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

     Appendix B -  Specification
     And Assessment Of Airshed
     Model Input Requirements

-------
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-450/4-81-031C
         The Sensitivity of Complex
Photochemical Model Estimates To  Detail
             In Input  Information
   Appendix B - Specification And Assessment
      Of Airshed Model Input Requirements
                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

-------
                                 CONTENTS


LIST OF TABLES	   1v
   1.  A REVIEW OF AIRSHED MODEL INPUTS	   B-l
   2.  ESTABLISHMENT OF CHARACTERISTIC LEVELS OF INPUT DATA
       DETAIL TO BE EXPECTED FOR A RANGE OF MODEL APPLICATIONS	   B-6
   3.  ESTIMATION OF DATA ACQUISITION COSTS ASSOCIATED WITH
       RAISING THE LEVEL OF DETAIL OF INPUT DATA	  B-12
   4.  ANALYSIS OF AIR QUALITY MODEL SENSITIVITY TO
       VARIATIONS IN INPUTS	-	  B-19
          a. Studies Focusing on Air Quality Inputs	  B-25
          b. Studies Focusing on Meteorological  Inputs	  B-26
          c. Studies Focusing on Chemistry Inputs	  B-29
          d. Studies Focusing on Emissions Inputs	  B-30
          e. Studies Focusing on Grid Specification	  B-31
   5.  ISSUES RELATED TO THE PREPARATION OF EMISSION INVENTORIES....  B-32
          a. Mobile Source Emission Inventories	  B-37
          b. Stationary Source Emission Inventories	  B-38
   6.  CONCLUSIONS	  B-39
REFERENCES   	   R-l
                                      111

-------
                                  TABLES
B-l     Input Parameters for the SAI Urban Airshed Model	   B-2

B-2     Summary of Data-Dependent Input Requirements of the
        SAI Urban Airshed Model (EPA-5 Version)	   B-4

B-3     Summary of Routine Air Quality Monitoring Activities
        in 14 Major Cities in the United States	   B-9

B-4     Levels of Detail in Data Used as Input to Grid-Based Air
        Quality Simulation Models	  8-13

B-5     Cost Categories for Air Quality Monitoring Systems	  B-16

B-6     Estimated Annual Cost to Augment An Existing Aerometric
        Monitoring Network with Various Instruments	  B-17

B-7     Summary of Sensitivity Study Results Obtained with
        Grid-Based Photochemical Airshed Models	  B-20

B-8     Control Measures and Emission Inventory Data Needs	  B-33
                                       iV

-------
                                Appendix B

     SPECIFICATION AND ASSESSMENT OF AIRSHED MODEL  INPUT REQUIREMENTS
     Systems Applications, Inc.  (SAI)  has been  engaged  in  a research study
sponsored by the Environmental  Protection Agency  (EPA)  to  evaluate the use
of non-data-intensive methods for assessing  the effectiveness of state
implementation plans (SIPs) for  controlling  photochemical  oxidants.  The
need for this work stems largely from  the revisions  to  the SIPs that are
required by 1982.  Because the Immediacy of  such  a deadline places serious
practical constraints, both upon those formulating the  revisions and those
asessing them, it appears worthwhile to explore the  use of complex air
quality simulation models (AQSMs) as a means of assessing  the adequacy and
accuracy of simpler oxidant prediction methods.

     This appendix reviews the input requirements of the SAI Urban Airshed
Model--a three-dimensional, time-dependent photochemical dispersion
model.  Levels of detail characteristic of model  input  data  are discussed
and estimates of data acquisition costs are  made.  (These  estimates are
subject to change caused by regional differences  in  construction and
maintenance costs, labor costs,  inflation, etc.)  Finally, the results of
recent photochemical grid model  sensitivity  studies  are presented  to
provide a perspective on the sensitivity results  discussed in the  main
body of this report.
1.   A REVIEW OF AIRSHED MODEL INPUTS

     The objective of this review is to define comprehensively the many
data inputs required by a complex photochemical grid model  and to inter-
pret the extensive model output information.  Although several photochemi-
cal dispersion models now exist, the model recently refined by SAI under
EPA Contract 68-02-2429 (the so-called EPA-5 model) was selected primarily
because it has the most extensive input requirements of any operational
photochemical model developed to date.

     Airshed model inputs can be broadly categorized as either data
related or nondata related.  The list in table B-l are non-data-related
model inputs.  Many of the inputs listed in table B-l can be prepared
                                   B-l

-------
TABLE B-l.  INPUT PARAMETERS FOR THE SAI URBAN AIRSHED MODEL*
     Control Parameters
Region description:
  UTM coordinates of
  grid origin
  Cell size; horizontal
  and vertical
  Grid size
  Number of vertical  layers
Simulation controls:
  Run identification
  Beginning and ending times
  Time step size
  Minimum time  step
  Convergence criteria
  Averaging interval
  Instantaneous output
  interval
  Print options
       Chemistry Parameters
For each species:
  Name
  Reactive or unreactive
  Steady-state initial conditions
  Steady-state boundary conditions
  Resistance to deposition
  Upper and lower bounds on
  numerical integration and
  steady-state calculations
For each reaction:
  Rate constant
  Photolysis rate
  Temperature dependence
  Activation energy
  Reference temperature
For each coefficient:
  Name
  Value
* Ames, J., et al. (1978) discuss the airshed model  inputs in considerable
  detail.
                                  B-2

-------
without recourse to air quality measurements, emission  information, and so
forth.  Specification of certain input  parameters, however, does require
some knowledge of the unique conditions surrounding  a given model applica-
tion.  For example, if the model is  to  be  applied to a  city containing
several large elevated point sources, then some  estimate of the diurnal
distribution of plume rises must be  made so that the top of the modeling
region is high enough to contain point  sources emissions.  Clearly, in
estimating the distribution of plume rises, one  must have some information
about the meteorology of the region  and the physical emission characteris-
tics of the sources.

     Data-dependent input requirements  of  the SAI Airshed Model  are
summarized in table B-2, which indicates the spatial and temporal resolu-
tion of each input.  The resolution  of  each input corresponds to the
present configuration of the model.  The indicated resolution is, in some
cases, greater than that required to operate the model  and to obtain
acceptable simulation results.  Moreover,  the resolution identified in
table B-2 does not necessarily represent the maximum level of input
information detail that could potentially  be used  in a  photochemical
simulation.  A few examples of this  point  are discussed next.

     Atmospheric stability is characterized as the Airshed Model by three
scalars:  the temperature gradient below the base of the inversion  (the
so-called "diffusion break"), the gradient through the  stable layer,  and
the exposure class.  (The exposure classification is similar  to, though
not exactly the same as, the Pasquill-Gifford  stability categories.)
These inputs are used  in the model diffusivity and plume rise algo-
rithms.  Focusing on the first two scalars, one  can  see that  for some
applications, the vertical gradients in ambient  temperature  vary from
place to place.  The Los Angeles air basin is  an example.  Surface
temperatures near the  coast are moderated by moist marine air,  whereas
near the eastern end of the basin hot,  dry, desert-like conditions
prevail.  Moreover, the rate of adiabatic heating caused by  air mass
subsidence varies across the basin,  in part because  of the higher  water
content of the marine  air.  Thus, the temperature gradients  in  the mixed
and stable layers are  expected to exhibit spatial as well  as temporal
variability.  Owing to the paucity of upper air temperature  soundings,
however, the temperature structure is currently treated in the  model  as
scalar quantities, varying only in time.

     Radiation intensity  is another meteorological variable  that is
treated as a scalar.   Despite  the findings of photochemical  model  sensi-
tivity studies, which  consistently demonstrate the important role of solar
radiation  in oxidant formation, radiation  is treated as a scalar value
rather than a three-dimensional, time-varying field.   For model applica-
tions  in which large portions  of the region experience partial  obscuration


                                        B-3

-------
TABLE  B-2.   SUMMARY OF  DATA-DEPENDENT  INPUT  REQUIREMENTS OF THE
                 SAI  URBAN AIRSHED  MODEL  (EPA-5 VERSION)
                                Spatial and Temporal
         description
   Meteorology

     Horizontal (u-«)
     «nndi  (n/tec)
     Reference height of
     Surface wind eenitor-
     Ing sutions («)

              break (m)
                                    »J>
     Top »f Modeling
     rvgton (n)

     (round-level tenpera-
     tures CO

     Atmospheric pressure
     (•*>

     Te«c>tr*tur« gradient
     belch diffusion break
     CC/ii)

     Tcnvertlure gradient
     above diffusion break
     CC/»)

     Miter concentration
     1n the aowiiphere
     (PP»)

     Exposure (tubilltjr
     class)

     ladiation intensity
     factor (per nin)

   Air ojuality

     initial conditions
     (pphm)
     •oundary conditions
     (pph»)

     Concentritieni above
     tap of Modeling
     region (ppha)

     Surface concentra-
     tions at several
     locations •Itnin
     Mdeling rtglon («*•)

   Surface characteristics

     Surface roughness
     (o)

     Veaitatlon factor
The vertical comanrct.  «. Is coxouted
by the airsKM lofltl. rentferinq thr
resultant wind field *ass consistent

Used in the diffutivUy  algorithtr
elevation  at uhlch the stability struc-
ture of the atmosphere changes narked)/
(e.g., an  inversion or thenval internal
boundary layer)
Mot absolutely essential  to awoel
operation
Used In plume rise calculations



Used in plune rise calculations



Used in kinetic andule



Used in diffutivlty algorithm


Used in kinetic nodule




tequired for NO. H02. 03, hMO;. H;D;.
olefins, paraffins, aldehydes. aroMt-
Ics. PAN. SO;. SOj. and CO

Required for same tpecles as above


Required for sine species as above
Required for  verification and evalu-
ation of nodel performance (sane
species at above)
Used In diffuvivlty. surface sink, and
•icroscale algorithws

Used in surface  sink algorithm
                                                B-4

-------
                               TABLE  B-2  (Concluded)
      Description
(anssions

  limped ground-level
  •*1ssions  Irar  tri«-
  fU. area  sources.
  airports.  etc.  (»»/hr)

  Elevated stationary
  POlnt SOurct  emis-
  sions (g»/hr)
  Elevated
  source awnsions
  location end  height
  ef elevitee point
  tource «ni*S10nS
  <*)

  locitien of drcrtft
  flight »re«i  (»)

  Meit flui fror. tie-
  vtted point sources
Spit 1*1  tnd Tempor»l
     Kcsolutlon
                                                                       tewrks
                          Acquired for HO,  K)?. 0). MNQ;
                          olffini. piriffins. «d«h/oei, iro
                          Mticv. r*N. S02. SOJ. tnd CO
                          fmssions from t*11  turn for the
                          • bo*» ipeoes are
                                    frot. •1rtr«ft takeoffl «n(J
                           landings for the above species are
                           required (as appropriate)

                           •eouired for confutation of effective
                           stack heights
                                      depends en nagnftude Of
                          aircraft missions

                          Usefl 1n plmit rite a Igor it tr
  rttes 1rat tutos
          •tcrotcile
  yartmtter («/iec )
                           Icquired for NO, NO.-   use(  1n the
                           treitwert of lubgrte-tcale <*ino-
                           »cale)
                                          by
                                                         •here 1 • vehicle tjrpe. ?>4 • i»««p*r
                                                         •thtclei in category  i. and v.
                                                         ef vehicles 1n category i
                                                 B-5

-------
 (as  a result of clouds,  haze,  and  so  forth),  spatial variation  in solar
 radiation may be  important.  Although current routine field measurements
 do not allow the  preparation of three-dimensional radiation fields, the
 model's formulation does  attempt to account,  at  least in  principle, for
 the  vertical attenuation  of solar  radiation caused by aerosol scattering.

     The prescription of  initial and boundary conditions  is another area
 in which the model could  accept more sophisticated input  information  if it
 were available.   Currently, total  hydrocarbon concentration measurements
 (or  alternatively, nonmethane  hydrocarbons) are  apportioned among five
 reactivity classes--olefins, paraffins, carbonyls, aromatics, and ethy-
 lene.  The relative fraction of each class is assumed to  be spatially and
 temporally invariant.  For a homogeneous hydrocarbon source distribution,
 this approach is  reasonable, as long as the estimate of each species
 fraction is accurate.  In the  absence of any hydrocarbon  speciation
 measurements, and particularly for a varied source distribution  (e.g.,
 refineries, automobiles, dry cleaners) such an apportionment scheme can
 potentially degrade model performance.  Specifically, it  might  increase or
 decrease the oxidant maxima and alter the phasing of the  formation of
 secondary pollutants.

     A final example of refined inputs that could be accepted by the model
 if they were available, concerns the treatment of ground-level  emis-
 sions.  Currently, the emission data file (EOF)  lumps (for each  ground-
 level grid cell)  all surface emissions into one  emission  rate,  regardless
 of whether the emissions  are from  airports, autos, refineries,  rendering
 plants, and so forth.  If the goal of the modeling effort is to  assess the
 level of model accuracy and precision, this lumping procedure is adequate
 only as long as the aggregate emission value is  correct.  But,  if the
modeling objective is to assess the effectiveness of a specific  control
 strategy (say, a 60 percent reduction in refinery hydrocarbon emissions),
 then it becomes necessary to modify the EOF to reflect such a scenario.
 If gridded emission inventories are available that delineate each source
 type, the model can readily accept the increased level of detail.
2.   ESTABLISHMENT OF CHARACTERISTIC LEVELS OF INPUT DATA DETAIL TO BE
     EXPECTED FOR A RANGE OF MODEL APPLICATIONS

     Efforts performed under this task were twofold.  First, a brief
review of the status of data acquisition activities was carried out in the
following 14 cities:

     >  Albuquerque, New Mexico

     >  Chicago, Illinois
                                   B-6

-------
     >   Denver,  Colorado

     >   Houston,  Texas

     >   Las Vegas,  Nevada

     >   Los Angeles, California

     >   New York, New York

     >   Philadelphia, Pennsylvania

     >   Phoenix,  Arizona

     >   Portland, Oregon

     >   Sacramento, California

     >   San Francisco,  California

     >   St. Louis,  Missouri

     >   Washington, D.C.

     The purpose of this  review was  to  develop  a  general understanding of
the spectrum of urban-scale  air monitoring  activities throughout the
United States.  The cities that were selected shared several  attributes.
They were:

     >   Moderate to large in size.

     >  Representative  of major geographical  areas  in the  contig-
        uous 48 states.

     >  Reflective of  a variety of emission source  activities,
        including emphasis on transportation  (Los Angeles),
        petrochemical  (Houston), light density  residential
        (Sacramento),  heavy  density residential (New  York),  heavy
        industry (St.  Louis), and other activities.

     >  Subject to exceedances of the one-hour  federal  oxidant
        standard,  in some cases by a factor of  2  or 3.

Furthermore, some of the cities that were selected  have been the subject
of previous or ongoing photochemical modeling studies,  thereby making it
possible to develop a more complete picture of  the  available data  base.

                                       B-7

-------
     Investigation of the data bases  of  these cities was made through
telephone discussions and correspondence with many people, principally
officials from the EPA offices, state agencies,  local  air pollution
control agencies,  and other  individuals  who had  modeling experience in
certain of the cities.  The  results of this brief review are presented in
three parts in table B-3: meteorological, air quality, and emission
inventory data.  Blanks in the table  indicate that the information was
either unavailable or not readily accessible.

     As expected,  a wide range of number and type of measurements was
encountered.  All  cities have at least a few surface wind stations, but
the range in the amount of available  upper level wind  and atmospheric
stability data is  very broad.  Contrast, for example,  St. Louis with
Phoenix or Houston.  The amount of air quality monitors also varies and,
to a degree, reflects the predominant air quality concern in each city.
Note that oxidant  monitoring in St. Louis and Los Angeles is extensive,
whereas in Las Vegas concern seems to focus more on carbon monoxide.
Insofar as this brief review could determine, none of  the cities investi-
gated routinely carry out hydrocarbon speciation experiments or airborne
air quality measurements, though these measurements were sometimes
reported during occasional special field studies.

     In table B-3, the area  of greatest  uncertainty is the chartactenza-
tion of the emission inventories. Often, the individuals responsible for
supervising the collection of air quality and meteorological data were not
involved in preparing emission inventories.  Because the scope of this
review ruled out a detailed  characterization for each  city,  it was
occasionally necessary to rely on the general understanding of certain
people of the emissions data base rather than speaking directly with the
person or persons  who prepared each one. From table B-3(c), it is clear
that great variation exists  from city to city in terms of the thoroughness
and complexity of  the emission inventories.  As  an example, the St. Louis
mobile source inventory accounts for  spatial variations in the diurnal
distribution of the percentage of cold automotive starts, whereas traffic
emissions in New York are determined  borough by  borough, based on the
vehicle miles traveled (VMT).
* One of the prominent difficulties encountered in the review was
  uncertainty on the part of certain officials  as  to  the  current status of
  the monitoring networks.  Often, systems were being dismantled,  brought
  on line, or used only during special  studies.  In addition,  some
  agencies (or local air pollution control districts)  were,  at times,
  unaware of the scope of monitoring activities carried out  by other
  groups, such as the National Weather  Service, airports, the military,
  and educational institutions.
                                       B-8

-------
TABLE B-3.   SUMMARY  OF ROUTINE AIR QUALITY MONITORING  ACTIVITIES IN
              14 MAJOR CITIES  IN THE UNITED STATES

   (a)   Number of Stations Performing  Routine  Air Quality Sampling
C1t»
Albwauerqu*. NM
CMc«go. 1L
Denver. CO
Neutun. IX
Lai Vcgat. NV
Lot Angelei. CA
Nw fork. NT
Mi1litfe1pM«, P*
•fcoenli. AT.
Portland. W
fecrawnto. CA
SJM Franeluo. CA
St. Loull. MO
Mining ton. B.C.
Oritiint
4
4
f
3
3
38
7
8
S
3
•
»
K
10
mt
3

3
3
2
27
7
S
2
3
•
U
ts
10
cc
s

9
3
24
23

8
S
U
•
K
K
10
3

1
3
0
17

e



21
11
10
0

2
3
4
11

3
3

4
U
K
10
Upper A1r
Ptrtlcu- „. Htiii/re-
lites *°} Bents
13 0 0

0 0 K
3
300
8 II

8


1 0 0
17 17 0
n 10 s
10
Hydro
urboi
tt»t1(
0



0
S






S

S •
• •
         ttudits.
HeU:  A I»r« entry IndlCltci ttet I pcrtlcwUr mtnurmrn\ U net Ukrn; • »Unk tn4«cttct
     imctrulnt/ (i tfi «h«thcr or te rfiit eittnt tht •ctvurwnt U tik»n.
                                   B-9

-------
                              TABLE B-3  (Continued)

     (b)   Number  and  Type  of Daily Meteorological  Measurements
Surface
M1nd
City »e1ociti
Albuquerque. W
Chicago. II
Denver. CO
Houston. T>
Las v*g«s. MV
Lot Angrlei, f>
ft*. York, NY
Philadelphia, PA
Phoenix, A!
Portland, OR
Sacramento. C*
Sen Francisco. CA
St. IMU. MO
MasMngton, D.C.
7
10
25
3
8
44
10
2
6
9
12
17
25
25
Upper Level
Surface AtBDS»nrr1c Hind
Te*p«rature Stability Velocity
7 •¥, IW,
10 »2 **2
2 U, U,
3 0 P,
1 AC?. AS, 0
9 We. AS, PXfl
10 ^
2 Pl
600
9 W, W,. P,
4 AS ^*
17 tU, U,
25 IDB 'H
25 W, M,
Solar
Insolation
1
3
1
1
1
2
3
1
1
1
1
17
6
2
Hu»1di
1
3
1
1
1
B
3
1
1
1
2
17
20
2
AC •  acoustic Mwnaer.

AS •  aircraft spiral.

RD •  radiosonde.

to •  raoinsonde.

 • •  plbal.
Motes:  Subscripts refer to the nimber of Measurements taken each day.  A zero entry Indicates
       that • particular Measurement 1s not taken; a blank Indicates uncertainty as to whether
       or to what extent the Measurement 1s taken.  In some cases, the meteorological measure-
       ments presented here were drawn from special studies conducted during the sumer smog
       Mason; In other cases only routinely collected data are presented.
                                         B-10

-------
                                 TABLE B-3  (Concluded)
           (c)   Description of Methods  Used for  Emission   Inventories
                                                      Point and Area Sources Emission Inventory

dtjr
Alb***,.. *H


Chicago. 11
Denver, CO



Houston. Tl

Us Vegas. HV



Les Angeles. CA



fa. York, BY

Philadelphia. PA

Pnoenii. A7

Portland. OR


Sftcraecnto. CA



ton Francisco. CA


St. Louis. PC



"•*"""•"• § '

Formit
llnk-node:
VKI

(ridded
(ridded



link-node:
VKT
(ridded



(ridded



VKT

(ridded

(ridded

(ridded


(ridded



(ridded


Variable
lite grid



•ridded

Seeeiet
R, H. C


X, H. C
«. H. 5,
P. C




N. H, C



N. H. S. C





H. C, N

*'. H. C

S. P. C


M. H. S.
P. C


H. N. S.


*. N, S.
P. t



«, N. S.
P. C

(rid Site
•/A


50 i JO:
30 > 30:
1 «i




30 i 40:
1 tar


100 i SO:
2 m\


krough b>
borough
46 i 48.
2 »'
1 Hi

20 i 30:
2 tor.

2b i 2S:
2 tor


120 * (0:
1 tor


ISO * 200:
1-10 br.



4«1
Not/Cold
Start
Arw-.lde
texpora 1
reiolutton

Aret-Mlde
tenpora 1
distribu-
tion


Are»-«ride
tetieoral
distribu-
tion
Area-Hide
tenportl
dlstribu-
tior.









Area*«idt
temporal
resolu-
tion
Area-«1de
tetporal
distribu-
tion
Not/cold
distribu-
tions
appl led
to each
grid cell


Format
•EDS


(ridded
(ridded



•y counties

(ridded



Cridded





NEDS

NEDS

ly dis-
trict

(ridded



(ridded


(ridded



(ridded

!0*eiet
•. H. S.
P. C

N. X. C
N. H. S. P



«, S. H. f

N



N. H. S. C



S. P

N. H. S.
P. C
N. H. S.
0, C
N. N


*. H, S.
P. C


N. N. S,
P, c


N. H, S,
P, C:
hydrocarbon
speclation


M. N, S.

Seatiil
Are»-*1d«


1 mi
1 fii



County -
vi dt
1 tor



2 eii





Area -vide

Are* -vide

De&ends or
site o<
districts
2 tor



1 tor


1-10 tor



« •!

T«^>o'*l
Atinuil
averagr


B or ?4
hour, plus
set sent 1







Hourlj





Annuil
trrrtgt
Annul 1
aver* gt



An null



Hourly


Hourly




II • nitrogen eiidcs.
K • hydrocarbons.
S • sulfur oildts.
P • »*rt*culat*i.
C • carbon •onotidt.
                                               B-ll

-------
     Delineation of characteristic  levels of detail  in the data available
for photochemical  modeling was  attempted, once the foregoing assessment of
present urban-scale data bases  was  complete.  With this review as a guide,
and realizing that certain measurements  are clearly  beyond the scope of
routine monitoring practices, table B-4  was formulated, yielding three
general "levels of detail" of data  input to a complex model.  Associated
with each of nine types of input  information are  statements reflecting the
type and amount of data one might expect for each level of detail.  The
"maximum practical level" corresponds  to the most exteiu.ve data base that
is currently available or that  might be  made available given present
funding constraints and the state of the art in photocheical modeling.   In
many respects, the St. Louis and  South Coast Air  Basin data bases are
examples of this category.

     At the other end of the spectrum  is the "minimum  acceptable  level."
Although a data base so characterized  might be  adequate for modeling
purposes, certain assumptions must  necessarily  be invoked  in preparing
model  inputs.  For example, Phoenix has  no  upper  air temperature sound-
ings.  To estimate mixing depths  over  Phoenix,  one might  assume that  the
atmospheric structure at Tucson  (where upper air  soundings  are available)
is reflective of conditions over  Phoemx.   Although  mixing  depth  estimates
might  be generated in this fashion, the  extent  to which they degrade  model
performance and thus confound model performance evaluation  efforts  is
unknown.  Between these two levels  of  detail lies a  third category
entitled "commonly used level."  Most  of the data bases  presented  in  table
B-3 fall within this category.   This does  not  suggest, however, that  these
data bases are well suited to model performance evaluation  and  applica-
tion.  Some of the measurements that are either lacking  or  in short supply
in the intermediate category are ones to which model performance  is quite
sensitive.
 3.    ESTIMATION OF DATA ACQUISITION COSTS ASSOCIATED WITH RAISING THE
      LEVEL OF DETAIL OF INPUT DATA

      The objective of this analysis was to derive preliminary estimates,
 where possible, of the costs entailed in improving the quantity, quality,
 or  both of various components of an AQSM data base over that currently
 being collected.  If, in fact, one can identify the improvement in model
 performance  achieved through data from an additional wind station or
 pyranometer, for example, then this could serve as the basis for quantify-
 ing the degree of improvement in model performance ascribable to a
 specific expenditure.

      Relatively  little  in the recent  literature serves as a guide in this
 endeavor.  One might expect that the  lack of guidance is in part a result
                                        B-12

-------
                 TABLE  B-4.   LEVELS  OF  DETAIL  IN  DATA  USED  AS  INPUT TO GRID-BASED
                                   AIR  QUALITY  SIMULATION  MODELS
            Input
 Atwspherlc itibMUy
     Ma«1mui» "ractlcil  Level

 Continuous nonltaring  of nit-
 Ing depths with  acoustic soun-
 der it one or nore locations

 Several (i-8) verticil  tem-
 perature soundings through.
 out the day it various  loca-
 tions within the Modeling
 region

 Numerous turf«ce temperiturt
 awasure«i>ents recorded  hourly
 •t virlous locations through-
 out the modeling region

 One or nore Instrumented
 Uwers providing continuous
 e*asurenents of  the ntied
       thernal structure
      Ce*nonlr Used level	

 * It* (3-5)  temperature sound-
 Ings it different times of the
 diy it one or two locttlons

 Several turfict  temperature
 Measurements recorded It var-
 1ous locations throughout the
 Modeling region
  Hlnlmup Acceptable  Level*

 Twice dally temperature
 soundings It in llrport
 «1th1n or neirbjr the  region
 being Modeled

 A fen (1-3) Surface temperi-
 ture Measurements with which
 to estimate tempo'il
 virlitlon
                                                                                                   Limited spatial  resolution
                                                                                                   or none It ill
 • 1nd fields
                                  Numerous fround-based monitor-
                                  Ing  stations reporting hourly
                                  average values

                                  Freouent upper i1r soundings
                                  •t several locations through-
                                  out  the Modeling region

                                  Continuous upper level measure-
                                  nents on one or i few e leva let)
                                  towers

                                  Wind, Inversion, temperature.
                                  ind  terrain data used is  input
                                  to the 3-D numeric*!  Model
                                  yielding the Mass conserving
                                  3-D wind field
                                  Interpolations fro* ground-
                                  based monitoring network and
                                  limited (3-5) number of upper
                                  level  soundings at one or two
                                  locations

                                  Resultant wind field rendered
                                  mass  consistent by divergence-
                                  free  algontnr
                                 Interpolations fro* 11»H«J
                                 (3-S stations) routine
                                 turftce wind data; thwet-
                                 Icall; derived vertical  pro-
                                 file assumed
Solar radiation
 Severil  (3-5) UV pyrinoneters
 located  1n the region, contin-
 uously recording UV ridlitlon
 levels

 Vertical attenuation of radi-
 ation it i few locitlons
 several  times dally determined
 by aircraft observations

 Spatial  (3-D) Insolation fields
 determined by Interpolation  of
neisurtnents
A single, ground-based net
radiometer; Insolation assumed
constant over the region

Vertical attenuation estimated
empirically as a function of
aerosol miss
No radiation measurapients
available; estimated theo-
retical values based en  uie
tolar tenlth ancle

Attenuation not accounted for
loundarv ind Inltla i
conditions
Hourly species concentrations
titripolited ind Interpolated
throughout the region  using
data fro* the eitenslve
ground-based monitoring  net-
work ; airborne data  also
aviltable; hydrocarbon nli
Obtained from qis chromato-
graphlc analyses it  several
times during the day

Sulfate concentrations avail-
able on in hourly basis  It
several locitlons
Hourly concentrations  eitripo-
lited ind Interpolated using
data from several ground-based
stations; hydrocarbon  nli
obtained fro* gas chrom*to-
graphic analysis It One or two
Stations one or a fe*  times
during the day

Sulfate concentrations based
en i dllly average and diurnal
eione curve
Hourly concentrations  utra-
polated and Interpolated  fro»
I Minimal routine mon1tor1n«
network; either hydrocarbon
Mia assumed or average value
Obtlined from I compilation
of available data taken in •
tlMllar irea

No data en concentration
variations aloft

Sulfate measurements Inferred
fro» values obtained 1n
Similar ar»«i
                                                          B-13

-------
                                              TABLE B-4 (Concluded)
            Input
 Stationary source emissions
            Practical Level
 Separate grldded Inventories
 for  point and irei stationary
 Sources; Characterization of
 organic composition. ind NO/
 N02  and SOj/SOi (mission
 rates  for major sources;
 diurnal and seasonal varla-
 tlons  In nominal mission
 rates  for each major source
 type
                                                                                     Level
Limped, grldded Inventory for
Stationary sources; «0 • pectes
fractlonatlon; seatoiill tr1
dlurnil variation In regional
emissions for each pollutmt
  Minimum »ecept«plt it»fl*

Lumped tutlontry tewrct
(mission  Inventory for Uie
region ti • wtiole. Halted
1nformit1on on the pcrtenUge
of eich source t/pe; ne te»-
poril «irt«t1on
Mydrocirbon spedes distri-
bution
Mli  obtained from ;ts chnjmj*
too'ighlc *nilys1s ef samples
collected throughout the
renton, pirtlculirly neir
liroe sources

Cold start factors applied
grid by grid when calculi-
ting mobile source emissions
M1i obtained fro* standard
emission! factors (AP-42) to-
gether with i dttifled source
Inventory, supplemented n1th
one or t«c 1i, and
traffic data for tntrazonal
trios
AP.I2 emission  factors,
aisjmet vehicle nil,  and
Intrazonal VKT| estimated peak
and off-peak speeds,  fewer
traffic counts available  for
verification, VKT available
for fever najor arterlals
•ridded VMT, emission
factors estimated fro* It
state nil, and averaoe |fOC)
drlvlno profile; assumed
regional speed distribution
Vehicular cold  start dlstH-
butt or
Spatial and temporal  distri-
bution! of cold starts
Inferred from actual  traffic
and demographic data
Cold starts temporally resolved
using traffic distribution;  no
spatial resolution or spatial
resolution only from estimates
of drlvlna patterns
Cold starts as a flied t«r.
centaoe of all driving--
traffic data are not  detailed
enough for spatial  resolution
of cold starts; cold  starts
estimated from demographic
data
Data for nodel performances
evaluation
Hourly averaoed spedes  con.
centratlons for NO, NO;, Oi,
SO?, NWC, sulfate, CO,  and
partlculates from an extensive
•round-based nenltorlno
network
Hourly averaged concentrations
of NO, N02, Oj. SOj.  N«WC.  CO.
and partlculates from several
ground-based stations

Dally averaoed sulfate i«easure-
rents available from  a limited
(3-5) number of stations
Hourly averaoed concentrations
Of NO,. 0). THC. Kb, and CO
fro* a minimal routine vnl-
tori no network
• Using aata  at  this level of detail necessitates numerous  assumptions.
                                                             B-14

-------
of broad geographical  variations  in  labor rates, operating costs, and so
forth, and the ever-increasing  cost  of capital equipment, parts, and
supplies.  An earlier  study for the  EPA  (Miedema et al., 1973) addressed
the cost of implementing air quality monitoring networks of various sizes
in metropolitan areas  where little  if  any monitoring previously existed.
This study, building upon earlier work by Hickey, Rowe, and Skinner
(1971), estimated monitoring costs  for each  state in the United States for
two scenarios:

     >  The required number of  monitors  based  on federal regula-
        tions

     >  The required number of  monitors  based  on state regula-
        tions.

The analysis carried out by Miedema et al.  considered  31 cost  elements
when formulating overall costs  for  a particular network.  These cost
elements are listed in table B-5; as is  immediately apparent,  many  cost
categories exhibit a wide range of  variation,  and most costs  increase in
time.  Accordingly, it is difficult to estimate many costs.   For example,
if one wind monitoring station  was  purchased to augment the already dense
surface network in Los Angeles, the incremental costs  required to train
personnel, accommodate the new  data in the  existing data reduction  and
analysis system, provide for calibration and inspection equipment,  and  so
forth would probably be low compared with the  analogous costs  in Houston,
Texas.  Along the same line, hourly labor costs  in  Albuquerque,  New
Mexico, are less than  those in  San  Francisco,  California.

     Notwithstanding the difficulties in formulating  cost  estimates,  an
attempt was made to identify typical costs  of  routinely acquiring  addi-
tional air quality and meteorological  data.  The  results of  this study  are
presented in table B-6.  In this  analysis,  fixed  hardware  costs were
amortized over a five-year period of time,  but interest costs were
neglected.  In some cases (surface  winds, for  example), the  variable  was
measured continuously; in other cases, the  measurements were routine  but
not hourly.  Twice daily radiosonde soundings  are an  example.   The
frequency of occurrence of each parameter is consistent with the maximum
level of detail, outlined in table  B-4.

     Various sources were consulted in developing the cost figures given
in table 6, including  published reports, equipment  manufacturers,  managers
of air quality monitoring networks, and researchers working  on special
studies.  The costs for the surface air  monitoring  stations  employing a
variety of instruments should be clearly viewed  as  lower bounds because
these figures are estimates made  five years ago  (Miedema et  al., 1973).
In discussing the cost estimates with the authors of the Miedema et al.
                                       B-15

-------
    TABLE B-5.   COST CATEGORIES FOR AIR QUALITY MONITORING  SYSTEMS
                         (a)   Fixed Costs
             Hardware
  Remote electronics
  Central electronics
  Test and maintenance equipment
  Other digital equipment
  Initial spares
  Site installation, physical
          Nonhardware
 Specification preparation and program
 management
 System engineering
 Software
 Documentation
 Training
 Site installation, APCD labor
 Site installation, vendor labor
                       (b)   Variable  Costs
       Nonpersonnel
           Personnel
 Recurring spares
 Utilities, site
 Utilities, communications
 Transportation, local
 Transportation, other
 Computer rental
 Supplier
 Facilities
Personnel, field technician
Personnel, sensor maintenance, corrective
Personnel, electrical  maintenance, correction
Personnel, data analyst, routine
Personnel, data analyst, special
Personnel, laboratory technician
Personnel, chemist
Personnel, engineer
Personnel, clerical
Personnel, administrative
Source:   Hickey, Rowe, and Skinner (1971).
                                  B-16

-------
         TABLE  B-6.   ESTIMATED ANNUAL COST TO AUGMENT
                     AN  EXISTING AEROMETR1C MONITORING
                     NETWORK  WITH VARIOUS INSTRUMENTS
   Parameter Measured

Surface wind velocity
(continuously)

Upper level  winds,
temperature, relative
humidity (twice daily)

Upper level  winds
(twice daily)

Mixing depth
(continuously)

Mixing depth (twice
daily soundings)
Upper level winds
and temperature
structure (continuous)

Solar radiation
(continuous)

S02, particulates,
wind speed, wind
direction

S02, particulates,
wind speed and
direction, CO, 03,
and N02

SO?, particulates,
wind speed and dir-
ection. CO, 03, N02,
total hydrocarbons,
temperature, relative
humidity  (continuously)
        Instrument
Remote recording cup
anemometer and vane

Rawinsonde
Pibal
Monostatic acoustic sounder
Light aircraft with digital
recording temperature
sensor

Instrumented tower
Pyranometer
Surface air monitoring
station
 Surface  air monitoring
 station
 Surface air monitoring
 station
Estimated
  Annual
   Cost

$  7,500


  71,000



   7,000


  13.500


  40,600



  45,000



    3,800


  44,000



   66,700
  100,000
                                 B-17

-------
                    TABLE B-6 (Concluded)
  Parameter Measured

Vertical S02 pollutant
burden (four sampling
days each week during
three-month smog season)

Hydrocarbon speciation
(twice daily, three
times a week during
three-month smog season)

S02, NOX, 03, particu-
lates, relative humid-
ity, bscat, turbulence,
(4-5 hours daily during
special field program)
       Instrument
Estimated
 Annual
  Cost
Correlation spectrometer    $ 81,000
Gas chromatograph
  22,700
Airborne air quality
monitors
   5,000*
  Estimated cost per day.
                                  B-18

-------
study, we found that no attempt has been  made to revise the estimates to a
more current time frame.

     In some instances, it is possible to estimate the cost of additional
monitors by examining current costs of data  acquisition, analysis,
management, and so on.  The California Air Resources  Board  (CARB), for
example, has found over the years,  in  comparing the overall cost of its
monitoring network with the total  amount  of  data collected, that a typical
per-unit cost of data acquisition  is about $1 per number.   Thus, the cost
of one additional hourly surface temperature measurement in an existing
network might be on the order of $8,000 to $10,000 per year.
4.   ANALYSIS OF AIR QUALITY MODEL SENSITIVITY TO VARIATIONS  IN  INPUTS

     At the outset of the study (December 1977),  a review  of  previous air
quality simulation model sensitivity studies was  performed.   In  the
following subsection, the results of this review  are presented.   Because
the photochemical grid models studied (1) represent different model
structures (though they are based on the same general concept),  (2)
represent different levels of model refinement, and (3)  were  applied  to
different urban areas (e.g., Denver, San Francisco, Los  Angeles),  the
sensitivity results are not directly comparable in a quantitative sense.
Rather, they are indicative of trends in model performance likely to  be
observed when certain inputs are varied.

     Within the last five years, a limited number of sensitivity studies
have been performed with grid-based photochemical models.   From a review
of the literature, we found that only two models--the Lawrence Livermore
National Laboratory LIRAQ model and the SAI Airshed Model—have undergone
extensive sensitivity analyses and have had the results  of these studies
published in the open literature.

     Table B-7 briefly summarizes recent grid model sensitivity studies.
Although it is likely that other sensitivity runs have been made,* the
ones identified in table B-7 are the only major sensitivity results that
have been identified by this review.  It is apparent from the table that
several studies have investigated the impact on model predictions caused
by variations in several model variables.  The eight studies  are aggre-
gated  according to four categories--air quality, meteorology, chemistry,
and emissions—which are discussed next.
   Indeed,  in  carrying out  a photochemical model simulation, iterative
   adjustments made  to initial conditions, boundary conditions, etc.,
   constitute  a  form of  sensitivity  analysis, but these results are seldom
   reported formally.
                                     B-19

-------
                                                   TABLE  B-7.    SUMMARY  OF SENSITIVITY  STUDY  RESULTS  OBTAINED  WITH
                                                                      GRID-BASED PHOTOCHEMICAL  AIRSHED  MODELS
                         Study  Group
                 MacCracken.  M.  C., and
                 G.  0.  Sauter (197S)
       Model  Version
       and Attributes
                                                                                   Sensit ivity Analysis
                                                                                        Variations
                                                                                                                  Influence on Model  Predictions
                                                                                                                                                                     Remarks
DO
ro
o
                  Deroerjian,  K.  L.  (1976)
                    "EPA  3" VERSION
                  Liu.  M.  K..  et  al.
                  (1976)
LIRAQ photochemical model

  Two-dimensional time-
  dependent  grid model

  Lumped kinetic mech-
  anism similar to
  Hecht-Seinfeld-Dodge
  mechanism

  Mass conserving wind
  field
  SA1 photochemical model:
  SAI photochemical model:
  "EPA 3" version
    ?5 x 25 x 6 grid

    15-step Hecht-Seinfeld-
    Oodge kinetics

    Price numerical method

    Empirical diffusion
    algorithm

    Two-dimensional wind
    field
Relative humidity was
reduced from 40t by ?0*

Nominal temperature was
increased from  ?85"K to
31)4'K

Light intensity was reduced
by b(«
                                                                               Iiqht intensit y was
                                                                               increased by ,i  factor
                                                                               of 2
                                                                                Initial hydroi arhons  ,irt>
                                                                                increased hy j fdc tnr  of  ?
Initial NO^ concentrations
were increased hy a  factor
of 2

Boundary conditions  were
reduced by SUt
Initial and boundary  con-
ditions were reduced  by
bOT

Hind directions were  random-
ly perturbed by
0 or t22.5*
                                                                               Wind speeds were randomly
                                                                               perturbed by 0 or tl mph
                                                                               Wind station measure-
                                                                               ments were:
Peak ozone  increased by 3% and peak
N0? decreased by  4X

Peak ozone  decreased by 2X and peak
NOj increased by  SI
Peak ozone decreased by  70% and
Nl)^ peak  magnitude remained
unchanged but  was delayed 4
hours

Peak ozone increased by  100* and
NO;, pe.ik  magnitude slightly
increased and  preceded base case
peak by 1-3/4  hours

NiK peak  increased hy fit arid was
ife 1 ayert jpprox imdte! y I  hour;
ozone peak was  not reported, but
the increase in ozone concentrations
was delayed by up to 3 hours

NO-, peak  increased by 10* and was
delayed slightly; 0-j remained
unchanged

"Minor" differences  occurred in
ozone prediction in  the  eastern
and northern portions of the L.A.
basin; "significant" differences
were observed  in the western
and central portions of  the basin

Predicted ozone in the northern
and eastern edges of basin were
reduced 20 to  30*

A b.9* average deviation for manu-
ally prepared  and 4.9t
for automatically prepared
wind fields (based on CO
predictions)

A 4.9* average  deviation for man-
ually prepared  and ?.6<  for auto-
matically prepared wind  fields
(based on CO predictions)

Maximum absolute deviation  from
the base case results for CO
were
LIRAQ sensitivity  runs focused
on the kinetic  module; accord-
ingly, sensitivity results are
more reflective of smog chamber
simulations than they are of airshed
simulations
 In the automatic.wind field  studies,
 perturbations were made to the
 monitoring station measurements  and
 then automatic procedures were
 employed to derive gridded wind
 fields.  In the manual wind  field
 cases, perturbations were made to
 the gridded wind fields after they
 had been prepared manually
                                                                      The response of the model  to
                                                                      variations in wind speed varies
                                                                      with each chemical species and
                                                                      is time dependent

-------
                                                                         TABLE  3-7  (Continued)
            Study Group
                                          Model  Version
                                          and Attributes
    Sensitivity Analysis
         V.ir idt ions
                                       Two-dimensional
                                       initial  conditions
Increased SOI

Increased ?5X

Decreased ?0.?t

Decreased SOX

Horizontal diffusion was
decreased to II and in-
creased to 500 m?/sec
CD
 I
ro
                                                                   Vertical diffuswity was
                                                                   decreased to 0.5 m'Vsec
                                                                   and  increased to 50 m?/sec
                                                                   Mixing depths were in-
                                                                   creased and decreased by
                                                                   35X
                                                                  Radiation  intensity  was
                                                                  increased  and decreased
                                                                  by 30*
    Inf luence on Model Predictions

 19.bt

 U.8t



 51./t

 for KH  = 0, the maximum abso-
 lute deviation for CO ranged
 between 0.52 and O.OZt from
 0600 to 1600 hours

 For KH  =• 500 mz/sec, the
 maximum absolute deviation for
 CO ranged between 4.4 and 12.91
 from 0600 to 1600 hours

 The effect of varying vertical
 diffusivity by an order of mag-
 nitude was about the same as
 that of varying the wind speed
by 25 to 501

 Maximum absolute percentage
deviations for the increase and
decrease,  respectively,  were:

   For CO,  8t and 12t

   For NO,  lit and 18.51

   for N0?,  B.5t  and  15.5X

   For 03,  11.5J  and  23t
                                                                                                                                                           Remark s
                               Maximum absolute percentage
                               deviations for the increase
                               and decrease,  respectively,
                               were:
                                                                       The base case value was 5 m?/sec
 The buildup of the mixing depth
 variation effect  is time-
 dependent

 Decreasing the mixing depth has
 a greater effect on the ground-
 level concentrations than in-
 creasing it; this result is more
 pronounced for reactive pollu-
 tants

 The effect of changing the mix-
 ing depth is not uniform over
 the modeling region; it varies
 from place to place

 The effect on ground-level  con-
 centrations of changing the
mining depth is roughly the same
 as that of changing the wind
 speed, as would be expected from
 a dimensional  analysis

 The effects of varying the
 radiation intensity are time-
 dependent

-------
                                                                               TABLE  B-7  (Continued)
               Study Group
                                             Model Version
                                             and Attributes
                                     Sensitivity Analysis
                                         Variations
                                                                       Emissions rate (ground
                                                                       based) Mas  increased and
                                                                       decreased by 15*
CO
 I
ro
ro
         Reynolds.  S.  0.. et  al.
SAI photochemical  model:
"EPA 3" version [see Liu
et al. (1976)]
Uniform wind velocities with
height were compared with
vertical variation in hori-
zontal winds given by a
power law formulation
         Anderson.  G.  E..  et  al.
         (1977)
SAI photochemical model:
"Denver" version
Wind speeds were reduced
by 3 it
                                   Influence on  Model Predictions


                                   For  NO,  17*  and  40*

                                   For  N02.  74*  and 55*

                                   For  03,  9*  and 11*

                                The effects  of  increasing
                                and decreasing  emissions
                                rates are almost identical;
                                peak basin-wide  average per-
                                centage changes  in  CO and
                                NO^ were about  the  same
                                (6-8*)
[he maximum average percentage
deviations were:  28.St for NO, IS*
for N0?. 24* for CO, and 14* for 03

The maximum average deviations in
pphm were:  -0.35 for NO, -1.1 for
N0?, -4 for CO, and -2 for Oj

The maximum deviations in pphm
were:  7.5 for NO, 15 for NO;,, 30
for CO, and 26 for 03

Maximum predicted ozone increased
by 44; maximum area for which [Oj] >
0.08 ppm  increased by 12*
                                                                                      Remarks
                                                                                                                                               The effect  of  changing  light
                                                                                                                                               intensity is as significant
                                                                                                                                               as that  of  changing wind speed
                                                                                                        The study results are summar-
                                                                                                        ized by the following ranking
                                                                                                        of the relative importance
                                                                                                        of the input parameters (A -
                                                                                                        most important and D - least
                                                                                                        important):

                                                                                                        Parameter or
                                                                                                          Variable     CO  NO  03  N02
                                                                                                                                               Hind speed
                                                                                                                                                              A   A   A
                                                                                                        Horizontal     0   D   0    0
                                                                                                        diffusivity

                                                                                                        Vertical       C   C   C    C
                                                                                                        diffusivity

                                                                                                        Mixing depth   B   8   B    B

                                                                                                                       D   A   A    B
                                                                                                                                               Radiation
                                                                                                                                               intensity

                                                                                                                                               Emissions
                                                                                                                                               rate
                                                                                                                       B
                                                                                                                               B    B
The effects of including wind shear
were similar to those of increasing
surface wind velocities by roughly
25t because velocities within the
mixed layer are increased between 20
and 701 of the surface values as a
result of shear

-------
                                                                                 TABLE  B-7  (Continued)
                  Study Group
                                               Model Version
                                               and Attributes
03
ro
to
          Killus. J. P. (1977)
          (private communication)
          Anderson, G. E. (1977)
          (private communication)
          Attaway. L. 0., et al.
                                           31-step carbon bond
                                           chemistry

                                           3-D  Mind field

                                           Lower microscale  layer

                                           Lamb and Liu diffusivity
                                           algorithms

                                           30 x 30 x 7 grid

                                           SHASTA numerical method

                                           Suffice removal

                                           Three-dimensional initial
                                           conditions
          DeHandel et a\. (1979)
SA1 photochemical model:
"Denver" version [see
Anderson et at. (1977)
for mode) attributes]
SAl photochemical  model:
"Denver* version [see
Anderson et  0.08 ppm increased by 301

                                                                 No difference occurred in the time,
                                                                 location, or magnitude of maximum
                                                                 ozone concentration; differences
                                                                 among predicted ozone concentrations
                                                                 in all runs were not more than 0-010
                                                                 ppm in one or two grid cells at most

                                                                 No difference occurred in the time,
                                                                 location, or magnitude of maximum
                                                                 ozone concentration; differences
                                                                 among predicted ozone concentrations
                                                                 in all runs were not more than 0.010
                                                                 ppm in one or two grid cells at most
The coarser grid resolution led to
no noticeable change in the time
to peak" NO, N0?, and Oi concentra-
tions; the magnitude of peak con-
centrations was reduced for NO
(691), N0? (211). and 03 (131)
The maximum impact of increased
source emissions anywhere in the
model inq region was an increase
in hourly averaged NO and NO^ con-
centrations (I? and 5 pph, respec-
tively) and a decrease in 03
(-4 ppb)

The estimated maximum increment in
three-hour-average S0? (0900-1200)
concentrations was 70 ppb immediate-
ly downwind of the facility; concen-
tration differences dropped below
10 ppb at a distance of 24 miles
downwind of the source

Region-wide maximum ozone concentra-
tions were reduced by !0 percent
from a baseline peak of 0.20 ppm
                                        A synergisn exists between wind
                                        speed and mixinq depth
                                        In each scenario, no more than 71
                                        of the reqion-wide emissions were
                                        redistributed; changes of this
                                        size in the spatial distribution
                                        of emissions has little effect an
                                        secondary pollutants such as ozone
By the time ozone  forms,  its  pre-
cursors have been  distributed over
a much greater area than  their  source
regions; accordingly,  the influence of
increased grid size on ozone  predic-
tions should be less than that  for
primary pollutants such as NO

The effect was decidedly  local  and
did not influence  peak oiidant
concentrations
The 70 ppb value is an upper bound
because the mesoscale model over-
estimated ground-level concentra-
tions in the vicinity of buoyant
point sources
The time of occurrence of  the ozone
peak remained unchanged

-------
                                                                                 TABLE  B-7  (Concluded)
                Study Group
                                              ftxtel Version
                                              anj Attributes
                                     Sensitivity  Analysis
                                          »dr i il mns
                                                                                                           Influ^nci- mi Huilel I'rpJ ir. t ions
                                                                                                                        Memarks
                                       Compact kinetic mechanism
                                       similar to Hecht-Seinfeld-
                                       Dodge mechanism

                                       Mass-conserving wind field

         Sou ten, D. R.. et al. (1980)  SAI photochemical model:
                                       "IPA 5" version [see Reynolds
                                       et al. (1979)]
         Killus, J. P.. et al. (1980)  SAI photochemical model:
                                       •EPA-5- version [see Reynolds
                                       et al. (1979)]

                                       SAI photochemical node):
                                       "EPA-S" [see Ames et al.
                                       (1978)]

                                       Multiple-day simulation for
                                       Los Anqeles, CA
CD
ro
         Reynolds, S. 0., et  al.
         (1979)
SAI photochemical model:
"EPA-S" version [sec
Reynolds et al. (1979)]
                                 (missions  were  distributed
                                 according  to  the  demographic
                                 distribution
                                 Kea(tive  hydro* arbon  emis-
                                 sions  from bioqenic  sources
                                 (57  percent of  the RHC
                                 inventory)  were eliminated
                                 from the  inventory
                                 Three-dimensional  initial
                                 condition  field from monitor-
                                 inq data versus "clean air"
                                Keqion-Mide  ma«imum  o/one concentra-
                                tions  nere reduced by ?S percent;
                                the  predicted  peak ozone level
                                occurred  three hours after  the  base
                                case maximum

                                Area-wide n/one levels on the second
                                of  a multiple-day simulation were
                                reduced by no  more than 2 ppb
Background hydrocarbon 0.06
ppmC versus 0.18 ppmC

Wind fields for the airshed
model were prepared usinq the
following procedures:

  1.  An  interpolation
      algorithm

  ?.  A tMu-dimensional wind
      model

  3.  A three-dimensional Kind
      model
Essentially zero effect on second
day of simulation; some effect on
first day, especially in western
portion of mode I inq region
in the morning

No orone predictions above 0.? ppm
on first Jay; no o/one above back-
qrnund on second day

No effect on ofone peaks; minor
effects at some stations
                                                                                                                                    NO,
                                                                                                          Procedure
                                                                                                              I        -0.7   6.8   -?.0   3.5

                                                                                                              ?         1.4   B.6   -l.S   4.0

                                                                                                              3         0.1   7.6   -?.?   3.7
                                        The delay  in  the oione peak nas
                                        attributed to a redistribution of
                                        point  source  NO, emissions from
                                        industrial areas to residential areas
                                        For the meteorology studied,
                                        hioqenic hydrocarbon emissions
                                        had no major bearing on peak
                                        calculated  ozone  levels
                                                                        Effects on  the second day of twn-day
                                                                        simulation  are driven by emissions
                                                                                                                                                Effects on the second day of  two-day
                                                                                                                                                simulation are driven by emissions
Background HC has limited  effects
below certain point

Eiamination of the ozone results
reveals the following:

  Use of  interpolated *md fields
  leads to the qreatest bias toward
  underestimation at the highest con-
  centration levels

  Compared with the intercalated  and
  three-dimensional  wind field simula-
  tions,  the tKo-dimensional wind  field
  simulation exhibits a greater
  tendency toward overestimation  for
  most of the observed concentration
  range

  The three-dimensional wind field
  simulation exhibits less bias (posi-
  tive or negative) overall than  do the
  other two simulations
           Accuracy.
         * Precision.
                                                                         Grid restitution was related
                                                                         from ?«? miles tn 4 x 4
                                                                         miles
                                                                 taierally, d reduction in the max-
                                                                 imum ii/iine concentration occurs at
                                                                 the monitorinq stations together with
                                                                 d "broadening" of the diurnal ozone
                                                                 prof i le

-------
a.   Studies Focusing on Air Quality Inputs

     Sensitivity analyses in which air  quality inputs  have been varied
were reported by MacCracken and Sauter  (1975)  and  Demerjian  (1976).
Collectively, these studies examined perturbations in  model  predictions
from base case simulations caused by the following changes:

     >  Initial hydrocarbon concentrations  increased by  a factor
        of 2.

     >  Initial NC^ concentrations increased  by a  factor of  2.

     >  Boundary conditions reduced by  50 percent.

     >  Initial and boundary conditions reduced by 50  percent.

The measures of model performance that  were used in these studies  included
the percentage change in the magnitude  of the peak 03  and N0£ concentra-
tions and the time delay in reaching peak concentrations.   In each case,
the overall impacts on the spatial maximum  0^ and  N0£  concentrations  (in
percentage variation from the base case) were far  less than  the changes
made in initial or boundary conditions.

     These early studies represent an initial step in  analyzing the  impact
of variations in air quality inputs (i.e.,  initial and boundary condi-
tions) on grid model predictions.  Although they provide insight  into the
expected order of magnitude of changes  in model predictions  (at least over
the range for which the inputs were varied),  other issues need to be
investigated:

     >  What is the impact on predictions  caused by variations  in
        the assumed initial and boundary condition hydrocarbon
        species compositions?

     >  What is the impact on predictions  caused by variations  in
        boundary,conditions over a much wider range of concentra-
        tions than have previously been explored?  In  some
        simulations, uncertainties in boundary conditions upwind
        of the urban area, and pollutant concentrations in  layers
        aloft, have been much greater than  the range of  values
        explored in sensitivity studies to date.
                                     B-25

-------
     >  What is the impact on predictions caused by various
        procedures  for creating initial  and  boundary condition
        fields?

     >  What is the impact on model ozone predictions caused  by
        computer simulations of multiple-day  periods?

     To this point, the discussion of sensitivity analyses has focused on
airshed model simulations of one day or less.  As we point out in the main
body of this report, a reliance on single-day simulations as  a means of
revealing model sensitivities tends to overstate the importance of  air
quality data (used to specify initial and boundary conditions, and  to
understate the importance of other data,  i.e.,  meteorological and emission
inputs).  Recently, Killus et al. (1980)  reported results of  a multiple-
day simulation for Los Angeles.  Using this simulation as a basis for
comparison, Souten et al. (1980) conducted a  sensitivity simulation to
examine the influence of a 57 percent reduction in reactive hydrocarbon
initial conditions on predicted ozone maxima  on the second day of a smog
episode.  As indicated in table B-7, model ozone calculations were
perturbed by no more than 2 ppb on the second day.  These preliminary
findings suggest that the need for detailed air quality monitoring  data
may be reduced if it is possible to develop satisfactory multiple-day
simulations for a particular urban area.   Of  course, as the need for air
quality data is reduced by use of multiple-day simulations, the need for
improved meteorology becomes more pronounced.
b.   Studies Focusing on Meteorological  Inputs

     Sensitivity studies involving meteorological  inputs  have  investigated
variations in wind fields,  mixing depths,  and  diffusion rates.  For
conservative pollutants, it was found that the airshed model predictions
are noticeably more sensitive to reductions in wind  speed than to
increases (Liu et al., 1976).  Furthermore, in another study (Anderson  et
al., 1977), wind speed reductions appeared to  have a far  smaller effect on
secondary pollutant (ozone) concentrations than on primary  concentra-
tions.  Finally, the effect of including wind  shear  (vertical variation in
wind speed with height) in  place of uniform winds  was found to be  compar-
able to a 25 percent increase in surface wind  speeds (Reynolds et  al.,
1976).
  Linear interpolation of ozone concentrations  observed  at  street-side
  monitors may grossly underpredict  the magnitude of  an  area-wide ozone
  levels.  Other interpolation  schemes, for  example,  based  on mass
  balances or Poisson fitting routines, may  provide more realistic
  estimates (in some cases).
                                     B-26

-------
     The study by Liu et al.  (1976)  indicated that model performance may
be degraded more by a reduction  in the magnitude of an input variable than
by an increase in the magnitude  of an  input.  This trend was found to be
the case for mixing depths  in their  study.  Moreover, a concurrent
reduction in wind speed and  direction  revealed  that a synergism exists
between wind speeds and mixing depths  (Anderson et al., 1977).  Sensi-
tivity analyses in which horizontal  turbulent diffusivity was varied from
zero to probably an extreme  value for  the  urban atmosphere during smog
episodes (~500 m^/sec) showed only  a minor effect on conservative pol-
lutant concentrations.  The  effect on  reactive  species would probably be
even smaller.  However, the effect  on  ozone levels of varying the vertical
diffusivity an order of magnitude (below and  above a base case value) was
comparable to varying wind speeds by 25  to 50 percent.  Also, a decrease
in the vertical diffusivity had  a more pronounced impact on ozone predic-
tions than an increase in diffusivity.

     In short, the sensitivity studies carried  out to date  indicate  that
photochemical model predictions are more sensitive to overall reductions
in the magnitude of parameters associated  with  contaminant  dilution—wind
speed, mixing depth, and diffusivity--than to corresponding increases  in
the parameters.

     This review found only two studies  that addressed  the  impact  on model
predictions caused by alternative procedures for  preparing  meteorological
inputs, specifically wind fields (Liu et al., 1976;  Reynolds et al.,
1979).  Liu et al. (1976) investigated two procedures:

     >  Manual preparation of the wind field by smoothing and
        interpolating measurement data.

     >  Automatic  preparation of the wind field by numerical
        weighting  and smoothing routines.

The studies by Liu et al. involved  (1) randomly varying wind speed
measurements by 0  or ±1 mph,  and (2) randomly varying wind direction
measurements by 0  or 122.5*.  Wind  measurements so perturbed were used in
the manual  and automatic wind field preparation processes.  Neither type
of perturbation .had much influence  on grid average concentration devia-
tions  (about the base case).  However, the maximum local deviations (about
the base case) were  larger,  particularly  for the case of variable wind
direction.

     Reynolds  et  al.  (1979)  examined  the  influence on airshed model ozone
predictions caused by the use of alternative wind field generation
procedures.  Three approaches to the  prescription of wind field were
investigated:


                                        B-27

-------
      >   Use  of  an  interpolation scheme, together with an objective
         procedure  for minimizing wind field divergence aloft.

      >   Use  of  a two-dimensional, diagnostic wind model (Liu et
         al., 1974).

      >   Use  of  a three-dimensional, mass consistent, diagnostic
         wind model  (Yocke and Liu, 1978).

 Upon  examination of the ozone results, Reynolds noted the following:

      >   Use of  interpolated wind fields leads to the greatest bias
         toward  underestimation at the highest concentration
         levels.

      >   Compared with the interpolated and three-dimensional wind
         field simulations, the two-dimensional  wind field simula-
         tion exhibits a greater tendency toward overestimation for
        most of the observed concentration range.

      >   The three-dimensional wind field simulation exhibits less
        bias (positive or negative)  overall than do the other two
         simulations.

      Estimates of model accuracy and precision  were derived through
computation of the first and second  moments of  the distribution of
residuals (differences between hourly model calculations and observa-
tions).   The three wind field sensitivity runs  produced these results  for
ozone and N02:


                               Ozone (pphm)         	 NO? (pphm)
Simulation
Interpolated wind field
inputs
Two-dimensional model
wind field inputs
Three-dimensional model
Accuracy
-0.7
1.4
0.3
Precision
6.8
8.6
7.6
Accuracy
-2.0
-1.5
-2.2
Precision
3.5
4.0
3.7
wind field inputs
                                        B-28

-------
     Reynolds et al.  concluded that,  on  the  basis  of computed measures of
accuracy, precision,  and  bias, and  of precision  at upper percentile ozone
concentration levels, the three-dimensional  wind model  appears to offer
the best simulation results.   However, there were  several instances where
this procedure for supplying  wind inputs led to  poorer  model performance,
such as at a particular monitoring  station or over a particular range of
observed concentrations.

     With the exception of the two  studies just  discussed,  all of the
sensitivity studies to date have been designed so  that  the  perturbation to
diffusivities or wind fields  is uniform  across the modeling grid.  The
same is true for studies  involving  mixing depths.  With the results of
past sensitivity studies as a foundation, certain  additional  analyses
might be performed to investigate

     >  The impact on model predictions  of  using a fully  three-
        dimensional wind field rather than  a uniform  field  (x,y
        variations only)  or a uniform fielrf "extended  aloft",
        based on theoretical  arguments.

     >  The impact on model predictions  caused by horizontal
        variability  in the vertical diffusivity  fields.

In the first case, the extent to which model predictions  are influenced  by
the procedure for preparing wind fields  will undoubtedly be governed  by
the meteorological complexity of the urban area whose data base is used  in
the sensitivity analysis.  Model predictions might be much more sensitive
to wind field preparation procedures used in a Los Angeles application,
for example, than in Tulsa, Oklahoma.  Similarly, the horizontal  varia-
bility  in vertical diffusivity  is  greater over  an urban area exhibiting
irregular or complex topography than over smooth terrain.
 c.    Studies Focusing on Chemistry Inputs

      Sensitivity studies on the kinetic mechanisms of photochemical models
 have  centered on variations in ambient temperature, relative humidity, and
 solar radiation.  The first two parameters have been shown to be rela-
 tively uninfluential in affecting model predictions, at least for the
 ranges in  each  variable that were explored (MacCracken and Sauter,
 1975).  In contrast, variations in solar radiation, which affect the
 photolysis rates of NC^, aldehydes,  HNC^, and HgC^ have been shown to be
 quite significant.  For example, MacCracken  and Sauter (1975) found that a
 50  percent reduction in light  intensity reduced the peak ozone concentra-
 tion  by 70 percent.
                                         B-29

-------
     Other possible sensitivity studies  involving AQSM kinetic mechanisms
could be entertained that might comprise an examination of the effects of

     >  Attenuating the  intensity of  solar radiation with height
        instead of assuming uniform values throughout the depth of
        the modeling region.

     >  Prescribing the  individual photolysis rates for NOg,
        aldehydes, HNC^, and h^C^, instead of assuming that the
        photolysis rates of the last  three species are propor-
        tional to the nitrogen dioxide photolysis rate.

     >  Evaluation of alternative kinetic mechanism such as those
        proposed by Falls and Seinfeld (1978), Durbin and Hecht
        (1975), or Whitten and Hogo (1977).

Clearly other sensitivity studies focusing on chemistry inputs can  be
envisioned, (e.g., to vary chemical reaction rate constants).  However,
these are perhaps best reserved for the more complex photochemical  smog
chamber simulations (Whitten and Hogo, 1977) in which explicit rather than
condensed mechanisms are used.
d.   Studies Focusing on Emissions  Inputs

     Several basic sensitivity studies have been performed with  source
emissions:

     >  Overall increases or decreases in emission rates.

     >  Relaxation of the spatial resolution of the emission
        inventory to accommodate a  coarser airshed grid.

     >  Examination of the impact of single point sources or
        individual source categories on basin-wide oxidant or
        sulfate levels.

     >  Localized reductions in emissions with proportional
        increases elsewhere in the  region to give overall emission
        rates equal to those in the base case.

These first three sensitivity analyses are quite straightforward.  As
indicated in table B-7, studies involving small overall emission  increases
or reductions,  aggregation of sources into a slightly larger grid, and
examination of the influence of minor sources on basin-wide air  quality
have found that the impact on basin-wide model predictions is relatively
small.
                                   B-30

-------
     One sensitivity analysis  performed  by Anderson et al.  (1977) focused
on the influence of spatial  variations in emission rates.   They found that
a reduction in emissions of  25 percent in any  one of  eight  satellite
Denver suburbs did not influence  the  time, location,  or magnitude of the
region-wide maximum ozone concentration.  (In  each scenario, no more than
7 percent of the region-wide emissions were redistributed.)
e.   Studies Focusing on Grid Specification

     DeMandel et al.  (1979)  report several  interesting  sensitivity studies
that use the LIRAQ model developed at  Lawrence  Livermore  National Labora-
tory.  One evaluation reduced the model's  horizontal  resolution from 5 km
to 10 km.  In the single-day simulation, peak calculated  ozone levels were
reduced 10 percent from 0.20 ppm to 0.18 ppm.   This  reduction was
explained on the basis of "spatial smoothing".   The  emission densities of
precursor species were reduced by spatial  averaging  over  the larger grid
cell size.  This resulted in lower concentrations  of precursors and lower
reaction rates.

     Reynolds et al.  (1979)  compared airshed model ozone  predictions based
on grid resolutions of 2 miles (3.2 km)  and  4 miles  (6.4  km).  Comparison
of the temporal ozone profiles at the  monitoring stations indicated that,
for the most part, the profiles do not change  appreciably when the 4x4
mile simulation is introduced.  Four exceptions were the  Reseda,  Upland,
Azusa, and Pasadena stations.  Examination of  the  profiles indicates that
reducing the grid resolution to 4 x 4  miles leads  to:

     >  An increase in predicted concentrations of ozone  at Reseda
        by a few pphm and a broadening of  the  temporal  profile.

     >  A reduction of the predicted peak  ozone level at  Upland by
        roughly 5 pphm.

     >  A reduction in the predicted peak  ozone level at  Azusa by
        about 6 pphm.

     >  A reduction in the predicted peak  ozone level at  Pasadena
        by about 10 pphm.

Reynolds et al. concluded that a decrease in grid resolution may lead  to a
slight reduction in peak predicted concentrations, at least at certain
monitoring stations.  Furthermore, the 4x4 mile grid run yielded  results
that were more "accurate" over the entire concentration range, though  at
peak concentration levels it was  less accurate than was the 2  x 2 mile
grid simulation.
                                       B-31

-------
     The studies just discussed represent an Important  but  preliminary
step in understanding the sensitivity of  photochemical  grid models to
variations in emissions.  While an understanding of the model's  sensitiv-
ity to overall changes in emissions is naturally of interest, other issues
need to be addressed.  In the next subsection we consider certain analyses
that might be carried out to determine grid  model  response  to various
changes in the components of an emission  inventory.
5.   ISSUES RELATED TO THE PREPARATION OF EMISSION INVENTORIES

     Air quality models are generally used in two ways:   model  performance
evaluation and application.  Model  evaluation consists of tests of the
model using a data set or sets to determine the extent to which the model
replicates field measurements.  One of the objectives of  the  evaluation
phase is to ascertain whether biases exist in the model  performance that
might later be alleviated by a more suitable treatment of atmospheric
processes, alternative numerical methods, more accurate  and detailed model
inputs, and so on.  In evaluative studies, the disaggregation of  various
sources in an emission inventory by source type is seldom necessary.  What
is required is overall grid volume  emission rates for each pollutant
species.  Ideally, the temporal distribution of emission  rates  within each
cell is known or inferred from demographic, industrial, commercial, and
other types of data.

     In contrast, in an applications study, a model is typically  used with
an assumed set of "worst case" meteorological conditions  in conjunction
with an emission inventory that reflects a proposed or anticipated change
in emissions from some baseline level.  If the reduction  (or  increase) in
emissions is uniform, regardless of whether the concern  is region-wide or
within a given subarea, the emission inventory used for model performance
evaluation may suffice.  However, if the applications  study focuses on the
effectiveness of a particular emission control tactic  in  maintaining or
reaching a particular air quality goal, then a more detailed  emission
inventory may be essential.  In the following paragraphs  typical  emission
control measures are identified together with the corresponding level of
detail required of an emission inventory so that a complex model  could be
used to assess the effectiveness of the measure.

     To provide a structure for this discussion, table B-8 presents
various emission control measures and strategies/ which  were selected by
the San Francisco Association of Bay Area Governments  (ABAG,  1977) from
  A control measure is an individual emission reduction proposal;  a
  control strategy may entail  two or more control  measures.

                                       B-32

-------
  TABLE B-8.  CONTROL MEASURES AND  EMISSION  INVENTORY DATA NEEDS
        Control  Measure
Stationary source measures

  Restrictions on the type  of
  industrial  solvents used


  Closed organic storage
  Limitation on the maximum SO
  emissions of any source to a
  prescribed level
  Limitation on the maximum
  sulfur content in fuel
  Best available control tech-
  nology on new or existing
  sources

  New source review with or
  without offset
  Reduction 1n motor gasoline
  vapor pressure
        Data Needs in the
   Source Emission Inventory
Location, size, and operating
characteristics* of coating
facilities

Location, size, and storage
characteristics of facilities
handling organic chemicals and
fuels

Location, size, and operating
characteristics of all S02
emission sources larger than
a prescribed level

Location, size, and operating
characteristics of combustion
sources operating on high sulfur
fuel; emission rates given a
switch to low  sulfur fuel

Location, size, and operating
characteristics of new or
existing  sources

Location, size, and operating
characteristics of new source
as  well  as  the existing  source(s)
to  which the offset is to be
applied

Location, size, and operating
characteristics of all facil-
 ities handling significant  quan-
 tities of gasoline (see  also
mobile source emissions  measures)
   Operating  characteristics  of a  particular emission source may
   Include  such  factors  as  nominal  pollutant emission rate, emissions
   composition,  heat flux,  elevation of point of emissions, flow rate,
   diurnal  and seasonal  variations in emission rates, composition of
   fuel,  and  so  on.
                                 B-33

-------
                     TABLE B-8 (Continued)
        Control  Measures
  NOX control  of off-highway
  construction and agricultural
  activities
  N0j< limitations on new
  boilers and furnaces
        Data Needs in the
   Source Emission Inventory

Temporal and spatial  description
of construction and agricultural
activities (e.g., crop burning)
and characterization of emis-
sion rates

Location, size, and operating
characteristics of new boilers
and furnaces
Mobile source emission
measures

  Exhaust emission controls
  Evaporative emission controls
Gridded vehicular emission rates
embodying:

  Spatially and temporally resolved
  traffic flow characteristics,
  such as traffic volume, overall
  driving speed, cruise speed(s),
  acceleration and deceleration
  range, percentages of time
  spent at cruise and at idle,
  number of speed changes per mile,
  number of cold starts, etc.

  Vehicle mix (Including age dis-
  tribution of vehicle population)
  and model split (between motor
  vehicles and busses, trains,
  rapid transit, etc.)

  Emissions factors based on ele-
  vation, the "average vehicle in
  the region," EPA heavy duty
  vehicle emission estimates,
  unique terrain features (grades),
  etc.

Gridded estimates of the distribu-
tion of "hot soaks" (see reduction
in motor gasoline vapor pressure
measures)
                                 B-34

-------
                      TABLE  B-8  (Concluded)
       Control  Measures
  Operation of a retrofit
  program
  Emission standards for
  other mobile sources
  Motor vehicle inspection
  and maintenance programs
         Data Needs in the
    Source Emission Inventory	

Identification of the age distribu-
tion of the regional vehicle popu-
lation and emission rates result-
ing from evaporative emissions  and
catalytic exhaust emission retro-
fit dtvices

Emission rates embodying  spatial
and temporal resolution for mobile
sources, including motorcycles,
agricultural equipment, construc-
tion equipment, vessels,  locomo-
tives, aircraft* recreational
vehicles, and miscellaneous util-
ity engines (log splitters, tree
cutters, etc.)

Estimate of number of vehicles
inspected annually and percentage
emission reduction attributable
to vehicle maintenance, replace-
ment, etc.
Transportation control  measures

  Improvement in traffic flow
  (e.g., ramp metering)
  Reduction of peak-period
  traffic volumes

  Control over auto use and
  access (e.g., parking limi-
  tations, gas rationing, tolls)

  Encouragement of alternative
  travel modes (ride sharing,
  bicycling, etc.)
Similar to those under exhaust
emissions controls; in addition,
estimates of modal shifts and
changes in VMT due to the control
measure

Temporal and spatial resolution
of trip origins and destinations

Similar to improvement of traffic
flow measures above
Similar  to  improvement  of traf-
fic  flow measures  above
                                B-35

-------
proposals that might be adopted for controlling emissions from stationary,
mobile, and land use sources.  Though not exhaustive,  the measures do
reflect a range of possible control methods that might be investigated
using air quality models in future SIP analyses.

     Considering stationary source control measures first, table B-8
reveals that, for adequate testing of many of the measures, disaggregation
of stationary sources by type and size of operation is necessary.
Clearly, if one were attempting to assess the impact of controls imposed
on dry cleaners, for example, on basin-wide oxidant levels, it would be
necessary to locate and define the emission strengths  of these numerous
sources throughout the urban area.  Such a level of detail typically does
not exist in most conventional emission inventories.  Controls on refinery
operations might be easier to analyze given an aggregated emission
inventory because of the far fewer number of sources in an urban area and,
perhaps, because of a better estimation of overall refinery emission
rates.  (Note, however, that the distribution of reactive hydrocarbon
emissions from refineries is probably poorly known because of numerous
fugitive sources and hydrocarbon species.)

     Measures that attempt to reduce vehicular emissions are broadly
categorized in table B-8 under the headings "mobile source emission
measures" and "transportation control measures."  Examples of control
measures in these categories include

     >  Stringent exhaust and evaporative emission controls

     >  Inspection, maintenance, and retrofit programs

     >  Ramp metering

     >  Parking limitations and regulations

     >  Gas rationing

     >  Increased gas and parking taxes

     >  "Smog charges"

     >  Fare reductions on public transit

     >  Bus and carpool lanes

     >  Auto-free zones.
                                        B-36

-------
     As with stationary sources,  an  analysis of these or other vehicular
emission control  measures  is  complicated by the aggregation that takes
place in preparing conventional  inventories of the emission rates from
various sources into a composite  value.  The processes by which this
confounding takes place is summarized  next and then suggestions are
offered as to how the loss of detailed information on particular sources
might be avoided in the preparation  of new inventories.
a.   Mobile Source Emission Inventories

     Three general procedures are used in compiling mobile source  emission
inventories:

     >  Manual link-by-link summation

     >  Automated link-by-link summation

     >  Estimation based on gasoline sales.

The first method, a tedious one, requires estimation of emissions  from
each section of freeway and arterial streets on the basis of traffic
counts (available on maps from local agencies), peak and off-peak  speeds,
light versus heavy duty vehicle mix, and "minor" street traffic volumes.
Corridor inventories are generated through these analysis; regional
inventories are derived by apportioning the corridor emissions to a
regional grid and assuming that minor streets contribute some fraction of
the corridor emissions.

     Automated link-by-link emission inventories are based on regional
transportation models.  The transportation forecasting model is used to
simulate trip generation, travel on various roadway segments, peak and
off-peak speeds, total VMT, cold starts, hot soaks, and so forth.  These
estimates, when combined with appropriate emission factors [such as those
contained  in AP-42 (EPA, 1972) and the most recent supplements], are used
to generate emission rates that  are then "loaded" onto  a regional emission
grid.  More flexible than the previous method, the automated approach
(which uses a simulation model for traffic characteristics) sacrifices
some accuracy by  using the transportation model to calculate VMT rather
than using actual data.

     Finally, gross attempts  to  construct  a regional mobile source
inventory  can be  based on  an  inventory of regional gasoline sales.
Lacking temporal  and spatial  resolution, this  procedure is the  least
desirable  of the  three.
                                         B-37

-------
     Regardless of which of the three basic methods is employed,  some
details of the vehicular operations (and their spatial and  temporal
variabilities) that lead to emissions are lost when preparing a gridded
regional inventory.  Fortunately, in some cases,  information  concerning
vehicle mix, temporal distributions, and so forth can be retrieved.  For
example, if a modal shift was anticipated because of a particular control
strategy, one could go back through the calculations of a manually
prepared inventory and apply different light versus heavy duty vehicle mix
ratios.  Less tedious, with an automated transportation forecasting  model,
one could change model split factors and rerun the computer code, generat-
ing a new set of traffic estimates, which could then be used  to revise the
mobile source emission inventory.
b.   Stationary Source Emission Inventories

     This component of the overall inventory consists of major  point
sources (refineries, smelters, power plants, and so on)  and  "other"
sources.  The first category generally does not represent a  major  problem
in constructing an inventory because the main sources are usually  easily
identifiable.  However, frequently the emission characterization of major
point sources is made on an annual or "nominal" basis and thus  may depart
substantially from actual day-to-day emission rates.

     Lumped into the "other" source categories are  facilities such as
cleaners, gas stations, residential chimneys, coating and manufacturing
industries, and so forth.  Aggregation of these sources  into a  regional
inventory is often considerable.   For example, rather than identifying the
location and size of each dry cleaner in an urban area,  because of time
and resource constraints, the inventory may be prepared  by (1)  determining
the total number of dry cleaners  in the area (perhaps from the  telephone
directory), (2) estimating an average perchloroethylene  rate for a typical
dry cleaning shop (see EPA, 1972), and (3)  apportioning  the  total  emis-
sions on a regional grid according to a demographic distribution.  While
this procedure may be satisfactory from the model verification  point of
view, it is not acceptable if one is interested in  examining the reduction
in basin-wide oxidant levels caused in part by controls  on evaporative
emission sources that include dry cleaners.

     Several conclusions can be drawn from  the foregoing discussions and
the review of previous model sensitivity studies:

     >   Although the emission inputs required to operate a complex
        model are relatively straightforward (i.e.,  gridded
        emission fluxes of each pollutant),  procedures for
        compiling these inputs exhibit wide variability,  ranging


                                       B-38

-------
       from sophisticated traffic forecasting models to simple
       estimates based on regional fuel  sales.

       In general, emission inventories destined for use in
       control strategy evaluation must exhibit a greater  degree
       of detail and disaggregation of the various source  types
       than an inventory used in model verification.

       Existing emission inventories do not permit (without
       additional modification) the evaluation of many possible
       emission control measures  and strategies; only rather
       general analyses  (such as overall emission reductions or
       modifications of  large, stationary sources) are readily
       facilitated with current inventories.

       In the modification of existing, or the preparation of
       new, emission inventories, consideration should be given
       to the range of emission control strategies that are most
       promising for the region of interest;  in so doing,  the
       particular  source types amenable to control can be
       inventoried separately, thereby establishing a basis for
       future control strategy evaluation.

       Owing to the wide range in methods used to estimate
       stationary  source emission rates and  to develop traffic
       volumes  (and hence mobile  source emission rates),  it is
       difficult to estimate the  costs  entailed  in  enhancing the
       level of detail in emission inventories.  Accurate
       estimates of the  costs required  to  improve an  inventory
       for  a given city  can be made only after  an examination of
       the  distribution  of  source types and  the  procedures  used
        in forecasting traffic volumes.
6.   CONCLUSIONS

     This appendix presents a broad overview of the range in data  input
requirements of present generation photochemical  grid  models.   The  SAI
Urban Airshed Model has been used as the prototype for this discussion.
Review of the monitoring and data acquisition activities  at various urban
areas in the United States reveals a rather broad range in the quality  and
quantity of the data collected.  Only a very cursory attempt has been made
to estimate costs of data acquisition because of (1) the  wide geographical
differences in the cost of such activities, and (2) the rapid rate  at
which inflation is presently increasing the cost of these activities.
                                       B-39

-------
     Several model  sensitivity studies  have  been  performed  in the last few
years.  In the main,  studies  involving  the Lawrence Livermore National
Laboratory's LIRAQ  model  and  the Airshed  Model  have been the only ones
reported in the open  literature.  Most, but  not all, studies have involved
uniform reduction or  increase in the magnitude  of a model input.  Only
recently have sensitivity studies been  performed  that address the impact
on model calculation  caused by the selection of alternative procedures for
preparing model inputs.
                                       B-40

-------
                                REFERENCES
ABAG (1977), "Candidate Control Measures," Air Quality Maintenance Plan,
     Technical Memo 5, Association of Bay Area Governments,  Berkeley,
     California.

Ames, J., et al. (1978), "The User's Manual for the SA1 Airshed Model,"
     EPA-68-02-2429, Systems Applications, Incorporated, San Rafael,
     California.

Anderson, G. E., et al. (1977), "Air Quality in the Denver Metropolitan
     Region:  1974-2000," EPA-908/1-77-002, Systems Applications, Incorpo-
     rated, San Rafael, California.

Attaway, L. D., et al.  (1976), "Maintenance Shutdown of Tail Gas Treating
     Unit:  An Assessment of Potential SO? Concentrations and Related
     Health and Welfare Effects," TR-11700, Greenfield, Attaway & Tyler,
     Incorporated, San Rafael, California, and Systems Applications,
     Incorporated, San Rafael, California.

DeMandel, R. E., et al. (1979), "LIRAQ Sensitivity REsults", Bay Area Air
     Quality Management District, San Francisco, 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.

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.

Falls, A. H.-, and J. H. 'Seinfeld  (1978),  "Continued Development  of a
     Kinetic Mechanism  for Photochemical  Smog:  Environ. Sci.  Technol..
     Vol. 12, No. 13.
                                      R-l

-------
 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. 8W p. 327.

 Mickey. H. R., W. D. Rowe, and F. Skinner (1971), "A Cost Model for Air
     Quality Monitoring Systems." J. Air Pollut. Control. Assoc.. Vol. 21,
     No. 11, pp. 689-693.

 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.

 Liu, M. K., et al. (1974), "Assessment of the Feasibility of Modeling Wind
     Fields Relevant to the Spread of Brush Fires," Systems Applications,
     Incorporated, R74-15, San Rafael, California.

 MacCracken, M. C., and 6. D. Sauter, eds. (1975), "Development of an Air
     Pollution Model for the San Francisco Bay Area," University of
     California, Livermore, California.

 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., et al. (1979), "Photochemical Modeling of Transportation
     Control Strategies," report to Federal Highway Administration, EF79-
     37, Systems Applications, Incorporated, San Rafael, California.

 Reynolds, S. D., et al. (1976), "Continued Development and Validation of a
     Second Generation Photochemical Air Quality Simulation Model:  Volume
     II--Refinements in the Treatment of Chemistry, Meteorology, and
     Numerical Integration Procedures," EF75-24R, EPA-600/4-26-016b,
     Systems Applications, Incorporated, San Rafael, California.

 Souten, D.  R., T. W.Tesche, and W. R. Oliver (1980), "Evaluation of the
     Air Quality Impacts of Alternative Air Pollution Control Policies
     Utilizing the Airshed Grid Modeling Approach for the South Coast Air
     Basin," 305-EF80-220, Systems Applications, Incorporated, San Rafael,
     California.

Whitten,  6.  Z., and H. Hogo (1977), "Mathematical Modeling of Simulated
     Photochemical Smog," EPA-600/3-77-011, Systems Applications, Incor-
     porated, San Rafael, California.
                                     R-2

-------
Yocke, M. A., and M. K.  Liu (1978),  "Modeling Wind Distributions  over
     Complex Terrain," EPA-68-03-2446,  SAI  No. EF78-78,  Systems Applica-
     tions, Incorporated, San Rafael,  California.
                                      R-3

-------
                                    rECHMCAL RtPORT DATA
                             (Picase r<••*•.' t'-istntcrin:i'! •: tue -c'tvnv i^j^re f.
1. REPORT NO.
   EPA-450/4-81-Q31C	
4. TITLE AND SUBTITLE
  The Sensitivity  of Complex Photochemical Model  Estimate
  to Detail in  Input Information  ~  Appendix B:   Specifi-
  cation and Assessment of Airshed Model Input Requiremen
                                                             6. PERFORMING ORGANIZATION CODE
                                                             s
7. AUTHOR(S)

  T.  W. Tesche
                                                             8. PERFORMING ORGANIZATION KEPORT NO.
                                                               SAI  No. 332 EF81-4
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.
                                                             5. rIEFORT DATE
                                                             10 PROGRAM ELEMENT NO
                                                             11. CONTRACT/GRAN f NO.
                                                                68-02-2870
                                                             13. TYPE OF REPORT AND PERIOD COVERED
                                                             14. SPONSORING AGENC/ CODE
15. SUPPLEMENTARY NOTES
16. ABSTRACT
  Report  identifies key inputs to a photochemical grid model  (Urban Airshed Model).
  A literature review summarizing sensitivity tests performed prior to  1981 is also
  presented.   Costs associated with obtaining various kinds  of input data  in past
  studies  are also estimated.
 7.
                                 KE Y WORDS »ND DOCUMENT ANALYSIS
                  DESCRIPTORS
  Photochemical  grid models
  Urban Airshed  Model
  Sensitivity tests
  Model inputs
  Ozone
                                               b. IDEN1 IFIEflS/OPEN ENDED TERMS
8. DISTRIBUTION STATEMENT

  Unclassified
                                                                           c.  COSATi ! ielu.'Group
                                               I 19. FCURITY CLASS / / his Report 1    ,"1. NO. OF PAGES
                                               !_    	]   48
EPA Form 2220-1 (Rev. 4-77)   PRt'.iOU;.  &ci TION is OBsoi.trt

-------