United States
Environmental Protection
Agency
Office Of Air Quality
Planning And Standards
Research Triangle Park, NC 27711
EPA-454/R-99-004
May 1999
Air
& EPA
DRAFT GUIDANCE
MODELS AND O
YSES
IN ATTAEMME
FOR T
TRATIONS
ZONE NAAQS
-------
EPA-454/R-99-004
DRAFT GUIDANCE ON THE U
MODELS AND OT
IN ATTAINMENT DE
FOR THE 8-
AAQS
J.S. ENVIRONMENTAL PROTECTION AGENCY
Office of Air and Radiation
Office of Air Quality Planning and Standards
Research Triangle Park, NC 27711
MAY 1999
U.S. Environmental Protection Agency
Region 5, Library (PL-12J)
77 West Jackson Bputevard, 12th Floor
Chicago, IL 60604-3590
-------
DISCLAIMER
This document includes draft recommendations for using models and other analyses to
help identify strategies which are effective for meeting an 8-hour national ambient air quality
standard promulgated in 1997. Pending outcome of litigation regarding these Nil
recommendations contained herein may not be enforced and are subjectJjte£han|pr Readers
should review the "Foreword" section to the document to understand U&Sfelfltt within which
this information is being presented. Mention of trade na
document is not intended to constitute endorsement or i
' ..: u
-------
ACKNOWLEDGMENTS
We would like to acknowledge contributions from members of an external review group,
the STAPPA/ALAPCO emissions/modeling committee and U.S. EPA Regional Office modeling
staffs in providing detailed comments and suggestions regarding the draft version of this
guidance. In particular, we would like to thank staff members of the Lake Michigan Air
Directors Consortium (LADCO), Microelectronics Center of North Carolina (MQp?), South
Coast Air Quality Management District (SCAQMD), California Air Re^ucesjfird (CARB),
Texas Natural Resources Conservation Commission (TNRC£) and DiffiBfapporation for
testing our ideas for a modeled attainment test and sharing
,-A.
IV
-------
Table of Contents
DISCLAIMER iii
ACKNOWLEDGMENTS iv
List Of Figures jfr ix
List Of Tables '
Foreword
Executive Summary
1.0 Introduction
1.1 What Is The Purpose Of This Document?
1.2 Does The Guidance In This Document Ap
1.3 How Does The Perceived Nature Of O:
1.4 What Topics Are Covered In Thi|
2.0 What
2.
2.
2.
Is A Modeled Attainment Dafiastra&af-Anvgf» ..................... 27
1 What Is The Recommen^^plodele^Rttainnie^Test?~An Overview .......... 27
2 What Does A Recomn^^la WeLpf Of Eviddfee Determination Consist Of?~An
Overview ..... JT .... . .f. ............................. 28
3 Why Should A:
: .Analyses Bel
UsedfflkHvve" Sense And Why May Corroboratory
A WerailSRCTidence Determination? 29
imenc
lied Attainment Test? 33
For The Modeled Attainment Test? 34
2 How Do
3 Wliat-Xfe
4 What Is A s
Rmended Modeled Attainment Test? ............... 44
cations Of The Recommended Modeled Attainment Test? ... 47
est, And Why Is It Needed? .......................... 49
4.0 If I Use A Weight OjtEvidence Determination, What Does This Entail? 51
4.1 What Analyses Should I Consider In A Weight Of Evidence Determination? 52
4.1.1^^ Quality Models 52
; £v 44'S^.nalysis Of Air Quality And Emissions Trends 57
'' JiftS^f ^fffiO Use Of Observational Models 58
* 42 What If I Want To Consider Additional Corroborative Analyses? 60
5.0 How Can I Improve Modeling And Other Analyses In Weight Of Evidence Determinations?
63
-------
5.1 What Data Gathering Or Other Efforts Might Be Helpful To Support Current
Analyses Or Subsequent Reviews? 63
5.2 Why Is It Desirable To Plan For A Subsequent Review? 66
5.2.1 Integration With Attainment Strategies For 1-Hr Ozone NAAQS 66
5.2.2 Anticipated Modeling Principles For PM25 And Visibility, And Integrating
Ozone Strategies With Goals For PM2 5 And Visibility 66
6.0 What Documentation Do I Need To Support My Attainment Demor
7.0 References Cited In Part I And In Section 1.0 ...
8.0 How Do I Apply Air Quality Models?— An Overvii
69
10.0 What Does A Modeling/Analysis Protocq,
10.1 What Is The Protocol's
10.2 What Subjects Should Be
11.0 What Should I Consider In
11.1 What Prerequisites
» Attainment
1 L2'9S&at Factors
9.0 How Do I Get Started?
9.1 What Is A "Conceptual Description"?
9.2 What Sorts Of Analyses Might Be Useful Fi
Conceptual Description?
lodel? ...................... 95
Meet To Qualify For Use In An
.......................... 95
For A Specific Application? ...... 97
12.0 How Do 1^
12.1 W
12.2 Wat
Jes Of AlFQuality Models Which May Be Considered? . . 99
ical Episodes To Model? .................... 101
t Criteria For Choosing Episodes? ............. 101
econdary Criteria May Be Useful For Selecting Episodes? . . 104
13.0€What Should I ConaflejiWhen Selecting A Modeling Domain And Its Horizontal/Vertical
:;/ Resolution? .. *... T 107
; 13.1 How Do Choose Between An Urban Scale Or Regional Domain? 107
r 13.2 What Hjnzontal Grid Cell Size Is Necessary? 108
:\ 13.3 Hoj^tny Vertical Layers Should I Consider? 110
ilse Should I Consider In Choosing Finely And Coarsely Resolved Portions
Z" >; 7 Of Nested Regional Models? Ill
14.0 How Do I Produce Meteorological and Air Quality Inputs Needed By An Air Quality
Model? 113
VI
-------
14.1 What Approaches Are Available For Generating Meteorological Data? 113
14.2 How Do I Deal With Data Management And Computer-related Constraints When
Applying Dynamic Meteorological Models? 115
14.3 How Do I Quality Assure Results Generated By A Meteorological Model? .... 116
14.4 What Are Some Past Applications Of Dynamic Meteorological Models? 11.8
14.5 How Do I Address An Air Quality Model's Need For Air Quality Inputs? 118
15.0 How Do I Produce Emission Inputs Needed For An Air Quality'.
15.1 What Countywide Emission Estimates Are Needfed To Si
15.2 Can I Use the National Emissions Trends In
15.3 How Do I Convert Countywide Inventory
Models?
15.4 What Should I Do To Quality Assure Emisl
15.5 How Do I Estimate Emissions For Future
121
Quality Models?
122
16.0 How Do I Assess Model Performance And Make
16.1 How Can I Evaluate Performance
16.1.1 How Can I Use Graphi
Performance? .
16.1.2 How Can Ozone
16.1.3 How Can I Use Available Fjjscursor
Performance?^
16.1.4 How Can I
Help Edpte Air
16.1.5 What ItatlBases R
Analyses? 129
129
;sessment Of Model
130
'odel Performance? ... 131
ions To Evaluate Model
133
s With Observational Models To
Performance? 133
ges In Emissions Are Available To
erformance? 134
.6 Ho>£DcS&E&e Ratios offndicator Species To Evaluate Model
v.5S5\4KfS8r r
135
malyses Useful For Evaluating Model Performance?
16.1
lhese Performance Tests Have Shortcomings, So What Do I Do?
' °
136
- .
136
16.2 How Can I Mil|slGood Use of Diagnostic Tests? ......................... 141
16.2.1 UsI'Of Sensitivity Tests ..................................... 141
16.2.2 Use of Process Analysis ..................................... 142
H3Q:ReferencesJGitea In Part H 147
155
Vll
-------
List Of Figures
Figure 3.1. Relationship Between Grid Resolution And Grid Cells Considered
To Be In The Vicinity Of A Monitor-
Figure 3.2. Choosing Predictions To Estimate RRF's—
38
40
Figure 3.3. Mean Relative Reduction As A Function Of Mean Predict
Current 8-Hour Daily Maxima
Figure 4.1. Examples Showing Use Of Trend Analysis
vin
-------
List Of Tables
Table ES. 1. Recommended Core Analyses For A Weight Of Evidence Determination 5
Table ES.2. Recommended Documentation For Demonstrating Attainment Of
The 8-Hour Ozone NAAQS
-8
Table 2.1. Guidelines For Using A Weight Of Evidence Detenrn'nation J^-r-j^- 28
Table 3.1. Example Monitoring Data For Nonattainment ACTEA"---^IJB^BB&g 34
Table 3.2. Example Illustrating Selection Of Current
Table 3.3. Default Recommendations For Nearby GridlBU^d To Calculate R^SP-—37
Table 3.4. Example Calculation Of A Site-Specific Futu^S^^^toe (DVF), 46
Table 4.1. Recommended Core Analyses For A Wj|y
Affecting Their Credibility Anc
The NAAQS
Table 6.1. Recommended Documentation For
8-hour NAAQS For fejfie--
ination, Factors
Meeting
53
nmentOfThe
70
Table 11.1. Some Air Quality Mfwels UsedJli ModeJjSzone 100
:' ^
Table H.fcjSome Past Of D^^W^eteorological Models 119
Table 15.1>tSoi^Bmission^^^^^nd Example Applications 125
*\'»iCV *
Table 16.1. Simimaifl^Memodslfipvaluate Performance Of Air Quality Models 137
Table 16.2. PotentiallyW^^aiagnostic Tests At Various Stages Of Modeling 144
IX
-------
Foreword
en the
ce and revise
an extensive
numerous
t of the
Readers should note that the U.S. Court of Appeals for the District of Columbia Circuit
issued a decision on May 14, 1999 which prohibited enforcement of the 8-hr ozone national
ambient air quality standard (NAAQS). Therefore, this draft guidance is being distributed at this
time to document the position of the Agency when the Court Decision was issued. Moreover, the
concepts developed in the guidance are applicable to other multi-hour standards.
litigation issues have been addressed and resolved, we will review this
it as appropriate. Distribution is being made since this
dialogue with the stakeholders from the modeling commuc
requests for release of the current version. Stakeholders j
include representatives from State and local governmenlj
industry, and environmental organizations in addition t<
Distribution of this document should not be const
standard; it is merely provided for information to intere
the guidance at the time of the court decision. Thus,
draft guidance document to State and local agenci
actions should be taken to implement this draf^
resolved and final guidance has been issued.
Jementation of the 8-hr
to document the status of
to provide this
•rs. However, no
issues have been
\*
-------
Executive Summary
Readers should note that the U.S. Court of Appeals for the District of Columbia Circuit
issued a decision on May 14,1999 which prohibited enforcement of the 8-hr ozone national
ambient air quality standard (NAAQS). Therefore, this draft guidance is being distributed at this
time to document the position of the Agency when the Court Decision was issued. Moreover, the
concepts developed in the guidance are applicable to other multi-hour standardsJj&hen the
litigation issues have been addressed and resolved, we will review this dmfi gummce and revise
it as appropriate. Distribution is being made since this.document is ti
dialogue with the stakeholders from the modeling communmmad sin
requests for release of the current version. Stakeholders^
include representatives from State and local govern.
industry, and environmental organizations in addition
Distribution of this document should not be cons
standard; it is merely provided for information to intere
of the guidance at the time of the court decision. Th
draft guidance document to State and local agenci
actions should be taken to implement
resolved and final guidance has been issue
of an extensive
numerous
fomentation of the 8-hr
to document the status
>priate to provide this
-------
2. Develop a modeling/analysis protocol.
3. Select an appropriate model to support the demonstration.
4. Select appropriate meteorological episodes to model.
5. Choose an appropriate area to model with appropriate horizontal/vertical resolution.
6. Generate meteorological and air quality inputs to the air quality model.
7. Generate emissions inputs to the air quality model.
8. Evaluate performance of the air quality model and perform diagnostic
After these steps are completed, the model is used to simulatfeeffects
strategies. Model applications require a substantial effort, jffips sho
appropriate U.S. EPA Regional Office(s) in executing eajpiniep. This
likelihood of approval of the demonstration at the end ojjne proces
summarize Part n of the guidance.
te control
psely With the
te whether selected
.QS, and (b) an
tions. This guidance
ES 1.0. What Is An Attainment Demonstratioi
An attainment demonstration consists of (a)
emissions reductions will result in ambient concengjajKHis that me
identified set of measures which will result in U^^^^^^onissionS
describes how to use air quality models and^petM^^^^^termineif results of a simulated
control strategy indicate attainment. Detegilining reductions may be done by
relying exclusively on results obtained jph air qujflty include the outcomes of a
modeled attainment test plus a screej|iip test tojjpfimate winner a proposed emission reduction
suffices to meet the NAAQS. Oth^^^lyses^pcluding^Pids analyses, observational models,
etc. may be used to supplement j^^nodelejgtainmej^d screening tests.
A modeled attain
t compc
«it.. *•
boncentration predictions with the ozone
NAAQS.JThfNAAQS is
averaged ov&^sjjjj^ut
maximum concenfitenas called
passed if predicted^&^teesign v,
applied at other iocaf^fi^^feconsistently high model predictions, may also be needed.
highest 8-hour daily maximum ozone concentration,
fcO.OS ppm. The average 4th highest 8-hour daily
value" for ozone. The modeled attainment test is
icar all monitoring sites are < 84 ppb. A screening test,
Provided the mo^il^pttainment test and supplementary screening test are passed or
close to being passed, States may use a broader set of analyses to estimate if attainment is likely.
This is called a "weighitrf evidence determination". A weight of evidence determination
combines results oftne modeled attainment and screening tests with other results obtained with
air quality mode.Ls»!as well as conclusions drawn from analyzing monitored air quality data,
emissions estimates and meteorological data. Results of each analysis are considered in concert
to detemine'whether or not attainment is likely.
-------
ES 2.0. What Is The Recommended Modeled Attainment Test?
The recommended modeled attainment test uses monitored design values in concert with
model-generated data. The test uses model results in a "relative" rather than "absolute" sense,
and is applied near ozone monitoring sites. Eight (8)-hour daily maximum ozone is predicted
near a monitor for each day in the test. Each day's prediction is then summed and averaged, first
with current emissions and then with future emissions. The ratio of future to curreppredicted
i J$ijiff'
mean 8-hour daily maxima is then computed. This ratio is called the "rgHtive Auction factor",
or RRF. The test consists of 4 steps.
1. Compute a current site-specific design value fro;
2. Use air quality modeling results to estimate a si
3. Multiply the relative reduction factor obtained in
value in step 1. The result is a predicted site-i
84 ppb, the test is passed at the monitor site bein
the site-specific design
value. If this value is <
4. Repeat steps 1 through 3 for each moni
value approaches or exceeds 84 ppb.
ppb at each site, the test is passed.
The modeled attainment test is de
ES3.0 What Is The Sc
monitored design
design values are ^ 84
fr-
Needed?
Th& modeled attaaimeairtest does ripfflffiSs future air quality at locations where there is
' "~- > -~* v*t *^ee T- ^Iglii^V "*"
no nearby ozone%monitor.*5Khlsaffi8quality model consistently predicts 8-hour daily maximum
•* * . ,**;«, ^$%ati$m^&3£&im$f •* ^ A •*
lonitored location which are substantially higher than
lie should perform an additional screening test. The
ozone concentrations at a p
any predicted neagKmojdtoring
screening testjs|p^ffliii¥4he
specific mpnitore
receptor;location(s). If
3e monitored design value (i.e., the highest of the site-
s) times the relative reduction factor(s) predicted at the suspect
_:s,-^-,__-ig estimated future design Value(s) is <, 84 ppb, the outcome is
not inconsistent with attainment of the NAAQS. The screening test is discussed in Section 3.4.
ES4.0. What Is A Weight Of Evidence Determination?
. A State should always utilize available air quality, meteorological and emissions data to
complement amodeling analysis. These data are used to develop a conceptual description of an
area's npnattainment problem. This description is useful for guiding a modeling analysis. If the
modeled attainment and screening tests are passed or nearly passed a State may choose to use a
weight of evidence (WOE) determination to estimate if attainment is likely.
-------
A weight of evidence determination is a diverse set of technical analyses performed to
corroborate findings of the modeled attainment and screening tests. If a weight of evidence
determination is used, a State should consider a recommended core set of analyses consisting of
(1) a set of air quality model results which includes the previously described tests plus additional
analyses of estimated concentrations, (2) an analysis of observed air quality and estimated
emissions trends, and (3) an analysis of outcomes produced by observational models.
We identify factors which enhance credibility of evidence produj
analyses, as well as outcomes which would be consistent witkthe like!
demonstrates attainment. This is illustrated in Table ES.l Jsbe 3
A State may include other types of analyses, in
of evidence determination. For another analysis to be
satisfied:
of the core
a strategy
core analyses.
(1) a State should discuss why the proposed analysis is
(2) a State should identify the procedure to be usedjmtfe&e data bl
ssing attainment,
ie to support it, and
TJ.
concerned
years. It is di
inconsistent w,
whether
different years.
(3) a State should identify (in advance) outcjiiKS w
that a proposed strategy will lead to attainjpent.
Weight of evidence and its use is di
ES 5.0. Why Do We R<
Option To |erform A Weight
be consistent with a hypothesis
Attainment Test And Offer An
^related tolhe form of the NAAQS. The NAAQS is
maximum concentration averaged over 3 consecutive
led exceedance in a particular episode is or is not
, using a model by itself to rigorously assess
Id require modeling a substantial number of days in three
Furthe^^^Mfeve relatively resource-intensive models are needed to simulate
effects of reducing precUjES«3^inissions on ozone. Thus, the test uses observed design values to
"anchor" model predicti|ns to the form of the NAAQS. Design values are, by definition,
calculated consistentlyjpith the form of the NAAQS, and their use allows a State to apply a
resource-intensive molel to see how they might be changed by a control strategy.
recognizes uncertainty in model predictions. Problems in interpreting
model resiilts posed by uncertainty in the predictions may be greater for addressing the 8-hour
NAAQS than was true for the 1-hour standard. The 8-hour NAAQS is closer to continental
background values. Further, design values tend to be closer to the specified level of the NAAQS
than is true for the 1-hour standard. As a result, the signal (i.e., the change we wish to effect in
-------
Table ES.l. Recommended Core Analyses For A Weight Of Evidence Determination
(1)
Type Of Analysis
(2)
Factors Increasing Credibility Of
The Analysis
(3)
Outcomes Consistent With
Hypothesis That A Candidate .
Strategy Will Lead To Attainment
Air Quality Models
-good model performance
-extensive observational data ba
available
-short projection periods
-carefully quality assured j
-confidence in meteorologil
inputs
-the modeled attaii®ent test is
3S6- w'&S
tes for future
s w/o monito
nearly passed, the
uires additional reductions and
are underway to
ntly review/refine the
-good ability to pose and address
questions about a s
adequacy
-commitment is made to deploy
monuors at locations not passing
ning test
-other an
conclusi
-substantial modeled improvement
in air quality is predicted using
several measures described in
Section 4.1.1.
-similar conclusions are reached
with other peer reviewed models
-------
Table ES.I. Recommended Core Analyses For A Weight Of Evidence Determination
(concluded)
(l)
Type Of Analysis
(2)
Factors Increasing Credibility Of
The Analysis
(3)
Outcomes Consistent With
Hypothesis That A Candidate
Strategy Will LeadJb Attainment
Analysis of Air Quality and
Emissions Trends
-current or future (air'quality
model) predicted design value i§
within a few ppb above 84 pp£
-extensive monitoring nei
exists
-both ozone and precurso
are available
-statistical model used to
trend for meteorological
explains much vari
iward
I exists in the site-
value at all sites
ater than 84,
•end line
:p the required attainment date
jcates an 8-hour daily maximum
ition <. 84 ppb.
ved air quality trend
also show a substantial
ment.
ntinued, comparable relative
ans in emissions are
for
i$^;
- ",--:>&f
Use of Observational M
-an extensive monitoring network
exists
-precursor and indicator species are
measured using instruments with
appropriate sensitivity
-monitoring sites appear spatially
representative
-data have been quality assured,
and results are self-consistent
-plausible physical explanations
exist for findings
-Findings indicate sources
controlled in the candidate strategy
are important causes of observed
high ozone
-Analysis of indicator species
suggests the direction of the
strategy (e.g., emphasis on VOC or
NOx) is appropriate.
-------
the design value) to noise (i.e., uncertainty in predictions) ratio may be smaller than heretofore.
The recommended modeled attainment test reduces uncertainty (i.e., "noise") in three
important ways. First, monitored data (i.e., current design values) are incorporated directly into
the test. These data are likely measured with greater accuracy than an absolute model prediction,
and precision of the measurements is known. Second, the outcome of the test is based on a
composite set of calculations from several modeled days rather than a single day. Jphs reduces
the risk of choosing an inappropriate strategy .on the basis of a single outcome jprch is subject to
uncertainty. Third, if the outcome of the test is close to pass/fail, a
determination may be used to see whether other model outegpand omf analyses
provide corroborative evidence for conclusions drawn frgiPine test.
ES 6.0. What Documentation Is Needed To
Demonstration?
are enumerated in
ry which addresses
A conceptual description is
problem. For example, is the pro
also important? Do sites violati
way?, etc. A.conceptual
meteor°ip4ifnd
perfonnealpptefcshoul
r /stripes**
are seeking^
later decisions^
conceptual descrijpfiOTlaii?found i
A State should address 9 subject areas in its d
Table ES.2. Documentation should be accompan
each of the 9 areas shown in the table. Docum
dressed in Section 6.0.
ES7.0. What Is A "Conceptual Dsc rip t
mf™~~
terizing an area's nonattainment
Inated byjjibal emissions or are regional factors
reflecypfpatial or temporal pattern in some
;adily available air quality,
_ refined later as additional analyses are
uconceptulif description of each nonattainment problem they
developing a solution. It serves as a means for guiding
sling analysis. Suggestions for developing a
ion 9.0.
ES8.0. What Does A Modeling/Analysis Protocol Do And What Should It
„ , „;• _ * -v~
Contain? *
* A modeling/analysis protocol is a document which identifies methods and procedures to
be used in the analyse!? The protocol also identifies ground rules to be followed in undertaking
estirna^mission reductions needed to meet the NAAQS. Ground rules include a
^i;-.o||S&w affected stakeholders in the modeling/analysis process will be encouraged to
participate, the process by which decisions will be made, means used for communicating issues
and decisions, and the methods, data bases and procedures to be used to obtain results. As the
name implies, the protocol should address use of other analyses as well as air quality modeling.
The document is usually prepared by the State/local agency(ies) having lead responsibility for the
-------
Table ES.2. Recommended Documentation For Demonstrating Attainment Of The
8-hour Ozone NAAQS
Subject Area
Purpose of Documentation
Issues Included
Modeling/Analysis Protocol
Communicate scope of the analysis
and document stakeholder
involvement
Names of stakeho
in prepJigng ani
prot
participating
lementing the
Emissions Preparations and
Results
Assurance of valid, consis
emissions data base. Ap
procedures are used to deriv
emission estimates neede^plbr air'
quality modeling.
Data base used and
surance methods applied;
sing used to convert
model-compatible
Deviations from existing guidance
underlying rationale;
VOC, NOx, CO emissions by
State/county for major source
categories.
Air Quality/Meteorology
Preparations and Results
Extent of data base and procedures
used to derive & quality assure
inputs for analyses used in the
weight of evidence determination;
Departures from guidance and their
underlying rationale;
Performance of the meteorological
model, if used to generate
meteorological inputs to the air
quality model.
-------
Table ES.2. Recommended Documentation For Demonstrating Attainment Of The
8-hour Ozone NAAQS (continued)
Subject Area
Purpose of Documentation
Issues Included
Performance Evaluation for Air
Quality Model (and Other
Analyses)
Show decision makers and the
public how well the model (or other
analyses) reproduced observations
or otherwise performed on the days
selected for analysis
Summary of observational data
base available for c.ajnparison;
n (performance tests
suits;
iplies.
Diagnostic Tests
Ensure rationale used to adj
model inputs or to disco
results is physically justified and
the remaining resultsjaalfieasense
Its from application prior to
:nts;
'with scientific
ing and expectations;
performed, changes made and
ompanying justification;
Short summary of final predictions.
&
-------
Table ES.2. Recommended Documentation For Demonstrating Attainment Of The
8-hour Ozone NAAQS (continued)
Subject Area
Purpose of Documentation
Issues Included
Description of the Strategy
Demonstrating Attainment
Provide the EPA and the public an
overview of the plan selected in the
attainment demonstration.
Qualitative description of the
attainment strategyu
, NOx, and/or
each major'
or each
irrent (identi
ctions
predicted 8-hr site-specific
jesign values for the selected
riario and identify any
h fails the screening
ibed in Section 3.4;
itification of authority for
iplementing emission reductions
in the attainment strategy.
Evidence that emissions remain at
or below projected levels
throughout the 3-year period used
to determine future attainment.
Data Access' -• v'e -Sfci
'-
EPA or other interested
replicate model
" 'V.&f'^ff
performance and attainment
simulation results, as well as results
obtained with other analyses.
Assurance that data files are
archived and that provision has
been made to maintain them;
Technical procedures for accessing
input and output files;
Identify computer on which files
were generated and can be read, as
well as software necessary to
process model outputs;
Identification of contact person,
means for downloading files and
administrative procedures which
need to be satisfied to access the
files.
10
-------
Table ES.2. Recommended Documentation For Demonstrating Attainment Of The
8-hour Ozone NAAQS (concluded)
Subject Area
Purpose of Documentation
Issues Included
Weight of Evidence
Determination
Assure the EPA and the public that
the strategy meets applicable
attainment tests and is likely to
produce attainment of the NAAQS
within the required time.
Description of the modeled
attainment test and pbservational
data base used;
Outcome of each;
including the modeled attainment
it of the credibility
ith each type of
this application;
•ative describing process used
conclude the overall weight of
available evidence supports a
hypothesis that the selected strategy
is adequate to attain the NAAQS.
Review Procedures Used
•ovide as;
ie public
an the attai
Jieflect sound practice
A and
performed
iemonstration
Scope of technical review
performed by those implementing
the protocol;
Assurance that methods used for
analysis were peer reviewed by
outside experts;
Conclusions reached in the reviews
and the response thereto.
modeling/analysis, ^consultation with stakeholders. The protocol should be kept up to date to
reflect;majpr subsequent decisions made after the initial version is completed. Specific topics
wMcBtlhbiild^^included in the protocol, are identified in Section 10.0.
ES9.0. What Should I Consider In Choosing An Air Quality Model?
Several prerequisites need to be met for a model to qualify for use in supporting an
11
-------
attainment demonstration.
1. The model has received a scientific peer review.
2. The model can be demonstrated to be applicable to the problem on a theoretical basis.
3. Data bases needed to perform the analysis are available and adequate.
4. Available past appropriate performance evaluations ha#e showgfflBMiffel is not biased
toward underestimates.
5. A protocol on methods and procedures to be foil
6. The model is available to users for free or at a re!
1. Nature of the air quality proble
first be assessed, and the
with the perceived nature of
2. Availability, docu
To select a qualifying model for a particular
what attributes are needed for a qualifying model to
problem, and then choose among models possessi
considered in selecting an air quality model fo:
approximate order of importance.
should first determine
area's ozone
tors should be
are listed in
ling mionatta^gprof the ozone NAAQS should
shsafid have afmbutes and capabilities consistent
glgpsr
sice should be satisfactory.
3. Relevaht~experiencejOi^^|able staffand contractors should be consistent with choice of
a
4. Time and:
nstrainT^3iiIy be considered.
5. Consistency of tfiraDio3p|*with what is used in adjacent regional applications should be
considered.
Choice of a model should be concurred with by the appropriate U.S. EPA Regional Office and
U.S.3BPA Model Clearinghouse.
;<, ir^rior torusing model results in a specific attainment demonstration, a State should show
that me model performs adequately in replicating base case observations available for that
demonstration. Further discussion of model selection occurs in Section 11.0.
12
-------
ES10.0. How Do I Choose Meteorological Episodes?
States should consider four primary criteria when choosing meteorological episodes for
modeling. Tradeoffs among these may often be necessary. Such tradeoffs need to be resolved
on a case by case basis.
1. Choose frequently occurring episodes containing days reflecting a variety QpHnd
orientations observed to occur when 8-hour daily maxima exceed 84n3Db atfone or more
monitors.
2. Choose episodes containing days with observed 8.
concentrations close to (e.g., ± 10 ppb) the average
monitoring sites during a 3-year period straddling
drawn (i.e., days approximately as severe as impli
3. Choose episodes containing days for which mi
indicator species and/or precursor measurements
4. Choose a sufficient number of days so
modeled attainment test for each monitc
1. Givejpreference to
- " " 4
2. Give preference to
1e for use in the
is violated.
States may be able to resolve coJHicts amgpg the p^^^p:riteria for selecting episodes
by considering one or more secondaj|Jpteria. Jtie followjpgare identified as secondary criteria.
States may identify, document and&jraent the$ationale ^criteria in addition to these.
sly modd
\
jxcurringfcfuring the period corresponding to the current
3. Give preferc^f»W%|jisodes^na3nrnizing the number of days and sites observing 8-hour
daily maxima^o»1»Jle level of severity specified in the NAAQS.
A-'.' V^Ti^i.*-
4. Include weekend&an^ong the selected days, especially if daily maxima exceeding 84 ppb
are observed on sucfiMays.
. i /
-;u5. If applying ajegional model, choose episodes meeting the other primary and secondary
criteria in as^riiahy nonattainment areas as possible.
•-•' v^^tf.^^^^--«i-*w**^*j'"-r5s*1'*'
Episode selection is discussed in Section 12.0.
13
-------
ES11.0. How Do I Select A Modeling Domain And Its Horizontal/Vertical
Resolution?
States should review available air quality, meteorological and emissions data to help
select a domain size which is consistent with a nonattainment area's problem but which is not
unnecessarily resource intensive. We suggest a procedure for comparing regional (upwind)
observations with local design values which may be useful in choosing between regional and
urban scale domains. A typical urban domain may be about 300 km onjlgde.^pfypical regional
domain exceeds 1000 km on a side.
Choice of horizontal/vertical resolution presents
base management vs. scientific rigor. Sensitivity of resii
by case basis. For urban scale analyses and for the fine!
grid cells as small as 4 km may be preferable. Howevel
flexibility in choice of grid cell size, so long as the cells;
of regional applications may use 36-km grid cells or sma
for an urban analysis or in a "fine" portion of a nested^
sensitivity of conclusions drawn simulating a i
cells, if feasible.
Section
Air Quality MI
east
Carejpould
be ernated as
iested
ical reasons, we*
or smaller. Coarse portions
large as 12 km are used
irform a test to assess
of such large
Models should ordinarily include
layer (PEL), and 1-2 layers above the P.
the maximum afternoon mixing hei
sensitivity of conclusions drawn
should be performed, if feasible
More detailed suggest}
ithin the planetary boundary
:b place vertical layers so that
isely as feasible. Tests assessing
trategy to use of more vertical layers
fain size and grid resolution are contained in
rological And Air Quality Inputs Required By An
We recommend'ttafptites use a dynamic meteorological model with four dimensional
data assimilation (FDD^ta^Qfe principal means for generating meteorological inputs required
by air quality models used in ozone attainment demonstrations. Any such meteorological model
which has received a scientific peer review may be used. As with the output from emissions
models, it is critical.that results of meteorological models be quality assured. We identify several
potentially usefuliineans for doing so:
~1. comparison with upper air measurements "held back" from use in FDDA;
2. comparison of calculated trajectories with observed air quality patterns;
14
-------
3. use of computer graphics to discern spatial discrepancies;
4. simulation of inert tracers to identify discontinuities or mass balance problems;
5. comparing results obtained with different meteorological models;
6. calculating and comparing divergence and/or dimensionless parameters amttebrnparing
these with expected ranges;
7. comparing spatial ozone patterns obtained with a grj
M
mf *r~ -wp
8. using process analysis to flag contributions mad^ftf unexpecjeo ozone
attributable to meteorological factors.
Applying meteorological models over extensive
can be very resource intensive and present data base m
for reducing such problems.
|th fine scales (i.e., 4-12 km)
•jJems. We suggest means
Air quality inputs are needed for initial
a modeling domain. There is no satisfacto:
specify initial conditions. Thus, States s
simulation one or more days prior to th;
days earlier for regional application
generating boundary conditions
attainment demonstration. If anman seal
that emissions occurring injhejJjpter of
until the endaofcthe same^^^Har-day. TH
v,- - •- ,",/,-.;v j.Sa™^*1'-''^!!**^-.. *
specified for suchapplicaifi|
values at the edges of
ity observations to
ice by beginning a
applications and two or more
Is are the usual preferred means for
ional domain which is the focus of an
1, the domain should be large enough so
pit before sunrise remain within the domain
reduce importance of boundary conditions
Issues rela^cll^ineteoroldp^il^odels and procedures for accounting for initial and
boundary conditioaSs^iscussed nraection 14.0.
ES13.0. How Do I Produce Needed Emissions Inputs?
Producing needea.erriissions inputs requires several steps. First, compile Statewide and
then countywide estimates for VOC, NOx and CO emissions for point, area, mobile and biogenic
emissions. Second, ^quality assure the outputs. Third, convert the resulting estimates into
speciated, griddediiourly emissions using emissions models. Fourth, once again, quality assure
the results. Finially, project gridded, speciated hourly emission estimates to a future year which
corresponds to two years prior to the deadline for meeting the NAAQS.
The U.S. EPA has prepared a series of guidelines relating to these steps as a part of the
Emission Inventory Improvement Program (EIIP), as well as a guideline for developing
15
-------
emissions inventories. States should be familiar with these guidelines. States should use the
most recent emission estimates commonly available when applying the modeled attainment test
used to support an attainment demonstration. If available in a time frame compatible with
completing modeling in a timely manner, the National Emissions Trends Inventory (NET)
compiled for 1999 is the preferred source of information for State and countywide estimates in
portions of the modeling domain for which States who are stakeholders have no better
information. Otherwise the NET compiled for 1996, or derived for 1997 or 1998.jj|tay be used.
Different means are used to obtain emissions info]
quality models. Ideally, location and daily/weekly emissi
for point sources. Spatial distribution of surrogates for
major area source categories and for mobile sources in
Diurnal and weekly activity patterns are also useful. O
profiles are desirable for point sources and major area
then used to characterize emissions from other point sou
biogenic sources. We identify several commonly used
nt demonstration to
tion
eded by air
irectly available
stimated f
ty factors
to esti
t, locjBy applicabl
gories. Emissiowinoaeis are
ary area, mobile and
Is.
Quality assurance of emissions estimates is,
be credible. We recommend that it be perfo:
derive required emission inputs to an air qu
different studies, computer graphics and cjfrnpanso]
are useful means for quality assuring erjjpision estates. T$^ffi|itmg emission inputs is
addressed more fully in Section 15.
eral stage||ppthe process needed to
aring inventory estimates made for
f speciated air quality data
ES14.0. HowDoIEvs
Analyses?
•?\'.-
In Section 16.0,
1. Perform^
Modelllerformance And Make Use Of Diagnostic
5J meansTor evaluating model performance.
ising graphics.
.,'izis
2. Use "ozone metn^^m?statistical comparisons.
3. Compare predicteffiaijd-observed precursor or species concentrations.
jj
A. Compare predicted source attribution factors with estimates obtained using observational
. . ,<:>'
models.
5;:tGompatB?observations and predictions on weekends vs. week days.
6. Compare observed and predicted ratios of indicator species.
7. Use retrospective analyses in which air quality differences predicted with models are
16
-------
compared with observed trends.
Most of these approaches are only able to address how well a model replicates a past set
of observations. While this is useful, the key question is, "how well does a model forecast
changes in ozone accompanying changes in precursor emissions?". The 5th (weekend/week day
comparisons), 6th (use of ratios of indicator species) and 7th (use of retrospective analyses)
approaches have the potential to address this key question. However, each requir^Mditional
efforts to make certain measurements or to perform additional analyses .^JS^e dissatss these
additional efforts in Sections 5.0 and 16.0.
All of the identified means for evaluating model j^Bormance
weaknesses. Thus, we recommend that as many of rnesjlpproache
evaluate model performance. Assessment of whether
properly done by considering evidence produced by all Hmfelsvaluation technic
same way as a weight of evidence approach is used in an
identi: sens
ostcsts, State
onstra. Tha
Diagnostic tests should be applied throughout
key stages for use of these tests: (1) during model
evaluation, (3) during the process of choosing/
help estimate uncertainty in the resulting airJpDI
to study predictions at specific times and
Two types of diagnostic tesj
designing and evaluating results
be used to support an attainment.
sense. Thus, diagnostic tesj^hfild <
predicted changes in ozon^awe^fifected
We identify several
ormance
ol strategies, (4) to
ing a strategy, and (5)
tests and process analysis. In
ould be aware of how models are to
odels should be used in a relative
reduction factors (RRF) or other
17
-------
-------
1.0 Introduction
Readers should note that the U.S. Court of Appeals for the District of Columbia Circuit
issued a decision on May 14, 1999 which prohibited enforcement of the 8-hr ozone national
ambient air quality standard (NAAQS). Therefore, this draft guidance is being distributed at this
time to document the position of the Agency when the Court Decision was issued. Moreover, the
concepts developed in the guidance are applicable to other multi-hour standardsjjjjffhen the
litigation issues have been addressed and resolved, we will review this gujjiance and revise
it as appropriate. Distribution is being made since this document is of an extensive
dialogue with the stakeholders from the modeling communjjjijjed sin^^^^^^had numerous
requests for release of the current version. Stakeholdersilmlved in of the drjji
include representatives from State and local govem
industry, and environmental organizations in addition EPAegional Of
Distribution of this document should not be consi
standard; it is merely provided for information to intereA
of the guidance at the time of the court decision.
draft guidance document to State and local agencL
actions should be taken to implement this drafy
resolved and final guidance has been issue*
A
1.1 What Is The Purpose Of This Document?
veil as
til the
Cementation of the 8-hr
nd to document the status
te to provide this
However, no
issues have been
This document has two
modeling and other analyses suj
quality standard (NAAQS)^
second purjjpseas to des
support an attainment dei
to help demonsteatoijattainmeii
these results. .
Th
a cone,
lour dail
to appl
'* Part I a
st is to efplain how to interpret if results of
jn thatJBainment of the national ambient air
~w
izone concentrations will occur. The
piality model to produce results needed to
is document provides guidance for using results
[^provides guidance on how to apply models to produce
•»*.
With few exT"
iSMuidance herein should be viewed as recommendations rather than
-*.:>
requirements. States m^^^^pternative procedures if these are justified to the satisfaction of the
appropriate U.S. EPA Re^ontfOffice. Generally, an attainment assessment which leads to
greater protection of thejenvironment than that recommended in Part I of this guidance may be
used if a State choosesjto do so. Although this guidance attempts to address issues that may arise
in attainment demonstrations, situations which we have failed to anticipate may occur. These
shouldbe resolv^dioh a case by case basis in concert with the appropriate U.S. EPA Regional
"3*. "'jx,.,»»v^*Ar-^ ', ,.-'- "* 1. i IT C>
1.2 Does The Guidance In This Document Apply To Me?
This guidance applies to all locations which are not attaining the 8-hour NAAQS for
19
-------
ozone according to data reported to the US EPA's AIRS data base. This includes both
"traditional" and "transitional" nonattainment areas. Qualifications for a "transitional"
nonattainment area are defined in a 1997 Presidential Directive (Clinton, 1997). Under this
directive, States receiving this classification may not have to perform additional modeling under
some circumstances. States which have one or more "transitional" nonattainment areas should
consult U.S. EPA (1999a). U.S. EPA (1999a) identifies prerequisites to qualify for no additional
modeling. That reference also describes what a State needs to do instead, if it elecjflto do no
additional modeling. 'Traditional" nonattainment areas and other "tran|fconal^onattainment
areas are subject to the guidance in this document.
State implementation plan (SIP) revisions desigm
hour NAAQS in traditional nonattainment areas could
designated "nonattainment" (e.g., July 18,2003, if desi
this scenario, attainment demonstrations supporting th
2002 to allow States sufficient time to complete rulem
demonstrations to support a 2003 SIP revision would n
1.3 How Does The Perceived Nature Of i
Demonstration?
Guidance for performing attainmejPflemons
perceived nature of ozone. In this sectigmf we idejpry seve
We then describe how the guidance^pmmoda^ each.
correct
ue as e
on ooffirs on July
ns should be co
Thus, work underlying
later than 1999.
a variet
iccompa
perfe
ture. Thu
ig model'predictions. "Uncertainty" is the
j&Sroserved air quality at each receptor
/ill be a distribution of differences between
ing from comparisons on different days at each receptor.
BIS, including limits in the model's formulation which
to make computations tractable, data base limitations
reflect incomple^tniderstandin
and uncertaintyjinltdireiaiStine futurBSiterminants of emissions. States should recognize these
"- ' '•"^ "- "'"'*' fXf^i^Zwkffi*^ ^*''
limitations when prepar|Qgf||eir modeled attainment demonstrations, as should those reviewing
the demonstrations.
We recommend several qualitative means for recognizing model limitations and resulting
uncertainties when preparing an attainment demonstration. First, we recommend using models in
a relative sense in concert with observed air quality data (i.e., taking the ratio of future to present
predicted air quati^and multiplying it times a monitored design value). As described later, we
believe ibis approach should reduce some of the uncertainty attendant with using absolute model
predictions alone. Second, we recommend that a modeling analysis be preceded by analyses of
'Timing may differ for "traditional" and "transitional" nonattainment areas. See U.S. EPA
(1999a) for further discussion of "transitional" nonattainment areas.
20
-------
available air quality, meteorological and emissions data to gain a qualitative description of an
area's nonattainment problem. Such a description should be used to help guide a model
application and may provide a reality check on the model's predictions. Third, we offer the
option for States to use several model outputs, as well as other analyses besides models to
provide corroborative evidence concerning adequacy of a proposed strategy for meeting the
NAAQS. Outcomes of modeling and other analyses are weighed to determine whether or not the
resulting evidence suggests a proposed control strategy is adequate to meet the NA^QS. Finally,
we identify several activities/analyses which States could undertake, if Ute^ so choose, to better
apply models and corroborative approaches in subsequent rejaews/ana^^g||Kcontrol strategy.
These subsequent reviews may be useful for determining vfflSper a as expected. ^
A State has the responsibility to prepare a subsequent Sl^p/ision, if thPtSSSSufmds that
SIP is substantially inadequate to achieve the NAAQS.
Premise 2. Resource intensive approaches may oftei
attainment demonstration. This follows from the regie
approaching 0.08 ppm in large portions of the U.S. WhJ
NOx emissions should reduce ozone in much of the ea
.08 ppm level specified in the NAAQS will affect.
the remaining nonattainment areas.
•ategies
If regionally is a problem, this
Regional modeling applications requi
bases covering large areas of the coj
generating meteorological and e:
substantial. States facing the ne
intensive techniques may wj$b. t
cerec
ordinaupi, qu
Resources used
fed to support an^adequate
.of ozone concentrations
|hat regional reductions in
itions approaching the
ain the NAAQS in
egional modeling domains.
trance and management of data
?run recommended models for
id the aafquality model itself can be
attainment demonstration requiring resource
arces in some manner. Examples might
- r—JLS ;0nSide^
include delegating responsibilities for certafflfpHll>f the analyses to a single State which can
"specialize" in that kind of an|!i|£& Anothef example might be formation of a regional center of
some kind to perform andyses»fffijpected by its client group of States.
' ' ;- i^. i,*-" .. . f t " #*
.;'-^*fc - ry- *
Premise 3. There willfbe a widespread need to use nested regional models. Available air
quality data suggest ozpn||concentrations approach levels specified in the NAAQS throughout
much of the eastern U.SisanS^parge parts of California. Near nonattainment areas, more
detailed attention may nee&fplle paid to atmospheric mixing of nearby emissions than is
necessary for emissions in locations which are more remote from the nonattainment areas. This
is consistent with use ofcnested regional models.
j.-i
s', -
This guidance identifies several modeling systems2 with nesting capabilities, including
2A modeling system includes a chemical model, an emissions model and a meteorological
model. Terms, such as this one, which are introduced using italics are defined more fully in a
glossary at the back of this guidance. "Modeling system" and "air quality model" are used
interchangeably. "Air quality model" means "modeling system" in this guidance.
21
-------
one for which the US EPA will provide user support (MODELS3/CMAQ) (U.S. EPA, 1998a).
Support includes documentation, ready access to the code, training, updates and limited
troubleshooting. We believe it is premature to identity MODELS3/CMAQ or any other nested
regional modeling system as the "guideline model" for ozone. States may use
MODELS3/CMAQ or an alternate modeling system provided certain criteria, identified in this
guidance, are met. These criteria apply equally to MODELS3/CMAQ and alternative air quality
model(s). The guidance also provides recommendations for developing meteorolo^feal, air
quality and emissions inputs used in nested regional modeling systems, j&d makpfsuggestions
for quality assuring inputs and evaluating performance of emissions, ral^^fel^ical and air
quality models.
2J
Premise 4. Problems posed by high ozone, PM
commonalities. Ozone formation and formation of
common reactions and reactants. Secondary particulai
sources contribute precursors to both ozone and PM25
ozone and secondary particulates have been observed u:
conditions. Reducing PM25 is the principal controllab
U.S. EPA policy is to encourage "integration" of
haze to ensure they do not work at cross purpoj
costs.
similar
yens of the U.S., high regional
;ts of meteorological
^ing regional visibility.
'Mj 5 and regional
:otal benefit for lower
Integration of strategies to reducJozone, Elf25 haze is complicated by
different dates likely needed for SIRglisions 2003 fjff ozone, circa 2007-2008 for PM2 5
and regional haze, etc.). One a subjiquent review of a strategy selected to meet the
ozone NAAQS is to check its cojpatibilityj||i planjjlhose details become known later) to
meet goals for PM25 and rej
iaze.
useful information for a
considering>effects«of contra
1.4 What Topics Are Covered In This Guidance?
identifies activities which could yield
t revieWi^HSpStates prepare for such a check if they so
uch as MODELS3/CMAQ, which also have the capability of
PM2 5 and regional haze is desirable.
This guidance
other analyses to help den
ro broad topics: Part I, "How do I use results of models and
late attainment?", and Part n, "How should I apply air quality
models to produce resuljssneeded to help demonstrate attainment?". Part I is divided into 6
sections (i.e., Sections;Ji)-7.0). Part H consists of 10 sections (Sections 8.0-17.0).
ontains an overview of the procedure we recommend for using results to
helpj|teinoistrali;attainment of the 8-hour ozone NAAQS. The recommended approach is to
firstiise an air quality model to estimate current and future ozone concentrations. Next, use the
predicted relative changes in ozone in concert with measured data to estimate future ozone
3PM2 5 are particles having aerodynamic diameters less than or equal to 2.5 micrometers.
22
-------
concentrations. We refer to this exercise as a "modeled attainment tesf\ If the test is passed and
a similar screening test, applied at selected locations without monitors, is also passed, these
outcomes suggest attainment will occur if the simulated control strategy is adopted. States have
an option to use a suite of model predictions as well as several additional data analyses for
corroborating conclusions reached with the modeled attainment and screening tests.
Corroborative analyses use air quality and emissions data plus additional interpretation of model
results. Results of the modeled attainment/screening tests and corroboratory analysis are
considered together in a weight of evidence determination to assess whejner orjpt a proposed
control strategy is likely to be successful in meeting the NAAQS. "W^^H^p'idence'* may be
used either to require more or to require fewer control meajfi&than ^^^Hbd attainment test
suggests is necessary.
consistent wim.«oncludi
examples of^ptional ana
Section
undertake to erihano
of these activities is re
credibility of the modeling
Section 3.0 describes the recommended model
detail. The Section includes examples illustrating use
tests we recommend.
Section 4.0 describes how a weight of evideno
State chooses to use evidence produced by corrobg
modeled attainment test and screening test.
modeling plus a series of other core corroboj
Modeling is generally the most reliable
because it integrates a diverse set of infjpnation
transport and decay. Evidence that
makes its predictions more com]
of evidence determination. Eaci
identified, along with condi
ening
mild be performed, if a
analyses^^^^&ment results of a
ence deifpination consists of
^ali ""
additional, optional analyses.
fzone concentrations,
Ascription of ozone formation,
ice a detailed observed data base
ways be included as part of a weight
of evidence determination is
its outcome. Outcomes which are
rategy will work are also identified. Several
1, along with recommendations for accompanying
severaBflata gathering activities and analyses which States could
and corroborative analyses to support subsequent reviews. None
ie current ozone SIP revision. However, they would increase
ysis exercise. A subsequent review will be desirable to diagnose
whya strategy is or isn'tavorking, or to relate the chosen strategy to others which are later
considered to reduce PJ^2.5or regional haze. The less extensive a data base underlying a modeled
attainment demonstration, the greater the potential need for a subsequent review.
; Section 6.0 identifies the necessary documentation describing the analyses used to
demonstrate attainment of the ozone NAAQS.
Section 7.0 lists the references cited in Part I and in this introduction (Section 1.0).
23
-------
Part n ("How should I apply air quality models to produce results needed to help
demonstrate attainment?") begins in Section 8.0 with an overview of the topics to be covered.
Section 9.0 identifies a series of meteorological, emissions and air quality data analyses
which should be undertaken to develop a qualitative description of an area's nonattainment
problem prior to a model application. As we describe, this qualitative description should be used
to guide the subsequent model application.
Section 10.0 identifies the need for a modeling/anal
protocol's function as well as what subjects should be
use
;o model for an ozone
NAAQS and its
Section 13.0 identifies factors whic
and horizontal and vertical resolution for
Section 14.0 addresses how
boundary air quality data for use i
Topics covered include use of dwarnic mei
assimilation,del.
assurance, appli
Section 11.0 addresses what should be conside
attainment demonstration of the ozone NAAQS. Sevei
of a model for this purpose.
Section 12.0 provides guidance for selecting suii
attainment demonstration. Topics include a discussio
resulting implications for episode selection.
>sing a model domain
ivelop nlfteorologjll inputs as well as initial and
lin^xercise sjjpporting an attainment demonstration.
»logicaj|ll>dels, four dimensional data
br "ramp-up" days.
emissi
i develop appropriate emissions estimates for use in the
|jh)de use of available inventory estimates, quality
and estimating future emissions.
Section 16.0lBe£i5!mfetopics of model performance evaluation and use of diagnostic
analyses. * *
The guidance concludes with Section 17.0, which lists references cited in Part n, and a
glossary of important trams which may be new to some readers.
24
-------
-------
-------
2.0 What Is A Modeled Attainment Demonstration?~An Overview
A modeled attainment demonstration consists of (a) analyses which estimate whether
selected emissions reductions will result in ambient concentrations that meet the NAAQS, and
(b) an identified set of measures which will result in the required emissions reductions. As noted
in Section 1.0, this guidance focuses on the first component of an attainment demonstration-
interpretation and conduct of analyses to estimate the amount of emission reductiojpfeeded to
reduce ozone concentrations to a level which is consistent with meetingjiie NAjjpQS. Emission
reduction strategies should be simulated by reducing emissions from sj^^^^Ource categories
rather than through broad "across-the-board" reductions frgflpl soi
States should estimate the amount of emission
attainment by using a modeled attainment test plus usi
without an ozone monitor. In addition to these tests, a consider a broa31iiPF6f model
results plus perform a set of other corroboratory ^^y86^^^^^^06 whether the "weight of
evidence" produced by the tests and additional analysesjdBc^Pgg^i proposed emission
reduction will lead to attainment of the NAAQS.
A modeled attainment test is an e:
current and future air quality. If future gfemates
is passed. Our recommended test i
than "absolute" sense. That is,
ozone monitors. We call each
ozone design values are estj
reduction factor at locati
design value, Thfcresulting|
If all such futuresite^specifi
2.1 What Is The Recommended Modi
ill Overview
model is used to simulate
value are < 84 ppb, the test
lates are used in a "relative" rather
si's future to. current predictions at
s, relative reduction factors. Future
sites by multiplying a modeled relative
les the observed monitor-specific ozone
^-specific "future design value" is compared to 84 ppb.
yjjues are <, 84 ppb, the test is passed.
tment tesfpenrecommend predicts whether or not all observed future
-; -* N-£, &•
>r equal to the concentration level specified in the NAAQS for
The modeMi
design values will r>e*i
ozone under meteorologlS^^litions similar to those which have been simulated. By itself, the
test makes no statement rf>o1|Miture ozone at locations where there is no nearby monitor. Thus,
we require a supplementary screening analysis to identify other locations where passing the test
might be problematic ifsmonitoring data were available. Like the test itself, this supplementary
screening test is described more fully in Section 3.0. Briefly however, it entails the following:
.X*T ' •
.','**• '"
-identification of areas in the modeling domain where "absolute" predicted 8-hour daily
maxima are consistently greater than those predicted in the vicinity of any monitor site,
and
27
-------
-computation of relative reduction factors for each identified unmonitored area with high
predicted ozone. These factors are then multiplied by the areawide design value to
obtain an estimated future design value for each such location.
2.2 What Does A Recommended Weight Of Evidence Determination Consist Of?--
An Overview
As we note later in Section 9.0, States should always Berform
air quality, emissions and meteorological data, and considglpbdelinj
results of the attainment and screening tests. Such analyjlpire instrur
conduct of an air quality modeling application. SometinS, resultsjpjorro
may be used in a weight of evidence determination to cora|iide tJpKattainment
modeled results which do not quite pass the attainment eenin tests. Thi
intary analyses of
ler than the
lading the
site
attainment or screening tests are from just being passed,
produced by corroboratory analyses must be to draw a cj
the test results. If a conclusion differs from the outcoi
subsequent review (several years hence) with moi
test is failed by a wide margin, we doubt that
evidence determination can be sufficiently
attained. Table 2.1 contains rules of thui
determination may be appropriate.
Table 2.1. Guidelin
ofthe^
etc data!
impelling contrary evidence
ig from that implied by
|ests, the need for
sed. If either
its made in a weight of
e NAAQS will be
eight of evidence
Evidence Determination
lues 84
->5^V-> %^£'-^JW&&'' • «r ^£pv* *>• ftf\
FuturCfDesign Value 85 - 89
pph at «me or more sites
* f ^.'-.JX-^'-^V-iSS*^- ~ast'
Future Design Value > 90
_,, ^ t> —
ppb/at one or more sites
A Weight Of Evidence
Determination Be Used?
Yes
Yes
Not ordinarily. More
qualitative results are
unlikely to reach a
conclusion differing from
the outcome of an
attainment or screening test.
* Includes calculations at screening sites, if applicable
28
-------
In a weight of evidence (WOE) determination, States should review results from several
diverse types of analyses, including results from the modeled attainment test and, if applicable,
the screening test. States should next note whether or not results from each of these analyses
support a conclusion that the proposed strategy will meet the air quality goal. States should then
weigh each type of analysis according to its credibility as well as its ability to address the
question being posed (i.e., is the strategy adequate for meeting the ozone NAAQS by a defined
deadline?). Next, conclusions derived in the two preceding steps are combined to^pteke an
overall assessment of whether meeting the air quality goal is likely. Thiu&ist stejris a qualitative
one involving some subjectivity. If it is concluded that a strategy is demonstrate
attainment, a new strategy is selected for review, and the piie^ is re^^^Bl^tes should
provide a written rationale documenting how and why i
adequacy of the final selected strategy.
Results obtained with air quality models are an
art of a weight o
nether the NAAQS will be
osed as one of two
ility to integrate
ent and screening
adeq
determination and should ordinarily be very influential i
met. This follows from including ability to address the
criteria for weighing results from different analyses anWtrom a
information from scientific theory and observed datahe mod
tests are passed, this supports a hypothesis
included as one of several elements in a
adequacy. The further model results are pm
the more compelling results from ornenfralyses hare to
attainment. If either the modeled attjfimnent teigpr screenjtest produces one or more
estimated site-specific future desi^^Ries >Jp>ppb, it ijgpoubtful that other evidence will be
sufficiently convincing to conclu|lPtnat theJ^AQS jgpPbe attained. States should ordinarily
consider a revised control s
is information is
m to assess the strategy's
nment or screening test,
introl strategy to demonstrate
23 Why Should A Modd Be Used In A '"Relative" Sense And Why May
Corroboratory Analyses Be Used In A Weight Of Evidence Determination?
* ,i^^4f, o*£tts»^£8S8L..
The prowdur^we^ommendlforEstimating needed emission reductions differs from
that in pasfcgmdance'l^ozoneJn two major respects (U.S.EPA, 1996). First, we recommend a
modeled;attainment testan^wWcbanodel predictions are used in a relative rather than absolute
sense. Second, the role of the^weight of evidence determination, when used, has been expanded.
That is, results can now l>e used as a basis for requiring emission reductions greater than those
implied by the modeledjattainment test as well as a rationale for concluding that a control
strategy will meet thflNAAQS, even though the modeled attainment or screening test is not quite
passe&^.There: angpeveral reasons why we believe these changes are appropriate.
1. Thefonii of the 8-hour NAAQS necessitates such an attainment test The 8-hour NAAQS
for ozone requires the 4th highest 8-hour daily maximum ozone concentration, averaged over 3
29
-------
consecutive years, to be ^ 0.08 ppm at each monitoring site4. The feature of the NAAQS
requiring averaging over 3 years presents difficulties using the resource-intensive episodic
models we believe are necessary to capture spatially differing, complex non-linearities between
ambient ozone and precursor emissions. That is, it is difficult to tell whether or not a modeled
exceedance obtained on one or more days selected from a limited sample of days is consistent
with meeting the NAAQS. To do so would require modeling many days and, perhaps, many
strategies. This problem is reduced by using the monitored design value, calculatg|lfeonsistently
with the form of the NAAQS, as an inherent part of the modeled attainn^at tes
2. Current design values for the 8-hour NAAQS are gej
specified hi the NAAQS than is true for the 1-hour ^
ozone data reported in the U.S. EPA's AIRS data base sligests thz
have design values that are within about 40 ppb of the
NAAQS. Thus, the "signal to noise ratio" in model apfi
lower than is the case for the 1-hour NAAQS, if we cont
the basis for the modeled attainment test. This follows
trying to reduce to the level of the NAAQS is closer tdfflat ]
NAAQS. Therefore, difficulties posed by model uggggajnty
NAAQS than for the 1-hour NAAQS for ozone
eh
concentration
of 1994
specified rt
Related to the 8-hl
.QS is
^absolute model predictions as
; concentration we are
me for the 1-hour
for the 8-hour
maximum ozone concentration
day, should the target for
remains a prerequisite fa
disagreements pnian indiw
later, we have found that relal
the base value subject to
is follows for two reasons.
3. Starting with an observed rather th
improvement reduces problems hi interpretingjnodel
First, if a model under (or over) pred||f an obsji/ed dailyjpaximum concentration, the
appropriate target prediction is asjpght be desired. For example, if an 8-hour daily
0 ppb
icverthe
obsenssa and a model predicted 100 ppb on that
3b? Although good model performance
model ihiiiilliinment demonstration, problems posed by
"*$&&•" " " J
•e reduced by the new procedure. Second, as described
on factors reflecting predicted 8-hour daily maxima
averaged over seyeraplays are inseaisifiye*to the magnitude of a predicted current 8-hour daily
maximum conce^nti^6ff^eraged SvpFseveral days, unless the prediction is below about 70 ppb.
This finding may facillll|||piiig days with intensive data bases (for model evaluation) even
though such days are noBambijgsthe ones with the very highest observed concentrations of ozone.
4. Model results and projections will continue to have associated uncertainty. The
procedure we recommend recognizes this by allowing use of modeling plus other analyses to
determine whetheipweignt of available evidence supports a conclusion that a proposed emission
reduction-^wtllsuffice to meet the NAAQS.
4See 40CFR Part 50.10, Appendix I, paragraph 2.3. Because of the stipulations for rounding
significant figures, this equates to a modeling target of < 84 ppb. Because non-significant figures
are truncated, a modeling estimate < 85 ppb is equivalent to < 84 ppb.
30
-------
5. Focusing the modeled attainment test on monitoring sites could result in control targets
which are too low if the monitoring network is limited or poorly designed. We recommend
using a test for selected locations without monitors. This exercise provides a screening test for
identifying a possible need for more controls despite passing the modeled attainment test. As
noted in Table 2.1, a weight of evidence determination may also be used. A weight of evidence
determination includes several modeling results which are more difficult to relatejplhe form of
s- .^l&^
the NAAQS. These results address relative changes in the frequency anaimtenjpy of high
modeled 1-hour concentrations or high 8-hour daily maxima&n the garoroBBjillays selected for
modeling. If corroboratory analyses produce strong evidenjEpiat a is unlikely t<;
meet the NAAQS, a weight of evidence determination rnjpbe used is.
inadequate, even if the modeled attainment test is pass
Recommendations. States should estimate
attainment using a modeled attainment test and
sites without monitors. The modeled attainment
relative sense to compute relative reduction fat
factors should be multiplied by monitored
estimate future design values to com]
attainment test should be supplementea*wit
procedures using the areawide design value
actions needed troionstrate
test applied at selected
tse model predictions in a
ii^assocla||rith a strategy. These
values monitors to
, the modeled
; testpmich applies similar
ions without monitors.
States should undertake com]
emissions data. These addi
which underlies the a
tests are not failed
corroborative an;
det
and screenini
of air qualii
used eitherto
the NAAQS
resu
ntaryjahalyses ojpair quality, meteorological and
analyses are nejiid to design and focus modeling
t test tovidedjissults of the attainment and screening
evidence produced by
her withof the tests in a weight of evidence
*i$LS8$$$£'~ ~
idence determination includes the modeled attainment
isults of additional model outputs plus other analyses
rologicaliuadiemissions data. A weight of evidence analysis may be
~*yjji$^j$Ji$%C
or decrease emission reductions identified as sufficient to meet
attainment test
31
-------
jl>,^*T •*.; "-^ ; *
- ;-,«'-: ,.&=•-,' r -
s$t--'.<..-'- '' ••: '•
Irrr*
-------
3.0 What Is The Recommended Modeled Attainment Test?
In Section 2.0, we provided an overview of the recommended modeled attainment test.
However, there are several decisions which must be made before the recommended test can be
applied. In this Section, we identify a series of issues regarding selection of inputs to the test,
and recommend solutions. We next describe how to apply the test and illustrate this with
examples. We then identify some implications resulting from the test. We conclu^pwith a
further discussion of a screening test recommended for locations withoujgnonij||r for which
predictions are consistently higher than any near a monitorinmsite.
Equation (3.1) describes the recommended
monitoring site I.
(DVF), = (RRF), (DVC), 1)
where
(DVC), = the current design value (e.g.,
(RRF), = the relative reduction unit
The relative reduction factojfistrie daily
maximum concentration jjjfldicted nJEr a maBarapweraged over several days) to
the current 8-hour daitoffiiximun^roncentra^inpredicted near the monitor
(averaged over the sanwseveraj|iays), andjir
(DVF), = the estimatedJiituie design VJBjjJBSS&jffitint attainment is required, ppb.
- ,:.-:«. . X&J''
Equation (3.1) looks'sraffleenough. However, several issues must be resolved before
applying**' *****
^A^jpSKl XMS^vpr •**?»*•
(1) How is a "sitejs^^^urrenllpgn value ((DVC),) calculated?
(2) For which of the 3 yera^^eaised in the future to assess attainment from monitored data,
should future emissions bexHclilated for the modeled attainment test?
i; V
(3) In calculating the (RRF), , what do we mean by "near" site I?
^
(4) Several surfaceigfid cells may be "near" the monitor, which one(s) of these should be used to
calculafethe|RRF), ?
(5) Should any days be excluded when computing a relative reduction factor?
The preceding questions can be lumped under a single question, "how do I select appropriate
33
-------
inputs for the modeled attainment test?"
3.1 How Do I Select Appropriate Inputs For The Modeled Attainment Test?
Calculating the current site-specific design value (DVC),. The modeled attainment
test is linked to the form of the 8-hour NAAQS for ozone through use of current monitored
design values, calculated consistently with the form of the NAAQS. The current jjflNspecific
monitored design value (DVC), is calculated at each monitor by noting tjf^rnJpfhest daily
maximum concentration in each of 3 consecutive years.. Thejrithmet^^^^p these 3 values is
then computed. Table 3.1 illustrates the procedure (i.e., nojgpat get
truncated).
Table 3.1. Example Monitoring Data f<
(1)
Monitoring Site
Number I
(2)
4th High 8-hr
Daily Max.,
Yearl
(3)
4th High 8-
Daily
(5)
Average 4th
High 8-hr
Daily Max.
Monitor 1
94ppb
90.67 = 90 ppb
Monitor 2
89 ppb
82 ppb
84 ppb
Monitor 3
94 ppb
81 ppb
86 ppb
The site-specific current
zF'Tr "~ *<•
test are showJtTBn,,colum
j -L ""':~fl!**-&-'-ri!f-:'l A\
descnbed'injSerapii, 3.4)
ozone
alues the recommended modeled attainment
table. TfiStoeawide design value (used in the screening test
the highest of the average 4th high daily maximum
ifle^design value for Nonattainment Area "A" is 90 ppb.
less than 3 years of complete data?
We believe it is rctipjlr|irit to consider as much available air quality data as defensible in
the modeled attainment test!whus, States should include monitoring sites having less than 3
years of complete data for the current (e.g., 1997-99) period. As described below,/or use in the
modeled attainment test, a State should calculate current design values at sites having
obsecrations dura^lwo or more years during the current period. The following order of
prefeien^^oiuflplDe used for calculating the current design value for each site in the modeled
~f Jf j'Ss\/'1'%*'sS*''
attainmentrtesr
1. Three years of complete (i.e., > 74.5% days during the ozone season have valid 8-hour daily
maxima) data are available in the current period. Calculate the current design value as shown in
Table 3.1;
34
-------
2. Three consecutive years of complete data are available, but only two of these are within the
current period (e.g., complete data are available for 1996-98, but not for 1997-99). However,
year 3 (e.g., 1999) of the current period has some observations at the site in question. Calculate
the current design value as shown in Table 3.1 for
(a) 1996-1998 (in this example), and
(b) 1997-1999 (in this example), using the 4th high 8-hour daily ma
concentration observed in year 3 (1999, in this example^
Use the higher of "(a)" or "(b)" as the current design val
3. If the conditions in neither "1" nor "2" are met at a si
or 3 of the years in the current period, choose the high
"(d)" exceeds 84 ppb, do not calculate a current design v
modeled attainment test. Under these circumstances, i
an overly optimistic estimate of a future design value.
nor
site nor include it in the
in the test could result in
(c) Calculate the current design value by
maximum over 2 years at a site with 2
(d) Calculate the current design v
maximum over 3 years at a site
Choosing the
Ovepcpnsecutiv
heavily influeiacedyby mei
"current
high ozone we
monitored daj
available emissions
used to-designate an are
by takijlj'the m
3 yearsPobservati
hest 8-hour daily
ie 4th highest 8-hour daily
the attainment test.
ng -yef^w variability in observed design values is
variations. A State should seek some assurance that
Itgnment test do not reflect a period in which conditions for
ring site. Accordingly, States should review
straddling" the year represented by the most recent
|(e.g., 1995-97, for a 1996 inventory), and (b) the 3-year period
inment"5. The current design value used in the modeled
attainment and screeningjel^jpfs the higher of the two estimates obtained from (a) and (b). For
the modeled attainment Jest, this choice should be made on a monitor by monitor basis.
Selection of appropriatecurrent design values is illustrated below.
-I.'- ,£>'
' ,-,> "'": \. £&(,'
Example 3.1 ^'
: <" ,.*-- -**,~jf •• ,v5«««f ^^i,; '
', <: *««1W<-.VSV "i ,J
•^ntt*;^. •,4i«i"
Given:;Xi) An area is designated "nonattainment" for the 8-hour ozone NAAQS on the basis of
5For later, subsequent reviews, (b) becomes instead the most recent available 3-year
contiguous period of ozone observations
35
-------
1997-1999 monitored data.
(2) The most recent inventory reflects 1996 emissions.
(3) For purposes of illustration, suppose the area contains only 4 ozone monitors.
Find: The appropriate site-specific current design values to use in the modeled attainment test.
Solution: Since the most recent inventory reflects 1996, we need to examine
values for 1995-1997 and compare these to design values for 1997-
the two design values at each site. These are the values for QlfC in th
The procedure is shown in Table 3.2.
Table 3.2. Example Illustrating Selectio Af Curre
design
the highest of
attainment test.
CurrenfDesign
Value (DVC) Used
In The Modeled
Attainment Test,
ppb
1995-97 Design
Value, ppb
j*-;
Jf 81
screening test (see Section 3.4)
|ear to p'rfjiplflt|re emissions. States should project future emissions to
the year in whichlalliibntrol meH^^ft^led to meet the NAAQS are in place (U.S.
- ""'^^' :"-"^i?>v*v^ 'tf~ •^^^•X-**' •*•
EPA(1999a)).^Sihcevttife||G|.AQS incorporates observations from three years, the projection year
should be the first yeaTSflii^ear period used to judge attainment. For example, if the year by
life 2007, air quality monitoring data from 2005-2007 would
which attainment is
ultimately be used to jud^e^fSether the NAAQS has been attained. Thus, in this example, the
projection year used in the modeled attainment test should be 2005. As noted in Section 6.0,
accompanying documentation should show that emissions remain at or below amounts projected
to the initial year ojjdle 3-year period.
Identifying surface grid cells near a monitoring site. There are three reasons why we
believe it is appropriate, in the modeled attainment test, to consider cells "near" a monitor rather
than just the cell containing the monitor. First, one consequence of a control strategy may be
migration of a predicted peak. If a State were to confine its attention only to the cell containing
a monitor, it might underestimate the RRF (i.e., overestimate the effects of a control strategy).
36
-------
Second, we believe that uncertainty in the model's formulation and inputs is consistent with
recognizing some leeway in the precision of the predicted location of daily maximum ozone
concentrations. Finally, standard practice in defining a gridded modeling domain is to start in the
Southwest corner of the domain, and reckon grid cell location from there. This is often, and
indeed should be, many kilometers from monitoring sites in a modeled traditional nonattainment
area. Considering several cells "near" a monitor rather than the single cell containing the
monitor diminishes the likelihood of quirks in the test's results resulting from geqaaltry of the
superimposed grid system.
Earlier guidance (U.S. EPA(1996)) has identified
also consistent with the broad range of intended represei
monitors identified in 40CFR Part 58, Appendix D.
refine our default definitions for the arr
doing so is documented.
Table 3.3. Default R
For ease in computation, States may assume thai
which it is located and that this cell is at the center of an
Figure 3.1, the number of cells considered "nearby" (i.e
function of the size of the grid cells used in the modelt
recommendations for defining "nearby" cells for ^^te^ems ^a
Thus, if one were using a grid with 4 km grid defi
with the monitor located in the center cell.
demonstrated mesoscale flow patterns (e.dpfand/se,
tell in
earby" cells. As shown in
*15 km of) a monitor is a
ides a set of default
various sizes.
a 7 x 7 array of cells,
of topographic features,
s) or other factors to
ovided the justification for
Grid Cells Used To Calculate RRF's
Size of the Array of Nearby Cells, unitless
7x7
5x5
>8
3x3
1x1
37
-------
Figure 3.1. Relationship Between Grid Resolution and Grid Cells
Considered to be in the Vicinity of a Monitor
(a) 36 km grid resolution
15 km radius
Monitor Site
(b) 12 km grid resolution
(c) 4 km grid resolution
38
-------
Choosing model predictions to calculate a relative reduction factor ((RRF),) near a
monitor. Two decisions need to be made. First, given that a model application produces a time
series of estimated 1-hour ozone concentrations (which can be used to calculate running 8-hour
averages), what values should be chosen from within the time series? We recommend choosing
predicted 8-hour daily maximum concentrations from each modeled day (excluding "ramp-up"
days) for consideration in the modeled attainment test. The 8-hour daily maxima should be used,
because they are closest to the form of concentration specified in the NAAQS.
r daily maxima
lend choosing
with cur
Ed
on each day for
several reasons.
imum
ell with
eacrLsav in the tes
The second decision that needs to be made is, "which
predicted in cells near a monitor should we use to calculat
the nearby grid cell with the highest predicted 8-hour daij
emissions for each day considered in the test, and the;
daily maximum concentration with the future emission^
any given day, the grid cell chosen with the future emis
chosen with current emissions.
We believe selecting the maximum (i.e., peak)
subsequently calculating the relative reduction fac
First, it is likely to reflect any phenomenon w!
migrate as a result of implementing control^PBcorii
produced by a finely resolved modeling asliysis.
one
ions within a plume to
better advantage of data
The relative reduction facto|4||3£F) usedpm the mojlled attainment test is computed by
taking the ratio of the mean of the^Jliaed maximmJPpredictions in the future to the mean
of the selected 8-hour daily maxjpmi pred^^ns witj|ffirrent emissions. The procedure is
illustrated in Example 3.2.
Example 3.2
Given:(1)
days ha^SxiiiSamulated using current and future emissions;
. r-^ss^^-. ^K
(2) The horizontal dimS||D^|or each surface grid cell are 12 km x 12 km;
(3) Figure 3.2 shows preMctfliS-hour daily maximum ozone concentrations in each of the 9 cells
"near" a monitor with (a) future emissions, and (b) current emissions.
Frnd:.The site-specific relative reduction factor for monitoring site I, (RRF),
39
-------
Day1
figure 32. Choosing Predictions to Estimate RFFs
(a)PredctionsWtthR*ueEniss«ons
Day2
Day4
7B
78
71
81
TO
80
74
76 \73
82
77
81
Futue Mean Peak 84r Daily Max.=(87+82+77+81) / 4 =81 ppb
(b) Predctions With Orrert Emission
Day1
Day2
Day3
Day4
100
100
96
99
as!95
82
80
88
78
82
81
91
80
79
91
90
79
86 ( 88 [ 90
81
81
88
83
87
79
98 100 91
Curst Mean Pet* Wr Daly Max=(98+100+91 +90)/4=94ppb
90
t
if
40
-------
Solution:
(1) For each day and for both current and future emissions, identify the peak 8-hour daily
maximum concentration predicted near the monitor. Since the grid cells are 12 km, a 3 x 3 array
of cells is considered "nearby" (see Table 3.3). The numbers appearing beneath each 3x3 array
in Figure 3.2 are the peak nearby concentrations for each day.
(2) Compute the mean peak 8-hour daily maximum concentration for
emissions.
Using the information in Figure 3.2,
(a) (Mean peak 8-hr daily max.)^,^ = (87 + 82 + 77 +
(Note that we have truncated the insignificant fi
(b) (Mean peak 8-hr daily max.)cunent = (98 + 100 + 91 +A
(3) The relative reduction factor near site I is
id (b) current
(RRF), = (mean peak 8-hr daily max.),
= 81/94 = 0.86
Limiting modeled 8-ho
modeled day, meteorology
concentratioiH5i$.e., near,
coul
current
JiT-'
chosen'to calculate RRF. On any given
iilar to those leading to high
Sue) at a particular monitor. If ozone
lar day is very much less than the design value, the model
gonsive to controls (e.g., the location could be upwind
ment area on that day). Using equation (3.1) could
from most of the emissions in _._, ___ _ ___
, , , ^S^illfe, , . , ,. , - , .
then lead to an erroneouslvshigh projection of the future design value.
,, ,•;'.;. .«.«.»||»JS«'.lSfc, %-
Figure 3.3 shows^^fflSlfrom a study in which we modeled current and future emissions
for 90 days during 1995 ttsnpa grid with 12 km x 12 km cells and 7 vertical layers. One
purpose of the study wasio assess the extent to which a relative reduction factor (RRF) is
dependent on the magnitude of predicted current 8-hour daily maxima. We examined RRF's
computed near eacK«6rl58 monitoring sites in the eastern half of the United States. Sites reflect
a variely.pf surajujSiings and reductions in surrrounding VOC and NOx emissions to varying
degr^iHFhe^arves depicting the relationship between mean current 8-hour daily maximum
concentrations and RRF averaged over 10 days for the two sites shown in the Figure are typical.
Generally, the RRF is not a strong function of the predicted current 8-hour daily maximum ozone
concentration averaged over several days when these averages are > 70 ppb. We would expect
relationships like those in Figure 3.3 to be more variable if they reflected averages over only 1-2
41
-------
1.00-1
ago
aeo-
igUB3*
n
i
30
x ivEanrBcnverBaucaonaBarincuoncji
Mean Redded Orrert frhoir Daily Maxima*
* Msncf ttttxfetedOayB
'n D NwirtbkOtySteKBOne
\^d'D ""D"'-..« . n
\ \j LJ
•
i i i 1 i 1 [ | | |
50 70 90 110 130
ft
42
-------
days. Thus, it is better to simulate several days so that RRF values are less likely to be affected
by how closely a model's predictions match observed 8-hour daily maxima at individual sites on
a given day.
The episode selection procedure recommended in Section 12.0 should help focus
modeling on days with observed concentrations near a nonattainment area's design value.
Nevertheless, there will inevitably be some modeled days where the predicted Srhplr daily
maximum ozone concentrations near a monitoring site do not reflect co^|ipnsjfeading to
observations near its design value. To illustrate with a simpl^xampljJBiraer a city with 2
monitors, one north of the city and one south. We would |fipP the *™BBKBltLthft city to
observe high ozone, at or near the design value on the selgpd days to the
north. However, on days when the wind is to the soumJle norme
from the control strategy. If local emissions are influen^^m afjdng observe
we would expect to predict concentrations well below site's design vl
with winds to the south. Presumably, there would be se^
analysis needs to provide assurance that a strategy will
the nonattainment area, including the site south of the ttoly.
sons,
a day
iodeled days, since the
e NAAQS at all sites in
As a rule of thumb to avoid overestima
excluding some modeled days from consider
near a monitoring site. More specificallj^flates ne
predicted current maximum 8-hour daihfmaximuprconcens
ppb. Example 3.3 illustrates what low qarrent predjptions occur near a monitor on a day
:sign vaTBlpwe recommend
ieledjptrainment test is applied
ly day for which the
Bat a nearby grid cell is < 70
(e.g., as might happen if the monij
Example 3.3
Given:
current em
predicted for tfii
mul
5 ppb'
Sipwiur on that jjy).
•^ r,x>K£&r,~
formed irfExample 3.2 yield low predictions near site I with
peak 8-hour daily maximum ozone concentration
Kan the 91 ppb shown in Figure 3.2).
Find: Thejelative
ictor near site I ((RRF),).
Solution: (1) Calculate !Hipan peak 8-hour daily maximum ozone concentration obtained
near site I for current ancfoiture emissions. Exclude results for day 3 from the calculations.
From Figure 3.2, f"
*<*
(a) (mean peak 8-hr daily max)future = (87 + 82 + 81)/3 = 83 ppb
^ f \ " ""
< * ^^^
(b) (mean peak 8-hr daily max)cunew = (98 + 100 + 90)/3 = 96 ppb.
(2) Compute the relative reduction factor by taking the ratio of (l)(a) over (l)(b).
43
-------
(RRF)i = 83/96 = 0.86
Recommendations. States should estimate current monitored design values (DVC) for
each monitoring site in the nonattainment area using the procedures illustrated in
Tables 3.1 and 3.2. States should consider modeled 8-hour daily maxima from all
surface grid cells near a monitoring site. Default recommendations for "near" are
provided in Table 3.3. Site-specific relative reduction factors (RRF)
calculated by taking the ratio of mean highest 8-hour daily maxima obtained for the
future and current emissions. A day may be excludedJfrom <
the nearby peak modeled current 8-hour daily maximlan ozoi
day is < 70 ppb<
3.2 How Do I Apply The Recommended Me
States should apply the modeled attainment test
nonattainment area observing a current design value of
be applied at monitors (with current design values ^ 7:
affected by transport from the area. If the focus of
needed to supplement national or large scale rej
those within one day's travel time of the nonJpEQnml
estimated using trajectory models. If a trajectory
within the nonattainment area shortly before sunrj|§ affect*
sunset or earlier on the same day, St^^should^Jlume thi
transpoif'from the nonattainmentJsePTor ujpan scale'
modeled attainment test generalplpplies toft ozone
domain having current monitorealdesign v^fcB^^Spb.
on at a site if
tration on the
itoring site within a
The test should also
onattainment area but
local measures
s, ''site9Hfected by transport" are
travel time may be
fat emissions occurring
:6r's readings shortly after
ie site is "affected by
lyses (i.e., see Section 13.0), the
itors within the urban modeling
Inputs described inSectiinS.l are applied in Equation (3.1) to estimate a future design
value at aliirti6m||||for whici^i^TOdeled attainment test is applicable. The 8-hour NAAQS is
met in an area ii^lote^xonseciit^^^&Sf the average 4th highest 8-hour daily maximum ozone
concentration obse^A^c&ch monftpr is < 0.08 ppm (i.e., < 84 ppb using rounding
conventions) .
lliresulting predicted future design values (DVF) are <, 84 ppb, the test
is passed. The modeledaffiflnment test is applied using 4 steps.
1 f~ Jn- ..-,-ys&2 .-:«*ws *,.s. - A * f i.
Step 1. Compute a site-specific current design value (DVC) from observed data.
" - f£~
This is done as Illustrated in Tables 3.1 and 3.2. The values in the right hand column of
Table 3^2 are sitefSpecific design values.
Step 2. Use air quality modeling results to estimate the relative reduction factor for that
site.
640CFR Part 50.10, Appendix I, paragraph 2.3
44
-------
This step begins by computing the mean peak 8-hour daily maximum ozone
concentration for future and current emissions. This has been illustrated in Examples 3.2 and
3.3. The relative reduction factor for site I is given by Equation 3.2.
mean peak 8-hr daily max)c
(RRF)i = (mean peak 8-hr daily
Using Equation (3.2), the relative reduction factor is calculated;
(5) in the last row of Table 3.4. Note that the RRF is calculated to twc
right of the decimal place. The last significant figure is objnfpl by
or more rounded upward. For the illustration shown in Tj|Pe3.4, we
days described previously in Example 3.3 have been sinjpated. Wf
monitored current design value at site I is 102 ppb.
(3-2)
the column
it figures to the
pth values of "51
L the same
that
Step 3. Multiply the observed current design value ot
reduction factor obtained in Step 2. If the estimated^
test is passed at the monitoring site being evaluat
fStep 1 times the relative
value is s 84 ppb, the
These calculations are illustrated in Table 3.4.
peak 8-hour daily maximum ozone concent
Section 3.1, predictions for this day are nj
last row of the Table.
wa
licted current modeled
>). As discussed in
mean values shown in the
In Table 3.4, we see (column (2)),
Using Equation (3.1), the predic
(DVF},=
•
Note that in this;
current observeaimesign value at Monitor I is 102 ppb.
ature ddign valued site I is,
ItariBr
i K ^oa^ar^^^^^^
= 88 ppb,f^f?
**«<.:£ **
Utainment test is not passed at monitor I.
Step 4. Repeat stejpsl-3jfor all ozone monitoring sites with current design values > 75 ppb
during the 3 years us^asthe basis for the current monitored design value. If the test is
passed at each site, the miodeled attainment test is passed.
•K-f,^ 4 -K- •-^i^: *•
' "
45
-------
Table 3.4. Example Calculation of a Site-Specific Future Design Value (DVF),
(1)
Day
(2)
Observed
Current
Site-specific
Design
Value,
(DVC),,Ppb
(3)
Current
Peak
Predicted 8-
hr Daily
max.conc
near
Monitor,
ppb
(4)
Future Peak
Predicted 8-
hr Daily
max.conc
near
Monitor^
ppb
(5)
Relative
Red. Factor
(RRF),
(6)
Future Site-
Specific
aue,
(DVF),, ppb
Recommendations. Th<
'sv^,^2^£j&-;
~-< i vfiir^
I. Computefa
,^<"';''' "'
2. Use air quality
attainment test is applied in 4 steps.
t design value from observed data.
results to estimate a site-specific relative reduction factor.
3. Multiply the relative reduction factor obtained in step 2 times the site-specific
^monitored designjralue obtained in step 1. The result is an estimated site-specific
jfuture design vjaftie. If this value is < 84 ppb, the test is passed at the monitor site being
evaluated. ^rf!rfp
4. Repeat steps 1-3 for all ozone monitoring sites with current monitored design values
> 75 ppb during the 3 years used to compute the current monitored design value. If
the test is passed at each site, the modeled attainment test is passed.
46
-------
33 What Are Key Implications Of The Recommended Modeled Attainment Test?
The recommended modeled attainment test for the 8-hr ozone NAAQS raises some
implications which warrant further discussion.
1. The attainment test focuses on monitoring sites. In this sense, the modeled
attainment test is identical to the monitored test, used to define whether or not attajphent occurs.
One shortcoming of this approach is that there are usually only a small nj^berjjPTnonitoring
sites compared to the area which is modeled. This could result in me j|g||jypdicting 8-hour
daily maximum concentrations in one or more locations wj
a monitor. If this result occurs consistently (e.g., predic
location is > 5% above any predicted near monitored 1
modeled), it should be investigated further. As descri
be completed if there is one or more locations with pre
any near a monitoring site.
If the attainment test and screening test are pas
predicted future design value is < 90 ppb), a State
analysis to assess whether attainment is likely.
determination includes an assessment of othejirnocle:
monitored information. Other importantjphsiderai
would be evidence that the State plans tJBeploy ajfoozone
the modeling suggests there could bjlpure prcdflems meei
perform a subsequent review to selSlF contro>
evidence approach for demonstrffing attainjplat is di;
predicted
lould
er than
passed (e.g., the
eight of evidence
3, a weight of evidence
less dependent on
I of evidence assessment
ng site in a location(s) where
g the NAAQS and is planning to
itegy, if necessary. The weight of
led in Section 4.0.
y maxii
intakes assun^tions about the shape of a distribution of
ter controls are in place. What is the basis for these
sumes that the distribution of 8-hour daily maxima for each
averageS-ho
assumptions'!
site will flatte
concentration^anci
day's ranking in the
distribution. The first
zd (e.g., the difference between the 90th percentile
concentration diminishes after controls are in place), but that a
control distribution is similar to its ranking in the post-control
*.™e™w. jssurnptions is common sense. Concentrations which are close to
background levels alreadyli|Slikely to be greatly affected by control measures (Lefohn, et al..
1998). This assertion islilso1x>rne out by results of a study by Pacific Environmental Services
(PES), Inc. (1997) sho>wng trends in observed distributions of 8-hour daily maxima in 9 cities
between 1980 and 1<$5. The PES study also shows that relative change observed in typical 95th
and 99th percentile 8-hour daily maximum ozone observations between 1983 and 1995 is
generallysimilar. A modeling study by Meyer, etal. (1997) lends further support by showing for
numerous cities that if a day is ranked as a severe one in the pre-control distribution of predicted
8-hour daily maxima, it will have a similar ranking in the predicted post-control distribution.
Our previously identified finding that a relative reduction factor tends to increase when current
predicted 8-hour daily maximum ozone is below about 70 ppb but is independent at higher
47
-------
predicted current values is consistent with these other studies.
3. The test adjusts observed concentrations during a current period (e.g., 1997-1999) to a
future period (e.g., 2010) using model-derived "relative reduction factors". It is important
that emissions for the base period used in the test correspond with the period reflected by
the current design value (e.g., 1997-1999). Large deviations from this constraint may diminish
credibility of the relative reduction factors. For example, if 1990 emissions were jiiripared to
emissions projected for 2010, the contrast would probably be greater thafef 193p$>r 1999
emissions were compared to the 2010 estimates. Presumably^this woj
relative differences in ozone between 1990 and 2010 vs. bjJpeii 19
resulting smaller relative reduction factors were applied t011K)97-1999
State may underestimate 2010 ozone concentrations.
larger predicted
id 2010.
Unfortunately, the constraint described in the pr
confusion about just what we mean by "emissions for the
case" is commonly understood to mean emissions correj
For example, if we were modeling a 1991 episode, "I
emissions. As described in Section .16.0, it is essei
with meteorology occurring in the modeled eph
However, once the model has been shown topnrTor
model the "base case" emissions. It nowjjeomes ijnpai|
to the period with the current observed design val|fr(e.g.
corresponding to the statutory attainrfyift date tgjj'., 2010).j
former as the "current" emissions.^^Rontinjt®to refer
evaluation as "base case" emissh
paragraph may int
That is, the term "base
episode we are modeling.
would be 1991
Emissions together
to evalulBipodel performance.
longer necessary to
!l emissions corresponding
>) and the period
this guidance, we refer to the
(missions used for model performance
it
It is desirable to
the current design rvalue i
which differsIromithe "basel
model simulation cclrespondmi
inventory. Hqw^^^^^des ne
design value, provic
occur when exceedance
to capture sensitivity of ]
occurring during the period reflected by
999). This avoids the need to derive a "current" inventory
sntory. It also avoids the need for an additional air quality
nt" inventory which differs from the "base case"
selected from the period corresponding to the current
iresentative of meteorological conditions which commonly
m occur. The idea is to use selected representative episodes
ozone to changes in emissions during such commonly
occurring conditions. There are at least two reasons why using episodes outside the period with
the current design valu^may be acceptable: (1) availability of air quality and meteorological data
from an intensive fielflistudy, and (2) availability of a past modeling analysis in which the model
performed well andiwith which the State is satisfied.
Recommendations. States should review absolute projected future model predictions
for 8-hour daily maxima to identify locations with consistently higher predictions than
any near a monitoring site. An additional screening test is needed for identified
locations. To apply the recommended modeled attainment test, States should use
48
-------
emission estimates which correspond (a) with the period represented by the current
design value (e.g., 1997-1999), and (b) with a year two years prior to the required
attainment date.
3.4 What Is A Screening Test, And Why Is It Needed?
An additional review is necessary, particularly in nonattainmen
monitoring network just meets or minimally exceeds the sizejrf the
data to AIRS. This review is intended to ensure that a contraptrate;
ozone at other locations which could have current desigryjflraes exo
monitor deployed there. The test is called a "screening' becau
were measured at a location identified in the test, mode^^esul^bggest it mi
sites with available measurements.
the ozone
fuired to report
luctions
were
4
\y at
The additional review is hi the form of a sci
in the domain where absolute predicted 8-houi
are consistently greater than any predicted^
for each identified area, multiply a locatit
areawide current design value for the
value". If the resulting estimates aidless'
locations, the screening test is j
In the first part of the sc
prediction which exceeds any m
which focuses on the 4th hi
years. Interpretation of "
protocol ^jHtowesrei:, in tfi
default critefii
-predic!
monitored
hould: (1) identify areas
izone concentrations
ring site, and (2)
ion factor tunes the
te a "future design
ppb at all flagged
stently" is important. An occasional
arily indicative of violating a NAAQS
intration, averaged over 3 consecutive
for those implementing the modeling
any stronger rationale, we recommend the following
.t the location in question is > 5% higher than any near a
% or more of the modeled days.
^%?^^3&i^?-
The "5%" difference is cgn^itent with the size of rounding differences at 0.08 ppm. Occurrence
of such a difference on 5(5% or more of the modeled days increases the likelihood that a
difference might showjip in a design value averaging observations over 3 years should a monitor
be deployed at the flagged location.
mean by "areas in the domain" in the first part of the screening test? For
eachlripdeled day, States should consider individual surface grid cells with predictions more than
5% greater than any "near" any monitoring site. An array of cells, centered on the identified cell,
should be considered "near" the monitor (see Table 3.3). As a result, several cells may be
identified for each modeled day. If any surface cell shows up within these arrays on 50% or more
49
-------
of the modeled days, a future design value should be estimated for that cell using screening
procedures described in the following paragraphs.
Once one or more locations is identified with current predictions consistently exceeding
those near any monitor, we recommend applying a screening method to estimate future design
values for such locations. The screening method applies an equation similar to Equation (3.1).
For location j,
= (RRF),
where
Recommendations. In ad
apply a screening
consistently highenoip®il|predi
DVFest = the estimated future design value obtai
ppb;
(RRF)j = the relative reduction factor at locatiolPf, conf
Section 3.1, unitless.
J = the design value for
This is the highest
example shown in Tj®e 3.2,
ign values. Thus, for the
be 94 ppb.
leled attainment test, States should
ign value at all locations with
daily maxima than those near any
monitoring site. Thfescreening test is passed if the predicted future design value is s 84
ppb at all flagged locations;^!*™'
,/,*-
50
-------
4.0 If I Use A Weight Of Evidence Determination, What Does This Entail?
In Section 9.0, we note that it is preferable for States to analyze available emissions,
meteorological and air quality data to provide insights concerning appropriate inputs and
assumptions to include in air quality modeling analyses. States may wish to use outcomes of
corroboratory analyses to provide support for results of a modeled attainment or screening test, or
to conclude that attainment is (un)likely despite the outcome of the tests. In Sectigpt.O, we
describe a weight of evidence determination, note key corroboratory anajbes wjfch should be
included, and identify criteria which should be met to use additional cgjjjffljglbTy analyses in a
weight of evidence determination.
A weight of evidence determination examines
including outcomes of the attainment and screening tes
outcome or set of outcomes consistent with a hypothesi
sufficient to meet the NAAQS within the required time
results of that analysis support the hypothesis that the
analysis is weighed qualitatively, depending on the ab
a strategy and on the credibility of the analysis.
produced by the diverse analyses supports
demonstrated with the proposed strategy.
is a document which describes analyses
outcomes of each analysis, and why a Sjife belief
supports a conclusion that the area
and screening test are from being
reach a different conclusion in
modeled attainment or scr
believe it j&tfoabtful thai
convincing ^Mgport a
ght of
st predi
itative
Each
data
nonattamment area,
analyses to perform
control strategy is
ch an outcome occurs, then
is adequate. Each
:o address adequacy of
>f) evidence
>f the NAAQS is
if evidence determination
ey assumptions and
, viewed as a whole,
te further a modeled attainment
incing other evidence needs to be to
[nation. As noted in Table 2.1, if a
tore future design values > 90 ppb, we
fs based on other analyses can be sufficiently
at attainment will occur.
will be subject to area-specific conditions and
facinay affect the types of analyses which are feasible for a
^r
significance of each. Thus, decisions concerning which
credence to give each needs to be done on a case by case
basisrby those implementinfte modeling/analysis protocol. In Section 4.1, we identify several
recommended core analyses which should be used to corroborate one another in a weight of
evidence deterrninationfPIt is appropriate to require considering a core set of analyses to reassure
reviewers and the pujwic that a selective set of analyses, supporting a particular viewpoint, has
not been .chosen^at is not feasible to perform one or more of the core analyses, a State should
documeflt^iiyjiot. In Section 4.2, we note that additional, optional corroborative analyses may
be performed. We provide several examples of such analyses, and identify conditions which
should be met for them to be considered in a weight of evidence determination.
51
-------
4.1 What Analyses Should I Consider In A Weight Of Evidence Determination?
At a minimum, a weight of evidence determination should consider the following 3 types
of corroborative analyses: application of air quality models, observed air quality trends and
estimated emissions trends, and outcome of observational models.
Table 4.1 addresses each of the 3 recommended core analyses. In the tablejjflt identify
factors which might cause those implementing the modeling/analysis pr^^oltgppve greater
credence to a particular set of results (column (2)). We also jlentify each analysis
consistent with a hypothesis that emission reductions imphj
demonstrate attainment (column (3)).
We discuss each of the recommended core corroboratoi
4.1.1 Air Quality Models
predicted rria3gmi|||8-hour
of the days used||6^»njgute the
difference" is
sufficient toipass
begin toiecome sensiti%<
Performance evaluation
Weight given to results obtained with
performance is as well as the rigor with
suggests that a relative reduction factor
mean 8-hour daily maximum ozone concfentratio
practical reasons, it may not be feas.
attainment test. It is conceivable
daily maxima could become noi;
credibility if-nearby maximuMrsidicted
monitoring^i^. It is nc
should bei :However, we'sl^^^Sbat modellxjsults may have higher credibility if nearby
how good the model
tested. Figure 3.3
ays is insensitive to the
days. However, for
s at every site in the modeled
n RRF and current predicted 8-hour
izes. Thus, model results have higher
axima agree with observed values at the
guidance on how close the agreement
agree within about 20% of observations on most or all
uction factors near each monitor. This "20%
ifference between the design value concentration
test (i.e., 84 ppb) and the concentration at which RRF
nt predicted 8-hour daily maxima (i.e., < -70 ppb).
ssed further in Section 16.0.
f Model applications for which an extensive observational data base exists have greater
credence, especially^fthe data base includes monitored values of indicator species and precursor
datajiFor ozone,«one of the most uncertain inputs to a modeling analysis is the emission
projections>pliicri must be made to a future year(s) of interest. This uncertainty is reduced if the
projeciion*petiod is short. Hence, weight of evidence provided by modeling is increased with
short projection periods. If rigorous quality assurance and review is provided for the model's
emissions and meteorological inputs, this may increase confidence that the model is yielding
correct answers for the right reasons. Thus, rigor used in preparing model inputs also increases
52
-------
Table 4.1. Recommended Core Analyses for a Weight of Evidence Determination, Factors
Affecting Their Credibility and Outcomes Consistent with Meeting the NAAQS
(1)
Type of Analysis
Factors Increasing Credibility of
the Analysis
-good model performance
-extensive observational
base available
-short projection peri
-carefully quality
inventory
-confidence in meteoro
inputs
-good ability tq
questions a
adequacy
-otheramaiyses
corroDorate co
'^selected epi
*»bservati
•value
(3)
Outcomes Consistent with
Hypothesis That a Candidate
Strategy will Leadlfr Attainment
Air Quality Models
nt test is
t-x
the attainment/scriipni tests
re nearly passed, the control
itegy requires additional
ions and efforts are
to subsequently
>e the strategy
-commitment is made to deploy
~~itors at locations not passing
ie screening test
-substantial modeled
improvement in air quality is
predicted using several measures
described in Section 4.1.1
-similar conclusions are reached
with other peer reviewed models
* i*t*£j&* ^ V v &ff*p< *
* ~4 "* "** *~ ^f *<- ^°
53
-------
Table 4.1. Recommended Core Analyses for a Weight of Evidence Determination, Factors
Affecting Their Credibility and Outcomes Consistent with Meeting the NAAQS (continued)
(1)
Type of Analysis
(2)
Factors Increasing Credibility of
the Analysis
unced, ffatistically
nt downward tre
^continued, comparable relative
(3)
Outcomes Consistent with
Hypothesis That a Candidate
Strategy will LeadAo Attainment
Analysis of Air Quality and
Emissions Trends
-current or future (Air quality
model) predicted design valuej
within a few ppb above 84 pj
-extensive monitoring ne
exists
-both ozone and prec
are available
-statistical model used
normalize trend for
meteorological diffej
explains much
-short proj
the a
-a
iward
fid exists in the
ign value at all
values greater
'sing projc
extrapolate the trend
to the required attainment
indicates an 8-hour daily
im concentration £ 84
rved air quality trend
parameters also show a
improvement
54
-------
Table 4.1. Recommended Core Analyses for a Weight of Evidence Determination, Factors
Affecting Their Credibility and Outcomes Consistent with Meeting the NAAQS
(concluded)
(1)
Type of Analysis
(2)
Factors Increasing Credibility of
the Analysis
(3)
Outcomes Consistent with
Hypothesis Thaj&ICandidate
Strategy will Lewrto Attainment
Use of Observational Models
-an extensive monitoring netwj
exists
-precursor and indicatory
are measured using ii
with appropriate sensit
-monitoring sites appear i
representative
itegy (e.g., emphasis on VOC
is appropriate.
te sources
candidate
t causes o
-data have been quality
and results are
credibility given to the results. G
values increases confidence that
appropriate.
S
if there is
(compared to
12.0, we suggest
Proper
attainment test.
itrations near site-specific design
f developed for use in the tests, are
lh have (IJeitrred recently is also advantageous. This follows
i observed ozone. Selection of a "severe" episode
b earlier year may not actually be so severe. In Section
Ing severity which circumvents this problem.
: days increases confidence in the results of a modeled
quantitative results of a modeled attainment test is greater if
corroborative, more quative, analyses yield supporting conclusions about appropriateness of a
strategy. Finally, of rnejlnalyses available, modeling reflects the most comprehensive attempt to
integrate emissions ajjd#meteorological information with atmospheric chemistry. As such,
moin has theaeatest capability to address questions about adequacy of a strategy to meet air
future. Thus, States should include modeling results in a weight of evidence
j •• , , u u j- ML. • n • i • j -j- -r • -11
detemmaUonf and these should ordinarily be very influential in deciding if attainment will occur.
The outcome from modeling supports use of a proposed strategy for attainment if the
modeled attainment and screening tests, described in Section 3.0, are passed. If the screening test
for future design values at locations without monitors is not passed, a commitment to deploy
55
-------
ozone monitors at such locations should be an important consideration in approving an
attainment demonstration. In general, the closer modeled output is to passing the attainment and
screening tests, the easier it is for other analyses to produce evidence which supports attainment.
If a modeled attainment or screening test is not passed, selection of a strategy which substantially
reduces precursor emissions and an agreement to perform a subsequent review (based on
improved data bases/tools) to refine the strategy, if necessary, can be considered in a weight of
evidence determination if other modeling outputs and other analyses support a coapbsion that
the current selected strategy may lead to attainment. Other model-produ^^ indiMtors that a
proposed strategy may be adequate are (1) model predictions^how maMBil^vements in air
quality using a variety of measures, and (2) other peer revij
suggest attainment occurs using modeled attainment test j
We recommend that at least 3 additional model
evidence determinations to provide assurance that passi
modeled attainment and screening tests indicates attainnl
reasons, each of these additional outputs reflects relativ<
may use other model outputs (not described herein) in
well.
1. Compute the relative change in
area.
This output reflects the f
concentration specified in the 8-
of the NAAQS. Further, if cu
output couldube misleadin
meteoroloililicondition
and the ^
the frequenVMHBiicted ho
conclusion stra
amined in
y passing the
inded
i tests, and for similar
;dicted air quality. States
determination as
b in the nonattainment
hourlyjJdhcentrations
NAAj^S. An example o|
hourly concentrations exceed the
ure is not directly related to the form
are subject to a systematic bias, this
odes are chosen to represent a variety of
exceeded at one or more monitoring sites,
measures described in Section 16.0, a large reduction in
Ujtrations of 85 ppb or more is consistent with a
leet the NAAQS.
consitutes a "large" reduction in the predicted frequency of
ubjective, since it is difficult to relate this measure to the
would be an "80%" reduction.
•
J • :;: 2. Compute therrelative change in the number of grid cells in the nonattainment
area with predictedJPhr daily maxima > 84 ppb.
it estimates reduction in the pervasiveness of estimated 8-hour concentrations
in excess Wthe concentration specified in the NAAQS. It is subject to the same caveats as the
preceding output. One additional complication may occur if there are not many surface grid cells
in which current emissions lead to 8-hour daily maximum ozone estimates > 84 ppb. An 80%
reduction in this measure may be regarded as an example of a "large" reduction.
56
-------
3. Compute the relative change in the total difference (ppb-hr) of hourly predictions
> 84 ppb hi the nonattainment area.
Although not the same, this output is similar in concept to the change in the "dosage" to
concentrations greater than 84 ppb. Since we are interested in estimating the likelihood that a
strategy will lead to attainment rather than estimating reduction in total dosage, this output
should be calculated differently from the procedure ordinarily used to calculate dojdige metrics,
like SUM06. We recommend using Equation (4.1).
G N
RD = -^—'-
G N
j '
Where
RD =
NJ
G
Thj^oetric is su
two we ^jggesttan,80%
'•Sr 'JL* -A
example o
4.1.2 Analysis Of Air Qus
Relative Difference
Predicted 1-hourq
Total number ofi
Total number
d grid cell j greater than 84
&1
the preceding two metrics. As with the other
fee., leadinfTo a value of < 0.20 in equation (4.1)) as an
ind Emissions Trends
This approach i^pao^^alize air quality trends observed over a period for meteorological
differences occurring from^gir* to year. The Cox/Chu approach, used extensively in U.S.EPA
(1996), is one example of how air quality trends can be normalized (Cox, et al.. 1993, 1996).
Other procedures can also be used. A curve is fit through the normalized trend and extrapolated
totthj^year in whichjhe air quality goal is to be met. Extrapolations are made by considering past
tren8iai§. wellt and projected emission reductions. If the trend is statistically significant,
e for the attainment year is at or below the air quality goal, and projected
relative emission reductions are comparable or greater than reductions occurring during the
period for which the trend is constructed, results of the trend analysis suggest a strategy will be
adequate. This procedure is illustrated in Figure 4.1 and by the following example.
57
-------
Example. Estimate the relative reduction in emissions (VOC, NOx or both) occurring
during the period corresponding to the observed normalized trend in the average 4th highest
8-hour daily maximum in the nonattainment area. Use the estimated emission trend in
concert with the normalized air quality trend to determine an "emission reduction sensitivity
factor" (e.g., (ppb)/(percent emission reduction)). Multiply the sensitivity factor times the
percent reduction in emissions projected between the current period and the required
attainment date. Subtract the result from the current design value to get a projjjlted design
value; If the projected design value is <, .84 ppb, the trend analysis oiothesis that
a proposed control strategy will suffice to reach attainment by th
The trend analysis we suggest assumes that a lin<
correspondence between monitored design values and ej
describe future air quality. This assumption probably
nearly meets the goal of being < 84 ppb. Weight given
other factors as well. The more air quality data available1
parameters which show major improvements, the more
availability of trends in ambient precursor data which
ozone trends also lend credibility to the results. Wj
higher if the procedure used to normalize the
of the variability attributable to these diffei
is not necessary to extrapolate very far intjihe ful
ctrapolatk
lated er
ibest
past
iely
consis
if evident!
le current
lyses dependslilillFeral
ter variety of trend
gults. In the case of ozone,
: emissions and
I by trend results is
Drologicalffifferences explains much
analysis more believable if it
4.1.3 Use Of Observational
lels
Observational models ta
relative importance of diffej
observed ozone. There ad
approach
1999)) and a
ldvantagj||| momtoji:! data to draw conclusions about the
5s of emissions as factors contributing to
jt 4 potential for doing this: receptor models
" I and Henry (1997, 1997a, 1997b)), indicator species
^produced algorithm approach (Blanchard, etal.. (1997,
Ity approach (Cardelino, et al.. (1995)).
potentially useful for assessing whether a proposed strategy is
orientejlioward source1i||||||pWhose emissions appear to be associated with current observed
high^zone. However, thieiMfflity to estimate how much control is needed is limited, unless one
can justify assuming an approximately linear relationship between precursor emissions and
observed 8-hour daily jp&aximum ozone. Thus, observational approaches are ideally suited to
c6n||borate results o|famed for ozone with more quantitative techniques, like air quality models.
LUc^l^flualitvjni^is, observational models are also subject to uncertainties. Thus, results
whiSiii^3blieaous should not be considered at odds with conclusions reached with an air
58
-------
Hgie41.
Qmrt
Qmrt
\fear
I
(b) Cases yieking resits inconBistertHfthattanrrert
BnHon MncHon
QflM
Maer
Camt
\feer
59
-------
Observational models which rely on use of indicator species can be used to show whether
or not ozone may be sensitive to the types of precursors (i.e., VOC or NOx) reduced by a
particular control strategy. Receptor models, like the chemical mass balance approach, may be
useful for confirming whether a strategy is reducing the right sorts of sources. Observational
models can be used to examine days which have not been modeled with an air quality model, as
well as days which have been modeled. The resulting information may be useful for drawing
conclusions about the representativeness of the responses simulated with the air quality model for
a limited sample of days.
Summarizing, if conclusions-drawn with one or
the types of sources to be controlled under a proposed s
with high ozone and/or are those to which observed oz
that the strategy is directionally correct.
Strength of the evidence produced by dbservatior
monitoring network exists and at least some of the moni
measuring pollutants to the degree of sensitivity requires by trii
observational models is more compelling if several|
-------
Identity of additional corroborative analyses is, in part, a function of the available data
base and analytical tools, as well as questions posed by the outcomes of the recommended core
corroboratory analyses. For purposes of illustration, we identify some additional analyses of the
sort which States might consider. None of these is required, and States may well choose to
consider other optional analyses or no optional analyses at all.
Quantifying uncertainty associated with air quality model estimates; Mitrus
guidance, we recommend that "uncertainty" be accounted for using a mc|||led attainment test
which uses models in a "relative" sense and by recognizing that use QJd8lHjP*tory analyses
may be desirable in a weight of evidence determination. TJjjfpve accl^^^gpcertainty in a
qualitative way, without actually estimating it.
States may find it useful to quantify estimates o|
qualitatively in a weight of evidence determination. In
tests which may be useful for this purpose. The first of I
d then ul
xO, we identify
Reynolds, et al.. (1997). This test is to prepare "alternati
reflecting reasonable alternative assumptions about current ei
or better model performance. Note .differences in njojecjed desi
current emissions. A second test is to assume
could reflect using differing growth rates orj|tecementiim!ii&sourc
probable, locations. Note the differencesjffproject
agnostic
which has been proposed by
Remission estimates,
h lead to comparable
nn these alternative
onablelliPvth assumptions. This
different, equally
For the different growth
assumptions. Combinations of the firstJPo tests afi also A third test is one in which a
control strategy under serious considj||lon is sjpulated wMan alternative grid resolution or
with different (reasonable) meteor^^pcal assttiptions. Mr example, due to resource
constraints, it might be necessar^p initiallv^ect a sgalegy using a grid with 12 km grid cells
(horizontal dimension). DiJ|ere|les in proj^g^K^ality obtained with a grid having 4 km
cells could men?be ascerti
Othef
§L, (1996)and
predictions
uncertainty have been described in the literature (Gao, si
of these approaches also assess sensitivity of model
ibles. For outcomes to be most relevant to the way we
recommendinode!
focus onsensitivity of
valuesto the variations i
attainment demonstrations, it is preferable that such procedures
lative reduction factors (RRF) and resulting projected design
or model formulations.
•>,"--
;J Once a range infprojected design values is obtained using tests like those previously
described, a qualitative" assessment can be made of how likely it is that a strategy will lead to
attajrapentjof JKej^AQS. For example, if most of the results lead to projected design values <
84*^^^G^^pbits a conclusion that the strategy, if implemented, will demonstrate attainment.
Choleejofielsts and interpretation of the outputs should be agreed upon beforehand in concert
with the appropriate U.S. EPA Regional Office.
61
-------
Compare monitored design values for the current period used in the test with those
measured hi other periods. The objective of this analysis is to assess whether current design
values used in the modeled attainment and screening tests are atypically high or low due to
natural or meteorological conditions. If the current design values are lower (higher) than normal,
the tests would yield overly optimistic (pessimistic) results.
An analysis of current design values is complicated by trends in emission
betwee
onditio
e curcdit period. A
r example,
current
s does not
emissions
al conditk
*«w
h
i; et al.. (1998) mlgRVPtried to
; terms of its meteorological
^ areawide
in the upper quartile). If
icalllnigh (low), this could be
deled attainment test is too
one would expect design values measured several years previously to
values if there has been an ongoing program to reduce precursor emi
necessarily imply that the current design value is atypicall^iHP The
trends can be addressed by examining statistical relation;
and 8-hour daily maximum concentrations to see whetfo
ozone occurred more or less frequently than usual duri
are analogous to that described by Cox, etal.. (1996) o:
see how the current 3-year period ranks with other 3-ye
ozone forming potential.
States may use results of an assessment of
design value qualitatively to see if the value is
the current monitored areawide design vahn
used to support an argument that the cont^Ptarget
restrictive (not restrictive enough).
Examine Basis for inclu
attainment test This analysis
conditions tajefine the basj
modeled da^Sbeen improj
monitoring
calculations made in the modeled
ir quality and meteorological
SFrelative reduction factors. That is, has a
"*'*»•
rom the calculations at a particular
the attainment test is to consider days with meteorological
when observed 8-hour daily maximum ozone
In Section 3.1, we suggest using model predictions
as the basis fpr«icS(u8®^|QproperlKysv(i.e., days with already low ozone) from the
calculations, 'fit may^^^^Ble to refine choice of days used for a site using available air quality
and meteorological datafjfefte^ample, States may examine days used to calculate the relative
e ||t|x»'iS*,*i>-. " J J
reduction factor to ensurettfieyiTeflect wind orientations corresponding with observed
concentrations exceedinjp84 ppb.
: ^ Recommendations. Optional analyses may be considered in addition to the 3
;:^ommend0|i%nalyses identified in Section 4.1. To use an optional analysis in a
weight of eTidence determination, a State should (1) explain the rationale for the
analysis, (2) identify the data base underlying the analysis, (3) describe the
methodology to be used in applying the analysis, and (4) identify outcomes which
would be consistent with a hypothesis that a proposed control strategy will suffice to
attain the NAAQS.
62
-------
5.0 How Can I Improve Modeling And Other Analyses In Weight Of Evidence
Determinations?
In Section 4.0, we identified a set of analyses which should be considered when
performing a weight of evidence assessment of whether a proposed control strategy will lead to
attainment of the 8-hour ozone NAAQS. To be most credible, modeling and many of the other
analyses rely on presence of good emissions and ambient data bases. Although cpjfflhitment to
undertake subsequent review of a SIP revision is not a prerequisite for
States should anticipate a need at the required time of attainment to cc
indeed been met or to diagnose why not. In this Section,
which provide better support for modeling and other an;
determinations. Resulting improved data bases may im
reasons for attainment or non-attainment of the NAAQ
control strategy, if necessary. We conclude by identify?
could benefit from prior efforts to improve available
the revision,
the NAAQS has
its and activitie
mga
tify
in wei
e reliabi
pro
5.1 What Data Gathering Or Other Efforts
Analyses Or Subsequent Reviews?
Efforts to improve the monitored airjpiaii
improve emission inventory estimates shcpa lead t
including modeling) and improved subjpuent re
monitoring which may prove helpfuJ|||Fe thenjaiefly dis
We conclude by identifying waysjjfpinch thisimprovi
'o Support Current
bases and to update and
ft of evidence analyses (i.e.,
(section, we identify types of
is efforts to improve the inventory.
^formation might ultimately be used.
type of additional monitoring which
3.0 and 4.0. This is to deploy
Deploying additional»air3quality
^^"jSSte&JsjL * "
should be considered has alreao&een
* l£»x.&#&**
additional ozone monitorsnnf ^fions which a screening test, described in Section 3.4, suggests
may have future design va
the NAAQS is being met at locandns^pire the model now consistently predicts concentrations
higher than any near existing monitqqig sites.
Measurement oujapr species" is a potentially useful means for assessing which
precursor category (VOC^n^^x) limits further production of ozone near the monitor's location
at various times of day and under various sets of meteorological conditions (some of which may
nothave been previously considered with an air quality model). Sillman (1998) and Blanchard,
efaL (1997, 1999) identify several sets of indicator species which can be compared to suggest
^s^ *
whethersmonitp|edlJ6zone is limited by availability of VOC or NOx. Comparisons are done by
looking, at ratios of these species. The following appear to.be the most feasible for use in the
fieldiy a regulatory agency: O3/NOy, O3/(NOy - NOx) and O3/HNO3. Generally, high values for
the ratios suggest ozone is limited by availability of NOx emissions. Low values suggest
availability of organic radicals (e.g., attributable to VOC emissions) may be the limiting factor.
For these ratios to be most useful, instruments should be capable of measuring NOy, NOx, NO2
63
-------
and/or HNO3 with high precision (i.e., greater than that often possible with frequently used
"routine" NOx measurements). Thus, realizing the potential of the "indicator species method" as
a tool for model performance evaluation and for diagnosing why observed ozone concentrations
do or do not meet previous expectations may depend on deploying additional monitors. States
should consult the Sillman (1998) and Blanchard, etal. (1997, 1999) references for further details
on measurement requirements and interpretation of observed indicator ratios.
:entially useful
ons for
Receptor models are another class of .observational approaches
for corroborating assumptions made in air quality models or
unexpected air quality observations in a subsequent revie
applications, receptor models require observations of V<
PAMS network. Receptor models work by noting a coi
which best explains speciated air quality observations
or by noting a limited number of species which track e;
variate statistical approach). Both approaches are limi
species. This prevents many distinctive source categori
inconclusive results concerning which source categori
Measuring more species is a potential means for
opportunity for doing this may exist as a result
PM25 monitoring program (U.S. EPA,
species (including some organic particul
collocate monitors collecting gaseous
with collocated gas and aerosol phaja||fganic rfl&uremenjf could increase the power of
receptor models as diagnostic too^^Kxpla^mg obseryjlons in subsequent reviews. In
Section 5.2, we note that "organ^pirbon" of thjgiey components of PM25. This
component is not so well
volatility .ofi&Htne speci
gaseous and^ojol phasdq^^^Mtt a site may lead to a better understanding of sources of
organic particula^&s well
iach)
from day to cfe|plulti
ity of many of the VOC
ratified or leads to
observed air quality.
initation. An
itation plan for the
rces for measuring PM
ations. A State could
sites. Availability of sites
s. This is true, in part, because of the
1 designed studies which measure both
Making measurements alof&rAlmost all measured ambient air quality and
meteorologicafdata ale'coiiected within 20 meters of the earth's surface. However, the modeling
domairugenerally exten(ls'jSOTeral'kilometers above the surface. Further, during certain times of
^ * •^
day (e,;g., at night) surfacenneasurements are not representative of air quality or meteorological
conditions aloft. Concentrations aloft can have marked effects when they are mixed with
ground-level emissionjpurmg daytime. Thus, weight given to modeling results can be increased
if good agreement i&Jnown with air quality measurements aloft. The most important of these
measurements arejaione, NOy, NO, NO2, as well as several relatively stable species like CO and
,* .-,',£• >#£•"* ' ~( ~A,>-^ t "".^t'j^X - '• *
selecteal^p(3|^)ecies. Measurements of SO2 may also be helpful.for identifying presence of
plurae$ifrom large combustion sources. Highest priority should be given to making
measurements near sunrise as well as during midday.
Measurements of altitude, temperature, water vapor, winds and pressure are also useful.
64
-------
Continuous wind measurements, made aloft in several locations, are especially important. They
provide a data base to "nudge" windfields, initially calculated with dynamic meteorological
models, so that these estimates are more consistent with observations. This provides greater
assurance that the air quality model correctly reflects the configuration of sources contributing to
ozone formation. Temperature, pressure and water vapor measurements aloft provide a basis for
assuring that the air quality model accurately reflects vertical exchange and mixing within the
planetary boundary layer. This is a key factor affecting dilution of emissions, as weu"as
atmospheric chemistry.
Collecting locally applicable speciated emission;
a library of default VOC emissions species profiles (U.S
www.epa.gov/ttn/chief/software.htmltfspeciate, some
reflect local sources. Use of speciated emissions data i
balance receptor model as well as to air quality models.
for local sources should thus enhance credibility of seve;
use in a weight of evidence determination.
Projecting emission estimates and comp,
estimates. States addressing traditional nonat]
find it worthwhile to project emissions to
use in two subsequent reviews. The first
2010 for ozone). The second is some i
projection could be useful to help
are inconsistent with earlier e;
projected emission data bases
with an inventory which is
inventory update for 20Q||^SSes availal
be possibl^yoE^gjjectiorisW
attainment
In Secj«
impacts in
-------
to determine whether differences are explained by revised emission estimates, poor choice of
meteorological episodes in the initial analysis or by changes which have occurred in the model
formulation during the intervening years. Insights from such comparisons should help a State
explain why changes in the strategy reflected in its 2003 SIP revision may or may not be
necessary.
5.2 Why Is It Desirable To Plan For A Subsequent Review?
Commitment to undertake activities supporting subs
a prerequisite for approval of the revision. Thus, "why do
about it now?" Subsequent reviews will be needed at th
is required for nonattainment areas. The purpose of sue]
has indeed been met, or to diagnose available informatij
attainment date for traditional nonattainment areas coul!
late as 2010.
IP revision is not
id why worry
>neNAA(
|ht be as
5.2.1 Integration With Attainment StrategieSlFor 1-
A 2010 attainment date for the 8-hour
or "extreme" nonattainment areas for the 1-!
given to attaining the 1-hour NAAQS. AJpRional
may be implemented after the attainmeofilate for
attainment date for areas with the mostiserious
dfffilimg
SIP revision may be due. Furtherfflsre is a 1
the 1-hr NAAQS in "severe" noispainmen
quality data so as to refine ananltal strate
uent revi
u
e attain
j*silsj^
review ispf conrra
determine why no
'classified as "severe"
"r such areas, priority is
o meet the 8-hour NAAQS,
S. Thus, the projected
performed:at:an .interim
NAAQS for ozone-are "i
a strategy for^meefing the 1-1
circa 2005-2007if
ne problems occurs well after 2003, when the
' 'AWthe required time of attainment for
ew and diagnose emissions and air
ig the 8-hour NAAQS. Such a review,
two su
to ensuf^r^pffategies for meeting the 8-hr and 1-hr
That is, an 8-hour strategy builds upon the consequences of
;gS, which are reflected in air quality and emissions data
iews are desirable for nonattainment areas with later
attainment datesi(«5^||afflbetirne w^ehSattainment of the 1-hour NAAQS is required in "severe"
nonattainment areiS^andP^Sbje&time the 8-hour NAAQS must be met). For nonattainment areas
with more immediate attaronirailPdates (e.g., 2005-2007), only one subsequent review is
appropriate within the tini^Eaihe required for attainment. This might occur at the time
attainment is required, f*
js~
fc
5.2.2 Anticipated Modeling Principles For PM2J And Visibility, And Integrating
Ozone Strategies With Goals For PM2 5 And Visibility
The U.S. EPA's policy is to encourage integration of control strategies to reduce ozone
with those designed later to meet NAAQS for PM2 5 and reasonable progress goals to reduce
regional haze. We believe such integration will reduce overall costs of meeting multiple air
quality goals. The desire to integrate strategies meeting air quality goals for ozone, PM2 5 and
66
-------
regional haze is another reason for subsequent review of an ozone SIP revision submitted in 2003
or earlier. Much of the data base used as a basis for later PM2 5 SIP revisions will be collected
during 2000-2002. Thus, the scope of the PM25 problem, if any, will not be fully known at the
time modeling and other analyses must be completed to support the 2003 or earlier SIP revisions
for ozone. We anticipate that SIP revisions for PM2 5 will be due about 2007-2008—about the
same time as a subsequent review of the sufficiency of the previously selected strategy to meet
the 8-hour ozone NAAQS. Periodic review of strategies to improve visibility is aJgjPanticipated
within this time frame.
Guidance for demonstrating attainment of PM2^ NJflpI* and4
reducing regional haze is not yet available (i.e., as of midj|p99). Dura
guidance will be subject to intense review. ConsequentMour
and regional haze could change. Nevertheless, we
to help States develop data bases and capabilities for ccHBfjomt effects of i
2. PM is a mixture of componej
which differ from those for o]
effects of a control strategy
can be assessed by notjSfetHpnet eff<
We raayjiecommeni
-^mass-associate
strategies for ozone, PM2 5 and regional haze in a subseqi
the 8-hour NAAQS for ozone.
1. Emissions and meteorological conditions vj|
annual PM2 5 concentrations should be as
season and using the resulting info;
for VOC, NOx, primary PM2 5, sulfuyfiioxide
of the initial SIP revision for
control strategy on
mean PM2 5 for each
ts. Emission estimates
needed.
ics. Jpch comppient may be attributable to causes
The njideled attamment test should separately estimate
ijor cgggpnents^puie mix. Effect of a strategy on PM2 5
on each major component of the mix.
separate consideration:
-mass«assocaated wi ,y_,_i_
- «\^^fe , . ,
-mass associiated with organic carbon;
jth elemental carbon;
tttrall other species.
;-mass
-mass ass<
3. The recommendeJlTnodeled attainment test for the annual PM25 NAAQS will focus on
monitoring sites will speciated data. Models will be applied in a relative sense to estimate
component- andJSite-specific relative reduction factors. Relative reduction factors will be
used with current speciated design values to estimate future design values. A weight of
- -* '"-,^-";' -i"^-.- -MX-^- '-i-,,4%s'f u a
will be identified as an alternative to using the modeled attainment test
Because the period of record for measurements is much less than that for ozone,
observational models will probably be relatively more important and trend analysis relatively
less so.
67
-------
4. Ambient air quality data should be reviewed to assess whether exceedances of the
concentration specified in the 24-hr NAAQS for PM2 5 is a hot spot problem significantly
influenced by nearby primary emissions, or a problem which is significantly influenced by
more pervasive high concentrations of secondary PM2 5. If the problem is a hot spot problem
and a model performs well predicting primary PM2 5 from the nearby source(s), a modeling
approach similar to that followed for PM10 may be appropriate. If the problem is more
pervasive, with important contributions from secondary components of PM2 c
performance predicting primary PM2 5 is poor, a relative approach sjlgilar tgipne approach for
the annual NAAQS is likely.
5. Visibility attenuation estimates will be obtained fi
previously identified major components of PM2 5.
will estimate relative reduction factors for the maj
used with speciated PM2^ concentrations represen'
in a Class I area to estimate representative future sp^QnHB^centrations of PM25. Current
and future visibility extinction coefficients will thenjjlRRm Easing procedures described
in Sisler (1996). Reasonable progress will be detelmined D| estimates of current
and future extinction coefficients to see whethjsi«anmdentified^^^Sment goal is realized.
-make continuous me
measurements aloft,
as during midda
-collocate suffi
1 ozone
>nito
the PM^ monitoring
~^S««-sl|i»
-improve locatspecia
'*"
Recommendations. The following
informative weight of evidence
-deploy ozone monitors in a
than any predicted near easting mo;
may lead to more
reviews:
fitly predict ozone greater
; aloft and air quality
iorning hours (near sunrise), as well
o measure NOy, NO2, HNO,. and NOx at
isitive i
sites;
gaseous organic species with selected monitors in
irkfused to estimate particulate organic species;
, 0C emission data bases;
-retain meteorological, current and projected emission files as well as output files
„• used in modeling the strategy reflected in the initial SIP revision for possible
-•^7 ^,-jf <&*£--% o»< M.
future diagnostfotests with newer data bases and/or models.
r *
A State need not include plans for a subsequent review of its strategy demonstrating
attainment of the 8-hour NAAQS for ozone as part of its initial SIP revision.
However, a subsequent review will be needed at the time of required attainment to
ascertain whether attainment has occurred. States with a protracted attainment date
for the 8-hour NAAQS may also wish to consider a subsequent review at the time the
1-hour NAAQS should be met (e.g, 2005-2007). Subsequent reviews may be helpful for
"integrating" strategies to meet the 8-hour ozone NAAQS with those for meeting the 1-
hour NAAQS and with those addressing air quality goals for PM2S and regional haze.
68
-------
6.0 What Documentation Do I Need To Support My Attainment Demonstration?
States should follow guidance on reporting requirements for attainment demonstrations in
U.S. EPA (1994). The first 7 subjects in Table 6.1 are similar to those in the 1994 guidance.
The 1994 guidance envisions an air quality model as the sole means for demonstrating
attainment. However, the current guidance (i.e., this document) identifies a weight of evidence
determination as a means for corroborating the modeled attainment test in an attainment
demonstration. In addition, feedback received since the earlier guidancelb emasized the
need for technical review of procedures used to identify a sufficient c
have added two additional subject areas which should be ingped in
accompanying an attainment demonstration. These are ajplcription
determination, and identification of reviews to which analyses use
demonstration have been subject.
shown hi Table 6.1 in the
Jhe documentation
in the table. More
Recommendations. States should address the 9
documentation accompanying an attainment d
should contain a summary section which add
detailed information should be included i
** v
I -t'-ar**-
69
-------
Table 6.1. Recommended Documentation for Demonstrating Attainment of the 8-hour
NAAQS for Ozone
Subject Area
Purpose of Documentation
Issues Included
Modeling/Analysis Protocol
Communicate scope of the
analysis and document
stakeholder involvement
Names of stakeholders
participating in preparing and
j(BS^*
implementing theprotocol;
performed;
each type of
Emissions Preparations and
Results
Assurance of valid, coi
emissions data base. Appf
procedures are used to i
emission estimates needgpbr i
quality modeling.
Data base used anl
irance methods applied;
recessing used to convert
to model-compatible
ions from existing
ince and underlying
tionale;
VOC, NOx, CO emissions by
State/county for major source
categories.
Air Quality/Meteorology
Preparations and Results
rotative air
logical
iuts are used in analyses
Extent of data base and
procedures used to derive &
quality assure inputs for analyses
used in the weight of evidence
determination;
Departures from guidance and
their underlying rationale.
Performance of meteorological
model if used to generate
meteorological inputs to the air
quality model.
70
-------
Table 6.1. Recommended Documentation for Demonstrating Attainment of the 8-hour
NAAQS for Ozone (continued)
Subject Area
Purpose of Documentation
Issues Included
Performance Evaluation for Air
Quality Model (and Other
Analyses)
Show decision makers and the
public how well the model (or
other analyses) reproduced
observations or otherwise
performed on the days
for analysis
Summary of observational data
base available foroftmparison;
J? ^Ir*
Identii
performance ev;
implies.
Diagnostic Tests
Ensure rationale used to
model inputs or to disco;
certain results is physically
justified and the
results make
Its from application prior to
nts;
with scientific
Iding and expectations;
performed, changes made
accompanying justification;
Short summary of final
predictions.
71
-------
Table 6.1. Recommended Documentation for Demonstrating Attainment of the 8-hour
NAAQS for Ozone (continued)
Subject Area
Purpose of Documentation
Issues Included
Description of the Strategy
Demonstrating Attainment
Provide the EPA and the public
an overview of the plan selected
in the attainment demonstration.
reductions and o
Qualitative description of the
attainment stratee
C, NOx, and/or
Tom each major
for each
nm current
ructions;
predicted 8-hr site-specific
.design values for the
ntrol scenario and
location which fails
ing test described in
[3.4;
lentification of authority for
implementing emission
reductions in the attainment
strategy.
Evidence that emissions remain
at or below projected levels
throughout the 3-year period
used to determine future
attainment.
Data Access
Enable the EPA or other
interested parties to replicate
model performance and
attainment simulation results, as
well as results obtained with
other analyses.
Assurance that data files are
archived and that provision has
been made to maintain them;
Technical procedures for
accessing input and output files;
Identify computer on which files
were generated and can be read,
as well as software necessary to
process model outputs;
Identification of contact person,
means for downloading files and
administrative procedures which
need to be satisfied to access the
files.
72
-------
Table 6.1. Recommended Documentation for Demonstrating Attainment of the 8-hour
NAAQS for Ozone (concluded)
Subject Area
Purpose of Documentation
Issues Included
Weight of Evidence
Determination
Assure the EPA and the public
that the strategy meets applicable
attainment tests and is likely to
produce attainment of the
NAAQS within the required tij
Description of the modeled
attainment test andsobservational
data base used; „'
Outcome of each
including the modeled attainment
at of the credibility
iiwith each type of
i this application;
itive describing process
to conclude the overall
'weight of available evidence
supports a hypothesis that the
selected strategy is adequate to
attain the NAAQS.
Review Procedures Used
eEPA
land the pubBc;fluiiBnalyses
terformed in the attainment
^demonstration reflect sound
ipractice
Scope of technical review
performed by those implementing
the protocol;
Assurance that methods used for
analysis were peer reviewed by
outside experts;
Conclusions reached in the
reviews and the response thereto.
73
-------
-------
7.0 References Cited In Part I And In Section 1.0
Blanchard, C.L. and P.M.Roth, (1997), User's Guide Ozone M.A.P.P.E.R., Measurement-based
Analysis of preferences in PJanned Emission Reductions, Version 1.1, Final Report
prepared for the U.S. EPA pursuant to Contract 68-D6-0064 Work Assignment,
W.M.Cox, EPA Work Assignment Manager.
Blanchard, C.L., F.W.Lurmann, P.M.Roth, H.E.Jeffries and M.Korc,
Ambient Data to Corroborate Analyses of Ozone <
Environment, 33, pp. 369-381.
Cardelino, C.A. and W.L.Chameides, (1995), "An Ob;
Ozone Precursor Relationships in the Urban A
pp.161-181.
Clinton, W.J., (July 16, 1997), Memorandum to the A<
>c 45,
Protection Agency, Subject: "Implementation
Ozone and Paniculate Matter".
Cox, W.M. and S.Chu, (1993), "Meteorolo,
Probabilistic Approach", AtmosphjlKc Envin
Cox, W.M. and S.Chu, (1996), "As:
from a Climatological Peri
Deuel, H.P. and S.G.DougL
Visibility and Acii
e Environmental
Standards for
in Urban Areas: A
.p.425-434.
;rannuaet)zone Variation in Urban Areas
"ivironment 30, pp.2615-2625.
for the Integrated Analysis of Ozone,
ion for f^S^Kern Appalachian Mountains, Draft Technical
Report-Systems App1i
-------
Henry, R.C., (1997b), "Receptor Modeling Applied to Patterns in Space (RMAPS) Part ffl.
Apportionment of Airborne Paniculate Sulfur in Western Washington State", J.AWMA
47, p.226.
Lefohn, A.S., D.S. Shadwick and S.D. Ziman, (1998), "The Difficult Challenge of Attaining
EPA's New Ozone Standard", Environmental Science and Technology, Policy Analysis,
32, No. 11, (June 1,1998), pp.276A-282A.
Meyer, E.L., K.W.Baldridge, S.Chu and W.M.Cox, (1997), Choice oj
Considering Effects of Control Strategies on
Days", Paper 97-MPU2.01, Presented at 97th
Ontario, (June 1997).
Pacific Environmental Services, Inc., (1997), Draft Tec
Ozone Values: 1980-1995, Report prepared undo
Assignment ni-88, Edwin L. Meyer, Jr., Work
to Model:
ive Episode
norandum
No.68D30032, Work
Reynolds, S.D., H.M.Michaels, P.M.Roth, T.W.Te^h^.McNaI^^^gner and
G.Yarwood, (1997), "Alternative Their
Construction, Role and Value", Ingress.
Sillman, S., (1997), "The Method of Ph^chemicJ5Tndica^^KBasis for Analyzing O3-NOx-
ROG Sensitivity", NARSTOjjgSical reyjiw paper, Jpfbe published in Atmospheric
Environment.
Sillman, S., (1998), Evaluating jtie RelatioiSjj^^ijizone, NO, and Hydrocarbons: The
Sisler, J.F. (Edifor^|1996), Spatialand Seasonal Patterns and Long Term Variability of the
.r«y *Mfc.. Haze^nj^te:Jjfnited States: An Analysis of Data from the IMPROVE
Instila^ior Research in the Atmosphere Report ISSN: 0737-5352-
32, Colorado^jffi33mversity, Fort Collins, CO.
U.S. EPA, (1993), Volattte^jfganic Compound (VOC)/F'articulate Matter (PM) Speciation Data
System (SPECIAXE),Version 1.5, EPA/C-93-013.
', '.. .$£
U.S. EPA, (1994), Guidance on Urban Airshed Model (UAM) Reporting Requirements for
"j! Attainmenii^bemonstration, EPA-454/R-93-056.
U.S. EPA, (1996), Guidance on Use of Modeled Results to Demonstrate Attainment of the
Ozone NAAQS, EPA-454/B-95-007.
76
-------
U.S. EPA, (1998a), EPA Third-Generation Air Quality Modeling System, Models-3 Volume 9b
Users Manual, EPA-600/R-98/069(b).
U.S. EPA, (1998b), Implementation Plan PM25 Monitoring Program, Available from EPA
Internet website, "http://www.epa.gov/ttn/amtic/files/ambient/pm25/pmplan3.pdf'.
U.S. EPA, (1999a), Implementation Guidance for the Revised Ozone and PaniculdjjPMatter
(PM) National Ambient Air Quality Standards (NAAQS) and Rejigmal l&te Program.
Watson, J.G., (1997), "Observational Models: The Use
Prepared for the U.S. EPA Source Attribution W
July 16-18,1997. Available from EPA Internet
"http://www.epa.gov/ttn/faca/stissu.htmF' and
workshop materials".
Yang, Y.J., W.R. Stockwell, J.B.Milford, (1995),-"Uno
of Volatile Organic Compounds", Environmen
pp.1336-1345.
of
jmental Reactivities
tology 29,
77
-------
-------
-------
X
-------
8.0 How Do I Apply Air Quality Models?- An Overview
In Part I of this guidance, we described how to estimate whether a proposed control
strategy will lead to attainment of the ozone NAAQS within a required time frame. We noted
that air quality models play a major role in making this determination. We assumed that
modeling had been completed, and discussed how to use the information produced. We now
focus on how to apply models to generate the information used in the modeled attamrnent
demonstration. The procedure we recommend consists of 8 steps:
1. formulate a conceptual description of an area's non
2. develop a modeling/analysis protocol;
3. select an appropriate air quality model to use;
4. select appropriate meteorological episodes to m
5. choose a modeling domain with an appropriate
sized grid cells;
6. generate meteorological and air quality inputs to
7. generate emissions inputs to the air quality mo
8. evaluate performance of the air quality mo
-------
methods described in Section 4.0 (e.g., trend analysis, observational models) may be used. Other
times, these types of analyses may be deferred until after a team is in place to develop and
implement steps following a modeling/analysis protocol. The initial conceptual picture may be
based on less resource intensive analyses of available data.
2. Develop a modeling/analysis protocol. A protocol describes how modeling will be
performed to support a particular attainment demonstration. Its direction and
stakeholders are influenced by the previously developed conceptual desc
be resolved. The protocol outlines methods and procedures
subsequent 6 steps needed to generate the modeling result
modeled attainment and screening tests as well as other cJpSboratingj
evidence determination. It does this by: a) identifying i
modeling, b) identifying those who will review each steBlas it ocjgfrs, c) identil
atmg
ie problem to
perform the
to
be used to consider input/suggestions from those potent
"stakeholders"), and d) outlining how decisions will be
needed to complete each step in the modeling procedu
plan" and the "rules of the game".
outcof
earning technical analyses
arotocol defines the "game
3. Select an appropriate model for use.
insight about the nature of a nonattainment
the Guideline for Air Quality Models (U
of those performing the modeling. Ide:
the process, since it may affect howj
model. It could also affect size ot
resolution considered.
4. Select appropriate
r quality data to gain
ing rules established in
ring experience/expertise
to be used is an early step in
sgical information are input to the
Dice of the horizontal/vertical
jical episodesScfmodel. Like the preceding step, this step
data. K also requires a thorough understanding of the
^standard and of the modeled attainment test described in
of meteorological conditions which have been
form of the nationalambieni
•
Sections 3.1
observed to accompaawlffiionitored fclteeedances of the concentration specified in the NAAQS
, ; - '/^'.^^^-.jfs&.'jijgg^^,.. *•%' ^ ^
(i.e., > 85 ppb). Throl^ec£!cf|hese reviews is to select episodes which a) include days with
observed concentrationsucldi^SJsite-specific design values so that all sites with current design
';•• , V-' •?«&''• id o
values^ 75 ppb can be considered in the modeled attainment test, and b) reflect a variety of
meteorological conditions which have been commonly observed to accompany monitored
exceedances. This lattt;! objective is desirable, because it adds confidence that a proposed
strategy will work under a variety of conditions.
5. Choose a modeling domain with an appropriate number of vertical layers and
appropriately sized grid cells. Appropriate domain size is influenced by the choice of episodes
modeled. Meteorological and air quality (i.e., ozone) data corresponding to these episodes and, if
applicable, to other, plausible episodes, need to be reviewed prior to choosing size of the area
modeled. Presence of topographical features or mesoscale meteorological features (e.g., land/sea
82
-------
breeze) near or in the nonattainment area of principal interest are factors to consider in choosing
size of individual grid cells and the number of required vertical layers for that portion of the
modeling grid. Another factor affecting choice of grid cell size is the available spatial detail in
the emissions data used as input to an emissions model. Finally, feasibility of managing large
data bases and resources needed to estimate meteorological inputs and air quality in many grid
cells are factors which cannot be ignored in choosing size of a domain and its grid cells.
6. Generate meteorological and air quality inputs to the air quality^
Unlike emissions, meteorological inputs remain constant durjag "base
"future" periods simulated with the air quality model. Nej
specifying these, as they may affect relationships predict
modeling may have to consider large geographical area
modeling has shown that meteorological conditions ale
predicted ozone. Finally, meteorological monitoring is!
especially, aloft. Thus, we recommend that meteorologi
meteorological inputs. Application of meteorological
resolution in the preceding step are closely related. Mi
less cr
stween ozj
many instances.
ha\»Sn important
model,
nt" and
be taken in
issions.
si
Ozi
sparse outside o
and,
which is the focus of an attainment demonstration
the other hand, cost and data management diffj
Thus, those implementing the protocol will
cost/feasibility of running air quality andjniteorola
be most desirable to treat dispersion of jfiarby emissions.
ordinarily be used to generate
|ce of model grid
itions near the area
ictate theTCqunicauspatial resolution. On
jrolol
finely resolved grids.
5ff between
ssolution at which it might
Air quality inputs consist
domain. Importance of initial c
time prior toythe period whj
in decidingliow large tojnf!
generateluturelxmndaiy
those impiemenmigthe proto
r ,f .-irA^vK., * ^$&*I *
of data base management vs. a cl
the modeling exerci
ial coalitions and&undary conditions to the model
tions be diminished by beginning a simulation at a
if interes^^^to^Koundary conditions is an important factor
.size of tio^WBambdeled. The most satisfactory way to
'B^.'ttJs-.-SffKjSK » »
is through use of a regional air quality model. Therefore,
Lce again be faced with a tradeoff between cost/feasibility
imit the importance of an arbitrarily specified input to
7. Generate emissions inputsio the air quality simulation model. Emissions are the central
focus in a modeled attainmehiilemonstration. That is, they are the only input to an air quality
model which those implementing the protocol can control. Hence, they are the major input
which gets changed between the present and future. Emissions which are input to an air quality
model are generatec|*6sing an emissions model. Applying such a model is as complicated as the
air quality modeWtself, and demands at least as much attention. In current emissions models,
emissions fronisome of the major source categories of ozone precursors are affected by
meteorological conditions. This requires an interface between meteorological inputs and
emissions. Emissions which are input to the air quality model are also affected by the latter's
horizontal/vertical resolution and, of course, the size of the area modeled. In short, treatment of
emissions is a central and complex one which, itself, involves several steps. These include
83
-------
deriving emission inventories, quality assuring results, applying results in an emission model(s),
and (again) quality assuring results. Emission inputs may be needed for as many as 3 periods: (1)
a "base case period" corresponding to that of the selected episodes, (2) a "current period",
corresponding to that represented by the current monitored design value, and (3) a future period,
corresponding to a time two years prior to the required attainment date.
8. Evaluate performance of the air quality simulation model and perform diagnostic tests.
To an important extent, credibility of a modeled attainment test's resultsimd otJiSsr modeled
outputs is affected by how well the model replicates observeckair qualj^H&||jiiating model
performance and conducting diagnostic tests depend on prij
and specification of model inputs. Hence, this is general]
to support an attainment demonstration, as described in i
In the past, performance evaluation has relied all
comparing predicted and observed ozone, or visual ins
These are still important tools. However, photochemics
possible to get similar predicted ozone concentrations'
inputs. There is no guarantee that ozone will resp
different combinations of inputs. Thus, we plaoi
than was true in past guidance. These inclujjpile or
species, use of corroborative analyses wi^plbservj
analyses.
finite
ie last stei
I.
jlusively on numel
;ling exercise
; the mode
sts
ithdif
same wa
Diagnostic tests are separa
of a model's ozone predictions t|
purposes, including selectic
quality assurance and ass
""""'< "^
such tests^ States should •
recommended infection 3.C
edictions and observations.
|have many inputs, and it is
jinations of these
pels with these
basis cMffltional kinds of tests
pbseraanons, use of indicator
fflnase of retrospective
»mulatio|Sr which acpiperformed to determine the sensitivity
ious to the^aodel. This can be done for a variety of
fective c^^^^lSpcgies, prioritizing inputs needing greatest
acertaint^aBBipefflied with model predictions. In performing
low modef results are used in the modeled attainment test
w||ral, model results are used in a relative rather than absolute
sense. In partiaalaf|^tiSe:modeleXiafl^Minent test requires use of relative reduction factors (RRF),
generated by mod^.%TKu§, diagnos^ctests should be used to consider how RRF, as well as
absolute ozone prediction^are affected by changes to model inputs.
RecommendationsifeStates should follow eight steps in applying models to generate
information required for use in modeled attainment demonstrations.
1. Formulate a conceptual description of an area's nonattainment problem.
2. Develop a modeling/analysis protocol.
3. Choose an appropriate model.
4. Choose appropriate episodes.
5. Choose a modeling domain with an appropriate number of vertical layers and
appropriately sized grid cells.
6 .Generate appropriate meteorological and air quality inputs.
84
-------
7. Generate quality assured emissions inputs.
8. Evaluate model performance and undertake diagnostic tests.
Execution of subsequent steps should be performed in accordance with procedures
identified in the protocol. Rationale and outcome of the steps should be documented as
described in Section 6.0. To increase the likelihood of an approvable demonstration,
States should carefully coordinate development and execution of steps wit&the
appropriate U.S. EPA Regional Office(s).
'. 4
" ,
85
-------
\
-------
9.0 How Do I Get Started?
A State should start developing information to support a modeled attainment
demonstration by assembling and reviewing available air quality, emissions and meteorological
data. Current design values should be calculated at each ozone monitoring site, as described in
Section 3.1. If past modeling has been performed, the emission scenarios examined and air
quality predictions may also be useful. Readily available information should be ujpH>y a State
to develop an initial conceptual description of the nonattainment problejfei theprea which is the
focus of a modeled attainment demonstration. A conceptualj^escriptip^Bli^mmental for
identifying potential stakeholders and for developing a i
influence a State's choice of air quality model, modelingJjHnam,
quality assuring and refining emissions estimates and chjlce of initi
potentially effective control strategies. In general, a coJIttrtual ddlcription is
State to identify priorities and allocate resources in perf^n^l^iodeled attainrri
demonstration.
ing a
In this Section, we identify key parts of a conc^
examples of analyses which could be used to descrj|
analyses may be complemented later by additic
protocol, and that many of the analyses we
data bases.
sh of the!
jrforme
lore cj
|We then present
note that initial
3se implementing the
/incing with improved
9.1 What Is A "Conceptus
scriptk
A "conceptual descriptions a qua^pve wavgrcharacterizing the nature of an area's
nonattainment problem. key components of a description.
Examples .aielisted belp^^^^pxamples^^^^Kessarily comprehensive. There could be
other featuresof an area's*^i|^fevhich are important in particular cases. For purposes of
illustrationiatetjafethe discusmffl^irbave answered each of the questions posed below. Our
. . f-,Jr,j,;j,^i. '^» ^afe,s»vAA>, ^ ~
responses appear^if^npntheses;^
.-•-i5l»fj^g|jS?%,. V1
-1. Is the nonattaraf^ppjblem primarily a local one, or are regional factors important?
"fe.
(Surface me|IuKfipQents suggest transport of ozone close to 84 ppb is likely. There
are some other nonattainment areas not too far distant)
*?;,
4
-2. Are ozone and/or precursor concentrations aloft also high?
are no such measurements.)
-3. Do violations of the NAAQS occur at several monitoring sites throughout the
nonattainment area, or are they confined to one or a small number of sites in proximity to
one another?
87
-------
(Violations occur at a limited number of sites, located throughout the area.)
-4. Do occasions in which observed 8-hour daily maximum ozone concentrations exceed 84
ppb occur often or just on a few occasions?
(This differs for different monitors from 4 times up to 12 times per year.,)
-5. When 8-hour daily maxima in excess of 84 ppb occur, is there aj||ccomjianying
characteristic spatial pattern of these, of is there a variei
(A variety of patterns is seen.)
-6. Do monitored violations occur at locations subj|
coastline) which may differ from the general wind
(No.)
-7. Have there been any recent major changes^
nonattainment area? What?
(Yes, 4 measures believed to
implemented in the last 5
Dx in or near the
VOC have been
-8. Are there discernible treni
accompanied a change in ernjssions?
4
;r air quality indicators which have
'out 10% at 4 sites, smaller or no reduction
-9. Is therejm^'ajjparent spatlatpatton to the trends in design values?
dt). Have ambient plreqirsbr concentrations or measured VOC species profiles changed?
^V,v *"
"•' %&. -
(There are.no measurements.)
%. What pasferhodelmg has been performed and what do the results suggest?
i^4?i*2?y*l|K'-
*6S;t;-:; (A regional modeling analysis has been performed. Two emission scenarios were
modeled: current emissions and a substantial reduction in NOx emissions
throughout the regional domain. Reduced NOx emissions led to substantial
predicted reductions in 8-hour daily maximum ozone in most locations, but changes
88
-------
near the most built-up area in the nonattainment area in question were small or
nonexistent.)
-12. Are there any distinctive meteorological measurements at the surface or aloft which
appear to coincide with occasions with 8-hour daily maxima greater than 84 ppb?
(Other than routine soundings taken twice per day, there are no measi
There is no obvious correspondence with meteorological
daily maximum temperatures are always ^ 85F qa these da
Using responses to the preceding questions in thi,
initial conceptual description of the nonattainment area'^
questions 1 and 11 suggest there is a significant region;
problem. Second, responses to questions 3,4,7 and 8 f
component to the area's nonattainment problem. The res'
that high ozone concentrations may be observed under
The responses to questions 7, 8 and 11 suggest that ozone in
may be responsive to both VOC and NOx controls^
spatially. The response to question 6 suggests'
using a model with 12 km grid cells rather thfflFlTneT
finer resolution on a limited basis.
lents aloft.
Fs other than
construct i
ses
ment
izone proj
iponjif to the areiP
Jere is an impor
juestions 4, 5 and 12 indicate
leteorological conditions.
lonattainment area
at the ex!Qggg&tresponse may vary
> develop a strategy
and latlr check its adequacy for
The preceding conceptual
area in this example will need to
implement a modeling/analysis
analysis wilLbe needed to
of epis
don it
vests
col. It
ie pro!
idition
lies that til State containing the nonattainment
Elders froprother, nearby States to develop and
suggesliFthat a nested regional modeling
r, it may be necessary to model several
emissions sepatStelysand in t
*-,---, -.*~j^.&Bri*"
will be needed to select episodes. Finally,
be needed to assess effects of reducing VOC and NOx
It should be;cl§ar|from the
•"""' • -if~^-^. iff*«$/%• 3ft *'•
ling example that the initial conceptual description of an
area's nonattairinTient^)ble^^ay draw on readily available information and need not be
detailed?:*It is intended t^p^Siiunch development and implementation of a modeling/analysis
protociBl in a productive idl^tion. It will likely be supplemented by subsequent, more extensive
modeling and ambient analyses performed by or for those implementing the modeling/analysis
protocol discussed in Section 10.0.
Recommendations. States should begin an analysis to support a modeled attainment
; ?ll|^^O]pis^Btibn by developing a conceptual description of an area's nonattainment
problem. This description is based on use of readily available air quality,
meteorological and emissions information. It may be refined later as additional
analyses are performed by those implementing the modeling/analysis protocol.
89
-------
9.2 What Sorts Of Analyses Might Be Useful For Developing And Refining A
Conceptual Description?
Questions like those posed in Section 9.1 can be addressed using a variety of analyses
ranging in complexity from an inspection of air quality data to sophisticated mathematical
analyses. We anticipate the simpler analyses will often be used to develop the initial conceptual
description. These will be followed by more complex approaches or by approacheplequiring
more extensive data bases as the need later becomes apparent. In the fojtawing||)aragraphs, we
revisit key parts of the conceptual description identified in Section analyses which
may help to develop a description of each part. The list seg^^s an It is not
necessarily exhaustive.
1. Is regional transport an important factor
-Are there other nonattainment areas within a day's trans
-Do "upwind" 8-hour daily maximum ozone concentrairons ap
or all of the days with observed 8-hour daily maxira^^^ ppb i
-Are there major sources of emissions
-What is the size of the downwind/upwj
concentrations compared to the upw^iKPvalues?,
-Do ozone concentrations aloft bfifewithm
l^g^"
ppb at night .or in the momingJhcrars prior tola;
nonattainment area?
ceed 84 ppb on some
inment area?
iy maximum ozone
:>undary layer approach or exceed 84
" the nocturnal surface inversion?
-Is there a significant posl^ffeSpelation between observed 8-hour daily maximum ozone
e- -, :»..-„,,- " y%§f«||(|>-'S- J
concentrationsatfinost monitbnngfslfes within or near the nonattainment area?
, "-,"/'"f:- r&j'^t-"- ^^Tt-^-.X^V -r
^^jf^ N^tr^
-Is timing of high^o]5s«n'ea.>ozone consistent with impacts estimated from upwind areas using
trajectory models?
** ** -i*i««:^',^£t •'i'•': '»?f'
.... ~,: . .-**;.•?*
-Examine spatial patterns of 8-hour daily maxima occurring on each day for which a value > 84
ppb occurs to try to identify a limited number of distinctive patterns.
-Review synoptic weather charts for days having observed concentrations > 84 ppb to identify
classes of synoptic scale features corresponding to high observed ozone.
90
-------
-Perform statistical analyses between ozone or 8-hour daily maximum ozone and meteorological
measurements at the surface and aloft to identify distinctive classes of days corresponding with
observed daily maxima > 84 ppb.
-Apply indicator species methods such as those described by Sillman (1998) and Blanchard, etal.
(1999) at sites with appropriate measurements on days with ozone exceedances. Identify classes
of days where further ozone formation appears limited by available NOx vs. classeffbf days
where further ozone formation appears limited by available VOC.
3. Is ozone limited by availability of VOC, NOx
sorts of source categories may be important?
-Apply receptor modeling approa
(1994) and Henry (1997, 1997a,JJp97b) to
VOC on days with high observeftozone.
not high?
-What are the major source categories of VOC and NO
the most recent inventory?
-Review results from past modeling analyses to assess
nonattainment area will be more responsive to VOC o
different locations?
-Apply indicator species methods at sites
maximum observed ozone at these sites i;
conclusions differ for different days?
tat ozone in the
conclusions vary for
to assess whether
f VOC or NOx. Do
lose described by Watson (1997), Henry, etal.
'• categories contributing to ambient
differ on days when measured ozone is
g exanipfe^e jisted analyses to help describe three major components of
so thatthejeasestanalysis, relying on available data, appears first. The
to perform; the more complete and more accurate the description of
In the
a conceptual
more analyses
an area's nonattainmerif|)i^ob][em may be. As noted in Section 5.0, the most complete description
will depend on use of resGnba^Sa bases which supplement routinely collected data. For
example, statistical models%etween meteorological variables and observed ozone will probably
better describe relationships if meteorological measurements are available from aloft.
£ •
,«';-'
Some of the analyses may be identified as desirable as issues arise in implementing a
modeling/analysis|protocol. Their function is to channel resources available to support modeled
attammenideipphstrations onto the most productive paths possible. They also provide other
pieces of information which can be used to reinforce conclusions reached with an air quality
model, or cause a reassessment of assumptions made previously in applying the model. As noted
in Section 4.0, corroboratory analyses may also be used in a weight of evidence determination to
help assess whether a simulated control strategy is sufficient to meet the NAAQS.
91
-------
Recommendations. States should analyze ambient air quality, meteorological and
emissions data in concert with an air quality modeling analysis. These analyses
perform at least 3 functions. First, they are needed to help develop a conceptual
description of a nonattainment area's problem. Second, they help guide application of
a model hi an ah* quality modeling analysis. Third, analysis of air quality,
meteorological and emissions data generates corroborative information vpich may
confirm conclusions drawn with an air quality model or cause §|me ojpfe underlying
assumptions in the modeling to be reexamined.
92
-------
10.0 What Does A Modeling/Analysis Protocol Do, And What Does Developing One Entail?
Developing and implementing a modeling/analysis protocol is a very important part of an
acceptable modeled attainment demonstration. Much of the information in U.S. EPA (1991)
regarding modeling protocols remains applicable. States should review the 1991 guidance on
protocols. In this document, we have revised the name of the protocol to "Modeling/Analysis
Protocol" to emphasize that the protocol needs to address all types of analyses cor^Bered in a
weight of evidence determination, not just modeling.
10.1 What Is The Protocol's Function?
The most important function of a protocol is to:
communicating how a modeled attainment demonstrati|
The protocol is the means by which States and other sta
default recommendations described herein and develop
to widespread participation in developing the demonst
spending time and resources on efforts which the appr
believes are unproductive or inconsistent with Agej
The protocol also serves several im]
who will be helping the State or local airajlfflity agi
or evaluate analyses needed to support apefensibl|^emotf
Second, it identifies how communicant! will occur amon
various issues. Third, it identifie;
Fourth, the protocol describes
Fifth, it describes how ch;
upon and communica
Major steps taken in imp!
EPA Regional Ofl^s) as
decisions are madelconcerning
\ -A 1•>- . *^
ire as a:
/ill bej
for!
formed be
"can assess appli*S®lify of
A good protocol should lead
also reduce risk of
sgional Office(s)
ns. First, it identifies
lead agency) to undertake
i.e., the stakeholders).
eholde
akeholders to develop consensus on
used to support the demonstration.
:o key steps in the demonstration.
or in the protocol itself are agreed
ppropriate U.S. EPA Regional Office(s).
protocol should be discussed with the appropriate U.S.
decided. States should update the protocol as major
,g analyses.
10.2 What Subjects Should Be Addressed In The Protocol?
States should addreifple following subjects in their modeling/analysis protocol:
1. Stakeholders participating in the process.
,
2. Management/communication procedures used, including those to amend the protocol.
• " -- :',tg:..'.- .'-ia .".«*'•
3. Choice of the air quality simulation model to be used and how it meets requirements in
40CFR51, Appendix W for using "alternative" models.
4. Assurance that proposed modeling procedures have been scientifically peer reviewed and plans
93
-------
for technical review of how procedures are used in the specific application and the resulting
outputs.
5. Types of analyses included in the weight of evidence determination, if used.
6. Outcomes for each analysis which will be considered consistent with suggesting a
selected strategy will meet the NAAQS.
7. Data base used to support air quality modeling and other
evidence determination.
8. Rationale for choice of air quality and emissions mi
meteorological inputs
9. Methods used to quality assure emissions inputs
10. Domain size and spatial resolution to be used.
11. Criteria/goals in selecting periods to mode
12. Performance evaluation procedures
13. Outcomes in the modeled attain
used in a broader weight of evidei
14. Procedur.es to be used t
15. Identification of spec!
Regional Office. '- *
.ecting episodes.
ts planned.
well as results of analyses to be
report results.
jverables anoschedule for delivery to the appropriate U.S. EPA
Recommendations. States should prepare a modeling/analysis protocol as part of an
acceptable demonstration of attainment. Generally, procedures recommended in the
1991 guidance andlfollowed for the 1994 ozone SIP revisions are appropriate. These
procedures should be augmented to include a discussion of all analyses to be included
in the weight of evidence determination, not just modeling. The protocol should also
include provision for review of key parts of the analysis and data base underlying the
attainment demonstration. The protocol should be kept up to date to reflect major
changes inJnitial plans.
94
-------
11.0 What Should I Consider In Choosing An Air Quality Model?
Photochemical grid models are, in reality, modeling systems in which an emissions
model, a meteorological model and an air chemistry/deposition model are applied. In this
guidance, we use the term "air quality model" to mean a gridded photochemical modeling
system. Some modeling systems are modular, at least in theory. This means that it is possible to
substitute alternative emissions or meteorological models within the modeling sys|ph. Often
however, choice of an emissions or meteorological model or their featuc^s neatly influenced
by the chosen air quality model (i.e., an effort is needed to develop sofiiHgpRnterface
combinations of components differing from the modeling sfiipn's de^|^^||bjnation). Thus^
choice of an appropriate air quality model is among the i
implementing the protocol. In this section, we identify;
quality model should meet to qualify for use in an attaij
NAAQS. We then identify several factors which will hl|
quality models for a specific application. We conclude
quality models which are available for use in attainment
emissions models are discussed in Sections 14.0 and 11
;st decisic
bt of gene|pr requ!
it demgnstration fo!
tie by those
"~
Jzone
josing among quaatprng air
by identifying several air
is. Meteorological and
rest
11.1 What Prerequisites Should An
An Attainment Demonstration?
fodel
• Qualify For Use In
A model should meet several gejpral criteriE for itTOgpSandidate for consideration in
an attainment demonstration. Thesejjppleral crilpria are cojpistent with requirements in 40CFR
Part 51, Appendix W (i.e., the "A^^^uidehrte") to be opposed in 1999. Note that, unlike in
previous guidance (U.S. EPA, 19£f$, we arjgiDt recoromending a specific model for use in the
attainment demonstration fc«h&-hour At present, there is no single model
I'll v' •'• • ^^^Plp^l^L . ^S^^^^^^S^^^^
which hasibeei|extensive}^t^^i.and showQgo|^€learly superior or easier to use than several
alternativeSi^ffius, at not anticipate that the next revision to 40CFR Part 51
" for use in attainment demonstrations of the 8-hour
Part 51 Appendix W, States should consider
scale air quality models as "alternative models" for
Appendix W|entify
NAAQS for
nested regionafeaiiiodels o
ozone.
an
4 The U.S. EPA ha|Ins«sted considerable effort to develop a nested regional model
(CMAQ) within a modeling system called "MODELS3" (U.S. EPA, 1998a). The U.S. EPA will
provide support, in the^form of documentation, user's guides, computer codes, updates, training
andiimited troubleshooting for the CMAQ model. The CMAQ model is designed to address
ozonq/PM2 5 andjcegional haze-related applications. However, this model has not, as yet, been
showfi/to»'^-iciearfy superior or easier to use than available.alternatives. Thus, use of the CMAQ
model is subject to the same review criteria as other "alternative models" proposed to support an
attainment demonstration of the 8-hour ozone NAAQS.
"Alternative models" may be used if they are non-proprietary. A "non-proprietary"
95
-------
model is one whose source code is available for free or for a reasonable cost. Further, the user
must be free to revise the code to perform diagnostic analyses and/or to improve the model's
ability to describe observations in a credible manner. Several additional prerequisites should be
met for an "alternative model" to be used to support a modeled attainment demonstration.
(1) It should have received a scientific peer review.
(2) It should be applicable to the specific application on a theoreticaj|>asis^
(3) It should be used with a data base which is adequaJUpsuppoT
(4) It should have performed in past applications injich a wayjpft estif
to be biased low.
(5) It should be applied consistently with a protocol
An air quality model may be considered to ha
each of the major components of the modeling
meteorological and emissions models) has beei
documented and reviewed by one or more
be the responsibility of the model develo]
behalf of a State to document that a "scj
reference this documentation to gai
attainment demonstration.
Should the U.S. EP.
model" may'Still be used
specific application. ThilS
obtained with ttiiB.<|pjDeferred
comparisons may:jbe|iie?drable,
11.2 may be
itific peer review" if
sition,
ie results have been
believe that it should
fhg an air quality model on
Ecurred. States should then
qffility model for use in a modeled
ffify a "pr
v r i
quent ap
if it
" at some future date, an "alternative
is shown to be more appropriate for the
emonstrated by side by side comparisons of predictions
ative" models with observations. While such
necessarily required. Criteria described in Section
an "fflfernative model" is more appropriate than a "preferred
model" fora specificiapmiicafi.on.
~T*?S. *w*s*..
Recommendations
air quality model to qualify as a candidate for use in an
attainment demonstration of the 8-hour ozone NAAQS, a State needs to show that it
meets several general criteria.
1. The model has received a scientific peer review.
r <"' '. *£•'
2. The model can be demonstrated applicable to the problem on a theoretical basis.
3. Data bases needed to perform the analysis are available and adequate.
96
-------
4. Available past appropriate performance evaluations have shown the model is
not biased toward underestimates.
5. A protocol on methods and procedures to be followed has been established.
6. The developer of the model must be willing to make the source code available to
users for free or for a reasonable cost, and the model cannot otherwise*be
proprietary.
11.2 What Factors Affect My Choice of A M
States should consider several factors as criteria
model to support an attainment demonstration for the 8
(1) nature of the observed air quality problem; (2) docunl
candidate models in similar applications; (3) experience
required time and resources vs. available time and resources;
applications, consistency with regional models
factors is used to identify attributes needed
help choose among candidate models:
selected model can be used in an attainme
havin
choosjpg a qualifyn
me NAAQS.
are:
jid past track record of
ailable contractors; (4)
:ase of regional
*The first of these
choserflSpors (2)-(5) are used to
iFinallj^efore results of a
should be shown to
application.
perform satisfactorily using the data b
Nature of the observed airlquality problem. This is the most important criterion for
selecting an appropriate model. Babr to seljtlbg a model to use in an attainment demonstration,
we recommend that those i^Weinenting th^^||eig^^iew available air quality, meteorological
and emissionsidata, and ta^^^unt of the^e^^tpnic location of the nonattainment area(s)
relative toJIhat of^precursorli^Ssions. Section*9.0 identifies some types of analyses which may
»=•;»... ' ~*s$*tj> ^^PSiPlftiW
be useful forde^eloging a coho|pipg|gescription of an area's nonattainment problem.
States shQUl|rpffiilake thiSapwew to decide whether it is best to use an urban scale
photochemical grid 'Sii^i^ejg., domain size ~ 200-300 km on a side) or a regional
photochemical grid mole]||e.Ig;|-domain size -1000 km or more on a side) with or without
nesting. Choice between 5ari|irban scale and regional application depends on answers to several
questions f
/''<
1. Is transport offbzone (or precursors) into the nonattainment area a major contributor to an
area's ozone^peoblem?
, ^.^^a^'-fA- .< >i*
2. Are nonattainment areas sufficiently numerous, and in relatively close proximity so that it
is more efficient to estimate control requirements for several nonattainment areas
simultaneously?
97
-------
3. Is the nonattainment area located near major sources of anthropogenic precursors and/or
topographical features requiring fine scale resolution to adequately characterize wind flow?
Answers to the preceding questions require a case by case analysis of available air
quality, emissions and meteorological data. Generally however, we anticipate that an urban scale
model may suffice for "isolated" nonattainment areas (e.g., in the West, outside of California).
Locations subject to transported ozone well above natural background (i.e., 8-hr.jdJpfy maximum
"natural" background is - 40-50 ppb) may need to use a regional modelj|fjhej|f^re major
concentrations of anthropogenic precursor emissions within
an urban scale or nested regional model (incorporating a
area) is advisable. An urban scale or nested regional maJpFIs also recj
sites of interest are located near a major body of water.
Documentation and Past Track Record of C
in an attainment demonstration, evidence should be presi
for estimating hourly or 8-hourly ozone concentrations.
exhibiting satisfactory past performance under a varie^bf corf
(including a benchmark example and outputs) andjs^taical desc
available.
area of concern,
id over a limit
receptor
odels. For a
used
Required vs. Available
first two criteria.are met
i has been found acceptable
iuld be given to models
inally, a user's guide
ie model should be
Experience of Staff and Availahjpcont
choosing among several otherwise accejlable alteratives.''
the air quality model itself, or with jgrai:eorolojKal or er
readily linked with one candidate^^^^lity nmel than
gitimate criterion for
;t experience might be with
ions model which can be more
then
This is a legitimate criterion provided the
Consistencyiof a Proposp^Model with Models Used in Adjacent Regions. This
criterion is applic^bfe|foi; region^^^^prpplications. If candidate models meet the other
criteria, this critefiol^iouljd be considered in choosing a model for use in a regional or nested
• ™-'--*l|'?*^t'jS%!.JV!'--' 'i • °
regional modeling appucabc
Demonstration that an "Alternative Model" is Appropriate for the Specific
Application. If an air quality model meets the prerequisites identified in Section 1 1.1, a State
may use the factors described in this section (Section 1 1.2) to show that it is appropriate for use
m*specific application. Choice of an "alternative model" for use in a specific attainment
demonstration pfefheT 8-hour NAAQS for ozone needs to be reviewed by the appropriate U.S.
and by the U.S. EPA Model Clearinghouse.
Satisfactory Model Performance in the Specific Application. Prior to use of a selected
model's results in an attainment demonstration, it should be shown to perform adequately in the
specific application. Means for evaluating model performance are discussed in Section 16.0.
98
-------
Recommendations. States should first determine what attributes are needed for a
qualifying model to address a nonattainment area's ozone problem, and then choose
among models possessing these attributes. Five factors should be considered in
selecting an air quality model for a specific application. Selection of an air quality
model should be concurred with by the appropriate U.S. EPA Regional Office and U.S.
EPA Model Clearinghouse. The five factors are listed approximately in order of
importance.
1. Nature of the air quality problem leading to
NAAQS should first be assessed, and the sel
ould be i
should be consistent with
5. Consistency of the model with
should be considered.
Prior to using model results i
show that the model perfo
available for that demons
113
. J^satt^ffi
Are Some
e ozone
attributes and
capabilities consistent with the perceived natjpfof the p
2. Availability, documentation and past pe
3. Relevant experience of available staff and
choice of a model.
4. Time and resource constraints may b
regional applications
emonstration, a State should
ing base case observations
fr Quality Models Which May Be Considered?
Table l«llfi|t|aeveral cra^spp|neration air quality models which have been used to
simulate ambieniozoE^S>ncentraWnlf *The list is not intended to be comprehensive. Exclusion
of a model .from"tKell^fepi^ot necessarily imply that it cannot be used to support a modeled
attainment demonstrationffoflffiSiDzone NAAQS. By the same token, inclusion on the list does
not necessarily imply th Jfllipoael may be used for a particular application. States should follow
the guidance in Sections*! 1.1 and 11.2 in selecting an air quality model for a specific application.
99
-------
Table 11.1. Some Air Quality Models Used To Model Ozone
Air Quality Model
References
Sponsors of Past Applications
CALGRTO
Scire,eiaL(1989)
Massachusetts Division of Air
Quality Control (New England)
CAMx
Environ (1997)
Texas NaturaljKesources
Conservatioiv<|ommission (SE
I environs),
Of Health &
City am
CMAQ
U.S. EPA (1!
I.S. EPA
Development (ea
.fine grids in northeast corridor
at Nashville and environs)
MAQSIP
MCNC
Odman,
i Carolina Division of
lotental Management
, most of NC and parts
f surrounding States)
SAQM
lifornia Air Resources Board
(San Joaquin Valley)
UAMV
. icatioi
International (1
LADCO (eastern U.S. with focus
on States bordering Lake
Michigan),
New York Dept Of
Environmental Conservation
(eastern U.S. with focus on
northeast corridor)
Kumar, et al.. (1996)
Georgia InsL Of Technology, Dr.
A.G.Russell (northeastern U.S.,
Southern Appalachian Mountain
Initiative, southern California)
100
-------
ler.
lese may i
ozone Jphcentfi
)ffs among the four]
|, there may onlj
afthe
12.0 How Do I Decide Which Meteorological Episodes To Model?
At a minimum, four criteria should be used to select episodes which are appropriate to
model. First, choose a mix of episodes reflecting a variety of meteorological conditions which
frequently correspond with observed 8-hour daily maxima > 84 ppb at different monitoring sites.
Second, model periods in which observed 8-hour daily maximum concentrations are close to
average 4th high 8-hour daily maximum ozone concentrations. Third, model perioarfor which
extensive air quality/meteorological data bases exist. Fourth, model a sufe;ier|pumber of days
so that the modeled attainment test applied at each monitor yjglating tiaBfertifS is based on
several days. The four criteria may often conflict with one
be a limited number of days with intensive data bases,
meteorological conditions which correspond with monit
specific design values during the base period. Thus, tr|
may be necessary in specific applications.
Those implementing the modeling/analysis prot|
criteria on a case by case basis. For example, prior ex
in its being chosen over an alternative. Another o
occurring during the 3-year period which
value. As we note in Section 3.3, this coul
consideration should be to try to ensure
with monitored ozone concentrations m
in a nonattainment area. If observedlSjHbur daUPmaxima
days should be included within s
several nonattainment areas sim|j|aneouslYJ^lE., wi
criterion is to choose episodj|.cfpitaining
areas. ../'.-'::-;.-».
ence
•ation she
In tn'«!Se|^|, we fi:
meteorological ep||8|^to mode
in specific apphca&o
r rjr '
ondary episode selection
ji episode, may result
hoose episodes
monitored design
:es/effort. A third
lat there are several days
alue at each monitoring site
4 ppb occur on weekends, weekend
s. If a State chooses to model
:ested regional model), a fifth secondary
Son interest to different nonattainment
ach of the four identified primary criteria for choosing
n discuss secondary criteria, which may be important
12.1 What Are The Most Important Criteria For Choosing Episodes?
\. -j- .Ml- -SM 3&te'>«'"-; *• ^ *
r Choose a mix of episodes which represents a variety of meteorological conditions
which frequently correspond with observed 8-hour daily maxima exceeding 84 ppb. This
criterion is important|because we want to be assured that a control strategy will be effective
uniieiyijvariety^piS>6nditions leading to ozone concentrations near current site-specific design
vali^jatsites^ere the NAAQS is violated. We believe the most important indicator of variety
is differing wind fields. This affects source/source and source/receptor orientations and,
therefore, the effectiveness of a strategy.
Those implementing the modeling/analysis protocol should describe the rationale for
101
-------
distinguishing among episodes which are modeled. The selection may reflect a number of area
specific considerations. Qualitative procedures such as reviewing surface and aloft weather
maps, observed or modeled wind patterns may suffice for distinguishing episodes with
distinctively different meteorological conditions. More quantitative procedures, such as a CART
analysis, to identify distinctive groupings of meteorological/air quality parameters corresponding
with high 8-hour daily maxima for ozone, may sometimes be desirable. An example of a CART
analysis applied to select episodes is described by Deuel, etal. (1998).
Choose episodes having some days with monitoi
observed average 4th high daily, maximum ozone concej
at any given site, the relative reduction factor (RRF) i
predicted current 8-hour daily maxima when these are
reflects relationships between current predicted 8-hour
future/current modeled concentration ratios are averagi
'.3
simulate enough days so that the test applied at each site^
10 days. Thus, we want to use episodes whose severityj
of the NAAQS (i.e., an episode whose severity is exce
time of the selected episode). Note that we said, "j
"base case period") rather than "current periodj|
choose episodes with days which are approx^ely"
daily maximum concentration specified injne
sding sen
:tical to
responses from as many as
that implied by the form
x>ut 3 times/year at the
tepisode" (i.e., the
The objective is to
the average 4th high 8-hour
Air quality measurements
characterize episode severity. Thi
modeled episode. For example,
at measured 8-hour daily
Using this information it
*the basejpase period can be used to
lone bylielecting ajBiyear period which "straddles" a
rom 19j^piwere modeled, we recommend looking
iattainment area during 1994-1996.
episc
it each
possible1t6%8eis the relative severity of the days chosen for
modeling at each site. Lirri|I^QBi|j.charactenzation to the three years straddling an episode
avoids problem&pdsed by loiSg^Stiafeends in emissions in assessing episode severity. However,
it leaves unansweredahe,questidn|S^ett»er the 3-year period selected to assess severity of a
modeled day is typic^Tor^ypical.^lftnere is an underlying long term trend in ambient ozone
attributable.to metebrolo-pcafccycles or other causes, it may not be appropriate to compare
different 3-year periodswiSh|meanother using air quality observations. Thus, if one uses a 10-
year old episode with an exceptional data base, there is greater uncertainty in ranking its severity
relative to the current period of interest than if the episode were drawn from the current period.
/»'
The problem Of dealing with longer term variations in meteorological conditions
producing high .ozone can be reduced by assessing the potential of meteorological conditions to
form high ozone*in concert with a climatological data base. An example of such an approach is
described in Cox, et al.. (1996). If such an analysis shows that the 3-year periods straddling each
selected episode day and the most recent 3-year period are not an extreme ones, this supports
using air quality directly to characterize episode severity.
102
-------
Note that if the episode is drawn from among the 3 years upon which the nonattainment
designation is based, days which are chosen are likely to have monitored observations very close
to the current design value. "Close to" could be defined in diagnostic tests in specific studies. In
the absence of such information, we suggest "+ 10 ppb" as a default recommendation for
purposes of prioritizing choice of episodes. If the base and current periods do not coincide,
"close to" is within ± 10 ppb of the design value during the base period straddling the episode. If
it is not feasible to meet this default criterion for all monitoring sites, meeting it atpfes with
current design values > 85 ppb should receive greatest priority.
Choose days with intensive data bases. Preferenj||pBuld 1
measurements aloft, available measurements of indicator^ecies (see
precursor measurements. These preferences result fromjliiesire toiporpora
performance evaluation as a part of the attainment dem^fcationjprhis reduce!
"getting the right answer for the wrong reason". Thus, l^P^ffiood of mischs
ozone/precursor sensitivity is reduced.
lainment test to be
Choose a sufficient number of days to enab
based on several days at each monitoring site
the relative reduction factor computed at any
response averaged over several days.
be more variable if based on an individu;
An air quality model may also have
space if comparisons are based on mjaji obser^iS over sefiral days. Therefore, States should
model as many days as feasible.
We offer the.fol lowing V-ste&pjacedure a
.,<•>% . f .J«Jj!*#& ,
primary critenajor selectmgtj>JS|)des to
"' "'
ire 3.3 indicates that
it if based on a mean
e reduction factor may
ilanchus. et al.. (1998V).
daily maxima matched in
isnay be useful in combining the four
\--%.~.',.;m^ v; TTSrjfv to
1. For each episoqejbeing considareolStates should examine observed 8-hour daily maximum
' •''«l!riijt^li> "s~ ^ ^'^^sfc-
concentrations a^p^i^tes withcdesjjg^alues < 75 ppb can be excluded) monitoring site during
the year of the epis^fe^^well as dialing the year before and the year after the episode. Thus, if
one is examining day^s|nM||>91 episode for suitability in the attainment test, severity of the
candidate days should belassessed relative to 1990-92 observations at each selected site.
2. For each of the three years, rank the top ten 8-hour daily maxima observed at each of the
monitoring sites selected in step 1.
xT*^--"
3. Compute the;ay,erage 1st high 8-hour daily maximum, the average 2nd high 8-hour daily
maximum, etedown to the average 10th high 8-hour daily maximum for each selected monitor.
4. Note a range of concentrations which are ± 10 ppb of the average 4th highest value at each
citp
site.
103
-------
5. Classify qualifying days from step 4 into meteorological regimes, using observed or computed
wind fields as the primary criterion for classifying the regimes.
6. Note days in the preceding sample for which intensive data bases exist.
7. Give priority to choosing a mix of episodes containing days with observations ±10 ppb of the
site-specific design values during the base period(s), drawn from a variety of mete^^ogical
classes identified in step 5, and for which observations aloft, indicator s||desjap/or precursor
measurements are available. Try to choose a sufficient number of dayjfufiM5everal days are
suitable for use in the modeled attainment test applied at ej^pte vid^^^^^^AAQS.
Recommendations. States should consider four
meteorological episodes for modeling. Tradeoi
Such tradeoffs need to be resolved on a case by
3. Choose episod
of indicator speci
*•i*-.s;^':,?j
1. Choose frequently occurring episodes coi
wind orientations observed to occur when
one or more monitors.
2. Choose episodes containing d
concentrations close to (e.g., +
observed at monitoring sites
which each episode is draw i.e., da
formoftheNAAQS).
fleeting a variety of
lama exceed 84 ppb at
Tally maximum ozone
:gh daily maximum
ddling the period from
itely as severe as implied by the
ich measurements aloft, measurements
ror precursor*measurements exist
•< ^
4. Choose a suffici
'- -> **&-$•** ^
the modeled attainme
, ,-f¥ *r*jjjjji
12.2 What Additt
of days so that several days are available for use in
ror each monitoring site where the NAAQS is violated.
ndary Criteria May Be Useful For Selecting Episodes?
In Section 12.1, we noted that there may often be conflicts among the 4 primary criteria
recommended as the basis for choosing episodes to model. Several additional, secondary
selection criteria mayjbe helpful for resolving these conflicts.
•'.-t Choose episodes which have already been modeled. That is, of course, provided that
past model performance evaluation for such an episode was successful in showing that the model
worked well in replicating observations. Given that the 4 primary criteria are met approximately
as well by such episodes as they are by other candidate episodes, a State could likely save a
substantial amount of work in evaluating model performance.
104
-------
Choose episodes which are drawn from the period upon which the current design
value is based. As we note in Section 3.3, fewer emission estimates and fewer air quality model
simulations are needed if the "base period", used to evaluate model performance, and the
"current period", used in the recommended modeled attainment test, are one in the same.
Following this criterion could also make the second primary criterion more straightforward. That
is, current air quality observations rather than episode severity estimated for a period several
years ago could be used as a basis for choice of episodes. We discuss choice of ajnrrent
period" in Section 3.1. A "current period" may be either (a) the 3-year
year of the most recent inventory (e.g., 1995-1997, when 199is the
(b) the 3-year period used as the basis for the nonatt
assume that the two choices very nearly coincide, so that
than the one straddling the year of the inventory, neededventoryptmer
and can be readily made for performance evaluation andRe in tbjlttainment i
straddles the
inventory), or
- 1 999). We
riod ot
Choose episodes having observed concentratioi
form of the NAAQS on as many days and at as
related to the modeled attainment test and to the fourti
The more days and sites for which it is reasonable
possible in the modeled attainment test.
irima
the test
implied severity of the
tie. This criterion is
>r episode selection.
;r the confidence
i, especially if
9,. Weekend days reflect a
It is desirable to include weekenjplays;
concentrations greater than 84 ppb ajpobservep-'on we
different mix of emissions than weekdays. ThisJjbuld also lead to different spatial
patterns of 8-hour daily maxima of Jpppb. Thar, for increased confidence that a
control strategy is effective it nefgfto testefflm^weekap as well as on weekdays. If emissions
and spatial patterns of highyozoipdo vs. weekdays, including weekend days
in the choice ofcepisode&j
to changes tin emissions. As
/ide a i
for evaluating accuracy of a model's response
in Section 16.0, such evaluations are highly desirable.
If a State chooses to modelSseveral nonattainment areas simultaneously, choose
episodes which meetSthe other criteria in as many of these nonattainment areas as possible.
As discussed in Sectio§l|:!0|^5tate or group of States may decide to apply a regional model or
a nested regional modelloMeliionstrate attainment in several nonattainment areas at once. Time
and resources needed fooffiisteffort could be reduced by choosing episodes which meet the other
criteria in several nonattainment areas which are modeled.
Recommendations. States may be able to resolve conflicts among the primary criteria
for selecting episodes by considering one or more secondary criteria. The following are
identified as secondary criteria. States may identify, document and present the
rationale for criteria in addition to these if they choose.
1. Give preference to previously modeled episodes.
105
-------
2. Give preference to episodes occurring during the period corresponding to the
current design value used in the modeled attainment test
3. Give preference to episodes maximizing the number of days and sites observing
8-hour daily maxima close to the level of severity specified in the NAAQS.
4. Include weekends among the selected days, especially if daily:
84 ppb are observed on such days.
5. If applying a regional model, choose episode
secondary criteria hi as many nonattainment
exceeding
rimary and
106
-------
13.0 What Should I Consider When Selecting A Modeling Domain And Its
Horizontal/Vertical Resolution?
A modeling domain identifies the geographical bounds of the area which is modeled.
Recommended domain size depends on the nature of the strategies believed necessary to meet the
air quality goal. This, in turn, depends on the degree to which air quality observations suggest
that a significant part of an observed exceedance is attributable to regional concentpftions which
approach or exceed levels specified in the NAAQS. Choice of domain sl^is aJp^affected by
data base management considerations. Generally, these are less demarJSli^Fsrnaller domains.
Horizontal resolution is a function of the size of i
is determined by the number of grid cells (i.e., layers) o
Choice of horizontal grid cell size and a suitable numl
variability in emissions, spatial precision of available e:
that mesoscale or smaller scale meteorological phenome?
precursor/ozone relationships, data base management cc
We begin this Section by discussing factor
size. Next, we address choice of horizontal
conclude by discussing factors affecting chojppo*l si!
scale grids considered in a nested modelirjglfflalysis
/idual gric
lidered injne vert
Lvertijglrlayers de
:al resolutk
lion.
lata, mixing h
a pronounced effect on
computer/cost constraints.
should i
numbel
slutior
»choosing domain
Ertical layers. We
" coarse scale and fine
13.1 How Do I Choose Between An Urban Scalejpr Regional Domain?
ixarnine
QS vs.
fied in
|gap bjgg»een a nonattainment area's design value
n observed regional (upwind)
If the former gap is less than the latter, an
o illustrateTor the case of ozone, if a nonattainment area had
jhour daily maxima were typically 60 ppb, the former gap
gap (24 ppb). Depending on the judgment of those
States may find it useful
and the leveLspecified in
*§$&
concentraticps«nd the 1
urban may
a design va1ug|^^^p,pb an
(11 ppb) is su'Ss^n|i^Wess th
implementing stralSjffor meeting the NAAQS may thus focus on local control
measures^An iarrja^S^M^PQain size may be appropriate. In contrast, if the local design value
were 9|jppb but corresp^n^^gpgional daily maxima were typically 80 ppb, the former gap
remajhs 11 ppb, but the latteps reduced to 4 ppb. Those implementing the protocol may wish to
consider using regional as well as local measures in such a case. This would necessitate using a
regional modeling domain. In general, if additional regionally implemented control measures are
expected to materiaUylaffect the amount of additional local controls needed to meet the air
quali|y«Qal, a regional modeling domain should be used. If not, an urban scale domain should
«U.T' .;-„'" *.}<•«-.•-«-•«•-< —-
su:
What do we mean by "urban scale" and "regional" domains? An urban scale domain is
one having horizontal dimensions less than ~ 300 km on a side. Assuming the nonattainment
area is located near the center of the domain, the domain should be large enough to ensure that
107
-------
emissions occurring shortly before sunrise in its center are still within the domain near the end of
the same calendar day. If recirculation of the nonattainment area's previous day's emissions is
believed to contribute to an observed problem, the urban scale domain should be large enough to
characterize this. If recirculation encompasses distances larger than about 300 km, an urban
scale model is probably not sufficient to address an area's problem.
A regional domain is one having horizontal dimensions typically exceedirygpOOO km on a
side. Data base management problems generally make it infeasible to uj||tiie same horizontal
grid cell size in urban scale and regional models. Nested regional modlSlletntended to address
this problem. A nested regional model is one whose dor
side. However only a portion of that domain (e.g., < 3(
similar to that recommended for urban scale models. St
monitoring sites considered in the modeled attainment i
size of individual cells comparable to that recommende
on a side,
ss shouldjpmanl
yithinjne area covel
LOGO km on a
ils with a si
all
*with
bin scale modeling
Recommendations. Selection of a domain size dc
strategies to be simulated. States should reviewiregio
those occurring in the nonattainment area to^detgrmine tl
regional vs. local controls. If this revie^^i^^^hat a reg
important component of an attainmegflln^S^Mia,. then th"
regional (>1000 km) in coverage. Qfnerw
suffice.
13.2 What Horizontal Grid Cell Size Is Necessary?
«*>*&.&''"• .I4£t& ,,\ ."•**
types of control
design values vs.
isis to place on
trategy is an
domain should be
alefuialysis (<~300 km) may
As we discuss in Se
models toprovide met&
commonlynse&orthese
km cells. ThuSfliie; issue ad
upper limit fortegfonalanodels
.0, we an
nputs nee
jQespread use of dynamic meteorological
iake air quality estimates. The most
et up ro produce meteorological fields for 108, 36, 12 and 4
;this Section is which of these sizes to recommend as an
. -iC;^;,
tSpStirban scale or fine portions of nested regional grids.
In past guidance^efhave recommended using horizontal grid cell sizes of 2-5 km in
urban scale modeling analys^lpkS. EPA (1991)). Sensitivity tests performed by Kumar, etal.
(1994) in the South Coas%fffilBasin compare hourly base case predictions obtain with 5 km vs.
10 km vs. 20 km grid cells. Results indicate that use of finer grid cells tends to accentuate
highest hourly ozone predictions and increase localized effects of NOx titration during a given
hour. However, statistical comparisons with observed hourly ozone data in this heavily
monitored area appear comparable with the 5 and 20 km grid cells in this study. Comparisons
between Jiourly ozone predictions obtained with 4 km vs. 12 km grid cells have also been made
in an Atlanta study (Haney, etal. (1996)). As in Los Angeles, use of smaller (i.e., 4 km) grid
cells leads to higher domain wide maximum hourly ozone concentrations. However, when
reviewing concentrations at specific sites, Haney, etal. found that for some hours concentrations
obtained with the 12 km grid cells were higher than those obtained with the 4 km cells. Since
108
-------
signs of the differences obtained with 4 km vs. 12 km grid cells vary for different hours, this may
suggest that 8-hour daily maximum ozone predictions are less sensitive to the selected grid cell
size than 1-hour daily maxima. Recent sensitivity tests comparing relative reduction factors in
predicted 8-hour daily maxima near 272 sites in the eastern United States indicate generally small
unbiased differences (£ .04, in 95% of the comparisons) using a grid widi 12 km vs. 4 km grid
cells (LADCO (1999)).
Intuitively, one would expect to get more accurate results in urban applical^s with
smaller grid cells (e.g., 4 km) provided the spatial details in the emission^nd meteorological
inputs support making such predictions. Thus, using 4 km grid cells fine portions of
nested regional grids and 12 km cells in coarse portions ofjjgpbnal gr1!i||Qi&rable goals.
However, extensive use of urban grids with 4 vs. 12 km^pcells with 12 vj
36 km grid cells greatly increases computer costs, rannipftimes andjolta baiSnanagement
needs. Further, elsewhere in this guidance we identify i
and several emission control scenarios. We also identif
would be desirable and suggest using more vertical layer
past. Also, there may be means of dealing with potentia
desired grid cells. For example, use of plume in grid algorithr
might be considered as an alternative with coarser thaffidesired
Relative importance of using a doma
weighed on a case by case basis by those jj
this guidance, we identify upper limits Jp^orizonjfi grid^
desired for some applications.
factors (e.g., number of modeled
given limits of time and resources^
^days,
1 large dor
r of diagnostic testsSwKich
Commonly been done in the
by using larger than
.point sources of NOx
4 km will need to be
analysis protocol. Thus, in
?which may be larger than
flexibility to consider competing
rforming a modeling analysis within
For coarse portion^^^||pnal grids^ej^ommend a grid cell size of 12 km if feasible,
but in no eVent&|er urbarfand fine scale portions of nested regional grids, it
may be desirab|||||aipe grid"cll^^^||4 km, but, in no event larger than 12 km. All ozone
monitor locatiom'i^m.a nonat^yBEJieoferea should ordinarily be placed within the fine scale
portion of a nesjra^egt^&jgrid if'lujfted models are used. States choosing an urban grid or fine
portion of a nesteH giiajwjffi|cells larger than 5 km should undertake several additional analyses.
First, States should apply^lj|mieir» grid algorithms to major point sources of NOx if they choose
an urban or fine portion of1a|regional grid with cells as large as 12 km. Once an emission control
strategy has been tentatively selected, States should test the current and the selected control
strategy with grid cells-4 5 km, if feasible, so that the outcome is available to be considered in a
weight of evidence determination.
Recommendations. Horizontal grid cell size in regional models should be < 36 km,
except in areas used to establish boundary conditions for the regional model (where
they may be larger). For urban scale analyses and the fine scale portion of a nested
regional model, cells which are 4-5 km on a side are preferred, if feasible. Cells should
not exceed 12 km on a side in these analyses. If cells as large as 12 km are used in
109
-------
urban areas, States should consider using plume in grid algorithms to deal with large
point sources of NOx. States should perform diagnostic sensitivity tests to see whether
using grid cells smaller than 12 km affects conclusions reached in the modeled
attainment test when the selected control strategy is simulated. If so, this should be
considered in a weight of evidence determination.
13.3 How Many Vertical Layers Should I Consider?
As described in Section 14.0, the preferred means
Accuracy of predicted base case ozone
a model is able to characterize dilution
precisely the model can estimate maxim
boundary layer). Precision of mixing h<
vertical layers aloft which are near
Because maximum mixing heigh'
numerous days and locations, in
layers considered by the
are not sensitive to the nj
namicme
often consider as
toJfterface me
fields for input to air quality simulation models is to use
four dimensional data assimilation (FDDA). Such mod]
vertical layers. To minimize a number of assumptions
quality models, it is better to use identical vertical resol
models. However, application of air quality models wirt
feasible nor cost effective. In this Section we identify fi
chosen for use in an air quality model, as well as the placemen
*d air
logical
prtical layers may not be
number of vertical layers
yers.
part, on how accurately
Si turn, depends on how
i.e., the planetary
the thickness of the model's
ght (Dolwick, et al.. (1999)).
ys and it is necessary to simulate
uenced by the number of vertical
have shown that base case predictions
above the planetary boundary. Thus, States
, < '* A^«fc- ' . JT — ., ^- >-
may assume-as Jew as oneoa^ca&abfl^e the highest conceivable maximum afternoon mixing
height with theisst of the vellSlflyKES occurring within the planetary boundary layer.
Placement oftveritical layers^witfiin the planetary boundary layer is also an important
issue. For practicarrl^ons|i^s besfto have an air quality model's vertical layer placement
coincide with layers cona^MmD'the meteorological model used to generate meteorological
inputs. So the placemenf^sJSlreally is, which ones of the boundaries between the
meteorological model's layers should one match with the boundaries between vertical layers used
in}the air quality modelf' Based on the discussion in the preceding paragraph, we recommend
highbred sion near tbiiranticipated maximum afternoon mixing height. In addition, observed 8-
hour3aily maximum ozone concentrations may well include some evening hours. Surface
concentrations during these hours may be affected by presence of a low level inversion whose
base is just above turbulence introduced by surface roughness or, .in some cases, by an urban heat
island. Thus, States should use a shallow surface layer, generally no more than 50 meters in
depth. In general, layers below the mixing height should not be too thick, or large, unrealistic
step increases in mixing may occur. States should try to avoid using layers within the planetary
110
-------
boundary layer thicker than about 200-300 meters.
Based on recent sensitivity studies (Dolwick, etal. (1999) and LADCO (1999)), it
appears as though as few as 7-9 vertical layers (including one above the planetary boundary
layer) may suffice in a modeling study if care is taken in specifying placement of these layers.
Prior to modeling, we recommend that States review available meteorological measurements
aloft to get a sense of where the maximum afternoon mixing height is likely to Ixsjpaays which
might be modeled. We recommend that the number of vertical layers cojiMderejipm coarse and
fine portions of a nested regional grid be identical.
Recommendations. An air quality model may coi
commonly considered in a meteorological nux
vertical layers used in an air quality model shottlp^pincKie with selecteptoumiaries in
the meteorological model. Care should be take^^^^^ure the vertical^pirs so that
as possible. The surface
50 meters deep, and no
ters thick. The
and
criteria, States
ry boundary layer
13.4 What Else Should I
Portions Of Nested Regional
Coarse Grid Domai
chemical/physical«lifetirne
which regiorialnic^ing is uf
assess effects of asregtlmal strati
i • 1.1•«Ar-1*-*i*sfe-,, . f »,-•
domain needs to be larger than if or
, _-,-«.Z.il,3&Wt •-«>M.'?M..-.,!
study.
the maximum afternoon mixing height is defined
layer considered hi the model should generally
layer beneath the mixing height should be mo;
minimum number of layers chosen dependjj^^e meteoi
characteristics of the area to be simulateJ^^^^^Jthe
should generally use at least 7-9 vertjgpliye]
and 1-2 layers above it
ely And Coarsely Resolved
of a cSppjpiirdomain should be consistent with the
mts to be modeled. It should also reflect the purpose for
For example, if a regional analysis is performed to
Itaneously for a number of nonattainment areas, the
Plimited number of nearby areas were the focus of the
; Lifetimes vary forozone and its precursors. Lifetime for NOx (i.e., NO + NO2) may be
less than a day. Region&analyses performed in the U.S. to date suggest that lifetimes for
sulfates and nitrates are|two days or less (Dennis, 1994). Sources of VOC are believed to be
ubigaitous, due to natural emissions. Many of these natural emissions are relatively reactive, so
that niuiti day traiisport of stable species of VOC or radical products resulting from oxidation of
more-reactiveispecies may not be a critical factor for selecting size of a domain for modeling
ozone, lifetime for ozone is notoriously difficult to estimate due to the recycling of this
compound with free radicals, concentrations of oxidized species of nitrogen and emissions of
fresh NOx and VOC precursors which occur in transit. Given information about the lifetime of
nitrates however, it is probably safe to assume a lifetime for ozone which is on the order of 2-3
111
-------
days. The foregoing information suggests that, ideally, the size of a regional modeling domain
should be large enough so that emissions occurring two days prior to the beginning of daylight on
a modeled day of interest are included within the domain. Thus, we suggest States focus on their
receptor areas of interest, perform some screening analyses with trajectory models to ensure that
major source areas within two days' travel time are included in the domain.
Fine Grid Domain. Size of the fine grid domain should be influenced i
(1) proximity of receptor sites to major sources of ozone precursors
presence of topographical features which appear to affect
limit resource intensive efforts needed to use numerical
an important concern for use of nested regional models.
smaller than that recommended for an urban scale anal)
domain is available to estimate impacts of sources loca|
receptor area, whereas this information is not available @^9^Iated urban
ral factors:
;(2)
id (3) desire to
last factor is
The
issue of how far to extend a fine scale grid is one which
case basis. We recommend that States examine the iss
Section 16.0). For consistency with the modeled attai
grid should initially extend 15 km (i.e., 3-4 4-km gtidmUs) beyo
Recommendations. Size of a coarse
important sources located two dayj
Applications which need to consi
apart therefore need to use larger do:
in close proximity to one anciraK Ex
receptor sites. States shoufipperfo;
resolved grid needs to extend. As a
finely resolved grid^sufficiently so
sites considered in theraodeled attainment test.
be resolved on a case by
sensitivity tests (see
end that the fine
tor of interest.
! enough to include potentially
>r sites of interest.
fes located some distance
iplications focusing on receptors
rid also depends on the number of
analyses to ascertain how far a finely
iption, we recommend extending the
aids at least 15 km beyond all monitoring
112
-------
14.0 How Do I Produce Meteorological and Air Quality Inputs Needed By An Air Quality
Model?
After episodes are selected for modeling, corresponding meteorological inputs need to be
generated for use in an air quality model. Although the resulting inputs remain constant, they can
affect outcomes of a number of the modeling outputs we have identified for scrutiny in Section
4.1. They may also potentially affect relative reduction factors used in the attainraMr and
screening tests. In contrast to meteorological data, air quality inputs ma»|hMgdpetween times
corresponding to "current" and "future" emissions used in rnj»modeledffl$iMlent test. This
presents a potential problem which needs to be addressed..
In this Section, we describe two approaches for
quality models, and identify advantages/disadvantages
dynamic meteorological models with output "nudged"
approach for generating needed meteorological data. Foi
for horizontal grid cells smaller than 12 km may present
diminish these, if they occur. It is important to qualil
being used in an air quality model. We next discuj
conclude by identifying the role of air quality ij
quality simulation, and note ways to reduce^
sparseness of these data.
14.1 What Approaches A
aerating
rats to.
sure IT
these ini
and bo!
imulatit
Two approaches have
quality mode|s for ozone.
and introduome ad
surface t
character!
sorolc
ith each, vmnnnpismg
tions is usuallyl&^pwferred
lications, use of these models
[ems. We identify ways to
|cal inputs prior to their
»evaluated. We
conditions for an air
resulting from
lable Jfor Generjpng Meteorological Data?
f
:e meteorological data needed in air
models) relies primarily on observed data
ow due to terrain features. Observed
are used to develop other information needed to
Most frequetitlyiused diagnositefwind models are described by Douglas, et al. (1990) and
"•»'*gBa«4««*||igfejgt, ° -.f>,- J o ' TT "«' v /
by Scire, etal.. (1998)Spij|^ain advantage of diagnostic models is that they are relatively easy
and inexpensive to appl0|^^ffiira', they make maximum use of wind observations. There are
several disadvantages, tiQJwejfCT. First, there are seldom enough observations to adequately
define a windfield, particularly aloft. Much of the input to the air quality model is derived
through interpolation orisubjective methods. Because of the sparseness of observations in many
areas; we do not encourage use of diagnostic models for generating inputs to regional scale air
quati^inodel applications. A second disadvantage is that the meteorological estimates derived
with»^|a^nJ|^
-------
PM2 5) are believed to be primarily affected by winds and urban scale source/receptor orientation,
the disadvantages are not serious enough to preclude use of diagnostic models.
The second approach for generating needed meteorological data is to use dynamic
meteorological models with four dimensional data assimilation (FDDA). These models attempt
to characterize theoretical relationships between meteorological variables and
topographical/terrain characteristics. Use is made of relatively sparse observationjlfioft to help
steer (i.e., "nudge") solutions so that they do not diverge from observed^^teoj^Dgical fields.
Wind observations aloft are typically used for this purpose. See SearnH) for a further
summary of the attributes of dynamic meteorological mod$|||phe
and Seaman, etal.. (1996)), RAMS (Pielke, etal.. (1992]^BLyons,:
SAIMM (Systems Applications International, (1996)) m®els are anting
most widely used with numerical air quality models. TnmajorjriPantage
meteorological models is that they provide a way of ch|
consistent with theory, terrain and each other at times an
exist. Disadvantages have been large required compute.
needed to apply the approach. Recent advances in co
increased use of dynamic meteorological models fj
is used as the default approach with the CMA
compatibility between candidate meteoroloj
use. We believe that use of dynamic metd|rologic;
preferable approach for generating mete$rologicaMhputs ti
iter te
llution^
[ODELS
air
g rneteorologicaKllfiaHions
where observations do not
considerable expertise
.ve resulted in
is. The MM5 model
;s need to consider
Ity model(s) chosen for
A is generally the
llity models for ozone.
Although improvements ii
meteorological models possible,.
increase dramatically as th
becomes finer»vFor example;
12 x 12kirigriaibellsisc
needed to processittieteorolo
may need to limit &;spatial
jutersjlave madejficreased use of dynamic
lave fpj^l that datt%torage requirements and CPU time
sntal gndf^wjB^esQuired of the meteorological model
° '$Sjg<3IK%>gff'
J timetntedeSrto generate meteorological data resolved to
; greater "than the expected factor of "9" increase in that
pSpmain with 36 x 36 km grid cells. This suggests that States
and the number of episodes for which dynamic
meteorological modeKjaiefUsed to'pTrbeess meteorological data for grids with horizontal
resolution-<12"kmr*iGeft»pli^; a finely resolved meteorological field needs to extend about 3 grid
cells beyond the bounds;ofeih^pne scale grid used to make air quality predictions. For example,
if 4 km grid cells were use3|in:the fine portion of a nested regional air quality model,
meteorological fields at this detail would need to extend 12 km beyond the bounds of the 4 km
grid used for air qualityipredictions.
Recommendations. States should ordinarily use a peer reviewed dynamic
meteorological model with four dimensional data assimilation as the means for
generating meteorological inputs to ozone models. Peer reviewed diagnostic models
may be used on a case by case basis. Grid cell size used in dynamic models should be
chosen considering factors discussed in Section 13.0.
114
-------
14.2 How Do I Deal With Data Management And Computer-related Constraints
When Applying Dynamic Meteorological Models?
States should ordinarily use dynamic meteorological models resolved to the same level as
desired for making air quality predictions. Occasionally, this may not be feasible, or may lead to
poor performance of the dynamic model. In this Section, we identify possible mejppfor
reducing one or both of these problems. The methods we discuss may ij||gasft|Be risk of
discontinuities at the bounds of a finely resolved grid. Thesejhould bj^^ftj^' and corrected to
the extent possible before proceeding.
The first approach is to use available results fror
coarse scale (i.e., 36 km for a desired 12 km estimate,
interpolate more finely resolved fields. An objective arjj
be used (U.S. EPA, 1991). This approach would be parti
desiring finely resolved meteorological estimates is rel
lynamic
forj
jlsol
;sired 4 krr
.4JM& ^aidK
estimates more finely. For example, in the case of ozoUe, fine g
accurately characterize the apparent detrimental effej|||LNOx
resulting from titration of ozone by nitric oxidejiealSic^fe of NOx!
:e bilinear interpollpin could
ful if the major reason for
resolve emission
ay be needed to most
^predicted ozone
A second approach for circumvenjjpfg majorjesi
dynamic models for finely resolved gridiRonsiderjiiopogri
land/water interfaces) and measuredcmSeorolo
by a dynamic model. This secondfiprbach
ments needed to apply
finely resolved meteorological ingots has t
which cannot be adequately^considered th;
the second^EJpJpach is t<|
<<*•
diagnostic
coarsely resolved dynamic model.
brmation (e.g., presence of
data topefine fields coarser fields generated
be preferred if the major reason for desiring
pjieeived importance of mesoscale features
tive interpolation procedure. In essence,
ibdel to the wind field generated by the more
Finally,«o^p[Ujences of using coarse grid cells (e.g., 12 km when 4 km might be more
desirable) can \3^^m^joy specifying~a land use for each cell that corresponds to usage near the
major portion of emis^OTf^thin a cell. This approach is most applicable at land/water
interfaces. By assumingffi^^lis entirely "land", vertical dispersion of fresh emissions is likely
to be better characterizedif^ffiis might also result in a better characterization of subsequent
transport of coastal emissions over adjacent large bodies of water.
P
Recommendations. Prohibitive computer-related constraints associated with applying
a dynamic meteorological model to derive a finely resolved (4-12 km) set of
meteorological data can be addressed in one of two ways.
1. Interpolate more coarsely resolved data using objective analysis.
2. Apply a diagnostic wind model using "observations" generated by the dynamic
115
-------
meteorological model for a coarser grid. Assume other variables remain the same
as for the coarser grid.
Consequences of using coarser than desired grid cells may be reduced by assigning a
land use factor for each surface cell which corresponds to the location of most
emissions within the cell (e.g., at cells including an interface between land and a large
body of water).
14.3 How Do I Quality Assure Results Generat
There are several ways to evaluate performance
desirable to evaluate meteorological inputs before air qi
means available for evaluating the meteorological mode
quality model is run. Important meteorological outputs
patterns, mixing heights (e.g., estimated by noting the vj
(KJ is suppressed), temperature, pressure, water vapof
output from a meteorological model include compj
derivation of trajectories, use of computer grat
results obtained with different models, use gjplmenS
patterns of observed and predicted daily maximum
these is briefly described in the followiojpparagrz
gal Model?
ithougffit is
id clo
.with sel
tions must wait unxynne air
scrutiny include wind velocity
^yhich vertical diffusivity
^lethods for evaluating
measurements,
ion-reacti
lets
acers, comparing
r; comparing spatial
>rocess analysis. Each of
IS. Thisjian be done by excluding selected
onal dataassimilation (FDDA) so that they can be
rperature, pressure and water vapor are
Comparison with upper
upper air observations from use j|p>ur di
used to assess model perfonnanip Wind
'; . r ,#wilKI*JE '
important variables to aloft mi
they can providermeans"^^^^g^ng how^well a model characterizes vertical exchange in the
lowest
its are available at more than one altitude,
bases (i.e., widely separated soundings taken twice per
day) are neededJtojiUjpnort FDDSlnlBBata base is probably insufficient to exclude data to
*" " "'-/•'-^T^ /^^ ^lilfti*^-^'
evaluate model^penoiptnlnce. In SlBiSn 5.0, we noted that it is desirable to increase
* - "V*''' ' ~ j$ji^~^'.%^&'''^~-W£''Z ' £'"
measurements aloft. 4dnel©ion for doing this is to provide better means for evaluating
X's "f •'$ -• !f^ ^yf-^'j
performance of meteorologtclttaaodels.
- W" '
Derivation of trajectories. A State could select several locations in the grid and use
trajectory models sucruis HY-SPLTT (NOAA, 1999) to derive back- or forward-trajectories from
the hourly wind fiel.d^generated by a meteorological model. If surface trajectories were limited
to dafpght hoursithe computed trajectory could be compared with observed surface air quality
otseiyations.;^Ethe timing of high ozone observed along the path of the trajectories is consistent
with expectations, given the configuration of sources, this would be an indicator that the
meteorological model is performing adequately. A State could also derive daytime surface
trajectories using observed wind data. These trajectories could also be compared with air quality
patterns. By comparing the two sets of trajectories with observed air quality patterns, it would
116
-------
then be possible to assess whether the meteorological model increases the skill with which ozone
plumes are oriented.
Use of computer graphics. Examining wind vectors for apparent discontinuities is
possible using graphics. It is also possible to construct difference diagrams between observed
and predicted temperatures and winds. Locations where agreement is poor may suggest areas
needing more finely resolved estimates. Geographical orientation between areas
agreement and locations of major sources or observed poor air quality ra^dbe jrfptted to judge
potential significance of any disagreement.
Simulation of inert tracers. This approach is tojj
(e.g., 10 ppb) of an inert tracer in an air quality model
and vertical size of the cells used in the meteorological j
unnecessary to consider atmospheric chemistry, deposit
constant boundary conditions should also be assumed. Iff
remain uniform, and there should be no systematic driftj|
the grid. Predicted concentrations of the tracer can the
major discontinuities in the concentration field orj
may suggest a problem with the meteorologies
to consider divergence/convergence predict
grid i
ay be f
reh emissions). I
concentration field should
laterial remaining within
whether there are
:e. If there are, this
' the air quality model
Compare results obtained wiMifferentfnodelS^^^pproach is to compare results
from two different models for a days bjang consiJpred. For example, MM5 and RAMS
results could be compared to note^^Sences^m predictedtfurface temperatures as well as wind
velocities at the surface and alof^leasons^^najor^prences would then need to be
diagnosed.
- X:
Compare estima
- 'c > £. ^
Calculations;can*beanade in
ce or dimensionless parameters with expected ranges.
ions of the grid to see whether they appear reasonable.
•,. ',;p "f%.
Comparespatial|patterns ofeair quality predicted with a photochemical grid model
with observed patteniionthedays of interest. If the predictions are systematically skewed
from the observations, ®s|c6||dfsuggest a problem with the meteorological outputs generated by
the meteorological
Use process analysis. Process analysis applies to the output generated by an air quality
model. It is describedlby Jeffries, (1997) and by Lo, et aL. (1997). Its use with air quality
modeisis noted,in|Section 16.0. Process analysis determines the relative importance of different
chemical or physical factors as contributors to predicted.ozone concentrations. If process
analysis suggests that a variable influenced by meteorological inputs, such as vertical exchange
(i.e., vertical diffusivity), plays a large, unanticipated role leading to a high ozone prediction, this
might warrant a closer examination of what led to such a prediction.
117
-------
Recommendations. To the extent possible, States should quality assure results from
meteorological models prior to using them in the intended air quality model. States
should select a mix of approaches for evaluating meteorological inputs to an air quality
model on a case by case basis. Candidate approaches include:
1. comparison with upper air measurements "held back" from use in FDDA;
2. comparison of calculated trajectories with observed air gualityj>atterns;
3. use of computer graphics to discern spatial
4. simulation of inert tracers to identify discdatinuities
5. comparing results obtained with dffierenif||||||plogical modelsf
less parameters and
vs. observed
pif unexpected ozone
6. calculating and comparing divergence an
comparing these with expected ranges;
7. comparing spatial ozone pattern
patterns, and
8. using process analysis to flag contributions
concentrations by meteorological factors.
14.4 .What Are Some Past Appli
lamic Meteorological Models?
ir qualiljTrnodeling applications using the two most widely
. Choice of a meteorological model may be influenced
l|il||model, as well as by past experience of those applying
14.1 is not comprehensive. Inclusion on the list does
ment for a specific application. Exclusion does not necessarily
for a specific application. States should consider using
available
by compatibiis
the air qualit
not necessanlylrnp
imply that an approachls^
methods such as those inflection 14.3 to determine whether the output generated by a
meteorological model is/adequate for use in a specific application.
14.5 How Do I Address An Air Quality Model's Need For Air Quality Inputs?
•quality inputs are needed in air quality models for two purposes: to specify initial
conditions, and to specify boundary conditions. There is no satisfactory way to specify initial
conditions in every grid cell. Thus, we recommend beginning a simulation at least one day prior
to a period of interest for urban scale applications, and two days prior to periods of interest for
118
-------
regional applications to diminish importance of arbitrary assumptions about initial conditions.
Table 14.1. Some Past Applications of Dynamic Meteorological Models
Meteorological Model
References Describing
Model Performance
Sponsors (Applications)
MM5
Seaman, etal.. (1995)
(1996b)
Tesche, et al.. (1992
(1993b)
oaquin Valley,
Basin
RAMS
Tesche, et al.. (1
(1993d), (1993e), (
LADCO (eas
?., with
emphasis on Lake Michigan
States)
Boundary conditions can be specified i
interest within a much larger domain.
models, as described previously. The ne
why we recommended in Section 13.0 $jii States
bound 2 or more days' travel time
practical to use a nested regional
relatively unimportant, a second
analysis. The domain sho
affecting loraLa;
occurringlii^^^nter o:
of the same'^entiaWay
l^ ^X:«^ ;*$$&> J
would need to'*
make use of monitored data and in
begs the question
:o nest the area of
ing nested regional
boundary conditions is
domain, with the upwind
focus of an analysis. If it is not
•undary conditions are believed to be
single domain in an urban scale
rical about the major local sources
d be large enough so that emissions
just before sunrise remain within the domain until the end
jon is thought to be part of the problem, the domain size
laded to bFtafflieconsider it. Use of a large, single domain requires one to
^^,^ M*"'*^-'*'?' '"$"•
ation to estimate boundary conditions. This approach
assume for future boundary conditions. It works best where
boundary conditions are;€oWiHifl;are expected to remain so.
':- ^
Recommendations! Simulations should begin at least one day prior to the period of
interest for urban applications and two days for regional applications. Use of nested
regional modelsis the preferred approach for addressing boundary conditions. Where
such an approach is not feasible, States should consider a single domain large enough
to ensurelthat emissions occurring in the center of the domain just before sunrise
remain within the domain until the end of the same calendar day or that next-day
recirculation (if important) can be considered.
119
-------
-------
15.0 How Do I Produce Emission Inputs Needed For An Air Quality Model?
Developing emissions inputs needed in air quality models requires several steps. First,
States need to compile statewide and countywide emission estimates for precursors of ozone, as
well as information subsequently used to spatially and temporally allocate emissions within each
county included in the modeling domain. The most recent commonly available emissions
estimates should be used in the modeled attainment and screening tests, described«S#1**^ • —"
• VoluffieJL Introduction and Use of EIIP Guidance for Emissions Inventory
z*.. Development (U.S. EPA, (1997a))
• . VbiurheH: ^ , Point Sources Preferred and Alternative Methods (U.S. EPA,
-' (1997b))
Volume nt" v Area Sources Preferred and Alternative Methods (U.S. EPA,
'(1997c))
Volume-iV: Mobile Sources preferred and Alternative Methods (U.S. EPA,
J- (1997d))
• Volume V: Biogenics Sources Preferred and Alternative Methods (U.S. EPA,
rivii.^O--" (1997e))
;-,'* Volume VI: Quality Assurance Procedures (U.S. EPA, (1997f))
Volume VII: Data Management Procedures (U.S. EPA, (1997g))
In addition, guidance exists or is being prepared on emission projections, the National Emission
121
-------
Trends inventory methodology, and temporal allocations, spatial allocations, and chemical
speciation of emission inventories (U.S. EPA, 1998d, 1998e, 1998f). The EIIP documents are
available electronically through the U.S. EPA Internet website at
http://www.epa.gov/ttn/chief/eiip/techrep.htm. States should consult these documents as they
prepare their emission inventories.
15.1 What Countywide Emission Estimates Are Needed To Support Air Quality
Models?
Statewide and countywide emissions need to be di
stationary point source emissions, stationary area source
road and off-road sources and biogenic/geogenic emissi
by SCC and have associated location information (e.g.,
diurnal and weekly operating schedules. Area source ei
reported by county. Surrogate factors, used to spatially
category within an air quality model grid superimposed
each area source category. Defaults for surrogates are
Examples of surrogate factors might be such thin
land use, etc. If information exists concerning
area source categories, this information
source emission estimates. On-road ani
using the most current version of the U
version of EMFAC) in concert witLg|Jgpity (i.ejfvehicle
mobile source emission estimatesjjpfKl be ajpbmpanie
spatially disaggregating the mobifl|Emissio
as
gridded, hourly estimates c
recommend&States disti
sources: Estimates for bi
(Geron, et
emissions on acoi
each county if a*Sl
manner.
sions, m
Point spirces
mde/lonfitude coor
ould be classiffiSiprSCC and
issions from the source
, should be identified for
emissions models.
iment by census tract,
ii patterns for different
^ and countywide area
fsions should be estimated
, in California, the current
traveled (VMT)) estimates. The
recommended surrogates for
al and weekly activity patterns so that
ission estimates in subsequent steps. We
activity levels for mobile and stationary area
ssions can be made using the BEIS2 emissions model
the U.S. EPA. A State should report biogenic
sgarding spatial pattern of land use is needed within
biogenic emissions within a county in a non-uniform
For model applicSSBfpiaddressing the ozone NAAQS, emission estimates for each source
category should include countywide VOC, NOx and CO estimates for each month of the year.
The/VOC estimates shduld be accompanied by a recommended speciation profile for each source
category. We recommend that States rely on local measurements to the maximum extent
pbssibleJor the^speciation profile estimates. However, default information on VOC species
profiles is available in U.S. EPA (1993), if needed. These data and updates can be obtained
electronically through the U.S. EPA's Internet website at
www.epa.gov/ttn/chief/software.htmltfspeciate.
122
-------
Recommendations. States should be familiar with guidance in U.S. EPA (1999c) and
with U.S. EPA Emission Inventory Improvement Program guidance describing
appropriate procedures for estimating statewide and countywide emissions needed to
support SIP revisions for ozone. Air quality models require emission estimates from
point, area, mobile and biogenic sources. In order to convert this information for use
in air quality models, VOC species profiles, rationale for suballocating emissions
within a county and for assuming diurnal and weekday vs. weekend variability in
source category, as
mission
ich are more
ites for
emissions is needed for each point source and for each major a
well as for mobile sources. Default assumptions for spatial/te
allocations are available in emissions models. HOWJ
appropriate for a specific area can be substitutedjpr these. E:
VOC, NOx and CO are needed for each month oflfhe yearJpsuppdi
regional model applications performed for waniiieathej^nties and
integrate ozone and PM2^ control strategies.
15.2 Can I Use the National Emissions Trent
Starting Point?
ies to sei^pi^a starting point, we
lh modfipFstarting from the
ic most recent NET reflects
However, the U.S. EPA
If there are no previously available moc
recommend that States derive an inventory j
National Emissions Trends inventory
statewide, annual emission estimates foffOC, N
plans to have a 1999 NET availablejfalmg mejjflfer half ooOO. If available on a timely basis,
the 1 999 NET is the preferred stajpnppbint f^pstimatin^missions needed to support modeling
underlying the 2003 SIP revisionjjphe EPA^»ntory^pance (U.S. EPA, 1999c) allows States
to select any^year from 199^toS9, is encouraged. If the NET is used, it
should be -fogtbe same by the State. Statewide emissions, by
county, are^^tie^JET an^^^^feble electronically through the U.S. EPA Internet website at
www.epa.gov/otxr/()agps/efig/^riiiiSs. If a State is performing a regional or nested regional
*• " - -.••.•«,*-«|!»ps|"!; " ° ^<|r*t^f "•$$¥• roo o
modeling anaIysi^ffli6|MET cari^^qsngserye to provide countywide estimates for locations far
removed from4tauKtph is the^olus of the modeled attainment demonstration. Closer in,
States should qualuyli|a^d improve emission estimates as necessary. The NET may be
used, at a State's discretioiSwhere there have been no previous State-sponsored efforts to
» * " Xlfff^^^Vv''
compile inventories. |H^r
&
15.3 How Do MBonvert Countywide Inventory Information Into Data Used In Air
Quality Models? jf
-"• ; .-,;>..-,. .,,,,^%??'
: / 3 iAir quality models predicting ozone require day specific hourly emission estimates for
VOC, NOx and CO for each cell of a grid superimposed over the area modeled. Typically, there
are thousands of grid cells in a model application. To utilize atmospheric chemistry in the air
quality simulation model, VOC emissions also need to have their component chemical species
identified. We recommend that source specific, local information be used for this purpose
123
-------
whenever possible. The U.S. EPA maintains the SPECIATE data base. SPECIATE can be used
when more source-specific information is lacking. It may be accessed electronically at
www.epa.gov/ttn/chief/software.htmMspeciate. Finally, emission factors for some sources are
dependent on meteorological conditions such as temperature. Thus, meteorological conditions
need to be known to estimate day specific emissions. Emissions models should be used to
account for the numerous and diverse factors which need to be considered to derive emissions
inputs to air quality models. Currently, separate models are used to prepare estirnattfe' from
anthropogenic stationary vs. mobile sources and from biogenic sources.j
Anthropogenic emissions from stationary sou
widely used to convert estimated emissions from station
simulation models for ozone-related applications.. The
used in past urban scale modeling applications for ozoi
in regional applications (Causley. et al.. (1990), U.S. E
emissions model which has had wide use (Alpine Geophr
used in the modeling underlying the U.S. EPA's rule to
EPA 1997h), as well as in other applications of nestedTCgion
voer
•urces foj^
it of thes
ut mojfrrecently hi
iels have been,
lity
pi)). EMS95 is trl
1995). EMS95 has been
|d NOx emissions (U.S.
models.
The version of EPS2 described in Causl
model applications for ozone. However, the
applications of a regional model for partipfnate
performed in the Gulf States (U.S. EP>
Operator Kernel Emissions (SMOK^tfBas had^
similar theoretically to EMS95. it,
newe
ited us
jK) may only for urban scale
sion i&ffich has been used in
ial analysis for ozone
ions model, Sparse Matrix
date (MCNC, 1999). SMOKE is
omputatienally more efficient, reducing time and
memory required to formulate inJjpFidual coj||gl strategies simulated in an air quality model.
Anthropogenic emissions from mobilesoiirces. MOBILES A is the most current
., _ , * ~ W.-M.^J'. *„•*>&&,•,£_,. %^ ^s
lission factors for ozone precursors from a vehicle fleet
representative o»||Specifiea^^^^. EPA, 1994a). The U.S. EPA's Office of Mobile
Sources (OMS)j^i3i^p[oping tr^^feiB^^6 model for highway vehicles as well as a
NONROAD moa||^^|ropve estiffiptil; for off-highway vehicles. These two models are
expected tolbe availa1fleft^tlie|end of 1999. Estimated emissions obtained with the new models
may differ from estimatespbjSined with currently available models. States may track the status
of MOB1LE6 and NONROAD at the following internet addresses:
http://www.epa.gov/omswww/m6.htm (MOBILE6) and http://www.epa.gov/oms/nonrdmdl.htm
(NONROAD model). Jr
••-';>•-,- M"
vf .Prior to the availability of MOBILE6 and NONROAD, States other than California
should%se4MQBILE5A.or .any.update to this .model identified as appropriate by the U.S. EPA's
Office*ofMobile Sources for highway and off-highway vehicles. The website
http://www.epa.gov/omswww/models.htm is a useful source of information on MOBILESa and
mobile source models in general. Resulting emission factors need to be combined with activity
levels (e.g., vehicle miles traveled) to estimate emission levels which have been suitably
124
-------
disaggregated spatially and temporally for use as inputs in air quality models. Methods for
estimating activity levels are included in U.S. EPA, (1997d).
Biogenic Emissions. The BEIS2 emissions model is the most widely used procedure for
estimating biogenic emissions (Geron, etaL 1994 and U.S. EPA, 1997e). This model requires a
mix of land uses to be specified for each county, as well as hourly temperature information. If a
State believes the average land use mix characterized for a county is inappropriate^ certain
gridded locations within a county, this may be overridden for the grid ce^yn ojpition on a case
by case basis. The model makes use of stored information rerarding distribution of
plant species, as well as the provided land use and tempera^^nform^^^^generate gridded^
biogenic emissions.
Table 15.1 summarizes available emissions mo
applications, and identifies some example applications.
Table 15.1. Some Emissions Models Mid
Emissions Model
Re
'*ftffigffiggijgiiS^v
Sponsors (Applications)
EMS95
Alpini
iDCO (eastern half of the
U.S.),
U.S. EPA, OAQPS (eastern
half of the U.S.),
NY DEC (eastern half of
the US.).
f T CS-WP
U.S. B/r.
U.S. EPA, Region IV (Gulf
States),
U.S. EPA, OAQPS
(nationwide)
MCNC, (1999)
NC DEM (Charlotte, most
of NC and parts of
surrounding States)
MOBILE or EMFAC with
'J i, ,:*
-^; activity estimates
' i Jr.' . "* *t ••&.*
U.S. EPA, (1997d)
MOBILE: Many sponsors
(throughout the U.S.
outside of California)
EMFAC: CARB
(California)
BEIS2
Geron, et aL. (1994),
U.S. EPA, (1997e)
U.S. EPA OAQPS (eastern
half of the U.S.)
125
-------
Recommendations. States should use emissions models to convert emission inventory
estimates into emissions inputs required by air quality models. Emission models
require additional inputs concerning chemical speciation, spatial and temporal
disaggregation. Choice of models depends on compatibility with the chosen air quality
model and the application at hand, as well as past experience oj^hose implementing the
modeling/analysis protocol. States should quality " " *—• —-
models prior to making air quality estimates.
15.4 What Should I Do To Quality Assure E
The most efficient means to quality assure (QA) 5
the initial emissions estimation process. The previous!
document, U.S. EPA (1997f), contains a number of Q
develop the basic countywide emission inventory.
modeling, there are three additional quality
first is to compare emission estimates to
from such comparisons to see whether
attention on portions of the inventory
locations, so that a State can confi
NET inventory provided by the U^SBPA maslbe useful
astimates is to apply QA during
quality assurance
ould be used to
ntory is ready for
lues appropriate. The
where^tates can use results
y way. This focuses
Estimates made for other
estimates are appropriate. The
this approach.
Displaying emissions estimates grafffiCallypisSalso a useful means for quality assuring
them. Emissions modelsidei^Bed in Sectiol'ISBican produce graphic displays useful for
quality assurance;,Por exammeilalile plot cifemissions made for a grid superimposed over the
^ J ..-•••- t*l*i»i«>*;«.:-*.. r ot-f
area to be modeledis an effecweanBans for identifying misplaced sources and for assuring
oneself that spatial patterns of emissions are consistent with where sources are believed to be.
i * , "-*'", ","• , '-V"-,'-*"'«•'"• K-.V
Other graphical:displaysinclude pie^charts and time series plots. Pie charts are useful for
assessing whether distribi&pnjof emissions among source types or categories is consistent with
expectations. Time series displays allow a State to look at estimated diurnal patterns in
-. '-if. ':7~tg
emissions to see whether^these appear logical. They enable comparisons to be made for
weekends vs. weekdays io see whether estimated differences appear reasonable.
Comparing emissions with monitored air quality is another means for quality assuring
emissions estimates! As we place increased emphasis on measurements of ozone precursor
species, comparison with monitored speciated data may become an increasingly important means
for quality assuring emissions estimates. Availability .of speciated VOC data, such as those in the
PAMS network or similar data, makes it possible to use monitored observations to apply source
attribution approaches (i.e., "receptor models"). A finding suggesting that air quality
observations are the product of a mix of emissions which differs greatly from that inferred from
126
-------
the inventory can point the way toward parts of the inventory which may need greater scrutiny.
Receptor models and their uses have been summarized by Watson (1997) as well as in Seigneur,
etal. (1997). Use of ambient data from the PAMS network to quality assure emissions estimates
is described in U.S. EPA (1996a).
Recommendations. Quality assurance of emissions estimates is an essential part of the
modeling process, and should be performed on a continual, ongoing basislpStates
should consider the following approaches to quality assurance:^vuna^^A emphasis
during the initial development of the basic emission entorvrafflmnrison with
available emissions estimates performed by ot
emissions model estimates, and comparison with spciated air j
The goal in making proj
that affect future year emissions
analysis the variables that
most, as weii:a%the changesjl
15.5 How Do I Estimate Emissions For Futu
Emissions projections for sources within a mod
nonattainment area will meet the ozone NAAQS by tr
ozone, we require States to estimate future precurs
years before the date of required attainment.
attainment date is distant (e.g., 2010) from tpaate
State may wish to consider projecting emjgfirons to airi
well.
a:e needed to determine if a
Si*
Attainment date. For
me future date—two
hen a required
imittal (e.g., 2003), a
(e.g., 2005-2007) as
to accounjior as many of the important variables
:h Statejas encouraged to incorporate in its
to affect its economy and emissions the
ce place over the next 10 to 20 years.
jSt^^^uld examine the source types that currently dominate its inventory and each
should perform^ffi^pugh calculations Jto see if that source distribution is likely to change much
in the near futur^^Mig|ould suggepthe emphasis that might be placed on projection methods
for predominant soufce^^gories (if there are any). There is normally a wide range of ozone-
precursor-emitting sourcpltypeSiiThus, it is probably only in exceptional cases where there are
one ortwo major sourceJtypes'ihat dominate the inventory. Large point-source emitters in ozone
nonattainment areas are Jlready subject to Reasonably Available Control Technology (RACT)
requirements and, in some cases, control technique guidelines (CTGs), which may be identical.
Therefore, there mayjbe many different emitters in ozone nonattainment areas whose emissions
need-to be trackedfwith time. In cases where there are a few dominant sources, special
- - <' -'", f; -,-t. - —,,(^. V i^^J-W: - * f
techniques should be used to ensure that those sources are modeled using more sophisticated
techniques than those used for the rest of the inventory.
A State's needs for inputs to a grid-based model are a factor in making projections. As
noted previously, grid-based models require source locations (coordinates) as input. Thus, a
127
-------
projection approach that makes its computations at this level is preferred. A less desirable
alternative is to assume that all growth and retirement occurs at existing facilities and that there is
no variation in growth or control within each source category.
Information detailing the different types of projections that might be required of a State or
local air pollution control agency can be found in the EPA publication "Procedures For Preparing
Emissions Projections" (U.S. EPA, 199la). In addition to the necessary types of Ejections,
methods for projecting changes in future air pollution generating activit^^qu^^^ing the
effects of current and future controls, and combining effects of growthJ^Bb^ol are addressed
in this document. Although last published in 1991, much
year emissions is still valid. There have been updates to
1991 guidance (BEA projection phase-out and EGAS
therefore States should review additional documentatioi
U.S.EPA(1998e).
timating future
ovided in
tc.
States may find it useful to examine techniques
where control strategy planning has been performed.
of emission projection preparation are recorded i
photochemical model. In the simplest sense,
and a control factor for each major source
applied in other areas
5 document, examples
a grid-based
Sping a growth factor
Recommendations. States
the U.S. EPA in 1991 an
States should concen
variables histori
uses of their
estimates. States stio
efforts to utilize exis
. .-. ^sJifet. .. •**
emission projections usi
temporal allocations, as
^ ,A -f^^a-i>^£ff^i'^:,;-Kff~,.trf
review pas
Id review guidance on emission projections issued by
itionalxevisions to this document (U.S. EPA, 1998e).
|use of dominant source type information and
conomy. States should be aware of the
MectionsMndlflctor relevant information into their
^"^^^
lew techniques previously used for emission projection
[actions information. States should quality assure their
methods designed to validate the spatial and
any speciation that may be calculated. States should
iriiprojection efforts as part of any subsequent review performed
for the reasons identified in Section 5.0.
128
-------
16.0 How Do I Assess Model Performance And Make Use Of Diagnostic Analyses?
Results of a model performance evaluation should be considered prior to using modeling
to support an attainment demonstration. Performance of an air quality model can be evaluated in
two ways: (1) how well is the model able to replicate observed concentrations of ozone and/or
precursors, and (2) how accurate is the model in characterizing sensitivity of ozone to changes in
emissions? The modeled attainment test recommended in Sections 3.1 and 3.2 u&fPmodels to
predict sensitivity of predicted ozone to controls and then applies result^krel^^e reduction
factors to observed (rather than modeled) ozone. Thus, while&both tyi^R^rormance test are
important, the second type is the most important. UnfortunfflHfc it is^if^Biyifficult to do.
Diagnostic analyses are potentially useful for
used to better understand why the air quality model
insight into whether or not the predictions are plausible^
information which helps prioritize efforts to improve/ref
can provide insight into which control strategies may bey
NAAQS. Fourth, diagnostic analyses can be used to
reasonjpp'irst,
;whajjPioes. This''
MSCS
>vide
J, diagnostic anal}
Inputs. Third, diagnostic tests
stive for meeting the ozone
i control strategy is.
That is, do I reach the same conclusion regarding adjggigy of a using a variety of
assumptions regarding current conditions?
In this section, we first identify
performance. We then discuss each of
no single method which offers a p;
performance be assessed by consi
evidence determination. We th
conclude by identifying sevj
various stagesiof the
.,-- '°"'-i-A ^- ,
Natfc
whi(
se methpas in
for evapating me
a varifi^ of met
[entifynjiodsfoi
for evaluating model
ail. We next note that there is
1 performance. We recommend that
s, much as is done in a weight of
brming diagnostic analyses. We
ntiallyj
Wostic tests which States should consider at
16.1HowiCan I Evalua
formance Of An Air Quality Model?
As notedJabo^Spdel perfomiance can be assessed in one of two broad ways: (1) how
accurately does the*rrf6^^»dict observed concentrations?, and (2) how accurately does the
model predict responseslp|tteftieted air quality to changes in inputs? An example of the latter
type of assessment is, "hgwScturately does the model predict relative reduction factors (RRF)?"
A
Given existingjata bases, nearly all analyses have addressed the first type of performance
evaluation. The underlying rationale is that if we are able to correctly characterize changes in
concentrations accompanying a variety of meteorological conditions, this gives us some
confidenceahatave can correctly characterize future concentrations under similar conditions.
Computer graphics, ozone metrics, precursor metrics and observational models are all potentially
useful for evaluating a model's ability to predict base case air quality.
The second kind of model performance assessment can be made in several ways. One
129
-------
way is by looking at predicted differences on weekends vs. week days, provided reliable
emissions estimates are available for both, and differences in weekend/week day emissions are
substantial. A second way is to examine predicted and observed ratios of "indicator species". If
observed ratios of indicator species are very high or very low, they provide a sense of whether
further ozone production at the monitored.location is likely to be limited by availability of NOx
or VOC. Agreement between paired observed and predicted high (low) ratios suggests a model
may correctly predict sensitivity of maximum (hourly) ozone at the monitored locajrons to
emission control strategies. Thus, use of indicator species methods sho^fepoteMiai for
evaluating model performance in a way which is most closelvyrelated t^^^models will be used
in attainment demonstrations. We recommend that greateraitae^SH^nf these methods.
in the initial demonstration and in subsequent reviews.
performance in predicting sensitivity of ozone to change
after the fact with observed trends.. One reason States
generated in simulating the control strategy selected for
analyses. As explained in Section 5.0, these analyses pr
diagnosing why a strategy did or did not work as expect
opportunity to evaluate model performance in a way
used to support an attainment demonstration.
Erd way fq
i ermssiojPis to
aid retidndata files
ichisc
to facilitate retrosj
|tially useful means for
>rovide an important
to how models are
States can assess model performano
precursor concentrations, corroborative ansflyses wi'
comparisons, ratios of indicator speciesjind retro;
emission trends. These methods arej||5cribed
ozone NAAQS, States should copJiiPl-hoi
and predicted 8-hour daily maxii
tive
ie follo>
iservath
rics, predictions of
lodels, weekend/weekday
•with observed air quality and
ig subsections. For the 8-hour
r and predictions as well as observed
"Big Picture" Assessment Of Model
iajefer to gulp5iot|n>U.S EPA (1991) regarding use of graphics to evaluate
plot prJSlctions and observations. The 1991 guidance describes the
Statessa
model perfo:
following,,graphical*di^p8iL^^rne series plots, tile plots, scatter plots and quantile-quantile (Q-
Q) plots;; Each of these%^j|l^^can also be used to display differences between predictions and
their paired observations^Si§phics are useful means for understanding how predictions and
observations differ. For;example, time series plots tell whether there is any particular time of day
or day(s) of the week when the model performs poorly. Tile plots reveal geographic locations
whem the model performs poorly. Information from tile plots and time series may provide clues
abouMvhere to focus quality assurance efforts for model inputs. Scatter plots and Q-Q plots
,-. ,,,jyA.-.4-«8%«-..v.!-*,-,<«,.".'SI'S,' * J r •* -»-•». r
show whetherthere is any part of the distribution of-observations for which the model performs
V „, * "^iX-v,^,.-. j . "
pooSyt^These plots are also useful for helping to interpret calculations of bias between
observations and predictions. For example, they could show large differences between
observations and predictions which just happen to balance, producing low estimated bias.
130
-------
16.1.2 How Can Ozone Metrics Be Used To Assess Model Performance?
Ozone metrics produce numerical comparisons between observations and predictions.
Appendix C in U.S. EPA (1991) identifies several metrics, as well as the mathematical formulae
for calculating them. We recommend that comparisons of observations and predictions for 1-
hour sampling times be used as well as comparisons of 8-hour daily maximum concentrations.
One-hour comparisons of metrics provide a much larger data base for assessing
performance than would otherwise be available.
States should calculate metrics which are closely
the recommended modeled attainment test (Sections 3.1
are used to calculate relative reduction factors (RRF) n<
taking the ratio of the mean highest 8-hour daily maxi
future to that estimated with current emissions. Thus,
mean 8-hour daily maxima is an important indicator of
States use the following set of ozone metrics to assess
corresponding to the selected episodes.7
suits are used in
test, modejr
ved
irmance. We recommend that
lance during base periods
to ho
.2). In t
onitorinSsites.
concentrations calc
dictions of highest
averages of 8-hour daily
ies should be taken from
1) Estimate bias between spatially paired
8-hour daily maximum ozone concentri
maxima observed and predicted ovey|everal d
grid cells "near" a monitor, as defijpain Tab,
maximum for each day is calculj|ua as illustrated in Fjfure 3.2. The comparison described
in this test leads to a separatej|jliriate ofjjperage biajtn predicted 8-hour daily maximum
ozone for each monitoring location.
nearby" 8-hour daily
2) Compute a correlajaojp^efficient andi^p[ay a scatter plot for the average observed and
pred&fedjyiuDur daily^^^Aised to estimate the average bias in test 1.
al correlafio!|ieoefficient of observed and nearby predicted 8-hour daily
tidily averaged by day. If there are a sufficient number of monitoring
3) Compute^
maxima whic
sites foftheanaTfsi|l^e?;,;rneaningful, it is also useful to group concentrations from those
monitors that repre^enj^jp|yand, downwind and center city locations. Include time series and
scatter plots of the rqpoltsT
f<
^4) Prepare quantilejquantile plots of observed and predicted 8-hour daily maxima using (a)
jail data pairs (Lpfsample size = (# of stations)(# of days)), (b) spatially paired mean 8-hour
maximafjfie., sample size = # of stations), and (c) temporally paired spatially averaged
axima (i.e., sample size = # of days).
7 "Bias" and "fractional bias" are calculated as described in Appendix C to U.S. EPA (1991).
In the text, we discuss 8-hour daily maxima. Similar tests could be performed for observations and
predictions of 1 -hour daily maxima.
131
-------
5) Calculate the fractional bias for the pairings described in tests 4 (a), (b) and (c) above.
It is not possible to provide a definitive set of performance criteria for the preceding
metrics, since we do not know how sensitive the model's response is to failure to replicate base
case observations. However, we suggest the following as a performance goal for tests 1 and 5.
The bias should be less than about 20% of the mean observed 8-hour daily maximum (test 1)
and the spatially paired fractional bias (test 5(b)) should also be les^foan about 20%. These
goals should be met at locations with monitored design jalues exo^^^^ne NAAQS and at
some of the other locations as well.
The "20%" goal is based on information like that sho
mean RRF values do not appear to be very sensitive to
unless these values are less than about 70 ppb. Seven
the lowest concentration which exceeds the level specifi
difficult one to meet, particularly if the mean predicted
This is one reason why we suggest simulating several
Performance goals for tests 2, 3 and 4
by case basis. This follows since the meanr
between paired observations and predicti
Assuming there is substantial variabilit^lti mean
a general goal is to have a moderate^
predictions which is statistically s;
5 ppb,
QS. This goal may be a
a single day's prediction.
J2.0.
established on a case
lation coefficients
y in the observations.
ily maxima from site to site,
between observations and
We emphasize that the«discussion inifhelpceclSIing paragraphs addresses performance
j *u M. • • * r u i j i-
goals ratherthan criteria.^3amasfto meet one«CKRmore of the goals does not mean that an
, , " ''*^ty£'%,^^ij%ipb , ' ^vvi-'" ^
analysis cannot proceed. TJffiaecfeion can 'only be made on a case by case basis. However,
" * " '**^W, vjj l^'J-Srj^^^aJ^A,-
success or faflurfc o|ra modert^m^g|«rformance goals should be noted, and may be considered
in a weight <
The precedrng":fiv(Ei:performance measures are oriented toward site by site comparisons.
Other useful measures mvdJfelpooling these data to calculate overall bias and gross error for 1-
hour predictions as well as€of$-hour daily maxima. The three most widely used pooled metrics
for ozone have been unpaired 1-hour daily maximum concentrations, normalized bias and gross
error. In past guidance we have identified performance criteria for these three measures. These
criteria are based on.results obtained in urban model applications (primarily in California) during
the 1980's. This,information may serve as one input in assessing how well a model performs.
If a State is primarily interested in showing that a strategy works for meeting the 8-hour
NAAQS within or downwind of a nonattainment area, it may be useful to subdivide the
monitoring sites into "downwind", "center city" and "upwind" categories on each modeled day
rather than pool the entire data base. Pooled ozone metrics could then be calculated for each
132
-------
category. Note that the identity of "upwind", "center city" and "downwind" sites could change
from day to day. However, the aggregated metrics for all modeled days would be valid for the
three defined categories of sites. This partitioning of sites may be considered using any ozone
metric, providing there are enough sites available to support making such a partition. An output
from such a grouping might be something like, "average bias or average fractional bias from all
upwind sites, from all center city sites and from all downwind sites".
16.1.3 How Can I Use Available Precursor Observations To
Performance?
Ozone models have many degrees of freedom.
predict similar ozone concentrations using a variety of
a comparison of observed/predicted ozone is not a defi
Testing the ability of the model to predict other species
increasing confidence in the results.
Js
that you can
States should include an assessment of how w
species treated explicitly in the model's chemical
data base permits. One concern however about^ontri
the monitored data represent? Monitored pri
and VOC species, could greatly depend species like NO2, or aggregates including
secondary speaes^ke NOx'ol^^'he rationale for this strategy is that time is required for
these secondary%pj|psjto form,^d«u^nnitting greater mixing on the scales assumed in the
model. A second^ra^gyj§ to consider ratios of primary (or secondary) pollutants which tend to
co-vary. For example^bblBwed ratios of one or more selected VOC species to CO are likely to
be less variable than concentrations of the individual pollutants. Therefore, the ratios may be
more representative of thj?sciles considered in the model. A third way for reducing
incommensurability is tosuse metrics which entail spatial averaging in some manner.
Comparisons between ^patially averaged observations of VOC at 3 monitoring sites with
spatially averaged model predictions "near" the 3 sites is an example of such an approach.
/*• ••"'"•
- ^--...-.f^W
16.1.4 How Can I Use Corroborative Analyses With Observational Models To Help
Evaluate Air Quality Model Performance?
Recently, techniques have been developed to embed procedures within the code of an air
quality model which enable users to assess contributions of specific source categories or of
133
-------
specific geographic regions to predicted ozone at specified sites (ENVIRON (1997), Yarwood, £t
aL (1997,1997a), Morris, sLal., (1997), Yang, et al.. (1997,1997a)). These source attribution
procedures characterize what the air quality model says are the effects of targeted areas or
sources on predicted air quality. Provided speciated VOC data are available at a site, source
attributions estimated with these approaches can be compared with those obtained using other
models which rely directly on observed air quality data.
The chemical mass balance model (Watson (1997)) is probably
applicable observational approach for this purpose, since it can focus
considered with the air quality model. Cautions raised pre^^ply ab
monitored data continue to apply. Available multi variatgpatistical
Henry, et al.. (1994) and Henry (1997, 1997a, 1997b)) rjf providejpfhore
assessing an air quality model's performance. Multi vaUle statisecal models
ptrectly
ic day(s)
atativeness of th
• example
examining temporal variability in monitored precursor
variability on one or a few occasions at many sites. A qul
can contrast observations on days when winds suggest
when a contribution is likely, or at locations where a source cat
it isn't. If the observational approach suggests a mjjarjc^iange in
and the air quality model also suggests that cat
wind conditions, the observational model let
a single site or s|
imparison is possible if one
ution is unlikely vs. days
irtant vs. those where
itegory contribution,
jortant under similar
ly model's predictions.
16.1.5 What Data Bases Reflecting Chi
Model Performanc
In
ions Are Available To Evaluate
Activity levels and patter
point sources, may differ
simulating*weekend as
model predicteAe-effects^Ql
compare mean
weekends fvs;."wee
leadinfrioilprecu
;nds vs.
;k days
ng emissions.
issions from mobile, area and some
these differences are substantial,
ride a means for evaluating how accurately a
Weekend^
iy informatioricoiild be used in one of two ways. The first way is to
'•^^w^^^,^ yam* - » •*
id mean^fopBrved 8-hour daily maxima at each monitoring site for
aysijifeere are\ sufficient number of monitors available, it is also
desirable to make these^coln^amons for categories of monitors, grouped according to whether
they .represent "downwinfl'^center city" or "upwind" conditions. Tests 1-5, described in
Section 16.1.2, as well as.other tests, could be applied first for week days and then for weekend
days. If the performance is adequate for both weekends and weekdays, this suggests that the
model is accurately characterizing composite effects of different meteorological conditions and
different emissions?
•''..,^: A second way for using weekend/week day information is to first screen the available
data to identify weekend days and week days for which meteorological conditions are "similar".
For example, for urban analyses, wind orientation, daily maximum surface temperature, presence
of precipitation and maximum mixing height might be considered for this purpose. If similar sets
134
-------
of meteorological conditions are identified for weekends and week days, changes in mean
observed 8-hour daily maxima can be compared with changes in mean predicted 8-hour daily
maxima for each monitor site, as well as for groups of sites characterized as "upwind", "center
city" and "downwind". Tests like those described in Section 16.1.2 may be used for this purpose.
If predicted changes generally provide an unbiased estimate of observed changes, this suggests
that the model characterizes effects of changing emissions accurately.
We need to mention several caveats regarding weekend/weekda;
changes in emissions between week days and weekends
associated with the weekend and week day estimates. S
into weekends and week days may mean that conclusion
identifying "similar" meteorological conditions may be
changes between weekend and week day emissions ma;
needed to meet the NAAQS in some, areas. Since the
concentrations may be nonlinear, the weekend/weekday
evaluations. Despite these reservations, weekend/week
relatively few means for evaluating a model's ability t
concentrations. States should include these comp;
performance, whenever feasible.
ions. First,
:o uncertainties
small) sample
in their
16.1.6 How Do I Use Ratios of
A performance evaluation
ratios of indicator species carries a larj
reveal whether the model is predJlsng sens.
correctly. That is, when rnetaJSel predic
... , •-*£ • f
sensitive u*changes in aj»a|ilir one of c
predictionsjan^ensitive W,
indicator specieJSftea that ol
. ' *';«"#W«
provides somexonfidence that
ozone
is may not be definitive
s provide one of a
Changes in ozone
aluate model
ite Model Performance?
:omparispns between modeled and observed
JtentiaLpvantage. Such a comparison may
of to VOC and/or NOx controls
a certain range, predicted ozone is
ors. Within another range of ratios,
anothefprecursor. If a model predicts observed ratios of
predicted ratios fall within the same range of ratios, this
change in ozone may be accurate.
For ozone modelingJappJications, uses for ratios of indicator species are described by
Sillmam(:1995, 1997, i%8^g|by Sillman, ejLaL (1997a). The authors of these references have
shown several ratios of inetiiSaior species to be good indicators of whether peak predicted (i.e.,
modeled) ozone is likelyjb b>e most sensitive to reductions in VOC or NOx. Many of the species
discussed require measurements beyond those which have been routinely made by most State
agencies. Of the ratkJS'discussed, the following involve compounds or mixtures most amenable
ten inelasurementbylState agencies: O3/NOy, O3/NOz8 and O3/HNO3. States should review the
Sillman and Sillman. et al. references for further details about measurement requirements.
Strength of the indicator species approach for assessing model performance depends on
"Note: NOz = NOy-NOx
135
-------
an assumption that the model is accurately characterizing the relationships between indicator
species and ozone. The validity of this assumption can be more readily tested in smog chamber
experiments than can absolute predictions of ozone. A second precaution is that there may be a
range of observed ratios for which the preferred direction of control is not clear. When this
occurs, agreement between predictions and .observations does not necessarily imply that the
model correctly predicts sensitivity of ozone to changes in precursors. Third, this method
requires more measurements than are commonly made. In some cases, it may be ^difficult to
achieve the required precision with routine monitoring. Finally, much Qj&he wjix done to date
with indicator species has focused on peak hourly concentrations of oraHlicabilit of the
approach to 8-hour averaging times has not yet been extensfflp|» teste
precautions, the approach of comparing predicted and obJPFed ratios
provides a means of assessing a model's ability to accui
ozone to changes in precursors. States should use the:
performance, whenever feasible.
air quality and estimated trends in emissions.
most interested in—does the model accura
as straightforward as it seems. Often, inrjgPEstimat
ambiguous and the emissions trends areJfu^itativjfFAlso?1
constant meteorology, which does ncrtjipppen. Jane of the j
described in Section 6.0 is to mak^^possible^r others
dates.
16.1.7 Are Retrospective Analyses Useful For
Retrospective analyses compare past model|iiaiuality pro
xxiel Performance?
ith observed trends in
ssment of what we are
Ility? However, it is not
s used in past studies are
studies generally assume
ses of the reporting requirements
plicate modeled analyses at future
Infection 5.0, weMptellihat a retroJieSiilPanalysis is an important means for
, •-.-• .s •tf-.-.-fif A i;t ^*2i5**v
diagnosing why%NAAQSiia^OE4ias not been attained. Such an analysis provides assurance that
i * >4 v"'*iv,ji' i ^ji.'V^^^*-^^^- ... . .
improved aiftquauj^|results froin;cnanges in emissions rather than meteorology and/or can
identify reasohsTi^t^atisfactor^B&jsessas not being observed. Retrospective analyses will
•' *•* •$•*,«"*' ^tpy&xj*. „ , J ^Sis? ^^^f '- fy f' ° ** "*
have an ancillary(bJnefiC«f^)rovidmgph additional means for evaluating model performance. In
, ;- ' ; ^^Wt^V^TCwt^^^i--- _ 1^'
order to ensure some^lannm^pr subsequent retrospective analyses and to promote some
uniformity in the methodJM|pp>r these analyses, they are probably best performed as part of a
subsequent review rather^hafeis supporting evidence in the initial SIP revision.
16.1.8 All Of TCiiese Performance Tests Have Shortcomings, So What Do I Do?
e is nosingle definitive test for evaluating model performance. All tests have
strengths ^and weaknesses. Credence given to model results is increased if a variety of tests is
appHelario' the outcomes either support a conclusion that the model is working well or, at least,
are ambiguous. Thus, one can think of a model performance evaluation as a "mini-weight of
evidence analysis" focused on the issue of how much credence to give model results in an
attainment demonstration. Table 16.1 summarizes the tests and their corresponding objectives or
136
-------
goals described in this guidance.
Table 16.1. Summary Of Methods To Evaluate Performance Of Air Quality Models
Method
Test(s)
Goals/Objectives
Big Picture Assessment Using
Graphics
-tile plots of observations &
predictions.
-tile plots of differences in
observations & predictio
-scatterplots & Q-Q plo
-time series plots
-complement ozone metrics
ether performance
low
erfonnance-is^m^Snce
better downwind uumnupwind?
i diagnostic tests on certain
cations
Ozone metrics
-bias pred/obs
hr) daily
monitor
E&i-
i^corr elation coefficient
~bias (8-hr daily max & 1-hr
obs/pred), all monitors
-gross error (8-hr daily max & 1-
hr obs/pred), all monitors
-partition pooled data base into
"upwind", "center city" &
"downwind" sites. Repeat
analyses
-scatterplots & Q-Q plots of 8-hr
& 1-hr metrics
monitors (8-hr
ns only)
most monitors (8-hr
omparisons only)
-moderate to large positive
correlations
-5-15%
-30-35%
-get a better idea of what parts of
the distribution of predictions &
obs agree or disagree & whether
there is any obvious pattern to
the model's performance
137
-------
Table 16.1. Summary Of Methods To Evaluate Performance Of Air Quality Models
(continued)
Method
Test(s)
Goals/Objectives
Precursor concentrations
Similar to ozone metrics. Focus
on
-secondary species (NO,, NOy.
NOx,NOz)
•ratios of co-varying sp
(VOC, or VOC species/'
or VOC species/NOx)
-spatially averaged predi
the above or of primary
-provide means fogassessing
whether model prformance
if model
is tam&mMX better ozone
met
Observational models
compare source attribution
estimates with >
models
i obsei
ttribution & CMB
lar source categories
iportant contributors
to observed precursor
Dcentrations
-day to day variability in air
quality model's source
attribution & observations or
mulit variate models is consistent
-these are qualitative
comparisons
138
-------
Table 16.1. Summary Of Methods To Evaluate Performance Of Air Quality Models
(continued)
Method
Test(s)
Goals/Objectives
Weekend/week day comparisons
-compare previously identified
ozone (& precursor) metrics on
weekends vs. weekdays
-if data base permits, partitij
data base into meteoric
classes. For each class co
differences in weekday v|
weekend predictions wit
differences in weekday'
weekend observations.
Objective is to tesUnodel's ability
to accurately reproduce effects of
,«%.
-pool data base to con
and gross error on weekends and1*
week days.
-if data I
pooled dat
"center^
bins i
Ratios of indicator species
ted and observed
•
ipare p:
^ratios at tii
observed
*
The following ratios are
recommended for comparison of
'predictions & observations
•* **
-Oj/NOz
-Oj/HNO,
Guidance refers to Sillman
references to identify ratios
where max.hourly ozone is likely
limited by NOx & ratios where
availability of VOC limits
max.ozone. Predictions &
observations should fall into the
same class (i.e., VOC-Iimited
cases, NOx-limited cases, cases
where it is too close to call)
139
-------
Table 16.1. Summary Of Methods To Evaluate Performance Of Air Quality Models
(concluded)
Method
Test(s)
Goals/Objectives
rform diagn
determine wheth
due to
Retrospective analyses
project ozone to a future
(preferably sooner than
attainment date) year
retain files
update emission estimat
future year & note obse:
future ozone
characterize future meti
& model in future
rily for a
igreement
in projected
emissions estimated
te
ences in assumed
orlogical conditions
-a combination of different
meteorological and emissions
assumptions
-one or more limitations in the
model.
Finally, we need to address:$!j|issue of adjusting model inputs to improve model
performanceii'®nem.lQiiBliKasons we^recommend a variety of tests for model performance is to
*" • ^'^^s^'^&^&Sife:; '
reduce the possibility 6fJ|gejtjng|he right answer for the wrong reason". We recognize however,
that many of the inputs Utlnoaels have associated (often unknown) uncertainties. It is acceptable
J r >-•<'," VH*'
to adjust inputs within reasonable bounds to improve performance, providing it does not result in
poorer performance in any of the several measures of performance which we recommend in
Sections 16.1.1 -. 16.yS*. If such an adjustment is made, it should be documented and
accompanied by anixxplanation as to why those implementing the protocol believe it is justified.
Recommendations. States should undertake a variety of performance tests. Results
from a diverse set of tests should be documented and weighed to qualitatively assess
model performance. Provided suitable data bases are available, greatest weight should
be given to tests which assess model capabilities most closely related to how the model
140
-------
is used in the modeled attainment test A narrative describing overall assessment of
model performance should be included among the material submitted to support a
recommended SIP revision requiring a demonstration of attainment
16.2 How Can I Make Good Use of Diagnostic Tests?
Diagnostic tests are performed using one of two broad approachj
consists of tests in which sensitivity of air quality predictions^) pertuj
combination of model inputs is examined. This is the me
has a longer track record. When it is applied, States shoij
the modeled attainment test recommended in Section 3.|
sense to provide relative reduction factors. Relative rec
ratio of mean 8-hour daily maximum concentrations
kThef a controlHsfirategy. For example, States can consider effects of
assumed boundarjicraSffiflons and iroporological assumptions on predicted effectiveness of a
control strategy If tn^jgntrp||Strategy appears to work for a variety of assumptions, this
increases confidence iri%ejmwel results. Second, models used to support ozone NAAQS
attainment demonstratioi^ajpresource intensive. Sensitivity tests provide a means for
prioritizing use of resources in applying the model. For example, how sensitive are relative
reduction factors to usejbf more vertical layers or smaller grid cells? Is using 4 km (rather than
12 km) grid cells mpcfimportant than simulating many days? Third, sensitivity tests may help
prioritize additional data gathering efforts so that a better subsequent review/diagnosis can be
performed at the time of required attainment. Finally, sensitivity analyses could be useful for
prioritizing control efforts or for noting sensitivity of predictions to uncertainties in the current or
future emission inventory.
Sensitivity tests can and should be applied throughout the modeling process, not just
141
-------
when model performance is being evaluated. Tests should be selected on a case by case basis by
those implementing the modeling/analysis protocol. We present a sequence of activities likely to
be followed in applying an air quality model. Under each activity we list some sensitivity tests
which might be useful to resolve certain issues which may occur in some locations. The list is
intended for illustrative purposes. The identified tests are not mandatory, nor is the list a
comprehensive one.
ler using a nested
asecase
Strategy Selection
-Simulate across-the-board reducti
combinations of the tw
strategy?).
-Simulate reductions in pointfvs. area vj
should my strategy, focus on?)
-Simulate teductions'in^bOTttidary condi
Model Setup
-Change boundary conditions (is domain size adequate?^lo I neec
regional model?).
-Alter initial conditions (do I need to extend the rarnjipp period IJ
m
Performance Evaluation/Troubleshooting
-Alter grid cell size and/or number of vertical layei
8-hour daily maxima affected?).
-Perturb specific inputs (e.g., mixing height, cloud
certain processes are identified as impo:
(are results affected by perturbations wjj
measurements should I try to make.
ich might explain why
is (see Section 16.2.2)
what additional
Ox emissions and
' be thinking of for my control
rce emissions (what types of sources
concert with reduced emissions in the
t;;i;non)^inment3area^all my strategy need additional help from regional controls in
*ve Use In Weirfit Of Evidence Determinations
Uncertainty Estinratelir*For Ou
-Simulate sele^l^!Stra|e^~starting with a different current inventory, reflecting reasonable
: uncertainties^fcupwnt emissions.
-Simulate selected strategy, but with different (reasonable) growth projections.
/^-Perturb meteorological inputs, like mixing heights or cloud cover, which may be poorly
„ '$* ,• .
:z;' characterized but which earlier analyses have suggested may be important in
; .;^ affecting-base case predictions.
Vf-Simulate selected strategy using different grid cell sizes and/or a different number of
I '•i/^^^T^'-^St''^^" J>-t|
layers.
16.2.2 Use of Process Analysis
Occasionally a review of a graphical display, like a tile diagram, may indicate a limited
142
-------
number of locations or incidents which bear further investigation. Diagnostic tests may be used
to perform focused analyses on these sites or incidents. These tests entail a more detailed look at
a time series of predictions and (if available) observations at or above a site, including chemical
species, winds and mixing. The examinations can be done qualitatively. However, more
quantification is possible using the second .type of diagnostic test described at the beginning of
this subsection. A procedure called "process analysis" is an example of the second type of
diagnostic test. Process analysis has been used to assess relative importance of vajpftis model
assumptions as well as simulated physical and chemical phenomena conjjtoutinjpo a predicted
ozone concentration at a particular time and location.(Jeffries^ 1994, l£HBP™ies> etal. (1996),
Jang, sL§I (1995), Lo, sLaL (1997)).
Process analysis requires a substantial amount o]
advantage. However, useful insights are also possible
focuses on selected grid cells. Process analysis then ti
grid model addresses physical and chemical factors affecl
example, a typical sequence followed in a model for
advection of ozone and precursors present at the beginning
emissions added during the time step, (3) vertical
emissions, (4) estimated cloud cover and its el
chemistry involving advected and diffused
certain compounds. Process analysis exanes inci
predictions from hour to hour attributabjlnto eachjflFthe
one gets a sense of how important e^lSrocess JFas a com
specified time and location.
If a focused diagno
chemicaiypHysl
*".'*• -
emissions m
prediction is
prediction in
;pertisett|ft>eintel
less detailed analy&qHHpceduie
ige of the fact tin^pmnerical
in a sequential manner. For
., 1 hour) might be (1)
(2) precursor
material and fresh
of
of the1
lysis rafl
missi
atmospheric
s, and (6) deposition of
changes in ozone
Sescribed above. In this way,
lutor to predicted ozone at a
ined with process analysis, suggests a
iodel assumption rather than a result of real
:atmosphOT^ps^sses, States may wish to go back to the meteorological or
ffKerify thlffipieSiiipUts and assumptions that have been used are correct. If a
•Mf.is "> * V^^V<>y* *^? -•**•'-
g£ an apparent^pFact which cannot be resolved, States may discount that
.jdemoristralfon.
Recommendations. States should include diagnostic analyses throughout the modeling
process used to help select a control strategy which demonstrates attainment. These
analyses should include sensitivity tests to assess robustness of a proposed strategy and
consequences of simplifying assumptions made in the modeling. Additional sensitivity
tests may be warranted on a case by case basis. Sensitivity of relative reduction factors
to input perturbations should be a prime focus of the tests. Provided capabilities have
' A'-rfc-if _J> *""'.iJ^Y',- > #^£tf> fff ' * *
been properly installed and tested, States may use versions of a model's code which
contain capability for tracing importance different phenomena as contributors to
predicted ozone concentrations at selected locations.
Table 16.2 shows examples of diagnostic tests which may be useful during different
143
-------
stages of a modeling analysis.
Table 16.2. Potentially Useful Diagnostic Tests At Various Stages Of Modeling
Stage of Modeling
Test(s) (Examples)
Purposes)
Model Setup
-change boundary conditions
-isdoi
icientiy large?
a nested
will an urban
Performance Evaluation &
Troubleshooting
ire ozone pr
-alter specific (unce:
(e.g., mixing heights,
cover).
-alter grid cell size or
vertical layers considei
•what is the effect oifother
rformance tests (e.g., precursor
ins, weekend/weekday
ices, indicator species
iorities should I assign to
kinds of improved
urements?
Strategy Selection
-simulate
across the board
,NOx
ns & combinations «f the
mobile vs.
. 9^gf3^8$&&$%l8&»~ .
point vs. area source categories
>
•perform the two preceding sets
of tests with and without
changing boundary conditions
-what sorts of strategies (VOC vs.
NOx, emphasis on which source
types) should I be considering?
-will additional regional
reductions in precursor emissions
be necessary?
144
-------
Table 16.2. Potentially Useful Diagnostic Tests At Various Stages Of Modeling
(concluded)
Stage of Modeling
Test(s) (Examples)
Purpose(s)
Estimating Uncertainty
-simulate alternative base cases
in emission estimates & project
AQ from the alternative bases
-simulate future AQ using
alternative (reasonable);
assumptions
--includes different grov
--different placement off
sources
-perturb meteorological ij
which cannot be well
characterized with available data
simulate select
smaller (e.
-assign a range (e.gi«± 1 std.dev.)
of predicted RRpif based on
iicted future
& cumntSoran 8-hour daily
values with the rang£of RRF's.
the preceding information
,. ively as an input in a
iiitetennination
Focused performance analysis
•proc
lo suspicious looking results
make physical sense?
145
-------
17.0 References Cited In Part II
Alpine Geophysics, Inc., (1995), The Emissions Modeling System (EMS-95) User's Guide,
Alpine Geophysics, Inc., Boulder CO.
Blanchard, C.L., F.W.Lurmann, P.M.Roth, H.EJeffries and M.Korc, (1999), "The Use of
Ambient Data to Corroborate Analyses of Ozone Control Strategies", Atm0jp>heric
Environment, 33, pp. 369-381.
CARB, (1996), Performance Evaluation ofSAQM in Cen
Demonstration for the August 3-6, 1990 Ozone
Causley, M.C., J.L. Fieber, M. Jimenez and L. Gardner]
Airshed Model, Volume TV: User's Manual for,
EPA-450/4-90-007D, U.S. EPA, Research 21111, (NTIS No.: PB91-
131250).
Chang, J.S., S.Jin, Y.Li, M.Beauharnois, C.L.Lu and|||Iuang, Quality
Model, Prepared by Atmospheric Scieno^fi9^Center,aHRMveisity of New
York (Albany).
Cox, W.M. and S.Chu, (1996), "Assessjjent cflnd^aStXSfvana&on in Urban Areas
from a Climatological Persrjsgife", Atmjipheric EJvironment 30, pp.2615-2625.
Dennis, R.L., (1994), "Using rnejRegional |ffl Depraffin Model to Determine the Nitrogen
Deposition Airshed^h|chesape^^^^Hrshed", Ch.21, Atmospheric Deposition
of Contaminants Jo^^eatLake^^^Ktal Waters, J.E. Baker (ed.), SET AC Press,
Deuel, H.P. andJS^G^iglas, (l%9^jjjj)isode Selection for the Integrated Analysis of Ozone,
Visibili^j^^^^Deposiiimyor the Southern Appalachian Mountains, Draft Technical
Report Systeil^^|cations International, Inc. (SYSAPP-98/07), prepared for the
Southern AppaIac^i|f|Mountains Initiative.
-•"
Dolwick, P.O., CJang and ilTimin, (1999), "Effects of Vertical Grid Resolution on
; PhotochemicaljSrid Modeling", Paper 99-309, Presented at 99th AWMA Meeting,
, ; StLouis, Mj^June 1999).
Douglas, S.G.^R!C. Kessler and E.L. Carr, (1990), User's Guide for the Urban Airshed Model,
Volume III: User's Manual for the Diagnostic Wind Model, EPA-450/4-90-007C, U.S.
EPA, Research Triangle Park, NC 27711, (NTIS No.: PB91-131243).
147
-------
ENVIRON International Corporation, (1997), User's Guide to the Comprehensive Air Quality
Model with Extensions (CAMx), Novato, CA, April 17, 1997.
Geron, C, A.Guenther and T.Pierce, (1994), "An Improved Model for Estimating Emissions of
Volatile Organic Compounds from Forests in the Eastern United States", J.Geophys.Res.
99,pp.l2,773-12,791.
Grell, G.A., J.Dudhia and D.R. Stauffer, (1994), A Description of the
State/NCAR Mesoscale Model (MM5), NCAR/TN-39&+STR,
Haney, J.L. and S.G. Douglas, (1996), Analysis of the
Simulation Results for the Atlanta Nonattainme
International Technical Memorandum SYSAPP,
Center participants, October 21,1996.
Henry, R.C., C.W.Lewis and J.F.Collins, (1994), "Vehi
Compositions from Ambient Data: The GRA
Science and Technology 28, pp.823-832.
ling
•ocarbon Source
Environmental
Henry, R.ep|>97b), "R
A^pbrfionment o
'
Jang, J.C., H.
Henry, R.C., (1997), "Receptor Model
Description", J.AWMA 47, p.216
Henry, R.C., (1997a), "Receptor M
Apportionment of Airbo:
p.220.
App
.APS) Part I. Model
Space (RMAPS) Part H.
eject MOHAVE", J.AWMA 47,
Patterns in Space (RMAPS) Part UJ.
IParticulafrSulfur in Western Washington State", J.AWMA
(1995), "Sensitivity of Ozone Model to Grid Resolution
Part B: DeTallepPrpSsss Analysis for Ozone Chemistry", Atmospheric Environment 29
Jeffiies, H.E., (1994), "Brocess Analysis for UAM Episode 287", memo to Brock Nicholson, NC
DEHNR, ApriL&f 1994. Available at the following Internet address
"ftp://airsiteMhc.edu/pdfs/esejunc/jeffries/projprpt/uammod", then download file
"panel287:pif".
-* f ,.«,*=*• x^w r^e-> «.»**.
148
-------
Jeffries, H.E., T.Keating and Z.Wang, (1996), "Integrated Process Rate Analysis of the Impact of
Nox Emission Height on UAM-modeled Peak Ozone Levels", Final Topical Report,
Prepared for Gas Research Institute, Chicago, IL. Available at the following Internet
address: "ftp://airsite.unc.edu/pdfs/ese_unc/jeffries/reports/gri", then download file
"uncgri.pdf".
Jeffries, H.E., (1997), "Use of Integrated Process Rate Analyses to Perform Sourcrffttribution
for Primary and Secondary Pollutants in Eulerian Air Quality Mj|fels",Jplsented at U.S.
EPA Source Attribution Workshop, Research.Triang^Park, ^fflfflfP^-18 1997.
Available at the following Internet websites,
"ftp://airsite.unc.edu/pdfs/ese_unc/jeffries/ipr
"sourcewksp.pdf, and "http://www.epa.gov/
file, "Source Att WS-Proc.analysis/mass track-
Kumar, N., M.T.Odman and A.G. Russell, (1994), "Mul
Application to Southern California", J.Geophysl
ity Modeling:
pp.5385-5397.
Kumar, N. and A.G. Russell, (1996), "Multiscale Ajsjj^iality Moaejing|6iaie Northeastern
United States", Atmospheric Envira
LADCO, (1999), Personal communicatioj
Lo, C.S. and H.E. Jeffries, (1997), •^guantitattfe Technijpefor Assessing Pollutant Source
Location and Process Con^^ffion injpotochernijti Grid Models", Presented at annual
AWMA Meeting, Toron^pntario^^>7). AJgftvailable at the following Internet
address, "ftp://airsite^H^idu/pdfs^^^^^ries/ipradocs", then download file
Lyons, W.A^j^^pemback^ip^Mgielke, (1995), "Applications of the Regional
Atmosp1jen<|iaodeling S^^Ml\MS) to Provide Input to Photochemical Grid Models
for theJaifcrfttinpgan OzoWsIdy (LMOS)", J.Applied Meteorology 34, 1762-1786.
MCNQi(i999), Intemetfweb'slte-describing the MAQSJP air quality model.
i| t'http://www.iceis.m^nc.org/products/magsip/"
Millmchus, M.L., S.T.,Rao, I.G. Zurbenko, (1998), "Evaluating the Effectiveness of Ozone
ManagemenjkEfforts in the Presence of Meteorological Variability", J. Air and Waste
KT 7ii3Managemenf49, pp. 174-188.
™
loris,ftS.tb.M.Wilson, S.B.Shepard and K.Lee, (1997), Ozone Source Apportionment
Modeling Using the July 1991 OTAG Episode for the Northeast Corridor and Lake
Michigan Regions, (DRAFT REPORT), Prepared for Mr. Dan Weiss, Cinergy
Corporation, Plainfield, IN.
149
-------
NOAA, (1999), website, "www.arLnoaa.gov".
Odman, M.T. and C.L. Ingram, (1996), Multiscale Air Quality Simulation Platform (MAQSIP):
Source Code Documentation and Validation, MCNC Technical Report ENV-96TR002-
v 1.0, 83pp.
Pielke, R.A., et al., (1992), "A Comprehensive Meteorological Modeling System-I
Meteor.Atmos.Phys., 49, pp.69-91.
Scire, J.S., R.J. Yamartino, G.R. Carmichael and Y.S. Ch;
Photochemical Grid Model. Volume II: User's
MA.
Scire, J.S., F.R. Francoise, M.E.Fernau and R.J.Y;
CALMETMeteorological Model (Version 5.0),
C"
>j >
Mesoscale
'., Concor
Seaman, N.L., et al.. (1995), "A Multi-Scale Four-Di
Applied to the San Joaquin Valley During
Performance Characteristics", JAppliedj
Seaman, N.L. and D.R. Stauffer, (1996)
San Joaquin Valleywide Air Po,
California Air Resources B
gnilation System
ing Design and Basic
761.
.*«-.-
liiModel. Final Report to
>3**-' *
ical Support Division,
'4pp.
Seaman, N.L., et al.. (1996b), "A^fication Jjihe MMJjpFDDA Meteorological Model to the
Southern CaliforniaJC^S-1997 gfgg^llggfiKiminary Test Using the SCAQS August
1987 Case", NinthWamfiGonferencemn^^jptications of Air Pollution Meteorology,
AinericaaMeteof61ogicd.|Society, Atlanta, GA., January 28-February 2, 1996.
Seaman, N.L., (199f|jZ'Use of Moae^Generated Windfields to Estimate Potential Relative Mass
ContributionsifrpniDiffereritlyocations'', Prepared for U.S. EPA Source Attribution
Workshop, ^(ssbafciisrriangle Park, NC, July 16-18, 1997. Available from EPA Internet
Website, "http://www.epa.gov/ttn/faca/stissu.html", then download file "source
attribution workshopmaterials".
Seigneur, C., P. Pai, J. Louis, P.Hopke and D.Grosjean, (1997), Review of Air Quality Models
for Particulate-Matter, Document Number CPO15-97-1 a, Prepared for the
:; AmericaniEetroleum Institute.
Sillman, S., (1995), "The Use of NOy, H2O2, and HNO3 as Indicators for O3-NOx-ROG
Sensitivity in Urban Locations", J.Geophys. Res. 100, pp.14175-14188.
150
-------
Sillman, S., (1997), "The Method of Photochemical Indicators as a Basis for Analyzing O3-NOx-
ROG Sensitivity", NARSTO critical review paper, to be published in Atmospheric
Environment
Sillman, S., D.He, C.Cardelino and R.E. Imhoff, (1997a), "The Use of Photochemical Indicators
to Evaluate Ozone-NOx-Hydrocarbon Sensitivity: Case Studies from Atlanta, New York
and Los Angeles", J.Air and Waste Mgmt. Assoc. , In press.
Sillman, S., (1998), Evaluating the Relation Between Ozone,
Method of Photochemical Indicators, EPA/600R-9J
Systems Applications International, (1996), Users Guid]
Model (UAMV), SYSAPP 96-95/27r, Documen
Internet website, "www.epa.gov/scram001/t22.
Tesche, T.W. and D.E. McNally, (1993a), Operational
Meteorological Model (MMS)for Episode 1:
Air Pollution Study Agency by Alpine Geo;
Tesche, T.W. and D.E. McNally, (1993b),
Meteorological Model (MM5)fc
Air Pollution Study Agency by
SARMAP
spared for the Valley
zrion ojie SARMAP
|||!f|D, Prepared for the Valley
" Butte, CO.
Tesche, T.W. and D.E. McNally, jjpc), Operational Equation of the CAL-RAMS
Meteorological Model fojfJMOS 1: 26jj$8 June, 1991, prepared for the Lake
Michigan Air Directs ^onsortiun^^^^^TCeophysics, Crested Butte, CO.
Tesche, T.W;ancLI>.E. Mc^jy||||?93d), Operational Evaluation of the CAL-RAMS
Meteorc^g^al ModS^orfM^ Episode 2: 17-19 July, 1991, prepared for the Lake
MichigatflffiBirectors Consorfiiim by Alpine Geophysics, Crested Butte, CO.
Tesche, T.W. and DTE^cNally, (1993e), Operational Evaluation of the CAL-RAMS
MeteorologicalMjodelfor
-------
U.S. EPA, (1993), Volatile Organic Compound (VOC)/Particulate Matter (PM) Speciation Data
System (SPECIATE), Version 1.5, EPA/C-93-013.
U.S. EPA, (1993a), User's Guide for the Urban Airshed Model, Volume IV: User's Manual for
the Emissions Preprocessor System 2.0, Part A: Core FORTRAN System and Part B:
Interface and Emission Display System, EPA-450/4-90-007d(R), NTIS Ag^jlsion No.
PB93-122380. ~
U.S. EPA, (1994a), Office of Mobile Sources, User's Gul
01.
rt,
cefor Emission
U.S. EPA, (1996a), Photochemical Air Monitoring Sy.
EPA-454/R-96-006.
U. S. EPA, (1997a), EIIP Volume I, Introduction and
Inventory Development, July 1997, EPA-454
U. S. EPA, (1997b), Point Sources Preferred
97-004b.
U. S. EPA, (1997c), Area Sources Prefejjred and jjjternail
97-004c.
y 1997, EPA-454/R-
's, July 1997, EPA-454/R-
U. S. EPA, (1997d), Mobile Sou/jes Prefer/jjjsand Aginative Methods, July 1997, EPA-454/R-
97-004d.
U. S. EPA,;{19f|e), Biojen&S^urces Preferred and Alternative Methods, July 1997, EPA-
rocedures, July 1997, EPA-454/R-97-004f.
U. S. EPA,
U. S. EPA, (1997g), Datijlaridgement Procedures, July 1997, EPA-454/R-97-004g.
U.S. EPA, (1998a), EPAThird-Generation Air Quality Modeling System, Models-3 Volume 9b
-/ r Users ManMa/,S>A-600/R-98/069(b).
V 'V
-•X:,:'l' ' ^f
U.1S^PA, (\99&d^National Air Pollutant Emission Trends, Procedures Document 1900-1996,
:, ^siEEA=4^/R-98-008, see "http://www.epa.gov/ttn/chief/eiJUitahtmWETDP".
U.S. EPA, (1998e), Guidance for making emissions projections, in preparation.
152
-------
U.S. EPA, (1998f), Guidance on temporal allocations, spatial allocations and chemical
speciation, in preparation.
U.S. EPA, (1999b), Proposed revision to 40CFR, Part 51, Appendix W.
U.S. EPA, (1999c), Emission Inventory Guidance For Implementation Of Ozone And Paniculate
Matter National Ambient Mr Quality Standards (NAAQS) and Regional H(jie
Regulations, EPA-454/R-99-006, (April 1999).
Watson, J.G., (1997), "Observational Models: The Use
Prepared for the U.S. EPA Source Attribution
July 16-18, 1997. Available from EPA Internet
"http://www.epa.gov/ttn/faca/stissu.html", the:
materials".
Yang, Y.J., J.G.Wilkinson and A.G.Russell, (1997), "
Multidimensional Air Quality Models for Sou:
Influence Identification", Prepared for the
Research Triangle Park, NC, July 16-1
"http://www.epa.gov/ttn/faca/stiss
Analys.-PowerP.-T.RusseH".
of
op
Yang, Y.J., J.G.Wilkinson and A.
Multidimensional Photoci
Technology, In press.
itivity Analysis of
ation and Area-of-
ton Workshop,
Internet website,
SAW-Direct Sensi
sell, (lifPa), "FasjlDirect Sensitivity Analysis of
i Mod©fe", Environmental Science and
Yarwood, G. and R.Morris, (1997), "Descf^^^lhe CAMx Source Attribution Algorithm",
Prepared for U.S. EPA Source Attribution Workshop, Research Triangle Park, NC, July
^-IS^l^^fiAvaila^leirroniTB^A Internet website,
"http://w^^a.gov/ttn/faca/s^ssu.html", then download file "source attribution
workshop materials'!, ^fl*
**• . '' "-At^if,. •<•,.•.
Yarwood, G., G.Wilson;"R;EMorris, M.A.Yocke, (1997a), User's Guide to the Ozone Tool:
&£•-_ '-i"- ,"' "
Ozone Source Apportionment Technology for UAM-IV, Prepared for Mr. Thomas Chico,
South Coast Air .Quality Management District, Diamond Bar, CA, March 28, 1997.
153
-------
Glossary
Areawide design value
Modeled attainment demonstration
Modeled attainment test
The highest design value monitored within a
nonattainment area. At a site with three years of
complete data, a design value is the average 4th
highest 8-hour daily maximum ozone concentration
observed over a consecutive 3 year
of a modeled attainment test.
jditional screening analysis
x>f model outputs and
Isrological data for
Modeling system
n consists of
igsion levels
anddfhst
!
sion
A modeled;
two parts:
consistent;
of measur
levels one
(analysis)
It may alscy
and a rev
emissi
cc
ividence determination
ngient of the NAAQS is
control strategy.
;st takesjpe ratio of mean predicted future and
;nt 8-hQBr daily maximum ozone concentrations
sragedjjSfer several days and multiplies this ratio
site-specific monitored design value at
lonitoring location. If the product is < 84 ppb
near all monitoring sites, the test is passed.
This is a group of models used to predict ambient
ozone concentrations. The group includes an
emissions model which converts countywide
emission information into gridded speciated
emissions which vary diurnally and reflect
environmental conditions. It also includes a
meteorological model which provides gridded
meteorological outputs and an air
chemistry/deposition model which takes
information provided by the emissions and
meteorological models and uses it to develop
gridded predictions of hourly pollutant
concentrations.
155
-------
Relative reduction factor (RRF)
The ratio of predicted 8-hour daily maximum ozone
averaged over several days near a monitoring site
with future emissions to corresponding predictions
obtained with current emissions.
Screening test
Weight of evidence determination:
A screening test is used to;
control strategyjyill be ef
at locations w^ppt an
attainment jjpnown throi
area. It coJlsts of tw<
examine
nonattainr
predictions
near a mo;
compute iftrelativi
locatioaszand multipl
re tflst a proposed
ah reducing ozone
lonitor so that
^attainment
is
~*:: the
:tionjjiiverywhere
:o identify locatioiK®Blving
:onsistently higher than any
The second part is to
factor for each flagged
rs times the
valuefc^Knonattainmentarea. If
at aHraagged locations, the test
is passe
Tins is a sey&f diverse analyses used to judge
Whether ^attainment of the NAAQS is likely. The
of each analysis is assessed and an
\V0Htedme consistent with an hypothesis that the
"NAAQS will be met is identified beforehand. If the
set of outcomes, on balance, is consistent with
attainment, then the WOE can be used to show
attainment. A weight of evidence determination
includes results from the modeled attainment test,
the screening test, other model outputs and several
recommended analyses of air quality, emissions and
meteorological data.
156
------- |