Overview of Regulatory 8-hr Ozone Modeling

1
Overview of Regulatory 8-hr Ozone Modeling

• T. W. Tesche
• Dennis McNally
• Alpine Geophysics, LLC
• Ft. Wright, KY
• Ralph Morris
• ENVIRON International Corp
• Novato, CA
• Farmington, NM
• 16 July 2003

2
Staff Credits Alpine Geophysics, LLC Dennis
McNally Cyndi Loomis ENVIRON Int. Corp. Ralph
Morris
3
Workshop Topics

• Role of Regulatory Photochemical Modeling in U.S.
• Review of Ozone Photochemistry
• Classification of Photochemical Models
• Overview of the CAMx Regional Photochemical Model
• Emissions Inventory Modeling
• Meteorological Modeling
• Model Performance Evaluation
• Future Year Emissions Control Strategy Modeling
• Ozone Attainment Demonstration

4
Ozone Concentration Trends for U.S. Cities Los
Angeles
Ozone ppb
2nd Highest Value
4th Highest Value
Year
5
Role of Regulatory Modeling

• Areas that are not in compliance with the Federal
  National Ambient Ozone Standards (NAAQS) are
  required to use modeling to
• Estimate the effects of growth (and expected
  control measures) on future ozone air quality
• Evaluate alternative/additional measures
• Demonstrate future compliance with 1-hr and 8-hr
  ozone standard(s)

Regulatory modeling must follow prescribed EPA
Guidance on model selection, episode selection,
data base development, model performance
evaluation, application of models, attainment
demonstration, peer-review, and documentation
6
Role of Regulatory Modeling (concluded)

• Modeling efforts for most areas include
• scoping study to assess the nature and extent of
  the problem and data gaps/needs
• collection and analysis of meteorological, air
  quality, emissions, and land-use data
• development and continued refinement of modeling
  databases and capabilities

7
Typical Regulatory Modeling Process
Conceptual Model Development
Select episodes and domains
Prepare/refine inputs
Model Evaluation
Apply emiss, met, ozone models
Compare model results to measurements
No
Performance OK?
Yes
Prepare future-year emissions
Conduct future-year evaluations
8
Ozone Photochemistry

• Sources and sinks for tropospheric ozone (bad
  ozone)
• In contrast to stratospheric ozone (good ozone)
• Ozone formation from VOC and NOx
• Control strategy implications
• Sensitivity to VOC and/or NOx
• VOC reactivity
• NOx suppression (inhibition or disbenefits)
• Condensed chemical mechanisms

9
Sources and Sinks of Tropospheric Ozone

• Sources
• smog chemistry involving VOCs and NOx with
  sunlight
• radical cycle among OH, HO2, RH2, etc.
• global methane and CO oxidation
• Stratospheric intrusion
• Sinks
• chemical reactions (e.g., NO, alkenes)
• deposition

10
Ozone Formation from VOC and NOx
no sunlight ? no ozone production no NOx? no
ozone production no VOC ? no ozone production
11
Control Strategy Implications ofVOC/NOx Chemistry

• Sensitivity to emission reduction
• VOC sensitive, NOx sensitive, or both
• VOC reactivity
• depends upon chemical nature reaction rate,
  radical yields, products
• approximated by reactivity scales, e.g., Carter
  MIRs
• NOx suppression or inhibition
• if ozone is highly VOC sensitive, NOx reduction
  may incur disbenefits
• two mechanisms operative
• (1) direct titration (or scavenging) of ozone by
  fresh NO emissions
• (2) scavenging of radicals by NO2 inhibits ozone
  production

12
Ozone and Precursor Relationships Ozone vs. NOz
reacted NOx (NO NO2)

• NOz NOy - NOx
• observed ozone/NOz relationship for a rural
  location in eastern U.S.
• interpret slope as production efficiency for
  ozone from NOx (i.e., amount of ozone formed per
  NOx reacted away)

NOz ppb
13
Key Photochemical Reactions

• NOx Reactions
• NO O3 ? NO2 O2 (NO-O3 titration)
• NO2 ?? ? NO O (NO2 photolysis -- sunlight)
• O O2 ? O3 (Ozone Formation)
•  VOC-Radical Reactions
• VOC OH ? VOC HO2 (OH to HO2 w/o NO2
  destruction)
• HO2 NO ? OH NO2 (NO to NO2 w/o O3
  destruction)
• HO2 HO2 ? H2O2 O2 (H2O2 Formation)
•  Secondary PM Reactions
• NO2 OH ? HNO3 (Nitrate Formation -- Day)
• N2O5 H2O ? HNO3 (Nitrate formation Night)
• SO2 OH ? SO4 HO2 (Sulfate Formation
  Gas-Phase)
• SO2 H2O2 ? SO4 (Sulfate Formation Aqueous)
• SO2 O3 ? SO4 (Sulfate Formation Aqueous)
• HNO3 (g) ?? NO3 (pm) (Equilibrium T, RH, NH3,
  SO4)

14
Classification of Photochemical Air Quality
Simulation Models

• Lagrangian employ a coordinate system that
  moves with air parcels
• Eulerian the coordinate system is fixed in
  space
• Hybrid incorporate features of Lagrangian types
  into a Eulerian framework

15
Photochemical Modeling Concepts

• All Air Models Solve Some Form of the Atmospheric
  Diffusion Equation (a complex differential
  equation) that Relates Changes in Pollutant
  Concentration to
• Advection (transport by the mean wind)
• Turbulent diffusion
• Chemical reaction
• Deposition
• Emissions

16
General Form of the SpeciesConservation
(Continuity) Equation
Mathematical solution (integration) of general
forms of the diffusion equation is difficult --
simplifying assumptions are required
17
Lagrangian Models

• Many Simplifying Assumptions
• Produce a simple, closed-form analytical
  expression
• for diffusion
• Do not require numerical integration
• Invoke the Assumption of Air Parcel Coherency
• Breaks down quickly not far from an emissions
  source, especially in complex wind flow
  situations
• Cost-effective solution at relatively close
  ranges for a relatively small number of sources

18
Lagrangian Models (continued)

• Chemical interactions between puffs, segments, or
  particles cannot be properly treated
• Readily produce source-receptor relationships
• Severe technical limitations especially for
• large numbers of sources
• regional-scale transport applications
• photochemically reactive pollutants

19
Lagrangian Models (continued)

• Gaussian Plume Models
• The earliest air models
• Invoke many simplifying assumptions to obtain
• closed-form analytical solutions
• Steady-state (i.e., time invariant)
• Spatially uniform (homogeneous) dispersion
• Plume coherency
• Inert or first-order decay
• Gaussian plume models are not capable of
  treating photochemistry

20
Gaussian Plume Model
21
Lagrangian Models (continued)

• Examples of Gaussian Plume models include
• ISC
• COMPLEX
• RTDM
• AERMOD

22
Lagrangian Models (continued)

• Gaussian Puff Models
• Fewer simplifying assumptions
• employ analytical solutions for each puff, but
• computers are required to track the large
• number of puffs
• still retain the plume coherency assumption
• a few have been developed for individual reactive
  plumes,
• e.g., RPM-IV
• numerical solution methods are needed to solve
  chemical
• kinetics equations

23
Lagrangian Models (continued)

• Examples of Gaussian Puff models include
• CALPUFF
• RPM-IV
• SCIPUFF
• SCICHEM

24
Eulerian (Grid) Model Concept
25
Processes Treated in a Grid Model

• Emissions
• Surface emitted sources (on-road and non-road
  mobile, area, low-level point, biogenic, fires)
• Point sources (electrical generation, industrial,
  other, fires)
• Advection (Transport)
• Dispersion (Diffusion)
• Chemical Transformation
• VOC and NOx chemistry, radical cycle
• For PM aerosol thermodynamics and aqueous-phase
  chemistry
• Deposition
• Dry deposition (gas and particles)
• Wet deposition (rain out and wash out, gas and
  particles)

26
Eulerian Grid Cell Processes
27
Coupling Between Grid Cells
28
Eulerian Models

• Generally considered to be technically superior
• allow more comprehensive, explicit treatment of
  physical processes
• chemical processes included
• interactions of numerous sources
• Require sophisticated solution methods
• employ discrete time steps and operator splitting
• computational grid (hence the term grid models)
• relatively expensive to apply for long periods

29
Eulerian Models (continued)

• Subgrid resolution can be a limitation
• as grid size and time step length are reduced
• accuracy increases, but
• computation time also increases
• advanced grid models employ variable grid spacing
  or nesting
• improves accuracy in critical locations
• allows cost effective application on urban to
  regional scales
• CAMx has flexi-nesting capability

30
Hybrid (Lagrangian/Eulerian) Models

• Incorporate features of Lagrangian models into
• grid model framework
• overcome many of the sub-grid model limitations
• Overcome many of the prior practical advantages
  of Lagrangian models, through the development of
• variable (nested) grid resolution
• source apportionment techniques
• Capitalize on the availability of low-cost high
  speed computers

31
Sub-Grid-Scale Plume Concept
32
CAMx GREASD PiG Concept
33
Hybrid Models (concluded)

• Examples of hybrid photochemical grid models
  include
• MODELS-3/CMAQ
• MAQSIP
• UAM-V
• CAMx

34
Modules in a Grid Model

• Emissions Modeling System
• EPS2x, SMOKE and EMS-2003
• Meteorological Modeling System
• MM5 and RAMS (SAIMM and CALMET)
• Preprocessors for Other Inputs
• TUV (photolysis Rates)
• Initial Concentrations and Boundary Conditions
• Air Quality Model
• CAMx/PMCAMx, Models-3/CMAQ (UAM-V, MAQSIP)
• Post-Processors and Visualization
• Model Performance Evaluation (MAPS, CAMXtrct,
  Excel, SURFER)
• PAVE
• Flying Data Grabber

35
General Model Summary

• Photochemical grid/hybrid models have evolved as
  the preferred means of addressing complex and
  nonlinear processes affecting reactive air
  pollutants in the troposphere
• These models invoke fewer assumptions but require
  high speed computers and sophisticated numerical
  integration methods
• Higher accuracies require more computer resources
• Modern hybrid models utilize variable grid
  spacing (nested grids) in combination with a
  Lagrangian puff sub-model to treat subgrid
  resolution of plume dispersion and chemistry

36
Overview of the CAMx Regional Photochemical Model

• Simulates the physical and chemical processes
  governing the formation and transport of ozone in
  the troposphere
• three-dimensional, Eulerian (grid-based) model
• requires specification of meteorological,
  emissions, land-use, and other geographic inputs
• output includes hourly concentrations of ozone
  and precursor pollutants for each grid cell
  within a (three-dimensional) modeling domain

37
CAMx Overview (continued)

• Mathematically simulates the following processes
• emission of ozone precursors (anthropogenic and
  biogenic)
• advection and diffusion (transport)
• photochemistry
• deposition

38
Example of a Multiply Nested CAMx Domain Denver
8-hr EAC Study
4-km grid
36-km grid
12-km grid

1.3-km grid
39
CAMx Model Formulation
Change in Concentration
Advection by Winds
Turbulent Diffusion
Ri Si Li
Emissions
Surface Removal/Deposition
Chemical Reaction
40
CAMx Modeling System Features

• Carbon-Bond-IV chemical mechanism with enhanced
  isoprene and toxics chemistry
• Two-way interactive nested-grid capabilities
• Plume-in-grid (P-i-G) treatment
• Accepts output from a variety of dynamic
  meteorological models

41
CAMx Input Requirements

• Meteorological Inputs
• Three-dimensional winds
• Three-dimensional temperatures
• Three-dimensional water-vapor concentration
• Surface pressure
• Three-dimensional vertical diffusivity (effective
  mixing height)
• Two-dimensional cover
• Rainfall rate

42
CAMx Inputs (continued)

• Emissions Inputs
• Low-level anthropogenic emissions
• Point sources
• Area sources
• On-road motor vehicles
• Non-road sources
• Elevated point source emissions
• Biogenic emission estimates

43
CAMx Inputs (continued)

• Air quality related input files
• Initial conditions (all grids, initial hour)
• Boundary conditions (outermost grid, all hours)
• Chemistry input files
• Chemical reaction rates
• Photolysis rates
• Geographic/other input files
• Land-use, Land-cover
• Albedo, turbidity, and ozone column

44
CAMx Probing Tools

• Probing Tool
• A module within the photochemical grid model that
  extracts information on source-receptor
  relationships, chemical and physical processes,
  mass flux, etc.
• Used to diagnose model to understand why it is
  getting the answer it gets and improve modeling
  performance
• Used to better define optimal emission control
  strategies

45
Probing Tools in CAMx

• Decoupled Direct Method (DDM) first-order
  sensitivity coefficients of ozone by emission
  source groups (also in URM and CMAQ models)
• Ozone Source Apportionment Technology (OSAT)
  ozone source apportionment by source groups
  (source category geographic region)
• Process Analysis output information on chemical
  processes and mass flux (also in CMAQ)

46
Probing Tool Comparison
47
Emissions Inventory Modeling
48
(No Transcript)
49
(No Transcript)
50
(No Transcript)
51
(No Transcript)
52
(No Transcript)
53
(No Transcript)
54
(No Transcript)
55
Biogenic Emissions

• VOC emissions from trees and other plants
• isoprene
• terpenes (e.g., a-pinene)
• oxygenated VOCs (e.g., alcohols)
• depend upon plant species, temperature, sunlight,
  season, etc.
• NO emissions from soils
• enhanced by fertilizer
• CAMx requires gridded, hourly emissions in UAM-IV
  file format

56
Biogenic Emission Factor Models

• EPA guidance
• BEIS2 recommended (BEIS3 emergent)
• BIOME an alternative for use with locally
  specific data
• BEIS2
• Fortran code distributed by EPA
• BELD land use/land cover (LULC) database for US
• Not easy to change LULC data or emission factors

57
Biogenic Emission Factor Models (concluded)

• BIOME
• Part of EMS-2003, requires SAS
• Can emulate BEIS2, or
• Can use local LULC data and alternate emission
  factors
• BEIS3
• improved emission factor models
• improved LULC data including satellite imagery
• GLOBEIS
• Incorporates BEIS3 factors and landuse data
• Has additional features including PAR data, heat
  stress, drought stress

58
Emissions Processing Systems

• EPS2x
• FORTRAN Based
• Modular and Flexible
• Designed primarily for urban rather than regional
  applications
• Widely used for a variety of applications inside
  and outside the US
• EMS-2003
• SAS Based, uses ARCINFO
• takes advantage of Unix operating system for
• control of job stream
• uses gridded temperature file for accurate MV
  emission factor adjustment
• used in the LMOS, OTAG, Models-3/CMAQ applications

59
Emissions Processing Systems (concluded)

• SMOKE
• FORTRAN based
• Computationally fast
• Emerging system, still being tested and module
  developed
• Supported by EPA, adopted by WRAP, CENRAP, etc.
• Current quality assurance (QA) not as extensive
  as for EPS and EMS

60
Meteorological Modeling
61
Model Performance Evaluation

• All air quality models have inherent limitations,
  uncertainties, and weaknesses
• limit the range of their applicability
• affect their ability to replicate actual
  conditions
• input data also have limitations, uncertainties,
  and inaccuracies that affect model performance
• It is difficult to separate the effects of these
  two sources of uncertainty

62
Model Performance Evaluation(continued)

• Operational Evaluation comparison of observed
  and measured ozone concentrations
• Scientific Evaluation comparison of precursor
  species, performance aloft, species ratios, mass
  budgets
• Diagnostic Evaluation to identify errors and
  examine the effects of input uncertainty and
  model physics
• Performance Appraisal consideration of how model
  results will be used in an attainment
  demonstration

63
Regional Scale Ozone Tile Plots

Daily Maximum 8-hr Ozone Concentrations (ppb) on
31 July 2001 Over the Eastern United States 36
km CAMx Grid Domain
64
Local Scale Ozone Tile Plots

Daily Maximum 8-hr Ozone Concentrations (ppb) on
3 Aug 2000 Over The San Juan Basin/Four Corners
Region 4 km CAMx Grid Domain
65
Ozone Time Series Plots

Hourly Ozone Concentrations (ppb) on at Mesa
Verde National Park For 31 July 4 August 2000
Over the San Juan Basin/Four Corners Region
66
Ozone Time Series Plots

Hourly Ozone Concentrations (ppb) on at
Substation For 31 July to 4 August 2000 Over
the San Juan Basin/Four Corners Region 4 km Grid
67
Hourly Ozone Scatter Pots

Scatterplot of Hourly Ozone Concentrations (ppb)
on 22 June 2001 Over Lower Lake Michigan 4 km
CAMx Grid Domain
68
Quartile-Quartile Ozone Plots

Q-Q Plot of Hourly Ozone Concentrations (ppb) on
24 June 2001 Over Lower Lake Michigan 4 km CAMx
Grid Domain
69
Vertical Diffusivity Slice Plot

Vertical Distribution of Eddy Diffusivity on 28
August 1991 Over Lower Lake Michigan 4 km CAMx
Grid Domain
70
CAMx 3-D Animation
71
1-Hr Ozone Statistical Measures

• EPA measures and performance goals for
  urban-scale regulatory ozone modeling
• Unpaired accuracy of the peak ( 20)
• Average accuracy of the peak ( 20)
• Nnormalized bias ( 15)
• Normalized gross error (lt 35)
• No criteria for regional-scale regulatory
  applications

72
Model Performance Evaluation

• A comprehensive three-phased evaluation of the
  base case emissions, meteorological and air
  quality simulation results should be conducted
  before a model is applied to sensitivity study
  analysis, control strategy assessment, or
  attainment demonstration
• Phase 1 -- Screening Assessment (e.g.,
  computation of EPA performance statistics
  cursory graphical analysis of overall features)
• Phase 2 -- Refined Statistical/Graphical
  Analyses using various statistical/graphical
  displays, probing tools, etc,
• Phase 3 -- Detailed Scientific and Diagnostic
  model performance evaluations as needed,
  including compensatory error analysis

73
Model Performance Evaluation(continued)

• Statistical model performance evaluation
• statistical comparisons between model predictions
  and observations
• provide quantitative measures of performance, but
• shed little or no light on the reasons for poor
  performance
• provide few indications of the robustness of good
  performance
• often leave users without a clear picture of a
  models reliability

74
Model Performance Evaluation (continued)

• Ability to replicate peak hourly ozone
  observations
• domain-wide unpaired accuracy of the peak
• station-wide unpaired accuracy of the peak
• average accuracy of the peak
• a cut-off value (60 ppb) is normally used

75
Model Performance Evaluation (continued)

• Ability to reproduce hourly ozone observations
• normalized bias and gross error
• fractional bias and gross error
• weight over- and under-predictions equally
• ratio of mean observation to mean prediction (no
  cutoff value)
• ratio of bias to mean observation (no cutoff
  value)

76
Model Performance Goals

• One-Hour Average Ozone
• EPA has developed 1-hour ozone performance goals
  for three statistical measures Guidelines for
  the Regulatory Application of the UAM
• (EPA, 1991)
• Unpaired accuracy of the peak concentration lt 20
  percent
• percent difference between peak observation and
  prediction unmatched in time and space

77
Model Performance Goals (continued)

• normalized bias lt 15 percent
• percent average difference between observations
  and predictions matched in time and space, over
  all hours above a cutoff value (usually 60 ppb)
• normalized gross error lt 35 percent
• percent average absolute difference between
  observations and predictions matched in time and
  space, over all hours above a cutoff value
  (usually 60 ppb)
• although there are no goals for fractional bias
  and gross error, it is useful to compare these
  with the EPA standards for normalized bias and
  gross error

78
Model Performance Goals (continued)

• Eight-Hour Average Ozone
• EPA recently published draft guidelines for
  8-hour ozone performance goals Use of Models
  and Other Analyses in Attainment Demonstrations
  for the 8-hour Ozone NAAQS (EPA, 1998)
• guidance is much more qualitative than for 1-hour
  performance evaluations
• suggests use of same statistical measures as for
  1-hour ozone
• with the exception of comparison of unpaired
  peaks

79
Model Performance Goals (concluded)

• some additional metrics are suggested
• computing statistics only for hours between
  8am-8pm local time
• adding a new comparison for hours above the
  standard (84 ppb)
• stops short of recommending specific statistical
• performance goals
• suggests evaluation of performance for precursors
  and
• secondary species
• emphasizes secondary species to reduce the
  effects of mismatch between predictions and
  observations regarding spatial-scale
  representation
• suggests using source apportionment results as
  corroborative evidence
• strongly recommends diagnostic evaluation

80
Model Performance Evaluation (continued)

• Diagnostic model evaluations supplement
  statistical evaluations
• conducted if
• statistics do not meet acceptable criteria, to
  assess the reasons why
• statistics do meet the standards, to confirm the
  robustness of
• good performance
• model predictions are compared with an
  observation-based
• conceptual model
• a general description of what probably occurred
  based on
• analyses of observational data
• experience
• physical/chemical considerations
• intuition
• provide a framework on which to unravel the
  cumulative
• modeling uncertainties

81
Model Performance Evaluation (continued)

• Diagnostic model evaluations supplement
  statistical evaluations
• reconciliation of differences yields
• a more informative assessment of numerical model
  performance
• and applicability
• an aide in designing model sensitivity runs
• more objective justification for altering model
  inputs to
• improve model performance

82
Model Performance Evaluation (continued)

• Diagnostic model evaluations supplement
  statistical evaluations
• conceptual models are imperfect and occasionally
  over-speculative, so they must be used with
  considerable discretion
• significant discrepancies between a predictive
  and conceptual
• model may indicate problems with
• the numerical model
• its input data
• the conceptual model

83
Model Performance Evaluation (continued)

• Diagnostic model evaluations supplement
  statistical evaluations
• graphical products are essential tools and
  provide useful insights
• station plots -- temporal alignment of
  predictions and observations
• isopleths plots -- spatial alignment of predicted
  fields
• scatter plots -- performance as a function of
  concentration level
• difference plots -- effects of model input changes

84
Model Performance Evaluation (continued)

• Diagnostic model evaluations supplement
  statistical evaluations
• model sensitivity runs are made to test
  hypotheses and explore
• model response to input modification
• helps prioritize effort to improve/refine model
  inputs within their
• range of uncertainty
• if feasible, sensitivity to emissions changes
• week-end vs. weekday
• retrospective
• trends confirmation
• VOC and NOx precursor sensitivity

85
Model Performance Evaluation (continued)

• Diagnostic model evaluations supplement
  statistical evaluations
• typical model sensitivity tests include
• scaling of emissions and/or emission components
• (mobile vs. biogenic vs. area, etc.)
• scaling of vertical diffusivity
• clean boundary conditions
• zero anthropogenic emissions
• systematic variation of wind fields

86
Future-Year Emissions Control Strategy Modeling

• Selection of a future year
• Projection of the emission inventory components
• Future-year sensitivity and control strategy
  simulations

87
Selection of the Future Year

• Application to attainment demonstrations and/or
  control strategy evaluation
• Emission inventory must be adjusted for
  conditions expected in a future year, to account
  for
• growth in emissions due to new sources and
  increased activities
• emissions controls expected as mandated by
  existing current regulations
• Model runs are made with the future base case
  inventory to predict future air quality

88
Future Emissions Control Scenarios

• If attainment is not predicted for the future
  base case, then additional controls must be
  considered
• A variety of control scenarios are developed and
  their effectiveness explored

89
Future Emissions Control Scenarios (continued)

• An emissions control scenario is a collection of
  generic or specific control measures on
  individual sources or groups of sources
• sources are often grouped by type or location
• examples of typical control measures are
• reduction of Vehicle Miles Traveled
• low NOx burners on power plant or industrial
  boilers
• lower auto tail pipe standards (LEV, ULEV, ZEV)
• fugitive VOC detection and control programs
• consumer product reformulation
• gasoline reformulation
• vapor recovery at fueling stations

90
Future Emissions Control Scenarios (continued)

• Since the number of combinations of feasible
  control measures is almost endless,
• a means of narrowing the list to a manageable
• number of scenarios is needed
• prior experience and economic feasibility
  usually
• enter into the selection process
• CAMx source apportionment is also an aide in
• identifying the most effective control scenarios

91
Future Emissions Control Scenarios (continued)

• Emission-sensitivity simulations provide
  information about
• sensitivity of the model/episode to different
  levels and types of emission reductions
• amount, type, source, and geographical
  distribution of reductions that reduce ozone
• Emission-sensitivity simulations used to guide
  the development of real-world control
  strategies

92
Future Emissions Control Scenarios (continued)

• Control-strategy simulations provide the basis
  for assessment of
• control-measure effectiveness
• separately
• in conjunction with other control measures
• viability of future-year attainment strategies

93
Future Emissions Control Scenarios (continued)

• Possible variables include
• species/emissions type (e.g., VOC, NOx, CO)
• amount of emissions reduction (e.g., 10, 20,
  30)
• general category (e.g., low-level, elevated,
  anthropogenic)
• source category (area, motor-vehicle, non-road,
  elevated or low-level point)
• geographical area (e.g., domain, region, states,
  counties, grid cells)
• time period (e.g., weekday only, weekend only,
  daytime, nighttime)

94
Future Emissions Control Scenarios (concluded)

• Identify possible controls based on
• available technologies
• feasibility of implementation
• Use the emission-sensitivity simulations to focus
  and refine selection of control measures
• Use the control-strategy simulations to
• examine the effects of specific control measures
• examine the effects of packages of control
  measures
• identify attainment strategy options

95
Examples of Control Measures

• Alternative fuels
• IM programs (various levels)
• NOx RACT
• VOC RACT
• Transportation measures (including ozone action
  days)
• Limits on agricultural burning
• Cool communities measures

96
Ozone Attainment Demonstration
97
EPA 1-Hour Modeling Attainment Demonstration
Procedures

• Statistical approach
• includes 2 benchmark tests
• accommodates episode severity
• attainment ozone concentration may be gt 125 ppb
• Deterministic approach
• all simulated concentrations must be lt 125 ppb
• Weight of evidence determination

98
Statistical Approach Benchmark Tests for 1-Hr
Ozone Attainment

• Number of simulated exceedances in each portion
  of the modeling domain must be less than the
  allowed number (which is determined by episode
  severity)
• Maximum simulated value for each day must be less
  than the allowed value for that day (also
  determined by episode severity)
• Simulated reduction in the number of exceedance
  grid-cell hours must be 80 percent or more
  (compare to the base-case simulation)

99
Weight of Evidence

• Possible elements include
• episode representativeness
• process analysis
• model performance issues
• observed ozone trends
• use of EPAs 8-hr ozone attainment demonstration
  methodology

100
EPA 8-Hr Attainment Demonstration Procedures

• Attainment demonstration for a specific area of
  interest includes three recommended elements
• attainment test
• screening test
• weight of evidence determination
• Key differences between draft 8-hr guidance and
  previous 1-hr guidance include
• modeled attainment test is based on relative
  (rather than absolute) use of the modeling
  results
• modeling comprises a part of the weight of
  evidence

101
EPA 8-hr Ozone Attainment Test

• Future Year D.V. Base Year D.V. x RRF
• RRF (Relative Reduction Factor)
• Future Year Mod. Value
• Base Year Mod. Value
• Ozone FY D.V. lt 85 ppb ATTAINMENT
• gt 85 ppb NOT ATTAINMENT
• PM2.5 FY D.V. lt 15 ug/m3 ATTAINMENT
• FY D.V. gt 15 ug/m3 NOT ATTAINMENT