

                 I  i  I  i  I  j  J  i   I
                  10    15    20    25    SO     35
                           CIMCENTRATIfM  (PPHH)
                                                    |   I
                                                           2.0
                                                           1.5
                                                           1.0
                                                           0.5
                                             40    45
                                                        ?0°
                                                            .0
                           (b)  4 August  1975
FIGURE  IV-2    ESTIMATES  OF ERROR  IN  COMPUTED OZONE  CONCENTRATIONS
               AS A FUNCTION OF MEASURED CONCENTRATION LEVEL
                                   IV-12

-------
           -20   -16   -12    -8
                      12    16    20
i °*4
tio.s
L
£0.2
i o.i
•e
o.oj
DATA PUNTS 213
_ _

— —
•B
n *nAtB" "A0*"'* i I i 1 i 1 J °ifiPi An*m6nAl'l6ni
j(f -76 -T2 -8-40 4 6 12 16 f
0.4
0.3

0.2
0.1

0°'°
                                CfNCENTRRTIlN (PPMM)
                                (a) 26 June  1974
            -20   -16    -12    -8
-4
12    16    20
£ 0.4
s
^O.S
h.
tO.2
(D
£0.1
o:
0.01
I | I | i | i | 1 | I | I | I | i | i
_ DftTfl POINTS 213
_
—
BOD
•"fto*m*irlpD°D° • ° • 1 ?D«P ? rfrmfimA 6 A 4
0.4
0.3
0.2
0.1
m.n
                                               4
                                 CSNCENTRRTIIN CPPHH)
                                (b)  4 August 1975
FIGURE IV-3.   DISTRIBUTION OF  RESIDUALS (PREDICTIONS  MINUS OBSERVATIONS)
               FOR THE OZONE SIMULATION RESULTS
                                          IV-13

-------
        concentration).   The mean value of the  frequency  dist-
        ribution of residuals has been  chosen to  represent the
        degree of accuracy demonstrated in a particular simula-
        tion.

     >  Precision is the degree of conformity of  the  ensemble of
        prediction and observation pairs to a specified residual
        value.  The standard deviation  about the  mean value  of the
        difference distribution has been chosen to  represent the
        degree of precision demonstrated in a particular  simula-
        tion.

     Overall accuracy of the ozone predictions  for  the two simulation  days
can be estimated from the mean (or first moment)  of the difference distri-
butions in figure IV-3.   The standard deviation about the mean (the  second
moment) is an estimate of model precision as defined  above and is also
obtainable from the distributions.  Analysis of the difference distribu-
tion yields the following estimates of  model accuracy and precision:
                                           0,  (pphm)
             Simulation Day          Accuracy     Precision


             26 June 1974                          ±  5  pphm

             4 August 1975           -3 pphm       +  5  pphm
     On the basis of these results,  one can conclude  that  the  overall  acc-
uracy of the model in predicting 0^  levels  was  greater  for 26  June than
for 4 August.
4.   Temporal Correlation

     The temporal correlation coefficients for ozone in  table  IV-3  indi-
cate a rather broad range of values for the individual monitoring sta-
tions.  At concentration levels exceeding 10 pphm,  average  values for the
coefficients are 0.423 and 0.686 for 26 June and  4  August,  respectively.
Of the two airshed simulations, the 4 August 1975 results exhibit the
better temporal correlation.  These results indicate that the  "timing" of
the ozone buildup and subsequent decay, when viewed over all stations, is
rather variable.  Peak model prediction at some stations are early,
whereas at others they are late (when compared with the  observed
                                    IV-14

-------
maximum).  Low temporal correlation coeficients  are  also  due  to  a more
rapid reduction in predicted ozone levels  than in  observed  values.
5.   Spatial Alignment

     Examination of the spatial  correlation coefficients  in  table  IV-3
indicates that these values are  generally smaller  than  the corresponding
temporal coefficients.  For instance,  at ozone  levels  greater  than  or
equal to 10 pphm, the spatial  coefficients are  0.380  (26  June)  and  0.565
(4 August).  Moreover, the spatial  correlation  decreases  with  increasing
concentration level, suggesting  that the location  of the  peak  predicted
"pollutant cloud" is displaced spatially from that indicated by the
monitoring network.  The 4 August 1975 results  also demonstrate the better
spatial correlation.

     An additional spatial measure of  model performance  is  the "distance
distribution."  To explore further this aspect  of  the  results,  every
monitoring station was examined  for each hour between  1000 and  1700 PST to
determine the distance from the  grid cell in which the  monitor  is  located
to the nearest grid cell at which a prediction  either  equaled  or "brac-
keted" the observed value.  This "distribution  of  distances" is presented
in figure IV-4.   In over half of the cases a prediction comparable  to a
given measurement can be located within two to  three grid cells—a  dis-
tance of 10 to 15 kilometers.   Such discrepancies  can  be  caused by  rela-
tively small errors in wind speed or direction  input data.

     Having discussed the base case simulation  results  in a  rather  general
fashion, attention is now focused on model performance  at the  various
monitoring stations.
C.   SIMULATION RESULTS FOR SPECIFIC MONITORING STATIONS

     An informative though often time-consuming method  for  analyzing mode'
performance is the evaluation of temporal  trends in  computed  and observed
values at the various monitoring station  locations.   From this procedure
one can develop a conceptual picture of the formation and transport of
  Distance distributions, as discussed  here,  are  not  intended  to be  used
  to determine whether model performance  may  be judged  satisfactory  or
  unsatisfactory.  Instead,  this procedure is used  to provide  possible
  further insight into the possibility  of biases  in certain data input
  (e.g., wind fields), the existence of steep concentration gradients,  and
  so forth.
                                   IV-15

-------
v     ft  o tn
ID  o  —  — —
                             o

                             £
                               a
r>
           er
           cc
       CD  CD
       (C
r>  en     tn en
    uj en  tu uj
    => uj  S S

              tr
..  ju«  «
5  ci
fe  fe  _J  fe
       cr
      tn
               o
               (O
                             en
                     V)  CO  (T
                     UJ  UJ  Z
                     U  (J  CB

                     o  a  a
                     •-  —  uj
                     ct  a:  «n
                     o  o  •  «n
                         ui  ~
                         a  »-

                     cc  a  ^
                     u  i—  «r>
                     z:  m  «
                        in
                        (M
                                                                      r>
                                                                         •a

                                                                         »-
                                                                         n
                                                                         o
                                                                         z
                                                                      o  ex
               o       in       o
               *>       «W       M
                                                           H>       O
                                                                                  O)
                                                                                  C
                                                                                  3
                                                                                  -D


                                                                                  VO
                                                                                  CM
                                                                                            o
                                                                                            u
                                                                                            o
                                                                                            M
                                                                                            O
oo  LU
                                                                                            •— > a.
                                                                                            OO HH
                                                                                            o
                                                                                               oo
                                                                                            o -z.
                                                                                            i— i O
                                                                                            a: H-.
                                                                                            O 00

                                                                                            a: 
                                                                                            LU r—
                                                                                            CO
                                                                                            z: en
                                                                                            => o
                                                                                            LU
                                                                                            Of.
                                    JB
                                              IV-16

-------
                                 o
                                 UJ
 N
 

 5?  0?
 >  >

 It-  li-
 CS  ea
o

ea      o
X      UJ
x      i-
tn      uj
UJ  O  X.
a:  uj  o
x  i-  et
i—  uj  cc
    *:  03
UJ  <_»
>•  ec  •-
«  tc  <9
CD  CD  Z
cc
    cn  cn
tn  LU  uj
UJ  =)  =3
         _j ea  ea
         cc
      .  ^_   .    .
     ca  ca cs  s
     z  »- z  z
                                 o
                                 cc
                                 a:
                                 en
                                 UJ
cn  OT   cc
_i  _j   •>
_i  _j
IU  UJ   Z
u  o   ea

o  a   a
—  —   uj
az  oz   tn
o  u   cc
         m
CD  CD
 •   •   cn
tsi  —•   o
         t_4
II   II    h-
    >•   en
    UJ   »
    O   I-
z       cc
tr  o   t-
lu  •—   cn
x  en   ^
                           in       o
                           rg       <\j

                                                                                           Oi

                                                                                en
                                                                                -j
                                                                                -i
                                                                                IU
                                                                                u

                                                                                o
                                                                                •««
                                                                                1C
                                                                                C9
                                                                                              O
                                                                                   cn
                                                                                lu
                                                                                =3
                                                                             n
                                                                                a
                                                                                a:
                                                                                oo
                                                                                IU
                                                                                ac
                                                                                ex
                                                                                IU
                                                                                tn
                                                                                •-«
                                                                                a
iiiil
                                     I  I  I  I  »  I I I I I I I  I  I  I  I  I  I  I
       in
       tn
                                    o
                                              LD
                                        J0
                                                  IV-17

-------
secondary pollutants, such as ozone,  throughout  the simulation  period.
The locations of the various monitoring  stations  and  certain  relevant
topographical features of the SCAB are identified in  figure IV-5.   Figures
IV-6 and IV-7 present the ozone results  for  the  26  June  and 4 August
simulations, respectively.

     Upon review of the plots of computed  and  measured concentrations
shown in figures IV-6 and IV-7, general  and  specific  comments can  be made
with regard to each of the two simulations.   (During  the following discus-
sion, the reader may wish to refer back  to figure IV-5,  which relates the
locations of the monitoring stations  to  general  features of the South
Coast Air Basin.)  The following comments  can  be  made on review of the
base case simulation results.

     >  For the 26 June 1974 simulation:

           In the San Fernando Valley predictions generally agree
           with observations; computed peak  concentrations  fall
           between 15 and 40 percent  of  the  observed  values.

        -  In the San Gabriel Valley  (Pasadena,  Pomona,  and Azusa)
           a slight (10 to 30 percent) overestimat>on occurs; pre-
           dicted levels increase more rapidly than the  observed
           concentration levels.

           In the San Bernardino Valley (Fontana, San Bernardino,
           Redlands) there is a "quenching"  of ozone  followed by a
           secondary ozone buildup after noon.  The model under-
           estimates the peak concentrations by  20  to 40 percent.

        -  Along the coast (Long Beach,  West Los  Angeles, Los
           Alamitos) predicted ozone  levels  build up  sooner and
           diminish faster than the observed concentration  levels.

     >  For the 4 August 1975 simulation:

           In the San Fernando Valley peak levels at  Reseda are
           underestimated by nearly 50 percent;  the peak at
           Burbank is overestimated by only  5  percent.

        -  In the San Gabriel Valley  (Pasadena,  Azusa) ozone is
           underestimated by 20 to 25 percent.

           In the San Bernardino Valley (Fontana, San Bernardino,
           Redlands) ozone peaks are  underestimated by  25 to 35
           percent.  However, only at the  Redlands  station  does
           the anomalous ozone quenching effect  appear.
                                   IV-18

-------
                                                          IS)
                                                          o:
                                                          o
                                                          o
                                                          o
                                                          UJ
                                                          or
                                                          f— z:
                                                            H«4
                                                          Q oo
                                                          as <
                                                          — i
                                                          o: cc

                                                          i— h-
                                                          
-------
  60
  50
              6
                   12
IB
i 30
  20
  10
      IIIIIIIIIII|1IIII|IT iIT_

     - ftZUSA
     - iBSERVED    •
       PREDICTED  —
      a in a
                                    a m a m n
                                         50
                                              40
                                         30
                                         20
                                         10
                        12
                   TIME (MiURS)
                              18
          24°
                      60




                      50




                      40

                    x

                    i 30

                    •>


                      20




                      10
     12
               18
                                                          i ill  i i  riT i   i i PI i   i  i i  I
                                                       -  BURBflNK
                                                       -  1BSERVEO
                                                       -  PREDICTED
                                                            n m n i a
                                                                              iltiiiiTTmam ti~
                                                                                                60
                           50
                           30
                           20
                           10
     12
TIME CH0URS)
               16
24°
  60
  SO
  40
t 30
  20
  10
                        12
                              18
_ I  I T I  ||  I I  I  I I |  I I I  I  I |  I  I I I  l_

- CHI KB
• iBSERVED    •
- PREDICTED  	
                   12
              TIME (HBURS)
                                  IB
                                         so
                                         40
                                         30
                                         20
                                         10
          24°
                                                    60
                      SO
                      40
                    i 30
                      20
                       10
                                             12
               18
                         24
                                                            _T  I II  T| IT  II T (  I I  I

                                                            -  OBWNTHH  Lft
                                                            -  iBSERVED    •
                                                            -  PREDICTED   	
                                                                                   a
                                                             n fn n i n_
     12
TIME (HBURS)
                16
                                                                                               a nj m
                                                                   60
                           50
                           40
                           30
                           20
                           10
24
          FIGURE   IV-6.  CALCULATED AND  OBSERVED OZONE  CONCENTRATIONS FOR  26 JUNE  1974
                                                    IV-20

-------
                                 12
                                           16
DU
50
40
?
L
±30
r>
B
20
10
0«
Ll
irn
50
40

:30
»
20
10
n1
Lo
- EL T0RB
- IBSERVED •
- PREDICTED —
- —
•
—
•
DO ~
D j
__H^n " O O —
3.f.l-»«T-ijBJ_l 1 1 1 1 	
> 6 12 16 2
TIME (HiURS)
> 6 12 16 2
- LENN8X
- iBSERVEO •
- PREDICTED 	
- -
"
~ -
:
— —
'a m n i Jrtfa m n m n iti n rnnaru^ti a m a m n~
6 12 IB 21
TIME (HSURS)
60 60
50 50
40 40
i
30 t 30
r>
20 20
10 10
n n1
4° °!
*60 60
50 50
40 40
i
30 i 30
m
20 20
10 10
n n'
P °3
. ' ' ' ' ' 1 ' 1 ' II 1 1 1 II 1 1 i 1 1 1 1 1
- FiNTRNfl
- iBSERVEO •
- PREDICTED —
— —
B
—
D 0
O
DO y\ ~
- / n o I.
A D D I
«Jr O . * o o~
> 6 12 16 2
TIME (HIURS)
> 6 12 18 2
- L0NG BERCH I
- iBSERVED •
- PREDICTED 	
— —
~ -
- -
~ —
7 -
-™ m fi i r» rinrfm | if?, , ^iSLJ" ? T ? m n~
6 12 18 21
TIME (H0URS)
60
50
40

30

20
10

4"
4
bU
50
40

30

20
10
°
FIGURE  FV-6 (Continued)
         IV-21

-------
12
               18
CD
so
40
30
20
10

°(
_ I I i i I | I I I i
- LtS BLUM IT
- IBSERVED •
• PREDICTED —
-
-
-
_
™
:, , , i-^r '  »  I  '
                                         50
                                         30
                                         20
                                          10
TIME CHBURS)
                   12         18
              TIME (HBURS)
SO
40
4.
t 30
•D
20
10

J
1 6 12 18 2
I I I i I | I I I I I | 1 I I i I | I I I I l_
• LTNH88D
• IBSERVED •
• PREDICTED 	
-
- -
" w
—
: ^v :
: yoDB^Yiv-p ;
"n o (j i |] -i^u i \ i i 1 i i i IT»J!) n T H n o*
) 6 12 18 2
'so
50
40
30

20
10

4°
60°



50
40
30


20


10,
°
6
_l I I I I | I I 1 I
• M8UNT LEE
- iBSERVED •
- PREDICTED — —

-
_.
~
-

- D/
- /
-_. BD/
:°0nDOffl I
:, , ,,J^/ , ,
) 6
12 18 2
I | I I I 1 I | I I 1 I l_
_
—
—
-
—
_^
o I
0 m
HI •_
B ° I
\ O D
\ °
\ BOOB-
V-^s ':
i 1 i i i i i 1 i i i i i"
12 18 2
'so



50
40
30


20


10
4°
TIKE (H0URS)
               TIKE  (HBURS)
                         FIGURE  IV-6 (Continued)
                              1V-22

-------
     - iBSERVED
     - PREDICTED
                                            - 10
                         12         16

                   TIME  (HBURS)
                                                         KO
                                                                      6
                                                            12
                                                         J!0
                                      KI
                                         _i  i  i i TI  n n i  |  i i

                                         - REDLfiHDS
                                         - BBSERVED    •
                                         - PREDICTED  —
                                                       IB
                                                      rprr
                                                                                       D
                                                                                                      50
                                                                        i  i i  i  i I  i  i  i »  i  I  i i  i  f t~
                                                                                                      30
                                                                                                      20
                                                                                                      10
                                                  6         12         IB

                                                       TIME (HBURS)
  60
  50
               6
     12
16
t 30
  20
  10
     _ I  I  I 1 T ]  I

     - RESEDR
     L iBSERVED
     - PREDICTED
Tl i j  ill  i \  |  ill
                     ED   D  D
      n m a i XL3r t  i  i  i _ i  1  i j i  i  i l i m n to a'
                           6C
                           50
                           30
                                              20
                                               10
              6         12         16

                   TIME (HBURS)
                       60
                       50
                       40


                    z

                    Q.
                    t 30
                                      20
                                      10
6
12
                                                                                          18
                            i  I  I  I |  I  I I I  l  j I  1  I 1

                            SRN  BERNflD
                            BBSERVED     •
                            PREDICTED   	
                                                                                       D
                                   6         12

                                        TIME  (HBURS)
                                                                      16
                                                                                                      60
                                                                                                      50
                                                                                                      40
                                                                                  30
                                                                                                      20
                                                                                                      10
                                       FIGURl  IV-6  (Continued;
                                                IV-23

-------
60
50
40
£30
20
10
                       12
                              18
_l  I  II I  |  I

- SIHI VflLLE
- iBSERVEO
- PREDICTED
                    i  I 1 I  i  I I I  |  I I I  i  (_
                  0 d
                                          SO
                                          40
                                             ;IO
                                          V.O
                                           !0
             6         12         18
                  TINE  CHBURS)
                                                     60
                                                          50
                                                          40
                                                   *; 30
                                                          20
                                                           10
                                                                     6
                                                                                 12
18
24
                                                             n\\ IT T | t T  I I 1 |  I  I I I  I  I

                                                             -  UPLRND
                                                             -  iBSERVEO     »
                                                                PREDICTED   	
                                                                                                        60
            50
            40
            30
                                                                                                   20
                                                                                                   10
                                                           "fl T I  I rt**T t i  i i i  I  i ' i  i  i  t T i  |  i  i~
                                                                  6         12         16
                                                                       TIME tHBURS)
                       12
                              18
                                              24
bu
50
40
X
t 30
V>
20
10
n'
_ i i i i i I i i i i i I i i i i i | i i i i l_
- WEST Lfl
• 8BSERVED •
- PREDICTED 	
_
-
_
_
h- 	 g „
S^*>& ffl ffl _
t^ n [j f flUfir^ IP*' '^ *^ '' ''TfflBPr^
                    12
               TIME  (HBURS)
                                 16
                                             50
                                             10
                                             30
                                             20
                                              10
                                     FIGURE  IV-6  (Concluded
                                                   IV-24

-------
50
40
i 30
20
10
              8         12
     _ I  I  '1 I  I | I I  i i i  |  i

     - AZUSR
        OBSERVED     •
        REDICTED   —
                                  18
                               "' > I
TT$
                                                                            12
                                                                                         18
24
                n             V  o
                i  i  i i  f  i i  i  i  p*J  i
                                              50
                                              40
                                            30
                                              20
                                              10
               6         12         18
                   TIME (HBURS)
                                                        so
                                                        40
                                                         IK)
                                                          I  I  I I  l J l I  I i J  |

                                                           BURBRNK
                                                           •BSERVED     •
                                                           PREDICTED  —
                                                                              12
                                                                         TIME CHtURS)
                                                                                            n ID n
                                                           50
                                                           *0
                                                                                                    30
                                                                                                    20
                                                           10
                                                                                       18
  60




  50




  40

x

tso

n

  20




  10
              6         12        18         M
      liiiijilll i  |  i i  i  [  i |  i  i i  i  i |

       CHIN0
       •BSERVED    •
       PREDICTED  	
                                                                            12
                                                                                         18
                                                          24
                                            50
                                            40
                                            30
                                            20
                                            10
              6         12         18
                   TIME (HBURS)
                                           4°
on
5ti
Itl
SO
20
10
[
IJc
_ 1 1 1 1 1 1 1 1 i 1 1 1 I 1 1 1 1 1 1
- DBWNTHN Lft
- BBSERVED •
- PREDICTED 	
-
-
-
\ /5\°B'OB
hl? f ? , p-fr-.' i i . r iTTTs^j ° '
) 6 12 16
I l i_
—
—
—
—
-
rj — m --
2
bU
50
40
30
20
10
f
                                                                       TIME (HBURS)
      FIGURE  IV-7.   CALCULATED AND OBSERVED  OZONE CONCENTRATIONS FOR 4  AUGUST  1975
                                                v  75

-------
«n°

50
40

SO

20


10
°C
_ 1
-
-
-
•*
-
—
:
-
—
"7
i

«n°
60

SO
40
30

20
10

«
_i
-
-
-
-
_
-
-
•
B 12 18 21. __0 6 12 16 24_
I I 1 I | I I I I i | i I i i i | i i i i i_
EL TIRB
•BSERVED •
PREDICTED — ^
—
-
^™*
-
—
a -
B °B -
B BO —
y n |" fl-*ti D f i i i 1 i i i i if T m i i i "
6 12 16 2
ou vw
50 50
40 40
X
30 t SO
s
20 20


10 10
1
4° °
I i i I i | i i i i I | i I I I i | I I I I l_
- FINTRNfl
- iBSERVED • . -
- PREDICTED 	
-
C -
" ^™
B
- >^™\? ~
I j \° n ;
- B / \ -
~ D J* \ ° ~
'- mmn vS/1 E B „ o"
^mol?2->. . i i i I i i i • i 1 i i i ° i-
) 6 12 16 2
DU
50
40

30

20


10
4*
TINE (HIURS) TIME (HBURS)
6 12 18 24.. ..0 6 12 16 24..
I i i I j i I I I t | I I I I i | i i I I i_
LENNBI
IBSERVED •
PREDICTED 	 ^

—
-
••
—
—
• P — i. "\ B „ D -
ra m a i_m^rl5n? a j i V i i FTTH re m n D o~
i 6 12 18 2

OU DU

50 50
40 40
X
30 £ 30
•0
20 20
10 10

f
I I i i i | i I I i 1 | I 1 I I I | 1 t 1 1 I _
- LING BERCH
- IBSERVED «
- PREDICTED 	
-
—
—
_
- —
—
I 	 n n 	 I
i"ii D B t II jjj-ff^C^ffl *? i 1 i 9 STlnsj}] p m n m n~
D 6 12 16 2


50
40
30

20
10

t
TIME (HIURS) TIME (HIURS)
FIGURE IV-7 (Continued)
         IV-26

-------
  60
t SO
  20
  10
                         12
              16
24
                         I '  '  ' '  '  I
       LiS HLflMIT
       •BSERVED
     - WED 1C TED
                  D
  DC    D

O    D    Q
        m n m n.
                          60
                                               SO
                          30
                          20
  10
               6         12         16
                   TIME  IHBURS)
                        24*
            BO
                                     50
                                                          40
           .
          i so
                                                                                 12
16
                     24
                                        Till

                                        i-  PflSRDENR
                                        -  iBSERVED
                                        -  PREDICTED
                                                                         i i i  i  i I i
                                                          50
                                                                                  40
                       30
                                   12         16
                              T1HE IHBURS)
                         12
              18
24
bU
50
40
ac
i 30
tt
20
10
°C
_ii II I] li II I | 1 I ii
- LYHH00D
- iBSERVED •
- PREDICTED 	
-
-
.
-
•| i i i j-J- 6 12
I | I i I i i_
—
—
—
—

18 2
50
50
40
30

20
10
4°
                                                          60




                                                          50




                                                          40

                                                        x
                                                        a.
                                                        S: 30

                                                        tn


                                                          20




                                                          10
12
18
24
                                           I  i I  i  |  i i  i  I  I |  I  I  I
                                        -  P0M0NR
                                        -  BBSERVED
                                        -  PREDICTED
                                                             'a m g  i
                   TIME  (MiURS)
                                                            12         18
                                                      TIME  (HBURS)
                                                                                  60
                                                                                  50
                                                                                  40
                                                                                  30
                                                                                  20
                                                                                                       10
                                       FIGURE IV-7 (Continued)
                                                   IV-27

-------
€0
SO
40
                      12
20
10
   ;7n-n-ri
   - PRfiDB
   " 1BSERVEO
   - PREDICTED
             6         12         16
                 TIME  tHSURS)

                                            ^
                                        m
                                              50
                                             30
                                              20
                                                       60
                                                       50
                                                         40
                                                         SO
                                                    20
                                                         10
                                                                t         12         18
                                                       _~J I  I i i  |  i i I  i  i |  i  I I i  i  |

                                                       - «EOLRNDS
                                                       - •BSERVED    •
                                                       - PREDICTED  —
                                                                 6         12
                                                                      TIME  IHBURS)
                                                                                            Q
                                                                                          18
           |60
                                                                                                 50
                                                                                                    30
                                                                                                    20
                                                                                                ho
          24°
                       12
                                   18
50
  40
S: SO
  20
  10
TT-rn-n

  RESEOfl
  •BSERVEO
- PREDICTED
                    I  I j I I  i I I  j  i
A\c
                       o
                                              50
                                         40
                                         30
                                         20
         6         12         16

              TIME CH0URS)
                                                        60
                                                        50
                                                        40

                                                      x

                                                      ^ 30
                                                         20
                                                        10
                                                                           12
18
24.
                                                                        i r~r i  i y i i i i i j  i
                                                           -  Sf)N BERNRO
                                                           -  *BSERVED
                                                           -  PREDICTED
                                                                      6         12
                                                                           TIME (HflURS)
                                                                                                      SO
                                                                                                      40
                                                                                                      30
                                                                                                     20
                                                                                                       10
                                                                                         16
                                     FIGURE IV-7  (Continued)
                                                IV-28

-------
  60
  SO
  40
t SO
  20
  10
                        12
16
     - SIHI VHLLE
     - iBSERVED
     - PREDICTED
                                         n
            50
            30
                                               20
            10
                        12
                   TIME (HflURS)
16
24°
  60


  50


  40

x
5: 30

S
  20


  10
                                             12
                          n n  i j i  i  i i  i
                          -  UPLRND
                          -  6BSERVED     •
                          -  PREDICTED   —
                                             16
                                         1 '  '  I
                                                                                      B D
                                                                           i  i  i I i  i  i  i t  I  i
                        12
                   TIME (HflURS)
                                                                    50
                                                         30
                                                                    20
                                                                                                      10
ou
50
40
I
i 30
20
10
I
°C
_ I I I I I | I I I I i | I I I I I | I i i i i_
- MEST Lfl
- gRSERVED •
- PREDICTED 	
—
"" ™
— — .
—
—
:' ^ ju^TT?^7T7>^ •» ? •? » B^
I 6 12 18 2
bU
50
40

30
20
10
C°
                   TIHE CHiURS)
                                        FIGURE IV-7 (Concluded)
                                                 IV-29

-------
        -  Along the coast and inland  toward downtown Los Angeles
           computed peak ozone levels  are typically underesti-
           mated.   Moreover,  predicted concentrations begin to
           drop before noon though  observed values remain moder-
           ately high (~ 10 pphm) until  1300 to 1400 PST.

     Overall,  the  model  appears to  have  underestimated ozone levels in the
western part of the Los Angeles basin.  However, the magnitude of the peak
ozone concentrations in this  region is typically much less than that of
the basin wide peaks.  The performance of the model appears to be the best
for the mid basin, between the San  Gabriel and San Bernardino valleys.
Farther to the east, there is again a  larger discrepancy between predic-
tions and observations.   The  trend  toward underestimation in that location
is attributed, in  part, to the current treatment of large NOX point
sources.  Computed peak concentrations for the area are typically within
30 to 40 percent of the observed values, and the timing of the peak is
offset by about two or three  hours.
D.   EVALUATION OF GROUND-LEVEL OZONE  CONCENTRATION  FIELDS

     The base case ozone results produced  by evaluation of hourly average
ground-level ozone fields are presented  here.   Figures  IV~8  and  IV-9
present ozone concentration  isopleths  for  both  simulations for the period
from 1000 to 1600 PST.  Examination  of the ground-level maps  leads to the
following general conclusions:

     >  The peak ozone levels of 4 August  (1200-1300) occur  two
        hours earlier than those of  26 June (1400-1500).

     >  Elevated ozone levels of 26  June extend over a  broader
        geographical  area than  do those  of 4 August.

     >  The Pomona-Upland area is the  region of highest ozone
        impact for both simulations.

     >  Concentration levels occurring over the western half  of
        the basin on  4 August are nearly one-half  of those occur-
        ring on 26 June.

     >  Both simulations exhibit steep ozone concentration
        gradients along a SW-NE transect of the San  Gabriel  and
        San Bernardino valleys.
                                   IV-30

-------
                                                    f—   vc
                                                         CVJ
                                                     10   O
                                                         u_
                                                    o
                                                    I—   to
                                                         IE
                                                     I/)   D-
                                                     t.   O
                                                     3   tn
                                                     o   •—

                                                         UJ
                                                     o>
                                                     c
                                                     O!
                                                         o
                                                         fsj
                                                         o
                                                     o>  >
                                                    CD  •—
                                                     10   CZ

                                                    *~*  C5
                                                         •—H
                                                         u.
IV-3'1

-------
1SB3
 o
                                         cn
                                               CSJ


                                               •o

                                               •o  -o
                                                    01
                                                o   c
                                                    o
                                                3
                                                O  CD
                                               Z   I
                                                   a:
                                                C  =5
                                                  ~
                                                O)
                                               CQ
   IV-32

-------
                                              10  -o
                                                   (U
                                              CM   3
                                              t—   C
                                              o   c
                                                   o
                                              •^  «_>
                                              L.  - ^

                                              O  CO
                                              Ol  C3

                                              I  C
                                              4->
                                              O)
                                              CO
IV-33

-------
1SH3
                                                 ID  13
                                                      O)
                                                 ro   3
                                                  o   c
                                                      o
                                                  I/I  O
                                                  t   -^Hl~

                                                  O  00


                                                  0)  •—

                                                  4->  UJ
                                                     QC
                                                  C  =5
                                                  flj  C3
                                                  O)
                                                  CD
 1S3M
     IV-34

-------
                                              •o
                                              c  —
                                              10  -O
                                                   Q)
                                              ff   =»
                                              i—   C
                                               O   C
                                                   O
                                               O   00
                                              3=   I

                                               
                                               OJ
                                               CO
                                               QJ
IV-35

-------
1SU3
 o
   ;||p
                                           c  *-»
                                           re  -o
                                              OJ
                                          LT)  3
                                          t—  C
                                           o  c
                                              o
                                           t  ~
                                           a
                                           O  00
                                           ft)  i—
                                           c
                                           O)
                                           OJ
                                          CO
                                              UJ
                                              a:
  o
  •«

 1S3M
   IV-36

-------
1SH3
 o
                                       o
                                       rv
                                               C  —•»
                                               «0  TO
                                                   CJ
                                               VD  XJ
                                               O  C
                                                   O
                                               tfl  O
                                               t   ._  ^
                                               3
                                               O  00
                                               Ol  •-•
                                               JC
                                                   CC

                                                   O
                                                QJ
                                               CO
  o

1S3M
 IV-37

-------
1C03

 o
                                                     IT)
                                                     f-
                                                     Oi
                                                     to

                                                     CJ
                                                 10   DC.
                                                     o
                                                 o   u.
                                                 l_  Q-
                                                 s  o
                                                 O  l>0
                                                 a:  ^

                                                 0)  LU
                                                 £  Z
                                                 •M  O
                                                     rvi
                                                 c  o
                                                 
                                                     ce
 IV-33

-------
•^rfr^r^rr^^^rrrt^* ••*!!» •••••»•••»*

                    m       s
"::;ma:;:;;;w::;H-; /  \      °-
    ..............  _  ....     ..    ...  .......    ... I  .............. .....  .......   ..... :v.x.-::
                                                                                                           •o
                                                                                                            C   *-*
                                                                                                            •o   -o
                                                                                                                  o>
                                                                                                            O   cr>
                                                                                                            C   =D
                                                                                                            
-------
                                     c
                                     rD


                                     C\J
                                     s-
                                     3
                                     o
                                     :r

                                     0)
                                     
-------
                                            J.SU3
                                              o

o
CM

                                              2
                                              T
                                           •vK5'
                                                                 |--:--;-':j--:-:. :,;•:• •y'jj-r;;:.^;:;:; .y::v'
                                              o
                                              •*«

                                             1S3M
                                                                                                     •O
                                                                                                      c  «—«•
                                                                                                      (O  T3

                                                                                                     CO   3
                                                                                                      O   C
                                                                                                           O
                                                                                                      (/>   <_>
                                                                                                      1-   —•
                                                                                                      3
                                                                                                      O   CTi
                                                                                                      c
                                                                                                      o>
                                                                                                      Ol
                                                                                                      CD
                                                IV-41

-------
                     XSW3
                      o
.*.>•.. 1...-.+ • • -.-1 . ...4  .. 1 • T i  4
                           i .. + •. i T. i.  T--1....+7
                                   <7mw.
                                     > ^rr r r rr r »r-*vrr rvrr*M^
                       *


                       5
. _•.• •.• •:•.• . • .-. . ;. . ^	.  .;..  i^-jl
                      :•? ;:::.::;::
                       ? ttX
                           ''•(•'" r •»' •' t":-• •!'• "i:':••:''t'•'•'••'*"'••
                                                 3
                                                 Jl
                                                      •o -o
                                                         0)
                                                      ^-  3
                                                      «*-  4->
                                                      O  C
                                                         o
                                                      IA  O
                                                      L.  ^^

                                                      O  »
                                                      OJ  ^«

                                                      *J  UJ
                                                         Q£
                                                      C  3
                                                      O)  O

                                                      I  C
                                                      4->
                                                      0)
                                                      m
                       O


                      1S3H
                       IV-42

-------
1SH3
 o
                                                  QJ
                                             If)   3
                                              o   c
                                                  o
                                              ifi  O
                                              L.  ^.^
                                              3
                                              O  O>
                                              QJ
                                             CD
1E3M
  IV-43

-------
1SH3

 o
                                               •o

                                                «J  "O^
                                                    
                                                a:   t
                                                    QC

                                                at   o
                                                OJ   «-•
                                                X   u.
                                                4->
                                                a>
                                                CO
1S3K
    IV-. •'

-------
E.   SUMMARY

     This chapter presents the results  of  the 26  June  1974 and 4 August
1975 base case airshed model  simulations.   Overall, both sets of results
agree favorably with ozone observations.
                                   IV-45

-------
                              SIMULATION RESULTS
A.   INTRODUCTION

     Figure V-l presents schematically the relationship  between  input
data, model input variables, and air quality model  predictions.   The
objective of this study is to determine the sensitivity  of airshed  model
ozone predictions to the amount (or "richness")  of  input  data.   This
objective clearly differs from that of classical  sensitivity analysis
studies.  The latter deal with the sensitivity of a mathematical  model's
predictions to variations in model parameters.  To  place  this study in
perspective, various approaches to sensitivity analyses of air quality
models are summarized.   The discussion then shifts  to how the sensitivity
of the airshed model to various levels of detail  in input data is ascer-
tained.
1.   Review of Sensitivity Analyses of Air Quality Models

     Mathematical model sensitivity analysis may be defined  as  the
evaluation of the deviation in model  output resulting  from a perturbation
in one or several model inputs.  One possible approach consists in
defining a mathematical relationship that  relates  the  extent of the model
output deviation to the perturbation in the input  variables  (or param-
eters).  Such sensitivity analysis methods have  been applied to some  air
pollution models (Seigneur, 1978; Koda et  al.,  1979; McRae and  Tilden,
1980); however, their application to complex urban air quality  models
would be too costly and cannot be considered at  present.

     Another approach that has been widely used  consists in  perturbing the
model input variables and parameters one by one.   This approach gives  less
information ( on a "per run" basis) than the more  complex mathematical
methods mentioned above, but the information obtained, if used  approp-
riately, may be very useful for understanding the  dynamics of the model.
For example, Liu et al. (1976) and Dunker  (1980)  used  this approach to
evaluate the sensitivity of the airshed model predictions to perturbation
in emissions, wind velocities, vertical eddy diffusivity and mixing
height.  The same technique was also used  to study the sensitivity of
atmospheric chemical mechanisms to individual reaction rates (  Dodge  and
Hecht, 1975; Duewer et al., 1977).

-------
     In this study, the latter approach was  necessary because  of the
complexity of the urban air quality model.   Specific  perturbations have
been introduced in the input data,  and the resulting  changes  in the model
predictions provide some information on the  model  sensitivity.
2.   Sensitivity of the Airshed Model  to Input  Data

     A detailed description of the airshed  model  is reported  by  Reynolds
et al. (1973).  A brief overview can  be  described  as  follows.

     The airshed model consists of a  simulation model  and  several  sub-
models that supply various input variables  and  parameters.  Figure V-l
shows that the simulation model is comprised  of a  set of continuity
equations that describe the advective  transport, turbulent  dispersion,
chemical transformations, pollutant emissions,  and removal  of chemical
species in the atmosphere.  Model input  variables  such as  wind fields,
eddy diffusivities, reaction-rate coefficients, and emission  rates are
defined either directly or by means of various  submodels.   Initial and
boundary conditions are estimated from air  quality data and historical
intensive field measurement programs.  The  airshed model configuration can
vary in terms of grid size and number  of grid layers.   Although  there are
basic theoretical constraints on the  grid size  (Lamb  and Seinfeld, 1973;
Reynolds et al., 1973), these features may  be varied  according to  the
airshed considered.

     As suggested in figure V-l, analysis of  airshed  model  sensitivity to
input data may be considered in two parts:  sensitivity of  the model
output to changes in the input variables (e.g., wind  fields,  diffusivi-
ties, emission rates, mixing depths)  may be evaluated  through successive
model simulations, each involving a prescribed  set of input files; and the
sensitivity of these input variables  to  perturbations  in the  input data
used to construct them.  For instance, wind field  maps will show the
effect of a change in meteorological  data on  wind  speeds and  direction.
Attention is therefore focused on (1)  the combination of the  sensitivity
of the model to its input variables and  on  (2)  the sensitivity of  the
input variables to the basic data.
B.   MEASURES FOR ASCERTAINING MODEL SENSITIVITY

     An important step in quantifying model  sensitivity is  the  definition
of specific measures.  Several measures  that appear  to  be useful  include
the following:
                                   V-2

-------
                                                  c/o
                                                  i—i
                                                  
-------
     >  Signed deviation.

     >  Absolute deviation.

     >  Temporal correlation.

     >  Spatial correlation.

     >  Overall maximum ozone level.

     >  Maximum ozone statistics (peak level difference, peak time
        lag).

     >  Dosage.

In addition, isopleths of maximum ozone deviation and ozone profiles  at
air quality monitoring stations are developed to provide information  on
the spatial and temporal perturbations in the model predictions.

     Note that these measures, to be defined more precise!;/ in the
following subsections, are quite similar to those used to evaluate model
performance in simulating the June and August oxidant episodes.


1.   Signed Deviation

     The signed deviation is calculated as follows:

                    N'         N     L        r

                                        '    "  b>i>j
                              11
where Cs ^ j and C^ ^ j are the ozone concentrations for the sensitivity
case and'tfie base case, respectively, at station (or grid cell)  i and  for
the hour j; N is the number of stations (or grid cells), and N1  the number
of simulation hours.

     Two signed deviations are calculated.  The signed deviation  is com-
puted for the ozone levels at the air quality monitoring stations and  for
the ozone levels in all ground-level grid cells (grid cells in mountainous
areas and over the Pacific ocean are not included).
                                   V-4

-------
2.   Absolute Deviation

     The absolute deviation in ozone levels is computed as follows:


                                                                   (5-2)
N
1 V
N' ^
1
N
N
Z
1=1
C . . - C, . .
b,i,j
Absolute deviation in ozone concentration levels (  i.e.,  base case  vs.
simulation) is computed for the air quality monitoring  stations  and for
the ground-level grid cells.
3.   Temporal Correlation

     The temporal correlation refers to the "timing"  of the  ozone  concen-
tration levels computed by both the sensitivity case  and the base  case at
a specified station or in a specified grid cell.   The temporal  correlation
at a given location is determined from the hourly concentrations predicted
by the sensitivity case and the base case at a given  location j.   A
correlation coefficient is then calculated for each station  according to
routine statistics.  These correlation coefficients  are normalized with
respect to the "perfect correlation line" (Hoel,  1962)  by performing the
following change of variable:
where r- is the computed correlation  coefficient  for  the  station or grid
                                    I

                                   N
cell j.  The mean value of the j's  is computed  for  all  locations:
where N is the number of stations or grid  cells.   Because  the  values of 4>j
are approximately normally distributed  (Hayes,  1978), the  average temporal
correlation coeYficient p is evaluated  from  the following  formula:
                                                                   (5-5)
                                   V-5

-------
Thus, the average temporal  correlation coefficient  is:


                                 (2? "  l     -                     (5-6)
                             exp (2 «) +  1

The temporal correlation is computed both from station  and  grid  statis-
tics.  Perfect correlation  exists  when p  =  1.0.


4.   Spatial Correlation

     The spatial correlation between the  concentration  fields  calculated
in a sensitivity run and those calculated in  the  base case  is  another
useful measure.  Hourly correlation coefficients  can be computed by consi-
dering the values of concentrations predicted by  the sensitivity case  and
by the corresponding base case for each station or  grid cell.  Then, the
estimation of the average spatial  correlation coefficient follows the  pro-
cedure described above for  the temporal correlation coefficient. Two
spatial correlation coefficients are computed from  station  statistics  and
from grid statistics, respectively.

     These sensitivity measures (signed deviation,  absolute deviation,
temporal correlation, and spatial  correlation) may  be evaluated  as  a func-
tion of concentration level.  They are computed in  this study  for ozone
levels > 12 pphm (the National Ambient Air  Quality  Standard) and for ozone
levels > 20 pphm.


5.   Overall Maximum Ozone  Level

     The maximum ground level ozone concentration in the basin is computed
for the sensitivity case and the base case  from station and grid statis-
tics.  The overall  maximum  ozone level and  the corresponding location
(station or grid) provide an additional measure of  model sensitivity.


6.   Maximum Ozone Statistics

     The maximum ozone levels occurring at  each monitoring  station  may be
computed for both the sensitivity and the base case.  The average peak
level normalized ozone difference may be  defined  as follows:
                                    *
                             N
                        1.
                        N
s.l
                              (5-7)
                                   V-6

-------
       *         *
where C   ^ and C b ^ are the maximum ozone concentrations for the  sensi-
tivity caSe and the'base case, respectively, at Station  i.  This measure
is computed for the 23 air quality monitoring stations within the computa-
tional grid.  Coastal stations are principally affected  by boundary condi-
tions and local emissions since the wind flow in both base cases is wes-
terly for the majority of the simulation time.  Accordingly, these
stations are removed from calculation of the maximum ozone statistics.
Peak level normalized differences are computed for the remaining 16
downwind stations.

     The average absolute peak time lag for maximum ozone level occurrence
is evaluated as follows:

                             N
                        1   ^                                     (5-8)
                        TT
                                   Ts-Tb
      T  and Tb are the time of the maximum ozone  level  in the  sensitivity
     ind in the base case, respectively.  This measure  is evaluated  for
where
case anc
the 23 stations and for the 16 downwind stations.
7.   Dosage

     Dosage measures were based on the gridded area with simulated  ozone
levels above 20 pphm.  Dosage is obtained by adding the number of grid
cells witn ozone levels above 20 pphm over the entire  19-hour simulation
period.  This measure is computed for both the base cases and sensitivity
cases.  The normalized difference of dosages provides  an additional
measure of the model sensitivity:


   Normalized  /Dosage of sensitivity case - Dosage of base case
   difference  \Dosage of base case

8.   Isopleths of Maximum Ozone Deviation

     Quantification of the spatial changes in predicted ozone levels  aids
in the interpretation of the sensitivity results.  Isopleths of the devia-
tions in maximum ozone concentrations are presented.   It should be  noted
  The coastal stations that are not included in this  analysis  are Costa
  Mesa, El Toro, Laguna Beach, Long Beach, Los Alamitos, Redondo Beach,
  and West Los Angeles.
                                     V-7

-------
that maximum ozone concentrations may occur at the same location at dif-
ferent times.  This isopleth presentation provides useful information
about the magnitude and location of changes in ozone concentrations.
9.   Ozone Profiles at Air Quality Monitoring Stations

     Comparison of calculated and observed ozone concentrations at various
air quality monitoring stations is another useful measure of model sensi-
tivity; accordingly, the time-varying ozone concentrations at the monitor-
ing stations are compared for the sensitivity and base cases.  This
provides information on the temporal variation and magnitude of the
perturbations in ozone concentrations at various locations throughout the
basin.

     To keep the display of simulation results to a manageable level, 6
stations were selected from the 23 monitoring sites for detailed discus-
sion of the sensitivity results. (Results for all stations are given in
appendix A.)  The six monitoring locations are:

     >  Anaheim                               >  Pasadena

     >  La Habra                              >  San Bernardino

     >  Lynwood                               >  Upland.

These stations were chosen because they typically present discernible
deviations in ozone concentration levels between the sensitivity and base
cases.  Moreover, the stations are aligned with the wind flow trajectory
that carries the photochemical plume eastward across the basin.  Where
indicated, additional monitoring stations that present interesting fea-
tures in the diurnal ozone profiles will be discussed in some sensitivity
studies.
10.  Summary of Sensitivity Measures

     These sensitivity measures are considered in the following discus-
sion:
     >  Signed deviation, absolute deviation, temporal correla-
        tion, and spatial correlation of ozone levels for the
        following selected station locations and grid cells.

        -  Monitoring station grid cells with ozone levels above
           12 pphm.
                                   V-8

-------
        -  Monitoring station  grid  cells with ozone  levels  above
           20 pphm.

        -  Grid cells with ozone  levels above 12  pphm.

        -  Grid cells with ozone  levels above 20  pphm.

     >  Overall maximum ozone  level  predicted by  the sensitivity
        case and the  base case and  the corresponding location of
        occurrence for:

        -  Station statistics  (i.e., statistics computed  based on
           ozone predictions  in grid cells  containing monitoring
           stations).

        -  Grid statistics (i.e., statistics based on ozone
           predictions in all  ground level  grid cells).

     >  Normalized differences and  peak time lag  of  the ozone peak
        level for:

        -  All 23 stations

        -  16 downwind stations.

     >  Dosages of the sensitivity case and of the base case, and
        the corresponding normalized difference.

     >  Isopleths of  maximum  ozone  deviations between the sensi-
        tivity case and the base  case.

     >  Diurnal ozone concentration profiles predicted by the
        sensitivity case and  the  corresponding base  case  at some
        monitoring stations.

Appendix A (bound separately)  presents the  detailed  results for each
simulation.
C.   SUMMARY OF SENSITIVITY RESULTS

     Major features of the sensitivity simulations  are summarized  in  table
V-l.  Signed deviations, absolute deviations,  temporal  correlations,  and
spatial correlations are presented in tables V-2 through  V-5  for ozone
levels above 12 pphm and 20 pphm, for both  station  and  grid statistics.
                                   V-9

-------
                TABLE V-l.  SUMMARY  OF  SENSITIVITY  STUDIES
Sensitivity
Study
1
2
3
4
5-1
5-2
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
Day of ^
Simulation
J
A
J
A
J
J
J
A
J
J
A
J
A
J
J
A
J
J
A
J
J
J
J
Change in Input Data
Reduced upper air meteorology
Reduced upper air meteorology
Reduced upper air and surface meteorology
Reduced upper air and surface meteorology
Reduced air quality
Reduced air quality and meteorology
Reduced upper air quality
Reduced upper air quality
Nitrogen oxides boundary conditions
Hydrocarbon initial condition
Hydrocarbon initial condition
Hydrocarbon speciation
Hydrocarbon speciation
Mobile sources— Older inventory
Mobile sources— Less detailed inventory
Mobile sources— Less detailed inventory
Point sources— Temporal resolution
Area sources— Spatial resolution
Area sources— Spatial resolution
Area sources— Temporal resolution
Larger grid size (10 km)
Reduction of grid cell layers --two-layer model
Reduction of grid cell layers-- single-layer model
* J = 26 June 1974, A = 4 August 1975.
                                  V-10

-------
TABLE V-2.  SENSITIVITY MEASURES FOR OZONE CONCENTRATIONS
            ABOVE 12 pphm
                       Station Statistics
Sensitivity
Study
1
2
3
4
5-1
5-2
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
Signed
Deviation
-0.177
-0.094
-0.172
-0.146
0.000
-0.208
0.105
-0.073
-0.141
-0.344
-0.450
0.046
0.121
0.027
0.064
0.115
-0.034
0.054
0.118
-0.005
0.148
-0.007
-0.250
Absolute
Deviation
0.300
0.164
0.320
0.189
0.028
0.329
0.106
0.073
0.145
0.344
0.450
0.046
0.125
0.037
0.067
0.127
0.035
0.055
0.127
0.007
0.193
0.053
0.377
Temporal
Correlation
0.023
0.912
-0.183
0.799
0.990
-0.183
0.981
0.980
0.914
0.869
0.680
0.989
0.969
0.988
0.988
0.977
0.990
0.990
0.977
0.990
0.848
0.975
0.787
Spatial
Correlation
0.622
0.638
0.310
0.648
0.986
0.313
0.975
0.953
0.927
0.800
0.578
0.985
0.818
0.983
0.983
0.738
0.990
0.982
0.789
0.990
0.634
0.972
0.347
                          V-ll

-------
TABLE V-3.  SENSITIVITY MEASURES FOR OZONE CONCENTRATIONS
            ABOVE 20 pphm

                       Station  Statistics
Sensitivity
Study
1
2
3
4
5-1
5-2
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
Signed
Deviation
-0.342
-0.034
-0.367
-0.073
0.003
-0.390
0.110
-0.048
-0.177
-0.384
-0.486
0.035
0.085
0.022
0.075
0.082
-0.034
0.057
0.103
-0.005
-0.037
0.006
-0.235
Absolute
Deviation
0.366
0.082
0.389
0.104
0.022
0.402
0.110
0.049
0.181
0.384
0.486
0.035
0.085
0.029
0.075
0.084
0.035
0.057
0.103
0.021
0.087
0.033
0.332
Temporal
Correlation
-0.009
0.768
-0.455
0.682
0.990
-0.402
0.983
0.982
0.933
0.984
0.280
0.990
0.961
0.987
0.989
0.955
0.990
0.990
0.963
0.900
0.982
0.989
0.790
Spatial
Correlation
-0.133
0.974
-0.185
0.937
0.986
-0.184
0.959
0.990
0.613
0.433
0.990
0.989
0.940
0.969
0.973
0.922
0.989
0.979
0.946
0.990
0.721
0.982
0.526
                           V-12

-------
TABLE V-4.   SENSITIVITY MEASURES FOR OZONE CONCENTRATIONS
             ABOVE 12 pphm
                         Grid Statistics
Sensitivity
Study
1
2
3
4
5-1
5-2
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
Signed
Deviation
-0.229
-0.046
-0.225
-0.107
0.006
-0.253
0.103
-0.071
-0.130
-0.317
-0.443
0.044
0.138
0.022
0.070
0.146
0.009
0.053
0.141
-0.008
0.100
-0.011
-0.067
Absolute
Deviation
0.329
0.168
0.347
0.198
0.032
0.355
0.108
0.073
0.139
0.318
0.443
0.044
0.142
0.039
0.073
0.160
0.010
0.055
0.149
0.010
0.200
0.055
0.207
Temporal
Correlation
0.134
0.878
-0.034
0.737
0.988
-0.067
0.977
0.981
0.925
0.807
0.722
0.986
0.966
0.984
0.986
0.963
0.990
0.988
0.969
0.990
0.831
0.975
0.776
Spatial
Correlation
0.376
0.396
0.247
0.450
0.980
0.265
0.061
0.965
0.896
0.800
0.806
0.985
0.923
0.983
0.9RO
0.906
0.990
0.984
0.919
0.990
0.397
0.965
0.701

-------
TABLE V-5.  SENSITIVITY MEASURES FOR OZONE CONCENTRATIONS
            ABOVE 20 pphm
                        Grid Statistics
Sensitivity
Study
1
2
3
4
5-1
5-2
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
Signed
Deviation
-0.366
-0.108
-0.392
-0.165
0.010
-0.416
0.110
-0.053
-0.165
-0.357
-0.463
0.041
0.090
0.018
0.085
0.095
0.007
0.060
0.103
-0.009
-0.053
-0.000
-0.086
Absolute
Deviation
0.378
0.171
0.443
0.187
0.032
0.459
0.113
0.057
0.169
0.357
0.463
0.041
0.091
0.034
0.085
0.108
0.008
0.060
0.107
0.012
0.150
0.041
0.162
Temporal
Correlation
0.012
0.321
-0.293
0.263
0.984
-0.253
0.972
0.950
0.928
O.RQ3
0.336
0.985
0.958
0.977
0.982
0.945
0.988
0.986
0.960
0.989
0.889
0.981
0.834
Spatial
Correlation
0.403
0.439
0.331
0.608
0.965
0.348
0.941
0.970
0.794
0.735
0.872
0.982
0.856
0.975
0.967
0.840
0.990
0.981
0.836
0.990
0.115
0.949
0.678
                             V-14

-------
The signed and absolute deviations for ozone levels  above  12  pphm  are pre-
sented graphically as well, in figures V-2 and  V-3,  respectively  (results
of sensitivity run 22 were obtained late and are  not included in these
figures).

     Simulations involving a reduction in the upper  air  meteorological
data or a change in the magnitude of reactive hydrocarbon  initial  condi-
tions result in the greatest signed and absolute  deviations.   Such devia-
tions range from 15 to 45 percent.  The change  in grid size results  in an
absolute deviation of about 20 percent.  Other  perturbations  in the  input
data result in absolute deviations of less than 15 percent.

     Reductions in meteorological data result in  the lowest temporal cor-
relations, with values as low as 0.023 and -0.183 for the  26  June  1974
simulation.  Changes in the hydrocarbon initial conditions also result in
a low temporal correlation coefficient for the  4  August  1975  simulation.
The same simulations (reduction in meteorological data,  change in  hydro-
carbon initial conditions) also lead to low spatial  correlation coeffic-
ients.  The change in grid size results in a low  spatial correlation
(i.e., p < 0.721), as is to be expected since the spatial  features of the
model have been modified.

     Table V-6 presents the ratios of the overall maximum  ozone level of
the sensitivity case to the base case, along with the location of  occur-
rence for station and grid statistics.  The largest  differences in overall
maximum ozone levels are obtained for the simulations that involve a
reduction in meteorological data or a variation in reactive hydrocarbon
initial conditions.  The lowest ratio of overall  maximum ozone levels is
obtained with the station statistics for sensitivity case  10  (hydrocarbon
initial conditions).

     Maximum ozone statistics are presented in  table V-7.   The peak  level
normalized differences and peak time lags are shown  for  the cases  where
all 23 stations and 16 downwind stations are considered.   The largest dif-
ference between these statistics is obtained for  simulation 5-1, which
involves a reduction in air quality data.  This result suggests that, for
the reduction in air quality data, the perturbation  is larger at the
coastal stations.

     The dosages and the corresponding normalized differences are  presen-
ted in table V-8.  In general, dosage is a sensitive measure  of model
perturbation; it is the most sensitive for simulation 20,  since the  change
in grid size affects the computation of the dosage noticeably.  Other
dosage-normalized differences appear to be consistent with the sensitivity
measures presented above.
                                   V-15

-------
                              O
                              CM
                          CT.
                          CO
                          CM

                          LO
                                              uo
                                              1/5   o
                                              i/)   t—

                                              z:   I—
                                              O   i/>
                                              i—»
                                              I—   Q
                                              <   i—'
                                              t—   cc
                                              1/1   C3
                                                                                                         Q.
                                                                                                         O.
                                                                                                         CM
O
CO
I/O
z
O
                                                                                                         UJ

                                                                                                    s-    z
                                                                                                    
-------
1 1 1 1


-
oo 	 	
H-t
(— oo
oo o
• — i i — i *•
I- i- 	
< oo
1— «-H
00 (-

O OO
1— t
oo cr
i — __
1
1

____.





,:



uu







i i i i
o «d- ro oo i — c
D CD O O 0

CNJ
(~ >
OO
CO
r^
LD
t
^1-
2

o


cn

CO
r~~
o
CvJ

i
LD
•a-

OO

C\J

r~~

3









n Number
o
4->
(0

J
00



















5:
CL
D-
C\J
UJ

CO
oo
ENTRATIOI^
* ^
•z.
o
LU
Z
O
l-v!

a:
o
a.

oo
C
i—t
i—
<=c
»-H
>
LU
O
LU
I—
=l
— 1
O
OO
m

ro
i
>
LU
Cr:
ra
i— t
u_







V-17

-------
TABLE V-6.  OVERALL MAXIMUM OZONE LEVELS
 Station Statistics
Grid Statistics
Simulation
Base case
26 June 1974
Base case
4 August 1975
1
2
3
4
5-1
5-2
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
Ratio Peak levels

1.000

1.000
0.718
1.038
0.710
0.988
0.997
0.683
1.109
0.927
0.804
0.632
0.523
1.023
1.128
1.020
1.101
1.139
1.000
1.067
1.171
0.988
0.909
1.019
1.054
Location

Azusa

Pomona
Upland
Upland
San Bernardino
Upland
Azusa
Upland
Upland
Pomona
Azusa
Upland
Upland
Azusa
Pomona
Azusa
Azusa
Pomona
Azusa
Azusa
Pomona
Azusa
Pomona
Azusa
Azusa
                                                  Location
                              Ratio Peak Levels     (x,y)
                                    1.000

                                    1.000
                                    0.814
                                    1.366
                                    0.755
                                    0.982
                                    1.251
                                    1.122
                                    1.019
                                    1.002
                                    0.832
                                    0.670
                                    1.018
                                    1.003
                                    1.027
                                    1.056
                                    1.033
                                    0.996
                                    1.046
                                    0.997
                                    1.002
                                    0.854
                                    1.020
                                    0.947
                33-14

                30-16
                30-16
                11-16
                30-16
                27-15
                19-15
                30-16
                19-15
                30-16
                33-14
                33-14
                30-15
                19-15
                30-16
                33-14
                19-15
                11-16
                33-14
                19-15
                11-16
                33-14
                20-14
                19-15
                14-15
                V-18

-------
TABLE V-7.  MAXIMUM OZONE STATISTICS
  Peak Level                 Peak Time lag
Sensitivity
Study
1
2
3
4
5-1
5-2
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
Normalized
23 Stations
0.253
0.128
0.249
0.143
0.061
0.294
0.162
0.057
0.188
0.320
0.394
0.114
0.303
0.028
0.070
0.243
0.029
0.064
0.254
0.008
0.155
0.050
--
uiTTerence
16 Stations 23
0.258
0.133
0.260
0.121
0.031
0.266
0.144
0.069
0.187
0.354
0.476
0.092
0.313
0.029
0.082
0.214
0.020
0.063
0.241
0.001
0.112
0.043
Not Available
^noi
Stations
1.77
0.57
2.17
0.82
0.52
2.25
0.56
0.04
0.87
0.96
1.00
0.30
0.25
0.35
0.13
0.12
0.13
0.13
0.12
0.00
0.54
0.22
_ _
ir)
16 Stations
1.87
0.50
2.37
0.81
0.00
1.93
0.12
0.06
0.75
0.56
0.81
0.44
0.39
0.44
0.19
0.26
0.19
0.19
0.26
0.00
0.75
0.25
__
                   V-19

-------
               TABLE V-8.  DOSAGES
                            Dosage    Normalized
	Simulation	   (km2)     Difference

Base case:  26 June 1974
Base case:  4 August 1975
           1
           2
           3
           4
           5-1
           5-2
           6
           7
           8
           9
           10
           11
           12
           13
           14
           15
           16
           17
           18
           19
           20
           21
           22
10,000
8,750
5,975
8,600
7,400
6,625
10,475
7,125
14,525
4,350
6,200
2,425
450
11,375
13,975
11,125
12,450
13,900
10,125
11,925
13,750
9,700
16,000
10,425
11,525
0.0
0.0
-0.4025
-0.0172
-0.2600
-0.2439
0.0475
-0.2875
0.4525
-0.5039
-0.3800
-0.7575
-0.9496
0.1375
0.5970
0.1375
0.1925
0.5900
0.0125
0.1925
0.5710
-0.0300
0.6000
0.0425
0.1525
                         V-20

-------
     Finally, table V-9 summarizes the results  of  the  22  sensitivity runs
in terms of the various measures  presented  in this chapter.   (These
results are drawn from those presented in tables V-2 through  V-8.)
D.   SENSITIVITY RESULTS

     In appendix A, results of the sensitivity studies  summarized  in this
chapter are presented and discussed.    Model  sensitivity  is examined by
means of various measures presented in  this  chapter  (ozone  level devia-
tions, temporal  and spatial correlations,  comparisons of  maximum ozone
levels, and dosage).   The largest  perturbations  caused  by reduction in
input data, manifested by changes  in  model input  files, are presented.
The sensitivity of model input variables,  such as wind  fields, mixing
depths, and emission  levels, to changes in available input data is ana-
lyzed to provide insight into how  restrictions in data  resources may
affect model performance.  Perturbations in  ozone concentrations through-
out the basin and at  specific monitoring stations are also discussed.

     In the following chapter, the sensitivity of the airshed model to
perturbations in available input data is interpreted based on the  results
presented here and in appendix A.
  For each sensitivity study,  corresponding  reduction  in  input data  is
  briefly introduced.   A more  detailed  presentation  of the procedure used
  to reduce the input  data has been made in  Section  III.
                                   V-21

-------
                                                         •O 9>
                                                         ai u
                                                         N C


                                    oo
                                     LU
                                     O

                                     O
                                     oo
                                     cr
                                     >-

                                     QL
                                     •=>
                                      u^  ^*>
                                                                  ro^-ro^-Orn-^C-^^^O— •OO^CO^HO^^O'*>

                                                                  ddddoddo'ddoodddddoodddd
                                                          Uj
                                                          C


                                                           "
                                                             
-------
                       VI   INTERPRETATION OF RESULTS
A.   RANKING OF DATA NEEDS THROUGH SENSITIVITY-UNCERTAINTY ANALYSIS

     A ranking of airshed model  data needs  must  take  into account the
sensitivity of model predictions to input data,  the cost of data acquisi-
tion, and uncertainties in the specification  of  model  inputs.  These
factors will be considered in defining  a sensitivity-uncertainty index
that will provide a basis for ranking data  needs.
1.   Definition of a Sensitivity-Uncertainty Index.

     The sensitivity of model  predictions  to the  level  of detail of  input
data is important in determining input  data  needs.   Several sensitivity
simulations have evaluated the effect on predicted  ozone levels of
perturbations in meteorological, air quality,  and emission data.  The
results of these simulations provide quantitative  information  on model
sensitivity.  Several measures of model  sensitivity, AJ, could be consid-
ered in developing a sensitivity-uncertainty index;  here consideration is
given to the absolute normalized mean deviation of  ozone concentrations
above 12 pphm and the absolute normalized  deviation of  the air-shed-wide
peak ozone level.  The former  has been  defined in equation (5-2).

     The absolute normalized deviation  of  the peak  ozone levels is
computed from the peak ozone levels of  the sensitivity  case and base case,
regardless of the grid of occurrence.  Let AJ be  the deviation of model
predictions resulting from input data perturbations.   In this  case AJ is
given in equation (6-1)
                       AJ =
                               C       - C
                                max,  s    max,b
                                     C
                                      'max,b
(6-1)
where C_,v c and Cm,v  h are the air-shed-wide maximum ozone  concentration
       IHaA, p      [MaA, u
in the sensitivity case and the base case,  respectively.

     Other measures, such as the deviation  in dosage,  might also  be
considered, but here we restrict our focus  to the  two  measures  given
above.
                                   Vi-1

-------
     In the following presentation  subscript  i refers to the meteorologi-
cal conditions (26 June 1974 or  4 August 1975) and subscript j to the
input data perturbed.

     A sensitivity index is  usually defined as the ratio of some model
output deviation to the corresponding model input perturbation.  If Alj is
the perturbation in the input data, the sensitivity index for the input
data j and meteorological  conditions  i  (26 June 1974 or 4 August 1975) is
defined as follows:

                                   AJ. .
                             rij •  IT1   •                        <6-2'
                                     J

This sensitivity index, r^., is  a measure of  the effect of a change in the
model input data on the model predictions.  In general, the deviation in
the model  predictions, AJ, may be expressed as follows:
                              AJ  =        ,                          (6-3)
where C is some model  output  measure  such  as the ozone level above 12 pphm
or the overall maximum ozone  level.

     Model input perturbations  differ in many cases, according to whether
meteorological, air quality,  or emission data were considered.  To compare
the various sensitivity studies, it  is  necessary to normalize these input
perturbations.  This may be done by  introducing a cost index that relates
the cost AKj of acquisition of  needed data (to raise the level of detail
in input data j from the sensitivity  case  to the base case) to the change
in input data:

                                  AK.
                                     i     .                         (6-4)
     Next, the sensitivity measure  is normalized with respect to a cost
difference corresponding to the  change  in  input data, because the ranking
of data needs clearly must include  such  information.  The normalized sen-
sitivity index is then defined  as follows:


                             S,j -  Ji     ,                         (6-5)
                                   VI-2

-------
that is,

                                   AJ. .
                                                                   (6-6)
     This sensitivity index is  a measure  of  the effect of the cost (of
improving the level of detail  of input  data  j) on  the model predictions
for meteorological  conditions  i.  It  is,  however,  a  "perturbation--
specific" measure of model  sensitivity.   That  is,  the value of the
sensitivity index has been  evaluated  for  a specific  perturbation.  The
classical sensitivity index, on the other hand, describes the sensitivity
of the system to any perturbation and is  a function  of the perturbation
level unless the sensitivity relationship is linear.  As shown in figure
VI-1, when the sensitivity  relationship is nonlinear, the "perturbation-
specific" and classical sensitivity indexes  are identical only at the
point where the former has  been calculated.

     Actually, the introduction of a  cost variable probably makes the
sensitivity relationship nonlinear, since a  minimum  threshold cost is
necessary before an input data file may be improved.  Before that
threshold, the model is insensitive to  the cost and  the classical sensi-
tivity index is zero; in that  range,  the  "perturbation-specific" index is
not representative of the model sensitivity.   Therefore, when applying the
results of these sensitivity studies  to a different  urban area, one should
consider variations in the  input data similar  to those variations con-
sidered in this work.  This will allow  for the best  use of the information
obtained from this sensitivity analysis,  which provides "perturbation-
specific" sensitivity indexes.   However,  the general methodology employed
here should be applicable,  in  principle,  to  other  areas with differing
characteristics.

     The need for increasing the detail  of input data should be weighted
by the accuracy available for  corresponding  model  input variables.  For
instance, an updated mobile source inventory will  provide detailed
information on mobile source emissions  of NOX  and  NMHC.  However, these
emissions are uncertain by  at  least 15  and 25  percent, respectively; thus,
  It should be noted that it would not be appropriate  to  conduct  the
  sensitivity analysis with respect to monitored  data,  since the  detailed
  effects of the model input variables are interrelated and complex,  and
  some misleading compensating phenomena could  occur during a particular
  sensitivity simulation.  For instance, the enlargement  of the grid  size
  gave better results (based upon few performance measures) than  the  base
  case when compared with experimental data.
                                   VI-3

-------
  .SENSITIVITY CASE
                                           SO

                            ^ - SLOPE  =  4
                                            \J
                                           ("PERTURBATION SPECIFIC"
                                           SENSITIVITY INDEX)


                                           (CLASSICAL
                                           SENSITIVITY INDEX)
 AK. = 0     AK.
   J           J
EXISTING   THRESHOLD
DATA BASE
                                         AK*

                                      BASE CASE
FIGURE VI-1.   COMPARISON OF "PERTURBATION-SPECIFIC"  AND
              CLASSICAL SENSITIVITY INDEXES FOR A
              HYPOTHETICAL CASE
                         VI-4

-------
the available emission data are clearly limited.  Since data for NOX
emissions are more reliable than data for RHC emissions, they would  be  of
more use in improving the accuracy of model predictions.

     Note that uncertainties in the model input variables as discussed
here (see figure V-l) result from limitations in a mathematical, physical,
or chemical procedure, such as the uncertainty in wind direction that
exists in the computed wind field or the uncertainty in emission data for
a detailed mobile source  inventory.  The uncertainty of model input
variables due to the level  of detail of the input data is not considered
in the uncertainty index.  It is taken into account in the sensitivity
analysis:  a change in input data introduces a change (or uncertainty)  in
the model input variables that leads to a perturbation in the model
output.  This uncertainty in the model input variables is therefore
implicit in the sensitivity index, which could be rewritten as follows:

                               AJ..  AP..

                          'u-ST-ir    •                     <6-7'
                                 1J    J

where AP^ is the change  in the model input variable P^, that is, the
uncertainty in PJ -• resulting from the level of detail of the input data.
                 ' J

     The following uncertainty index is defined:
                              ,,
                               J
                                   6P. .
                                   5-L1     ,                        (6-8)
where 6P^j is the uncertainty  in  P^ resulting from limitations in a
mathematical or physico-chemical  representation of a process.  It is
defined for the input data  j and  the meteorological conditions i.  The
model input variable P.,- ,•  is the variable that most affects model  perfor-
mance during the sensitivity study defined by i and j.  It may be differ-
ent, depending on meteorological  conditions.  For example, a reduction in
upper air meteorological  data  seems to modify the wind field for  the 26
June 1974 simulation most,  whereas the wind field for the 4 August 1975
simulation was not as affected; in contrast, for the August run,  the
mixing height is the most perturbed input variable.

     The uncertainty index  is  defined as a dimensionless number.   This
allows consistency among  uncertainties in input variables of a different
nature (meteorological, chemical, or emission variables).  Considering
dimensionless perturbations and indexes is a common procedure in  sensitiv
ity-uncertainty analysis  (Heinrich et a!., 1977; Dougherty et al., 1979).
                                   VI-5

-------
     The larger the uncertainty in the input  variable,  the  less  need for
additional  input data.   A very inaccurate  parameter of  the model  lowers
the need for improving the input data base to compute  its value  (which has
an inherent uncertainty).  Hence,  we weight the  sensitivity  index  S^ by
means of the uncertainty index U^,-,  and a  sensitivity-uncertainty inaex is
defined as  follows:
                                                                    (6-9)

     A large value of the index E..  corresponds  to  a  high  ranking for data
needs for the set of input data j  under  the  meteorological  conditions i
for the Los Angeles urban airshed  simulations.
2.   Cost Estimates for Data Acquisition

     Estimates of cost for acquiring data to  upgrade  a  data  base  from one
corresponding to the sensitivity case to  that  of  the  base case have been
made and are listed in table VI-1 for the various sensitivity cases
considered in this analysis.  The assumptions  made to derive these costs
are also listed.  Cost estimates are given as  a range of values because of
the uncertainty in data acquisition  costs and, in some  cases, because of
the nature of the assumptions (which may  define an upper and a lower
bound).

     These cost estimates should be  seen  as an illustration  of the proce-
dure for defining sensitivity-uncertainty indexes.  Application of this
analysis to a different urban area will undoubtedly require  updating and
modifying of these estimates.
3.   Uncertainty Estimates for Model  Input  Variable^

     The chemical species concentrations  are  computed  by  solving  a  set of
N continuity equations.
                   £ V c.  + Ri  + S.  + L.     i  =  1,  ....  N
. i   •  '  J ^ i    '  =  *  w i  '"i    "" i  ''"i     '   A»  •••»"     i     V O ~ •!• U J
                                   VI-6

-------
                        TABLE  VI-1.
             COST  ESTIMATES  FOR  DATA  ACQUISITION
              (FOR  THE  SOUTH  COAST AIR BASIN)
        Input Data
Upper air aeteorology



Surface meteorology



Surface air quality





Upper air quality




1C for RHC




RHC speciation
Mobile sources.
  Old inventory  simulation

Mobile sources •
  Gas sales  simulation
Point  sources


Area sources


Area sources
    Improvement of Data

 Upper winds, mxing depth
Surface winds, tempera-
ture
Ambient concentrations
of NO,, RHC, and 03
Upper air concentrations
of HO,., RHC and 03
Relationship to relate
RHC to THC
Distribution of HC
speciation according
to source categories
Update of an old
inventory

Creation of  an  inventory
More specific  resolution
of emissions  rates

Better spatial  resolution


Better temporal  resolution
                                                                           Assumptions
8 week program (10-12  sampling days):  3
acoustic radars,  4  daily  aircraft sound-
ings, 4 daily pibals  at 3  locations.

Surface meteorological network expanded
during eight week simmer  smog season.
6 meteorological  stations.

Surface air quality network expanded
during eight week summer  smog season.
6 air quality monitoring  stations (upper
bound for permanent stations, lower bound
for temporary stations).

Eight week program  (10-12  sampling days).
Early morning,  morning and noon soundings
at upwind and major source regions.   Night-
time sampling aloft over entire airshed.

Reactive hydrocarbon monitoring for an
eight week period during the summer smog
season.   Monitoring upwind and at major
source region sites.

Lower bound:   Breakdown of sources into
categories.   Assume known HC speciation
for each category.  Upper bound:
Literature review of hydrocarbon  emissions
speciation data.  Limited source  testing of
major categories  of emissions sources.

Old inventory available.   Acquisition of
updated  vehicle origin-destination data.

Routines such as  MOBILE 1 and EMFAC  5
available.  Creation of  vehicle  origin-
destination data files.

Characterization  of day-specific  emissions
rates.

Characterization  of spatial distribution of
area sources.

Characterization  of temporal  profiles of
all  major stationary sources.
    Cost  Estimate

S 50.000  -  100.000



  20,000  -  30,000



  75,000  -  150,000






  50,000  -  125,000




  75,000  -  150,000




  20,000  -  100,000







   5,000  -  20,000


 250,000  -  1,000,000



  10,000  -  50,000


  50.000  -  100,000


  60,000  -  150,000
                                                        VI-7

-------
where
              c.j  = the concentration of species i,
               v  = the wind vector,
               K  = the eddy diffusivity tensor,
              R.J  = the reaction rate,
              S.j  = the emission rates,
              L.J  = the removal rates,
               t  = the time,
               N  = the number of species considered in the model
Initial conditions and boundary conditions are prescribed.  Boundary con-
ditions must be given at the boundaries of the airshed both at ground
level and at the top of the mixing layer.

     Input variables in this system of equations are the initial and
boundary species concentrations, the wind vector, the eddy diffusivity
tensor, the rate parameters (which depend on irradiation intensity and
temperature), the emission and removal rates, and the depth of the mixing
layer.  For convectively mixed summertime conditions the eddy diffusivity
is not a parameter to which the airshed model is highly sensitive.  This
may be compared with wind velocity variations that may substantially alter
model predictions (Liu et al., 1976).  Reaction and removal rate coeffic-
ients are not affected in the sensitivity studies considered here.
Therefore, estimates for uncertainty are restricted to the following
variables:  wind velocity, species concentrations, emission rates, and
mixing depth.

     The importance of wind direction on model performance depends on the
spatial distribution of emission sources.  If the spatial distribution of
emission rates were uniform throughout the airshed, wind direction would
have little influence on the predicted air quality.   If, however, a
strong gradient exists in the emission source distribution, the wind
direction may be a major variable.  According to average emission rate
  It would have some effect, however, because the airshed has a finite
  size, and the amount of precursor transported to a receptor will vary
  according to the travel from the boundaries.
                                   VI-8

-------
isopleths of the Los Angeles basin,  an estimate of  the  gradient  in
emission sources corresponding to wind direction uncertainties   has been
made.

     Estimation of uncertainties in  chemical  species  concentrations was
restricted to nitrogen oxides and hydrocarbons,  since these  are  the main
precursors of photochemical  smog.  Uncertainties resulting from  averaging
these concentrations over each grid  cell  were not considered.  Only
uncertainties resulting from ambient monitoring were  used.   The  uncer-
tainty in the relationship between reactive hydrocarbons  and total
hydrocarbons was estimated from the  experimental data used to derive the
relationship.

     The model input variables considered in  this analysis are listed  in
table VI-2, along with the corresponding  uncertainties.
4.   Ranking of Data Needs

     The sensitivity-uncertainty indexes were  calculated  as  follows:   For
instance, the reduction in upper air meteorological  data  (set  1, j =1) for
the 26 June 2974 (i=l)  simulation leads to a relative  absolute deviation
in ozone levels above 12 pphm,

                              AJn  = 0.33.

A cost estimate for data acquisition AK^ = $50,000,  leads  to
                                   =6.6x10
When the upper air meteorological  data are reduced  on  26  June  1974, the
wind field is the most perturbed model  input  variable  and the  associated
uncertainty range is 10 to 50 percent.   For V^  =  10 percent,  the follow-
ing value of the sensitivity uncertainty index  is obtained:

                                1 1           -^
                          E    = 7^ = 6.6  x 10 °.
Similarly, a lower bound is obtained with  cost  and  uncertainty  estimates
of $100,000 and 50 percent, respectively:  £,, =  6.6 x  10~
                                                        -o
     The results of the sensitivity-uncertainty analyses  are  listed in
tables VI-3 and VI-4 for 12 sets of input  data  for  the  average  deviation
in ozone levels above 12 pphm and the  deviation in  air-shed-wide peak
  Uncertainties in wind direction and wind  speed  depend  on  wind  speed.
                                   VI-9

-------
     TABLE VI-2.  UNCERTAINTY ESTIMATES FOR MODEL INPUT VARIABLES
   Input Variable
Wind direction*
Emission sources
Wind velocity
Mixing depth
NO  concentration
  n
RHC concentration
RHC = function (THC)
NO  emissions
HC emissions
Uncertainty
20°-50°
10-50%
0.5-1 m/sec
50-70%

6%
20-60%
25-65%
10-20%
20-30%
          Reference
Tesche and Yocke (1978)
Present Study "
Tesche and Yocke (1978)
C. D. Unger (1976)
P. B. Russell and E. E. Uthe
(1978)
Burton et al. (1976)
Burton et al. (1976)
Present study
Present study
Present study
* Emission sources uncertainty is related to wind direction
  uncertainty.
                                 VI-10

-------
TABLE VI-3.   SENSITIVITY-UNCERTAINTY INDEXES:   SENSITIVITY TO OZONE LEVELS
                ABOVE  12 pphra

Input Data
Upper air meteorology


Surface meteorology

Surface air quality
Upper air quality
1C for HC

HC speciation

Mobile source
updating inventory
Mobi le source
gas sales
Point sources
Area sources
spatial resolution
Area sources
temporal resolution
Ozone
Level
Deviation*
0.33
0.17

0.007
0.06
0.03
0.11
0.32
0.43
0.045
0.14
0.04

0.07
0.16
0.01
0.055
0.14
.01

(J)
(A)

(J)
(A)
(J)
(J)
(J)
(A)
(0)
(A)
(J)

(J)
(A)
(J)
(J)
(A)
(J)

Cost for Data
Acquisition
(dollars)
50,000-100,000


20,000-30,000

75,000-150,000
50,000-125,000
75,000-150,000

20,000-100,000

5,000-20,000

250,000-1,000,000

10,000-50,000
50,000-100,000

60,000-150,000

Model Input
Component Affected
Hind field (direction)
Emission source
Mixing depth
Bind field
Emission sources
NOX, RHC
NOX, RHC
RHC

RHC

NOX, RHC emissions

NOX, RHC emissions

NOX, RHC emissions
NOV, RHC emissions
X *
NOX, RHC emissions

Sensitivity-Uncertainty
Index1. x 106
Uncertainty
25°-50°
10-501
50- 70*
10-50*

20-60*
20-60J
20-60*

20-60S

10-301

10-30*

10-30*
10-30*

10-30*


6
2


.6
.4

0.5
4
0
1
3
4
0
2
6

0.
0,
0.
1.
4.
(J-1
- (21) -
- (4.1) -

- (1-3) -
)
66
• 6.8

3.5
- (11) - 30
.33
.5
.6
.8
.75
.3
.7

23
,53
67
8
,6
0.22


- (0.82)
- (4.D -
- (8.7) -
- (11.7)
- (2.9)
- (9.0) -
- (23) -

- (0.8)
- (1-8)
- (2.6)
- (4.5) -
- (11.5)
- (0.6)

- 2
11
21.3
- 28.6
- 11.25
35
80

- 2.8
- 6.4
- 10
11
- 28
- 1.7

 J and A refer to the 26 June 1974 and 4 August 1975 simulations, respectively.

 Lower bound, geometric  mean value,  upper bound.
                                          VI-11

-------
       TABLE VI-4.   SENSITIVITY-UNCERTAINTY INDEXES:  SENSITIVITY TO
                    AIRSHED-WIDE PEAK OZONE LEVEL
      Input  Data
     Basinwide Peak
Ozone Level Deviation**
Sensitivity-Uncertainty
     Index § x 106
Upper air meteorology
Surface meteorology
Surface air quality
Upper air quality
1C for HC
HC spec i at ion
0.188 (J)
0.159 (A)
0.365 (J)
0.243 (A)
0.018 (J)
0.121 (J)
0.167 (J)
0.330 (A)
0.018 (J)
0.005 (A)
3.76
2.3
24.3
16.2
0.2
1.6
1.9
3.7
0.3
0.08
- (11.9) -
- (3.8) -
- (66.6) -
- (44.4) -
- (0.5) -
- (4.4) -
- (4.6) -
- (9.0) -
- (1.15) -
- (0.3) -
37.6
6.4
182.5
121.5
1.2
12.1
11.1
22.0
4.5
1.25
 Mobile source
 updating  inventory

 Mobile source
 gas  sales

 Point  sources

'Area sources
 spatial  resolution

 Area sources
 temporal  resollution
     0.026 (J)
     O.O(J)
4.3   - (15.0)  -
52
0.056 (J)
0.035 (A)
0.005 (J)
0.045 (J)
0.003 (A)
0.2 -
0.1 -
0.3 -
0.15 -
0.1 -
(0.7) -
(0.4) -
(1.2) -
(1.15) -
(0.25) -
2.2
1.4
5.0
9.0
0.6
          0.0
   The  basinwide  peak  ozone  level, deviation is the absolute relative difference
   between  the  sensitivity  and  base  case  basinwide peak  ozone levels.

 * J  and  A  refer  to  the 26  June 1974 and  4 August 1975 simulations, respectively.

 § Lower  bound, geometric mean  value,  upper bound.
                                   VI-12

-------
ozone levels, respectively.   Rankings  of  data  needs based on the sensitiv-
ity-uncertainty indexes for  the 26  June 1974 and  4 August 1975 simulations
are shown in figures VI-2 through  VI-5.   Note  that for the  average
deviation in ozone levels above 12  pphm,  only  the ranking of meteorologi-
cal input data (upper-air meteorology  and surface meteorology) is affected
by the meteorological  conditions.   Although the values of the sensitivity-
uncertainty indexes of the other input data  (air  quality, chemistry,
emissions) vary according to the meteorological conditions, their relative
ranking is not affected.  This suggests that meteorological conditions
should be taken into account primarily when considering the need for
meteorological input data.

     Consider the 26 June 1974 sensitivity results for the  average
deviations in ozone levels above 12 pphm.  Updating a mobile source
inventory is ranked the highest.  This is because of the relatively low
cost involved and because of the reasonable  accuracy of mobile source
emission rates.  Upper-air meteorology may be  important; this depends,
however, upon the meteorological conditions.   The initial conditions for
reactive hydrocarbons are also ranked  high,  despite the relatively high
cost of a large monitoring program. This, however, has been shown to be a
key input to the urban airshed model.  Model predictions may be greatly
affected by variations in hydrocarbon  initial  and boundary  conditions.
Reactive hydrocarbons, which are a  necessary precursor of photochemical
smog, are difficult to measure accurately in ambient air; they constitute
a small amount of the total  mass of hydrocarbons, which is  composed
primarily of methane.   They are usually determined via the  relationship of
their concentrations to nitrogen oxides or total  hydrocarbon concentra-
tions.  However, there is considerable uncertainty involved in these
empirical formulas.

     Spatial resolution of area sources,  upper-air pollutant concentration
data, detailed hydrocarbon speciation, and a detailed point-source inven-
tory are generally of similar importance  in  model performance.  The
importance of surface meteorology is of the  same  order, though it varies
with meteorological conditions.

     Surface air quality is ranked  relatively  low.  This is because we
assumed a 10-station network in the Los Angeles basin for the sensitivity
study.  The sensitivity relationship is probably  nonlinear  and, if only 2
or 3 stations had been considered,  the importance of surface air quality
probably would have been higher.  This underscores the caveat that the
results of this sensitivity-uncertainty analysis  must be considered in
light of the assumptions made in this  particular  sensitivity study.   It
would be misleading to assume that  surface air quality data have little
effect on model predictions in other applications.
                                   VI-13

-------
        10'
                 oo
                        I/)
                        <
                        o
 X
 QJ
•a
 c
        10
          -5
                      CO
                      
S_ C
3 >—"
O

   C
QJ -r-


•1^ (tj



II
                             cn
                             o
                                  ac
                                  o:
•r- r—     i-
*f. O     O
   i.     »4-
 J- O
 QJ OJ
                          Q. OJ
                                            C
                                            o
   3   i—

uo O   3
QJ in   cr
(J QJ
1-CCL   S_

Or-   <
oo    QJ
QJ (O   Q.
S- Q.   Q.
«S 00   ^)
                                              C
                                              O
                                              •r-
                                              •U
                                              (O

                                              U
                                              QJ
                                              Q.
                                              oo
to
QJ
O

3
O
co
                                                      O
                                                     a.
                                             £
                                             o
                                             'o
                                             o
                                             QJ
        QJ
        O
        fO
        4-
        t-
        3
        OO
                                                                              (O
                                                                              3
                                                                              cr
                                                                              OJ
                                                                              U
                                                     s-
                                                     3
                                                    oo
a; o
o -M
i- C
3 0)
O >
oo c
   t—I
a>

•i- (O
.a 4->
o o
                                                                                               c
                                                                                               o
en in
QJ QJ
u cc:

3 i—
O ns
00 i-
   O
ro O.
QJ E
S- QJ
           FIGURE VI-2.
                  SENSITIVITY-UNCERTAINTY  INDEXES--SIMULATION OF  26 JUNE  1974
                  SENSITIVITY TO  OZONE LEVELS  ABOVE  12  pphm
                                                VI-14

-------
            10
              -4
,-5
                      oo
                      
oo c
I-H
CU
•i- (O
o ^j
O O
FIGURE  VI-3.
SENSITIVITY-UNCERTAINTY  INDEXES—SIMULATION OF  4 AUGUST  1975:
SENSITIVITY TO OZONE LEVELS  ABOVE 12  pphm

                  VI-15

-------
      10
        -4
   X
   OJ
   •a
   c
      10
        -5
   a>
   o
   c
   1/1
   c
  oo
      10
        -6
      10
        -7
                        cr
                        o
cr
o
"o
O
OJ
•M
0

<4_
S-
3
oo
e Source
ing Inventory
f— 4_>
•i- 03
-Q "O
O 0.
z ^>
'o
O
QJ
•F™

o
2:
s_
0
«4-

f 1
1—4
>1
4->
3
o-
S-
•^-
S-
QJ
C.
Q.

Sources
4-^
C
• ^
o
a.
c
o
4-J

-------
          10
            -4
       X
       QJ
          10
-5
       QJ
       
                   o>
                   
                 
s- c
3 QJ
o >
00 C
   t—I
QJ
f—• r—
•r- CO
X) 4->
o o
c
o
                           o
                           QJ
                           c.
                           oo
             c
             o
                                (/> O
                                QJ on
                                (J QJ
                                s_ a:
                                Z5
                                o •—
                                fO -tJ
                                QJ ro
                                j- a.
                               et oo
FIGURE  VI-5.
   SENSITIVITY-UNCERTAINTY INDEXES—SIMULATION
   OF  4 AUGUST  1975:   SENSITIVITY  TO AIR-SHED-
   WIDE PEAK  OZONE LEVEL

           VI-17

-------
     A detailed specification of transportation  patterns  is  a  rather
expensive task, and this results in  a very  low ranking of the  need for
such input data.  One may compare this result with  the simple  updating of
the same inventory.  This shows  the  importance of the assumptions that
have been made and indicates that data needs will vary according to the
existing data base.  Quantification  of the  temporal  distribution of
stationary source emissions is rather expensive  to  obtain and  has little
effect on model performance.  This results  in a  low sensitivity-uncer-
tainty index.

     If the basinwide ozone peak level,  instead  of  the average ozone
level, is considered as sensitivity  measure, the ranking of  the input data
needs is slightly modified.  The main changes occur for the  surface
meteorological data:  The peak ozone levels appear  to be more  sensitive
than the average ozone level to the  number  of meteorological stations at
ground level, and surface meteorological  data are ranked the highest for
both simulation days (see table VI-4 and  figures VI-4 and VI-5).  The
ranking of the other input data are  modified within the uncertainties of
the sensitivity-uncertainty index values  and no  other notable  changes are
observed.
5.  Conclusions

     The sensitivity-uncertainty analysis that  has  been  presented  should
be seen as a procedure for defining input data  needs,  and the  results
obtained in this section should be considered  as  an illustration of  the
method.  It is strongly recommended that  revised  costs and  uncertainties
specific to the photochemical  model application area be  estimated..  These
city-specific cost and uncertainty estimates can  then  be used  with the
results of the sensitivity studies (Aj's) to compute the sensitivity-
uncertainty indexes.

     The types of results obtained are presented  in figure  VI-6.   The
hypothetical cost estimates shown in table VI-5 were used along with the
uncertainty indexes of table VI-2.  The priorities  for acquisition of
additional input data could be inferred from such information  on data
needs.

     It should be noted that some limitations  exist in the  results of the
sensitivity studies, though their extent  may be difficult to estimate.
These limitations result from the specificity  of  the Los Angeles bas^n,
the length of simulation time, and the model characteristics (such as the
treatment of point sources or the wind field submodel).  These issues will
be considered in the following section.
                                   VI-18

-------
       10
         -4
       10
         -5
 x
 0)
 (C
•M


 OJ
       10
         -6
OJ
oo
      10
        -7
I      I      I      I      I


    26 June 1974 Simulation
                                                    I
                            4 Auqust 1975 Simulation i

                                                    J.
                                                                         c
                                                                         o
o
o
OJ
QJ
^r~
J-
• ^
«t
i.
(U
Q.
CL
=>




O
1C
o:
j_
o
1^.

o
t—t
o
2
 O
QJ CO
O Q^
J- CC.
3
O i —
oo ra
•f—
fl3 4-^
ai ra
V. CL

CL
D.
o
"o
s_
o
4_)
^

OJ
u

V4_
i-
3
oo

4/1
O)
o

3
O
00
ai
^.
•r>
/"i
o


tn
QJ
O
S-
3
O
oo
•*->
c
•r—
O
D.
^^
«C
3
c-
S-
•r-

(D
U

-------
TABLE  VI-5.    HYPOTHETICAL  COST  ESTIMATES  FOR  DATA ACQUISITION
           Input Data
        Assumptions
     Upper air meteorology    8 week  program
                                                          Cost
Estinate    sensitivity-uncertainty
(dollars)   	Index x  10&*	
                             60,000     11 - (25)  -  55  0
                                        4.0 - (4.8)  - 5.7
     Surface meteorology      6 stations
                             20,000    0.7 - (1.6)  ••  3.5
                                       6 - (10) - 17.0
     Surface air quality      6 stations


     Upper air quality       8 week  program


     1C  for RHC              8 week  program
                             75,000    0.7 - (1.2)  -  3.0


                            100,000    1.8 - (3.2)  -  5.5
                            100,000     5.3 - (9.2)  -  16.0
                                        7.1 - (12.3)  -  21.5
     RHC Speciat ion
     Mobile sources
Known  HC  speciation
by categories


No inventory available
 40,000     1.9 - (3.2) - 5.6
           5.9 - (10.0)  - 17.4
250,000     0.9 - (1.6)  - 2.8
           2.1 - (3.7)  - 6.4
    Point sources
    Area Sources
     Area  sources
     SO; Emissions
Characterization of

emission rites


Characterization of
spatial  distrioution


Characterization of
temporal distribution


Data for point source
inventory available
                                                         40,000     0.8 - (1.4)  - 2.5
                                                         80,000
120,000
                                                         20,000
           2.3 - (4.0)  - 6.9
           5.9 - (10)  -  17.6
                                                                   0.3 - (0.5)  - 0.£
      Lower bound, geometric  mean value, upper  bound
                                       VI-20

-------
B.   GENERALIZATION OF RESULTS

     The sensitivity studies of the airshed model  have been  carried  out
based on one-day simulations of the Los Angeles  basin.   Because these
results may be of general interest to the photochemical  modeling  commun-
ity, it is helpful to define the specific features  of  the  simulations and
to evaluate the limitations of the sensitivity results.  Then  it  will then
be possible to extend the conclusions of this  study and  to outline general
recommendations for the application of the study results.
1.   Specific Attributes of the Simulations

     The results of the sensitivity simulations  depend  on  the  perturba-
tions introduced in the input data, the area chosen  for the  study, the
atmospheric conditions, and the modeling conditions  (e.g., type  of wind
model and length of simulation time).   Limitations of the  analyses have
been discussed in the previous section and the effect of aerometric
conditions has been considered in several  sensitivity runs carried out
under different meteorological and air quality conditions.   At this  point
it is appropriate to consider the effect on sensitivity results  both of
the specific attributes of the Los Angeles basin and of the  modeling
conditions.
a.   Specific Attributes of the Los Angeles Basin

     The meteorology of the western coast of the  United  States  is  strongly
affected by the presence of the North Pacific anticyclone.   The existence
of a subsidence inversion layer above the marine  layer,  resulting  from the
adiabatic warming of descending air, leads to the trapping of air  pollut-
ants under an inversion layer that lies typically at an  elevation  of  about
500 to 1000 meters.  The presence of a temporally and  spatially varying
mixing height in the Los Angeles basin affects the generality of the  model
sensitivity results, since it probably enhances the need for upper-air
meteorological and upper air quality data.

     The Los Angeles basin, because of its location, is  also characterized
by a coastal-desert meteorology.  The land-sea-breeze  effect is enhanced
by the presence of warm inland areas such as the  Mojave  desert, and this
recirculating meteorology largely determines the  atmospheric transport and
dispersion of pollutants in the basin.

     It should be noted that this land-sea-breeze effect is  observed  in
many locations.  It has been suggested, for instance,  that air pollution
episodes observed in Milwaukee are induced by a  land-lake-breeze effect
                                   VI-21

-------
(Lyons and Cole, 1976);  experimental  and  modeling  studies  in  the  Tampa/St.
Petersburg area have retrieved the main features of  the  land-sea-breeze
effect (Liu et al., 1979).

     The location of the Los Angeles  basin  is  such that  pollution observed
is a result of local emission of primary  pollutants.   This may not be the
case for other urban areas  where pollution  episodes  may  be, in part, the
result of pollutant emissions that occurred far upwind.  Usually,  the
importance of the boundary  conditions depends  on meteorological cond -
tions.  For instance, a  steady westerly flow can occur in the Los Angeles
basin for several days (e.g., the air pollution episode  of 4  August
1975).  In this instance, the boundary conditions  will have little
importance, since the air flowing into the  airshed should  have typical
background concentrations of pollutants.   During the  land-sea-breeze
regimes, however, polluted  air is carried offshore during  the evening and
the aged urban air mass  is  carried inland the  next morning.   In this case
the definition of the western boundary conditions  of the airshed  is
important.  For other urban areas, various  meteorological conditions can
also affect the importance  of air quality data at  the boundaries  of the
airshed.

     The topography of the  Los Angeles basin is characterized by  complex
terrain.  This feature will be discussed  in more detail  in the following
section, since it determines the selection  of  the  wind model.  In general,
the complex terrain increases the need for  meteorological data.

     The air quality network in the Los Angeles basin is dense and wide-
spread; such a rich air  quality data  base is not available in other urban
areas.  This has been emphasized in the previous section on the ranking of
data needs, and it should be taken into account in determining the
importance of air quality data from these sensitivity studies.

     Primary pollutants  in  the Los Angeles  basin are  emitted main'Iy from
mobile sources, which comprise about  44 percent of the reactive hydrocar-
bon emissions and 50 percent of the total  nitrogen oxides emissions.  This
should be considered in  the generalization  of  those  sensitivity stud es
that involve modifications  of the emission  inventories.
b.  Modeling Condition^

     The modeling conditions affect the performance  and  sensitivity  of  the
model.  Some issues, such as the grid  size  or  the  number  of  grid  layers,
have been addressed in the sensitivity analyses  (sensitivity cases 20 and
22).  Here some aspects of modeling conditions that  may  affect the
generality of the results are considered.
                                   VI-22

-------
     The wind model is an important component of the urban airshed  model,
and the use of various wind models (interpolated wind model,  two-dimen-
sional, and three-dimensional wind models)  has been investigated (Reynolds
et al., 1979).  In this study, a three-dimensional  wind model  developed
for complex terrain was used (Yocke et al., 1979).   The wind  field  is
obtained from the solution of a parameterized three-dimensional  boundary
value problem.  Therefore, this wind field  preparation procedure is more
sensitive to atmospheric data at the boundaries of  the airshed and  at the
top of the mixing layer than are other procedures based, for  instance, on
interpolation by inverse weighting.

     The mixing height and eddy diffusivities depend on time  and loca-
tion.  The computational procedures used to determine these atmospheric
parameters have been presented elsewhere (Reynolds  et al., 1979; Killus et
al., 1980).  Although other procedures could be considered, the  sensitiv-
ity of the air quality model to computational procedures for  the mixing
height and eddy diffusivities has not been  investigated.

     The limitation of model sensitivity results for emission inventories
stems primarily from the averaging of emissions over the grid cells.
Although an attempt to model microscale phenomena for line sources  has
been made (Reynolds et al., 1979), it was not used  in these simulations.
The scavenging of ambient ozone by nitrogen oxides  emitted from  power
plants and automobile exhausts is a well-known phenomenon in  air pollu-
tion.  The present treatment of these sources in the urban airshed  model
may obscure, in part, the need for greater  accuracy in the corresponding
emission inventories and, possibly, in other parameters (e.g., a finer
resolution of atmospheric dispersion.

     The length of simulation time has some effect  on the sensitivity
results.  For example, initial conditions at 0400 may represent  a sizeable
fraction of the total precursor emission burden between 0400  and the time
of the ozone peak (1200 to 1400).  If uncertainties exist about  the
boundary conditions of the pollutant concentrations, these uncertainties
may affect the model sensitivity more than  the uncertainties  in  emission
levels will affect it.  However, if a multiple-day  simulation were
performed, the influence of initial conditions on ozone levels would  be
markedly reduced (particularly for land-sea-breeze  regimes),  and the
emission inventories would have a dominant  role, since they would repre-
sent most of the precursor mass.  Clearly,  there is a need for analyzing
the sensitivity of the urban airshed model  for a multiple-day simulation,
to assess the importance of air quality data and emission inventories to
model predictions under such conditions.
                                   VI-23

-------
2.   Limitations of the Sensitivity Results

     Limitations of the sensitivity results  that derive  from the specific
attributes of the Los  Angeles  basin and  the  length of simulation time are
presented in table VI-6.   The  importance of  the air  quality data and emis-
sion inventories to model  predictions  should be strongly affected by the
length of simulation time.  For the reasons  mentioned above, when the
simulation is extended from a  single-day to  a multiple-day simulation, the
importance of air quality data will decrease, whereas the~effect of emis-
sion inventories on model  predictions  will increase.

     In sensitivity cases 1 and 2,  the number of upper air meteorological
stations was reduced.   The existence in  the  Los Angeles  basin of a strong
Inversion that varies  with time and location enhances the need  for upper
air meteorological data.   Although  the need  for upper air meteorological
data in other urban areas depends on local meteorology,  air pollution
episodes are often associated  with  strong inversions.  This would make the
results of this study representative of  a typical  urban  airshed simula-
tion.  Moreover, long-range transport  of pollutants  trapped in  the rear
part of an anticyclonic system above the inversion height is an important
air pollution phenomenon  in midwestern and eastern urban areas.  To
properly account for such phenomena, upper air meteorological and air
quality data would be  needed.

     Clearly, the spatial  variation of the mixing  height in midwestern and
eastern urban areas may be negligible  in comparison, and one upper air
meteorological station should  generally  provide the  information des;red.
In these sensitivity studies,  the ranking of meteorological input data
depends on the meteorological  conditions. The use of a  three-dimensional
wind model for the Los Angeles basin probably emphasizes the need for
upper air data as well.  This  model depends  on boundary  values, and the
reduction of input data at the upper boundary of the domain probably
affects the accuracy of this model.  A flat  urban  area would not require
the use of this wind model; consequently, a  simple interpolation wind
model would suffice, reducing  the need for upper air data.

     In sensitivity cases 3 and 4,  surface meteorological data  and upper
air meteorological data were reduced.  The density of the surface meteoro-
logical network in the Los Angeles  basin must be taken into account when
interpreting the sensitivity results.   In the previous section  on the
ranking of data needs, it was  pointed  out that the intensity of input data
in the base case and in the sensitivity  case was a major factor in the
sensitivity results.

     The same considerations apply  to  the reduction  in air quality data
intensity.  The density of the aerometric network  in the Los Angeles basin
                                                                330R/2

                                   VI-24

-------
           Of Q E O>
           QJ   -P- i—
           u.   to crt

           jf     -
           a      m
          ii;

          li
                                                                                                                                                  OJ 0
                                                                                                                                                  Q. «A
                                                                                                                                       £*
oo
z
UJ
                                                                                                                       O -i- tft
                                                                                                                       Q.X3 •—

                                                                                                                       ESP
                                                                                                                                                                    II
                                                                                                                                                                    •ft VI
                                                                                                                            isj O T3

                                                                                                                            L- C« QJ
                                                                                                                            O C •<--
                                              .CJ C C
                                               E - O
                                               re E •*- CD
                                                 t i_ CE
                                              T3 QJ QJ «E
                                               QJ 4-* Q.
                                              •— QJ    QJ
                                              ••- -O i/) .C
                                               fO    C •*->
                                              •*-» tJ O
                                               QJ Z r-  >,
                                              O CC 4J .0
                                                                                                                                                                    4J C *-"
                                                                                                                                                                     c •— c
                                                                                                                                                                     O  E t-
                                                                                                                                                                    a:  vi en
                                                                                                                              cn Q
                                                                                                                           — c  t-
                                                                                                                            
                                                                                                                                                                                   -       i—
                                                                                                                                                                                  u  o  c ai
                                                                                                                                                                                  a>  E  o —
O
HH


5
CO
OO
O
D-
                                 I —
                                —  O
                                f>  1-
           ,
       QJ cn
      -o o
QJ  U   W> OJ
                          m >
                            C
                          QJ «-
                                                                                                                                                                        o -•-
                                                                                                                                                                     t—    cn
                                                                                                                                                                     ID cr> O
                                                                                                                                  1C
CO
<:
•—      T3 Q. 3
   irt   ^ 01 ««-
^ 4>   QJ -O M-
oi —   ••-   —
•O £t   *+- CT»T3
b *o      c
SI -   T3 - 5s
   *.   c x -o
   *ol   — •- T3
                                         •  CT)TD
                                           c
                                        ^  —  X
                                        c  x -o
                                                       —     o
                                                       QJ -o -f-
                                                       >  C 4->
                                                       at  *a ••-
o —  x
1~ c  L.
(J -F-  *0
                         J3  C
                             O
                         •O •—
                          C 4-»
                     •r-  C
                      X  O
                      O  U
                                    O> U  tn
                                    o> *o  c
                                    O TD  O
                                               C ••- -O O
                                               3 *j  e -
                                               O —  3 *->
                                               l_ C  O -r-
                                                       -
 I  rt I   >,
-*- c o  c
 C - I-  O
 Cn**- ng _Q QJ
 •O QJ    L. C
 E — •  «3 QJ
   O t/i  (_> r- M
 OJ    C    >> C
 > **- -^   - .c O
•<- O **-  I/I 4-» •-
                                                                                                     O)  3 (D T  C
                                                                 QJ
                                                                 -o  c  i- c
                                                                 3  TJ  O O
                                                                 4->    *-* •«-
                                                                 •f   •  O i/i
                                                                 C  C  E irt
                                                                         •
•a o 4-> ^~
C irt C  tfl
10 e o r

«_al
3 T3 V-
4-» I- O  (U
— O    <_>
c Q. e  L.
Cn E O  3
QJ C
t- O
m -t-

**- u^

°E

s*
-i- qj
4-J U
« 1-
O 3
O O
                                                                                                                                                            :ei    t.
o.«I5    -SSC
C *J « ^OJ   -- u -O

5 ji £'
                ~ c  -
                  n O  3"1
                                                                 Q'O ^  C
                                                                 5 ID 4->  O

                                                                 (S cn E ^>
                                                                                                                                                   Of
                                                                  QJ —
                                                                  CL «j
                                                                  a. 3
                                                                                       *_i O
                                                                                       U 1-
                                                                                          -
                                                                                                                        _o  o c:
                                                                                                                        £  l/» -F-
                                                                                                                      O C   i-
                                                                                                                   «  C QJ   t>  -o
                                                                                                                   t)  3 >   *-  -
                                                                                                                   L.  O C   =  *-
                                                                                                                         ~       )
                                                                                             VI-25

-------
allows for a certain reduction in the number of air quality monitoring
stations without inducing any major perturbations in the model predic-
tions.  It would be misleading, however, to conclude that the need for air
quality input data is low, since the results of the sensitivity analysis
depend on that intensity of data in both the sensitivity and the base
case.  That is, the results depend on the number of air quality stations
used for the Los Angeles basin simulations.

     The need for upper air quality data depends strongly on the existence
and intensity of the inversion.  It is probable that areas with weaker
inversions would not be as sensitive to upper air quality data as is the
Los Angeles basin.  However, air pollution episodes are often induced by
the existence of a strong inversion, and the results obtained for the Los
Angeles basin should be fairly representative for the application of the
urban airshed model to the simulation of urban air pollution episodes.
The length of simulation time has a strong influence on sensitivities to
both surface and upper air quality data.  These input data determine the
initial conditions for the continuity equations of the air quality
model.  Hence, their importance will decrease if the simulation is carried
out over a longer period of time, and the model predictions will become
more sensitive to emission data.

     In sensitivity case 8, the boundary conditions for hydrocarbon and
nitrogen oxides were modified.  The length of simulation time would
probably affect the importance of this perturbation on the model predic-
tions, as mentioned above.  The location of the airshed will affect the
importance of the boundary conditions to some extent.  For the Los Angeles
basin, the model is particularly sensitive to the western boundary condi-
tions during land-sea-breeze regimes.  If the model is applied to an urban
area affected by pollutants transported from emission sources located
outside of the modeled area, it also will be sensitive to boundary condi-
tions, and air quality data needs at the airshed boundaries may be higher
than they are for the Los Angeles basin.

     The importance of the initial conditions for reactive hydrocarbon
concentrations has been investigated in sensitivity cases 9 and 10.  The
length of the simulation time strongly influences the importance of these
initial conditions.  It should be noted that the initial conditions for
reactive hydrocarbon (RHC) concentrations correspond 1) to the estimation
of reactive hydrocarbon concentrations from an estimated RHC/NOX ratio
(sensitivity case), and 2) to the more accurate estimation of the RHC
concentrations from detailed atmospheric chemical data obtained by gas
chromatography on reactive and total hydrocarbon (THC) levels (base
case).  Reactive hydrocarbon concentrations are difficult to obtain on a
routine basis, since they constitute a minor part of the total hydrocarbon
                                   VI-26

-------
concentration, which consists mainly of methane.   Consequently,  concentra-
tions of RHC, which are obtained from the difference  between  THC and
methane concentration (rounded off to the nearest  ppm),  are highly
uncertain.  Reliable data may be obtained by gas chromatography, but such
measurements are not presently made on a routine basis.   Detailed ambient
reactive hydrocarbon data are available for the  Los Angeles basin,  and
this may not be the case for other urban areas.  However,  this factor  is
taken into account in the cost index for the sensitivity-uncertainty
analysis performed to rank data needs.

     In sensitivity cases 11 and 12, the hydrocarbon  speciation  was
averaged over all source categories.  Clearly,  the effect  on  model  predic-
tions depends on the diverse nature of source categories.   In the sensi-
tivity results, it appeared that refinery emissions were perturbed,
whereas the dominant mobile source emissions were  not modified notably.
Higher levels of olefins, aromatics, aldehydes,  and ethylene  were observed
in the refinery emissions in the sensitivity case.  These  hydrocarbons are
oxidized faster than paraffins in the atmosphere,  and this leads to higher
ozone levels downwind of the refineries.  The importance of these input
data could be evaluated by careful identification  of  the various source
categories.

     The length of simulation time (single-day  or  multiple-day)  is  clearly
an important factor for this sensitivity study  and the following ones,
which involve emission inventories (cases 11 through  19).

     The mobile source inventory was perturbed  in  sensitivity cases 13,
14, and 15.  The importance of mobile source emissions in the Los Angeles
basin may enhance the importance of the corresponding inventories.
However, major urban areas present a similar pattern, and the sensitivity
results should be fairly representative of large urban communities.

     The point source emission inventory was perturbed in sensitivity  case
16.  The contribution of power plant emissions  to  the air pollution burden
is small for the Los Angeles basin. The importance of this inventory may
be slightly larger for an area with a large number of power plants  and
smelters.

     In sensitivity cases 17 and 18, the area-source  inventory was
spatially perturbed.  No specific features of the  Los Angeles basin appear
to limit the generality of these results notably.

     The temporal resolution of area sources was perturbed in sensitivity
case 19.  Mobile sources are dominant in the Los Angeles bas
-------
     The enlargement of the grid size in sensitivity case  20  leads  to  an
averaging of the model variables.   The spatially variable  meteorological
fields of the Los Angeles basin are probably affected by this averaging
procedure.  However, this sensitivity study should  be fairly  representa-
tive of the general effect of the grid size on  an urban  airshed model.

     The number of grid layers was reduced  from four to  two in sensitivity
case 21.  Specific features of the Los Angeles  basin are the  small  number
of elevated sources and the spatially varying mixing depths.   These
factors may affect the nature of the perturbations,  but  this  sensitivity
study probably gives a reasonable  evaluation of the  effect of the vertical
smoothing of concentration and meteorological variables  on air quality
predictions
3.   Generalization of the Results

     It has been pointed out that the nature  of  the  sensitivity  analyses
considered in this study is "perturbation-specific."   Therefore,  it  is
important to apply the sensitivity results  within  the  context  from which
they have been derived.   This is particularly apparent  for the sensitivity
case concerning surface air quality.   The low ranking  of  the need for air
quality data results from the density of  the  air quality  monitoring
stations in the Los Angeles basin; it does  not imply that air  quality data
are a negligible component of the model.

     In the presentation of the results of  the sensitivity-uncertainty
analyses, the use of cost estimates and uncertainty  estimates  specific to
the urban area should be considered.   This  provides  for optimal  use  of the
methodology developed in this study,  since  these estimates may vary
according to the urban area considered.

     The limitations of the sensitivity analysis that  result from the
specific attributes of the Los Angeles basin  and the modeling  conditions
have been discussed in the previous section.   These  limitations  should be
considered as a means for adjusting the results  of the  sensitivity studies
to different urban areas.  It is possible to  use the information obtained
for the Los Angeles area in another location  by  evaluating the specific
attributes of both areas.

     In conclusion, the results of the Los  Angeles sensitivity stud  es
using the airshed model  may be applied in a general  context as long  as
proper attention is paid to the features  both of the basin and the urban
area to be considered.
                                   VI-28

-------
                             VII    CONCLUSIONS
     Because of the large amount of input  data  (meteorology,  air quality,
and emissions) necessary to run grid-based photochemical  models, it  is
important to identify the input data that  most  affect model performance as
well as the amount and quality of data that should  be acquired  before
carrying out air quality simulations.

     This study investigated the sensitivity of the SAI  Urban Airshed
Model to the amount (and consequently the  quality)  of the input data in
order to develop a general  procedure for specifying the  meteorological,
air quality, and emission data most needed to achieve good model perfor-
mance.  The approach to sensitivity analysis used  in this study and  the
general procedure developed to define a ranking of  input  data needs  is
summarized in this chapter.  The major results  obtained  from  the simula-
tions of the Los Angeles basin and possible limitations  are then dis-
cussed.  Finally, suggestions on further studies to improve the knowledge
of photochemical model sensitivity to various levels of  details in input
information are presented.
A.   GENERAL PROCEDURE FOR SENSITIVITY ANALYSIS

     The sensitivity of the airshed model  to the  level  of  detail  of  the
input data was analyzed for the Los Angeles  basin  under  different meteoro-
logical and air quality conditions (26 June  1974  and  4  August  1975).   The
effects of perturbations in the meteorological, air quality, and  emission
data, as well as in the model  structure (grid size and  number  of  grid
layers), are analyzed and discussed in detail  in  Chapter V.  The  general-
ity of specific numerical results presented  in this study  may  be  limited
owing to the specificity of the Los Angles basin  and  modeling  condi-
tions.  However, the methodology used to analyze  the  sensitivity  of  the
model to the amount of input data is generally applicable  and  constitutes
a framework for the generalization of the  results  of  this  study and  the
definition of future sensitivity studies.

     The type of sensitivity analysis considered  in this study differs
from classical sensitivity analysis:  the  concern  is  with  the  sensitivity
of the ozone predictions of the model (which consists of the air  quality
                                   VII-1

-------
model and several input generation models, such as the wind model, eddy
diffusivity computations, and emission inventories) to the quality of
input data supplied (e.g., number of meteorological or air quality
stations, level of detail of emission inventories).  Classical sensitivity
analysis deals simply with the sensitivity of a mathematical model to its
parameters and initial conditions (e.g., Liu et al., 1976).

     It was shown in section V.A that the overall   sensitivity of the model
to input data can be considered as the combination of the sensitivity of
the model to the appropriate input variables (e.g., wind velocity, mixing
depth, emission term) and the sensitivity of these input variables to
perturbations in the input data.  For instance, the sensitivity of the
ozone predictions to the number of surface meteorological stations was
analyzed by considering the sensitivity of the wind field to the number of
meteorological stations and the sensitivity of the ozone levels to the
wind field (i.e., wind directions and wind velocities).  It was thus
possible to evaluate the dynamics of model sensitivity in detail and to
assess the possible limitations of the results.  Then, the analysis of the
sensitivity results was used as a basis for outlining general recommenda-
tions for the acquisition of input data (see section VLB).

     Several  measures involving ozone predictions were used to evaluate
the sensitivity of the model to perturbations in the input data:

     >  The signed deviation

     >  Absolute deviation

     >  Temporal and spatial correlations (for ozone levels above
        12 and 20 pphm and for both station and grid statistics).

     >  Overall maximum ozone level

     >  Maximum ozone statistics (peak levels and peak times).

     >  Dosage (area with ozone level above 20 pphm).

Whereas results obtained from these various sensitivity measures reflects
the specific features of model sensitivity (e.g., the dosage is a very
sensitive measure for the variation in grid size)  there is also consis-
tency among the results given by the measures.

     The general methodology developed to evaluate the level of detail in
input data required for grid-based photochemical modeling needs is based
on the definition of a sensitivity/uncertainty index that provides a means
for ranking input data needs (see section VI.A).  This index takes into
                                  VII-2

-------
account the sensitivity of the model  to  the  input  data  (e.g., deviation in
ozone levels above 12 pphm, deviation in maximum ozone  level), the
corresponding cost for input data acquisition,  and the  uncertainty  in the
model input variables affected by the input  data considered.  It  is impor-
tant to note that this sensitivity/uncertainty  index, which was used for
ranking data needs in our study of the Los Angeles basin,  is generally
applicable; it can be used to define  data needs for any urban area  once
the cost and uncertainty estimates have  been derived.   Therefore, this
index constitutes a basis for a general  approach to evaluate the  sens't v-
ity of the urban airshed model to the amount (and  thus  the quantity) of
the input data.  The development of a data base necessary  for performing
urban area model simulations would advantageously  involve  the information
provided by the sensitivity-uncertainty analysis.   Specific results
obtained for the Los Angeles basin are summarized  next.
B.   SPECIFIC RESULTS OF THIS STUDY

     The results of the sensitivity analyses  were  used,  along  with cost
estimates for input data acquisition and  uncertainty estimates, to define
a ranking of data needs (section VI.A).   The  sensitivity results depend
on the specific simulation conditions and on  the choice  of the cost and
uncertainty estimates (see tables VI-1 and VI-2);   a choice  of different
sensitivity measures or cost and uncertainty  estimates could lead to  a
different ranking of the input data needs, since the uncertainty bounds
shown in figures VI-1 through VI-4 display some overlap  in the estimated
values of the sensitivity/uncertainty index.   On the other hand, the
results of a sensitivity analysis are specific to  the model  considered,
and different factors (such as the magnitude  of the perturbation, the
urban area considered, the submodels--e.g., wind model—and  the length of
simulation time) have to be taken into account when generalizing the
sensitivity results.  The sensitivity results presented  below  should  not,
therefore, be interpreted as an absolute  ranking of data needs; rather
they (1) exemplify the general methodology developed, and (2)  provide
results for the specific case studied.

     The main results of the sensitivity/uncertainty analysis  of the
simulations of 26 June 1974 and 4 August  1975 for  the Los Angeles basin
are listed in table VII-1.  The relative  importance of upper air and
surface meteorology data depends strongly upon meteorological  condi-
tions.  Upper air meteorological data are ranked high for the  26 June 1974
simulation, but are only of slight importance for  the 4  August 1975
  That is, on the sensitivity measure (e.g.,  deviation  in  ozone  levels
  above 12 pphm, deviation in maximum ozone  level).
                                   VII-3

-------
    TABLE VII-1.  RESULTS OF THE SENSITIVITY-UNCERTAINTY  ANALYSIS FOR
                  THE LOS ANGELES BASIN (SENSITIVITY OF OZONE  LEVELS
                  ABOVE 12 pphm)

                      (a)  Simulation of 26 June 1974

                                      Sensitivity/Uncertainty
                                         Index Mean  Value
	Input Data	(times  1Q6)

Mobile Source—Updating Inventory              23.0
Upper Air Meteorology                          21.0
Initial Conditions for RHC                      8.7
Area Source                                     4.5
Upper Air Quality                               4.1
HC Speciation                                   2.9
Point Source                                    2.6
Surface Meteorology                             1.3
Surface Air Quality                             0.8
Mobile Source—Total  Inventory                  0.8
Area Sources—Temporal Resolution               0.6


                      (b)  Simulation of 4 August 1975

                                      Sens i ti vi ty/Uncertai nty
                                         Index Mean  Value
	Input Data	            (times  1Q6)   	

Inital Conditions for RHC                      11.7
Area Sources—Spatial Resolution               11.5
Surface Meteorology                            11.0
HC Speciation                                   9.0
Upper Air Meteorology                           4.1
Mobile Source—Total  Invenotry                  1.8
                                    VII-4

-------
simulation.  This can be related to the fact  that  the  wind field  is
strongly affected by upper air data for the 26  June  1974 case and not much
perturbed for the latter case.  On  the other  hand,  surface meteorological
data are more important for the 4 August 1975 simulation than for the 26
June 1974 simulation.

     The ranking of the other data  sets does  not  vary  with the  simulation
day.  The input data sets decrease  in  importance  in  the following order
(meteorological data are not included; see table  VII-1 for this simulation
day ranking):

     >  Mobile sources - updating inventory

     >  Initial conditions for hydrocarbons

     >  Area sources - spatial resolution

     >  Upper air quality

     >  Hydrocarbon speciation

     >  Point source emissions

     >  Surface air quality

     >  Mobile sources - total inventory

     >  Area source - temporal resolution

     A complete description, interpretation,  and  evaluation  of  the
possible limitations of these sensitivity results is presented  in chap-
ters III, V, and VI, respectively.   The sensitivity  results  depend on the
magnitude of the perturbation, unless  the model sensitivity  is  linear.  As
pointed out in section VI-A, the type  of sensitivity analysis considered
in this study is "perturbation specific"; the results  depend on the
perturbation considered, and caution is advised if the results  are to be
applied in a different context (i.e.,  with a  different perturbation  in the
input data).  For instance, in sensitivity case 5 the  number of air
quality monitoring stations was reduced from  23 to 10  and the sensitivity
results would probably be quite different had the  number of  stations been
reduced, say, from 10 to 5, since the  model sensitivity is not  expected to
be linear in this case.

     Specific attributes of the modeling conditions  also affect sensitiv-
ity analysis.  In this study, these conditions  are characterized  by the
features of the Los Angeles basin,  by  the submodels  considered  (e.g., wind
                                   VII-5

-------
model, emission inventories),  and  by the  length  of  simulation time.
Further sensitivity studies could  be carried out  to  investigate the effect
of these attributes on the model  sensitivity;  possible  studies are
outlined in the next section.
C.   FUTURE NEEDS

     Because of the complex nature  of  photochemical  grid models,  a
detailed sensitivity analysis  of  such  a model  is  an  enormous task.  Much
information has been obtained  in  this  study  about the  sensitivity of the
airshed model predictions to the  amount  (and thus the  quality) of input
data.  Further work could be carried out  to  improve  the knowledge base
regarding input data needs:

     >  Multiple-day simulation:  As the  simulation  time
        increases, the model approaches the more  realistic, case
        where emissions provide all the precursor mass and,  in the
        limit, the model  predictions become  insensitive to the
        initial conditions. Therefore,  in a multiple-day simula-
        tion, the influence of the  initial conditions  on the model
        predictions will  decrease as the  simulation  time
        increases, and the importance  of  the emission  levels (as
        well as meteorological description) will  accordingly
        increase.  We could therefore  expect the  ranking of air
        quality and emission data to be notably modified with a
        multiple-day simulation.  Useful  information would be
        obtained by carrying out  multiple-day  simulation stud es.

     >  Sensitivity studies for other  urban airsheds:  Specific
        attributes of the Los  Angeles  basin  (meteorology, aero-
        metric network, emissions,  topography) may limit the
        generality of some of  the results.   For instance, the wind
        model used in the Los  Angeles  basin was developed for
        complex terrain applications and  is  sensitive  to meteoro-
        logical data at the boundaries of the  airshed.  In an
        urban area with flat terrain,  another  type of  wind model
        that would show a different sensitivity to the input data
        could be used (e.g., interpolated model).  The choice of
        the urban area could be made according to its  specific
        attributes and the intended applications  of  the models.

     >  Sensitivity studies for various perturbation levels:  The
        sensitivity results presented  in  this  study  are "perturba-
        tion specific"; since  the sensitivity  of  the model to
        input data perturbations  is usually  nonlinear, the results
                                   VII-6

-------
        depend on the magnitude of  the  perturbation.   It would be
        of interest to investigate  for  some  specific cases the
        nonlinear nature of the model sensitivity.  For instance,
        the number of available air quality  monitoring stations
        was reduced in the sensitivity  simulation focusing on air
        quality data.  Further  sensitivity studies could provide
        additional information  on the minimum  number of air qual-
        ity stations needed to  obtain sufficient air quality data
        for the definition of adequate  initial  conditions and
        boundary conditions.

     Sensitivity studies involving  multiple-day simulations and other
urban areas would provide useful additional  information on the sensitivity
of photochemical models to the  richness of the information data base.
                                   VII-7

-------
                                REFERENCES
ABAG (1977), "Candidate Control Measures," Air Quality Maintenance Plan,
     Technical Memo 5, Association of Bay Area Governments, Berkeley,
     California.

Anderson, G. E., et al. (1977), "Air Quality in the Denver Metropolitan
     Region:  1974-2000." EF77-222, Systems Applications, Incorporated,
     San Rafael, California.

Attaway, L. D., et al. (1976), "Maintenance Shutdown of Tail Gas Treating
     Unit:  An Assessment of Potential S02 Concentrations and Related
     Health and Welfare Effects," TR-11700, Greenfield, Attaway and Tyler,
     Incorporated, San Rafael, California, and Systems Applications,
     Incorporated, San Rafael, California.

Blumenthal, 0. L., W. H.  White, and T. B. Smith (1978), "Anatomy of a Los
     Angeles Smog Episode:  Pollutant Transport in the Daytime Sea Breeze
     Regime,"  Atmos. Environ., Vol. 12, pp. 893-907.

Burton, C. S., et al. (1976), "Oxidant/Ozone Ambient Measurement
     Methods:   An Assessment and Evaluation," EF76-111R, Systems Applica-
     tions, Incorporated, San Rafael, California.

Demerjian, K.  L. (1976),  "Photochemical Air Quality Simulation Modeling:
     Current Status and Future Prospects," Paper 16-1, International
     Conference on Photochemical Oxidant Pollution and Its Control,
     Environmental Protection Agency, Raleigh, North Carolina.

Dougherty, E.  P., J. T. Hwang, and H. Rabitz (1979), "Further Developments
     and Applications of the Green's Function Method of Sensitivity
     Analysis  in Chemical Kinetics," J. Chem. Phys., Vol. 71, pp. 1794-
     1808.

Duewer, W. H., et al. (1977), "NOX Catalytic Ozone Destruction:  Sensitiv-
     ity to Rate Coefficients," J. Geophys. Res., Vol. 82, pp. 935-942.
                                 R-l

-------
Dunker, A. M. (1980), "The Response of an Atmospheric Reaction Transport
     Model to Changes in Input Functions," Atmos. Environ., Vol. 14, pp.
     671-679.

Durbin, Paul, and T. A. Hecht (1975), "The Photochemistry of Smog Forma-
     tion," internal paper, Systems Applications, Incorporated, San
     Rafael, California.

EPA (1972), "Compilation of Air Pollutant Emission Factors," AP-42,  U.S.
     Environmental Protection Agency, Research Triangle Park, North
     Carolina.

Edinger, J. G. (1973), "Vertical Distribution of Photochemical Smog  in  Los
     Angeles Basin," Environ. Sci. Techno!., Vol. 7, No. 3, pp. 247-252.

GRC (1974), "Air Quality Impacts of Electric Cars in Los Angeles,"
     Appendix A, RM-1905-A, General Research Corporation, Santa Barbara,
     California.

Hayes, S. R. (1978), "Performance Measures for Air Quality Dispersion
     Models," EF78-93, Systems Applications, Incorporated, San Rafael,
     California.

Hecht, T. A., J. H. Seinfeld, and M. C. Dodge (1974), "Further Development
     of a Generalized Kinetic Mechanism for Photochemical Smog," Environ.
     Sci. Techno!., Vol. 8, p. 327.

Heinrich, R., S. M. Rapoport, and T. A. Rapoport (1977), "Metabolic
     Regulation and Mathematical Models," Prog. Biophys. Mol. Bio!.,
     Vol. 32, pp. 1-82.

Hickey, H. R., W. D. Rowe, and F. S. Skinner (1971), "A Cost Model for  Air
     Quality Monitoring Systems," J. Air Pollut. Control Assoc., Vol. 21,
     No. 11, pp. 689-693.

Hoe!, P. G. (1962), "Introduction to Mathematical Statistics" (John  Wiley
     and Sons, New York, New York).

Husar, R. B., et al. (1977), "Three-Dimensional Distribution of Air
     Pollutants in the Los Angeles Basin," J. Appl. Meteorol., Vol.  16,
     pp. 1087-1096.

Kieth, R. W., and B. S. Selik (1977), "California South Coast Air Basin
     Hourly Wind Flow Patterns," South Coast Air Quality Management
     District, El Monte, California.
                                 R-2

-------
Koda, M., A. H. Dogru, and J. H. Seinfeld  (1979),  "Sensitivity Analysis  of
     Partial Differential Equations with Applications to  Reaction  and
     Diffusion Processes," J. Comp. Phys.. Vol.  30,  pp. 259-282.

Lamb, R. G., and J. H. Seinfeld (1973), "Mathematical Modeling of  Urban
     Air Pollution—General Theory," Environ.  Sci. Technol.,  Vol.  7,  pp.
     253-261.

Littman, F. E. (1978), "Regional Air Pollution  Study, Emission Inventory
     Summarization," Rockwell International,  St. Louis, Missouri.

Liu, M. K., T. C. Myers,  and J. L. McElroy (1979), "Numerical  Modeling of
     Land and Sea Breeze  Circulation along a  Complex Coastline," Math.
     Comp. Simulation, Vol. 21, pp. 359-367.

Liu, M. K., D. C. Whitney, and P. M. Roth  (1976),  "Effects of  Atmospheric
     Parameters on the Concentration of Photochemical Air Pollutants,"
     J. Appl. Meteorol.,  Vol. 15, ppp. 829-835.

Liu, M. K., et al. (1976), "Continued Research  in  Mesoscale Air Pollution
     Simulation Modeling:  Volume I—Analysis of Model Validity and
     Sensitivity and Assessment of Prior Evaluation  Studies,"  EPA-600/4-
     76-016a, Systems Applications, Incorporated,  San Rafael,  California.

Lyons, W. A., and H. S. Cole (1976), "Photochemical  Oxidant Transport
     Mesoscale Lake Breeze and Synoptic Scale Aspects," J. Appl. Mete-
     orol., Vol. 15, pp.  733-743.

MacCracken, M. C., and G. D. Sauter, eds.  (1975),  "Development of  an Air
     Pollution Model for  the San Francisco Bay  Area," University of
     California, Livermore, California.

Mayersohn, H., et al. (1976), "Atmospheric Hydrocarbon Concentrations
     June-September, 1975," California Air Resources Board, Atmospheric
     Studies Section, Division of Technical Services, Sacramento,  Cali-
     fornia.

Mayersohn, H., et al. (1975), "Atmospheric Hydrocarbon Concentrations
     June-September, 1974," California Air Resources Board, Atmospheric
     Studies Section, Division of Technical Services, Sacramento,  Cali-
     fornia.

McRae, G. J., and J. W. Tilden (1980), "A  Sensitivity and Uncertainty
     Analysis of Urban Scale Air Pollution Models--Preliminary Steps,"
     Proc. Second Joint Conference on Applications of Air Pollution
     Meteorology, AMS/APCA, 24-27 March 1980, New  Orleans, Louisiana.
                                 R-3

-------
Miedema, A. K., et al. (1973), "Cost of Monitoring Air Quality  in the
     United States," EPA-450/3-74-029, Research Triangle Institute,
     Research Triangle Park, North Carolina.

Reynolds, S. D., and P. M. Roth (1980), "The Systems Applications,
     Incorporated Airshed Model—A Review and Assessment of Recent
     Evaluation and Applications Experience," submitted to J. Air Pollut.
     Control Assoc.

Reynolds, S. D., P. M. Roth, and J. H. Seinfeld (1973), Atmos Environ.,
     Vol. 7, p. 1033.

Reynolds, S. D., et al. (1979), "Photochemical Modeling of Transportation
     Control Strategies, Vol. 1, Model Development, Performance Evaluation
     and Strategy Assessment," EF79-37, Systems Applications, Incorpo-
     rated, San Rafael, California.

Reynolds, S. D., et al. (1976), "Continued Development and Validation  of  a
     Second Generation Photochemical Air Quality Simulation Model:  Volume
     I I—Refinements in the Treatment of Chemistry, Meteorology, and
     Numerical  Integration Procedures," EF75-24R, Systems Applications,
     Incorporated, San Rafael, California.

Russell, P. B., and E. E. Uthe (1978), "Regional Patterns of Mixing Depth
     and Stability:  Sodar Network Measurements for Input to Air Quality
     Models." Bull. Am. Meteorol. Soc.. Vol. 59, pp. 1275-1287.

SCAPCD (1976),  "Fuel Use and Emissions from Stationary Combustion
     Sources,"  Southern California Air Pollution Control District,
     El Monte,  California.

Seigneur, C. (1978), "Mathematical Analysis of Air Pollution Models,"
     Ph.D. Thesis, University of Minnesota, Minneapolis, Minnesota.

Skelton, E. P., et al. (1977), "Methodology for Estimating Emissions  for
     the Sacramento AQMP," Sacramento Air Pollution Control District,
     Sacramento, California.

Souten, D. R.,  G. E. Anderson, and R. G. Ireson (1980), "National Commis-
     sion on Air Quality Los Angeles Regional Study SAI Task 1  Report:
     Review of the SIP Process and Framework (l.A.) Review and  Analysis of
     South Coast Air Quality Management District Air Quality Management
     Plan Modeling Methods (l.B)," Draft Final Report, EF80-17, Systems
     Applications, Incorporated, San Rafael, California.
                                 R-4

-------
Tesche, T. W. (1978), "Evaluating Simple Oxidant Prediction Methods Using
     Complex Photochemical Models," EM78-14, Systems Applications,
     Incorporated, San Rafael, California.

Tesche, T. W., and M. A. Yocke (1978), "Numerical Modeling of Wind Fields
     over Mountainous Regions in California," Conference on Sierra Nevada
     Meteorology, American Meteorological Society, 19-21 June 1978, South
     Lake Tahoe, California.

Unger, C. D. (1976), "Meteorological Input into Photochemical Models,"
     unpublished results, California Air Resources Board, Sacramento,
     California.

West, R. (1979), private communication from Richard West, CSBE, to William
     Rogers Oliver, Systems Applications, Incorporated, San Rafael,
     California.

Whitten, G. Z., and H. Hogo (1977), "Mathematical Modeling of Simulated
     Photochemical Smog," EPA-600/3-77-011, Systems Applications, Incorpo-
     rated, San Rafael, California.

Yotter, E. E. (1979), "Motor Vehicle Emissions Assessment Using Caltrans
     Transportation Systems Analysis (Modeling) Data," California Air
     Resources Board, Planning Division, Air Quality Maintenance Planning
     Branch, Sacramento, California.
                                 R-5

-------
                                   TECHNICAL REPORT DATA
                            {'Please read Instructions on the revers^ be fare i omplctingj
1. REPORT NO

   EPA-450/4-81-Q31a
4. TITLE AND SUBTITLE
The  Sensitivity of Complex Photochemical  Model  Estimates
to Detail  in Input Information
                                                          6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)                            . ,      .,
T. W.  Tesche, C. Seigneur, L. E.  Reid,  P.  M.  Roth,
W. Ro  Oliver, and J. C. Cassmassi
                                                           8. PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Systems  Applications, Incorporated
950  Northgate Drive
San  Rafael,  California  94903
12. SPONSORING AGENCY NAME AND ADDRESS
 U.S.  Environmental Protection Agency
 Office of Air Quality Planning  and  Standards
 Research Triangle Park, North Carolina  27711
                                                           3. RECIPIENT'S ACCESSION NO.
                                                             REPORT DATE
                                                           SAI  No.  330R-EF81-5
                                                           10 PROGRAM ELEMENT NO
                                                          11 CONTRACT/GRANT NO.

                                                             68-02-2870
                                                           13 TYPE OF REPORT AND PERIOD COVERED
                                                          14. SPONSORING AGENCY CODE
15. SUPPLEMENTARY NOTES
16. ABSTRACT
Using the air quality, meteorological  and emissions data  base  available in the
Los  Angeles area, two days with  distinctly different meteorology are simulated using  a
photochemical grid model  (Urban  Airshed Model).  The data base used to generate model
inputs is then degraded for  the  purpose of noting which data are most essential to
collect in order to have  a complex grid model perform adequately.  The results are  used
to  develop a more general methodology for prioritizing data needs.   The methodology
considers model sensitivity  to  input derived from data bases of varying detail, expense
in  collecting the data, and  the  uncertainty associated with deriving model input  vari-
ables from the data base.
                                KEY WORDS AND DOCUMENT ANALYSIS
                  DESCRIPTORS
 Photochemical grid models
 Urban Airshed Model
 Ozone
 Sensitivity studies
 Model inputs
                                              b.IDENTIFIERS/OPEN ENDED TERMS
                                                                        c.  COS AT I 1 icId/Group
                                              (9 bf.rjB,!' CLASS , I'lill Kefi-r.,    I 71 NO OF PAGES
a DISTRIBUTION STATEMENT


Unlimited

-------
                                                                                                                                                                                                                  OJ ftl —;
                                                                                                                                                                                                                      83 ^
                                                                                                                                                                                                                      Q) O
                                                                                                                                                                                                                          5
             .
yi  O n     en
"i  *        »
     3r*  -• O
     10  ~ O
       o  O -,
       i-  ft) ;-
    O  D  Q) ID
    S °
                                                                                                                                                                                                        m > TJ m T) -
                                                                                                                                                                                                        "              '"
                                                                                                                                                                                                                  3     Q.
                                                                                                                                                                                                                                         ooa
                                                                                                                                                                                                                                            3%
                                                                                                                                                                                                                                         cn  —  -
                                                                                                                                                                                                                                         oj  o  fl>

-------