United States
Environmental
Protection
Agency
Office of Air Quality
Planning and Standards
Research Triangle Park, NC 27711
450490006A
APRIL 1990
AIR
SEPA
URBAN AIRSHED MODEL
STUDY OF FIVE CITIES
(Summary Report)
-------
-------
EPA-450/4-90-006A
URBAN AIRSHED MODEL
STUDY OF FIVE CITIES
(Summary Report)
By
Richard D. SchefTe
OFFICE OF AIR QUALITY PLANNING AND STANDARDS
U. S. ENVIRONMENTAL PROTECTION AGENCY
RESEARCH TRIANGLE PARK, NC 27711
APRIL 1990
-------
-------
Disclaimer
This material has been subject to the agency's review, and it has
been approved for publication as an EPA document. Mention of
trade names or commercial products does not constitute
endorsement or recommendation for use.
Acknowledgments
This project has been jointly funded by the following
organizational groups within the United States Environmental
Protection Agency: Air Quality Management and Technical Support
Divisions within the Office of Air Quality Planning and
Standards; Office of Policy Planning and Evaluation; Atmospheric
Research and Exposure Assessment Laboratory within the Office of
Research and Development; and Regional Offices III, IV and VI.
The Office of Mobile Sources provided advisory assistance for
developing mobile source emission factors.
-------
-------
a
Abstract
The 5-City study embodied two major elements. These are (1) Q
demonstration of the "low-cost" application of the Urban Airshed
Model (UAM) and, (2) the use of the model to assess peak ozone
impacts in several cities (New York, St. Louis, Atlanta, Dallas-
Fort Worth and Philadelphia) due to various emissions strategies.
The Urban Airshed Model (UAM) is EPA's preferred model for
conducting ozone air quality simulation analyses. Faced with
nearly 100 ozone nonattainment areas, concerns were raised about
feasibility of widespread applications of a complicated model
like the UAM. The "low-cost" element of the project addressed
applications related concerns by outlining procedures which
reduce both the technical complexities and resources typically
associated with UAM modeling.
The project was conducted over a 15 month timeframe (July, 1988 -
November, 1989) covering a period during which various emissions
related issues drew significant public interest and scrutiny. At
the onset of the study, the project served as a timely tool to
address the ozone impact implications related to alcohol blended
fuels, mobile source evaporative running losses, and Reid Vapor
Pressure (RVP) reductions. Similarly, during the early project
stages the role of natural hydrocarbon emissions (biogenics) on
ozone precursor control calculations began to emerge as a major
ozone air quality issue. The later stages of the project
coincided with the Clean Air Act (CAA) amendment efforts which
included lengthy debates on alternative fuels, particularly
methanol, and numerous discussions on the relative merits of
hydrocarbon and oxides of nitrogen (NOx) control measures.
Throughout this period, UAM analyses were performed to provide
information for planning purposes and policy decisions on the
aforementioned topics.
The project is documented in this Summary Report and the
following 5 topic reports:
Report 1. Demonstration of Low-Cost Application of the Model to
the City of Atlanta and the Dallas-Fort Worth Metroplex
Region
Report 2. Evaluation of Base Case Model Performance for the
Cities of St. Louis and Philadelphia Using Rich and
Sparse Data Bases
11
-------
Report 3. Low-Cost Application of the Model to Atlanta and
Evaluation of the Effects of Biogenic Emissions on
Emission Control Strategies
Report 4. A Low-Cost Application of the Urban Airshed Model to
the New York Metropolitan Area and the City of St.
Louis (note, this report reflects Phase I efforts from
FY-88)
Report 5. Low-Cost Application of the Model to Future Year SIP
Control and Alternative Fuel Strategies for Dallas-Fort
Worth, Atlanta, Philadelphia, and St. Louis
The Summary Report condenses key material from each report and
synthesizes key results and conclusions for the entire study.
Reports 1 and 2 address the "low-cost" testing of the UAM.
Reports 3-5 document the results of the emissions strategy
analyses. This preliminary overview provides a synopsis of the
project structure and principal findings.
Low-Cost Applications (Reports 1 and 2)
As part of this study a set of procedures referred to as the
Practice for Low-cost Application in Nonattainment Regions
(PLANR) approach was developed to demonstrate UAM applications
using routinely available data. The approach relies upon
routinely available meteorological and air quality data to
formulate UAM input files. Existing meteorological, air quality
and emissions data are transformed to more highly resolved (in
space, time and chemical composition) data fields through the use
of default paramatization and surrogate substitution schemes to
drive the UAM. Historical UAM applications typically have relied
on special field studies to collect air quality and
meteorological data to supplement routinely available data.
Thus, major cost savings result from PLANR by eliminating special
field monitoring programs.
Although ozone concentrations were simulated for 5 cities, only
two cities - Atlanta and Dallas-Fort Worth - truly reflect
"PLANR" applications. Special field studies had been performed
previously on the remaining cities. More importantly, the PLANR
approach may be sufficient only for certain cities which are
considered relatively uncomplicated. For example, the
combination of simple terrain and prevailing climatological
features as well as a separation from major pollutant transport
corridors may permit simplified modeling approaches in cities
111
-------
such as Atlanta and Dallas-Fort Worth. Report 1 presents PLANR
base case applications to Atlanta and Dallas-Fort Worth. These
base case applications exhibited good performance for Dallas-Fort
Worth^and reasonable performance for Atlanta. Both applications
identified critical deficiencies in routinely available data
bases which must be addressed for future "low-cost" applications.
A PLANR approach was emulated for St. Louis and Philadelphia by
degrading the rich observational bases supplied by the original
field studies. The degradation was not systematic. Elements of
the rich data base considered to be "routinely available" were
retained, but the remaining elements which reflected measurements
based on intensive field studies were rejected. This design
permitted parallel comparison of model performance using
disparate data bases. Report 2 presents the comparison study
which suggested that the PLANR approach is not adequate for
cities like Philadelphia which are complicated by regional
pollutant transport.
Emission Strategy Results (Reports 3-5)
Emissions strategy results reflecting simulations considering the
importance of biogenic emissions and ozone impacts associated
with alternative fuels and various hydrocarbon and NOx control
measures are presented in Reports 3-5. Complete consistency in
emissions strategies among all cities was not adhered to as
budget and time restrictions forced certain city specific
emission strategy designs. For example, the design of
simulations for Atlanta was influenced by the fact that it was a
test city for biogenics sensitivity. Based on the sets of
emissions strategies among the 5 cities, several major findings
are reported:
1) biogenic emissions can affect ozone precursor control
calculations and must be included in modeling analyses,
2) hydrocarbon control measures in the absence of NOx
controls by themselves will not exacerbate peak ozone,
'base case applications refer to simulations using best
emissions estimates and comparing model results with observed
pollutant concentrations
IV
-------
3) effectiveness of combined hydrocarbon and NOx control
measures should be assessed on a case by case basis as
the direction of benefit on peak ozone appears to be
city dependent,
4) implementation of pure component alternative fuel
strategies (Ml00,' CNG) or proposed tightening of
tailpipe emissions limits will cause small but
beneficial impacts on peak ozone, and
5) conversion to alcohol blended fuels will have
negligible impact on peak ozone beyond the benefits
gained from reducing gasoline volatility.
v
-------
Contents
1 INTRODUCTION ...................................... i
2 DESCRIPTION OF PLANR PROCEDURES ................... 5
3 DEMONSTRATION OF LOW-COST APPLICATION
OF THE MODEL TO THE CITY OF ATLANTA
AND THE DALLAS-FORT WORTH METROPLEX
REGION (REPORT 1) .................................. 9
4 EVALUATION OF BASE CASE MODEL PERFORMANCE
FOR THE CITIES OF ST. LOUIS AND
PHILADELPHIA USING RICH AND SPARSE
DATA BASES (REPORT 2) ............................. H
5 LOW-COST APPLICATION OF THE MODEL TO
ATLANTA AND EVALUATION OF THE EFFECTS
OF BIOGENIC EMISSIONS ON EMISSION CONTROL
STRATEGIES (REPORT 3) ............... .............. 15
6 A LOW-COST APPLICATION OF THE URBAN AIRSHED
MODEL TO THE NEW YORK METROPOLITAN AREA AND
THE CITY OF ST. LOUIS - FY-88 RESULTS FOR
NEW YORK AND ST. LOUIS (REPORT 4) ................. 17
7 LOW-COST APPLICATION OF THE MODEL TO FUTURE
YEAR SIP CONTROL AND ALTERNATIVE FUEL
STRATEGIES FOR DALLAS-FORT WORTH, ATLANTA,
PHILADELPHIA, AND ST. LOUIS (REPORT 5) ............ 21
8 CONCLUDING REMARKS ................................ 26
VI
-------
-------
1 INTRODUCTION
BACKGROUND
EPA's 5-City UAM study applied the Urban Airshed Model (UAM) to
the cities of New York, St. Louis, Atlanta. Dallas-Ft. Worth, and
Philadelphia. The 5-City UAM study was motivated by the
perceived need that gridded photochemical air quality simulation
models would encounter widespread use in addressing ozone
precursor control requirements for post-87 State Implementation
Plans (SIPs). Previous applications of models like the UAM
generally were conducted with the benefit of an intensive site
specific field study designed to collect air quality and
meteorological data to drive and evaluate performance of the
model. Faced with over 50 ozone nonattainment areas, the
requirement to conduct separate field studies for each model
application could be prohibitively expensive. Consequently, EPA
initiated the 5-City project to determine the feasibility of
applying the UAM with routinely available data (i.e., without a
field study). For perspective, combined costs for a modeling
application with a field study would range from one to several
million dollars; costs for a model application with routine data
would range from $100,000 to $400,000.
Commencing in the summer of 1988, the study was sponsored
principally by EPA's Offices of Air Quality.Planning and
Standards (OAQPS), Policy Planning and Evaluation (OPPE),
Research and Development (ORD), and Regional Offices III, IV and
VI. EPA's Office of Mobile Sources (QMS) provided guidance on
mobile source emissions. Systems Applications, Incorporated
(SAI) performed the modeling analyses discussed in this report
Besides conducting a "paper" study, the project resulted in a
transfer of the UAM system to participating States -Georgia,
Maryland, and Texas. Previous UAM capability was largely limited
to California and New York. In addition to investigating the
feasibility of lower cost UAM applications, the project provided
data bases for assessing the impact on ozone concentrations due
to alternative fuels, biogenic volatile organic compound (VOC)
emissions, hydrocarbon and NOx control strategies. Overall
project objectives included:
1. Demonstration of UAM base case application with
routine data (referred to as the Practice for Low-cost
Application in Nonattainment Regions -PLANR).
-------
2. Comparison of model behavior based on rich (i.e., with
field study) and sparse (PLANR procedure) data bases.
3. Demonstrating use of the UAM for SIP control strategy
evaluations.
4. Determining impact of various alternative fuels on
ambient ozone levels.
5. Determining the effect of biogenic emissions on ozone
precursor control requirements.
6. Distributing UAM technology to participating States.
7. Developing an Emissions Preprocessing System (EPS)
which translates SIP/county level emissions inventories
into gridded, speciated and time variant UAM input
files.
Results of the project are compiled in 5 reports (listed below)
organized to reflect the major objectives.
Report 1. Demonstration of Low-Cost Application of the Model to
the City of Atlanta and the Dallas-Fort Worth Metroplex
Region
Report 2. Evaluation of Base Case Model Performance for the
Cities of St. Louis and Philadelphia Using Rich and
Sparse Data Bases
Report 3. Low-Cost Application of the Model to Atlanta and
Evaluation of the Effects of Biogenic Emissions on
Emission Control Strategies
Report 4. A Low-Cost Application of the Urban Airshed Model to
the New York Metropolitan Area and the City of St.
Louis (note, this report reflects Phase I efforts from
FY-88)
Report 5. Low-Cost Application of the Model to Future Year SIP
Control and Alternative Fuel Strategies for Dallas-Fort
Worth, Atlanta, Philadelphia, and St. Louis
The remainder of this summary discusses key results from each
report.
-------
OVERVIEW OF THE MODELING PROCESS
A brief overview of the modeling process will facilitate
comprehension of project results. The UAM is a gridded •
photochemical air quality simulation model which produces hourly
concentrations of ozone, hydrocarbons, nitrogen oxides and other
species for several vertical layers (usually 4 or 5) in grid
cells (having a horizontal dimension from 2 to 8 km) which
overlay the city and surrounding area of interest. Ozone
analyses conducted with photochemical grid models typically are
comprised of 5 steps:
1. Compiling air quality, meteorological and emissions
data to develop current year base case UAM input files
and provide air quality observations to evaluate base
year model performance.
2. Conducting diagnostic analyses. The principal purpose
of diagnostic analyses is to properly characterize the
system. Nevertheless, the visible product is enhanced
model performance (i.e., better spatial and temporal
agreement with observed data). Diagnostic runs refer
to simulations based on adjustments of input data
within the limits of known data uncertainty to better
understand physical/chemical conditions.
3. Projecting base case emissions to future year base case
emissions files.
4. Developing future year emissions control strategy files
which reflect chosen control measures such as basin
wide reductions in VOC or alternative fuel scenarios.
5. Performing future year simulations and assessing
effectiveness of various strategies.
The_PLANR approach refers to specific procedures for using
available data minimizing uncertainty associated with sparse data
sets. For example, the PLANR approach used in this project (and
described below) for Atlanta and Dallas relied on available
meteorological and air quality data from existing monitoring
programs. In contrast, specific field programs designed to
collect extensive data for model inputs have been set up in St
Louis, Los Angeles and other cities. The modeling protocol
followed for either approach resembles the steps listed above.
Existing modeling data bases from previous studies were available
-------
for New York, St. Louis, and Philadelphia; current base modeling
years associated with those cities are 1980, 1976, and 1979,
respectively. Current base year modeling data bases were
developed for Atlanta (1984) and Dallas-Fort Worth (1985). All 5
cities used 1995 as the future year for assessing impact on ozone
of various emissions control strategies. It should be noted that
simulations performed for this project are intended as
demonstrations and do not reflect the modeling effort necessary
to comply with post-1987 SIP or proposed Clean Air Act modeling
requirements. For example, future modeling applications are
likely to use emissions inventories superior to the 1985 NAPAP
inventory used in this project. In addition, only one ozone
episode was modeled for each city; typically two or more episodes
may be required for modeling demonstrations.
-------
2 DESCRIPTION OF PLANR PROCEDURES
The PLANR approach consists of applying the UAM without
conducting an intensive field monitoring program. Available data
are used as described below for each of three major data
categories.
Emissions The PLANR application utilizes the best available
emissions inventory, which ordinarily is county based and lacks
the required spatial, temporal and chemical resolution needed to
drive the UAM. Spatial, temporal and chemical speciation
surrogates are used to transform the available inventory into the
more highly resolved UAM input files. This "top-down" approach
is facilitated with an emissions preprocessing system (EPS),
developed as part of the project. The EPS is a group of fortran
programs which cross reference source category codes with user
supplied surrogates (e.g., population distribution, land use
data, species profiles) to develop the gridded, hourly and
speciated UAM emissions inputs.
Air Quality The UAM must be provided initial concentrations of
all non-steady state chemical species (currently 23) for every
grid cell and boundary concentrations over the duration of the
simulation. Generally insufficient routinely available data
exist to adequately characterize these concentration fields.
Thus, to minimize the influence of initial concentrations, the
UAM simulation time is extended from 1/2 to I day before the
actual period of high ozone. Similarly, the upwind modeling
domain is expanded to lessen the influence of boundary
conditions. Available data are utilized when possible for
initial and boundary concentrations; however, default
concentrations are substituted in the absence of adequate data.
It should be recognized that these approaches for minimizing
uncertainty related to sparse air quality data place greater
reliance on emissions data. In effect the emissions are
developing the boundary and initial concentration fields of
interest. Accordingly, compilation of accurate emissions
inventories is a crucial step in applying the UAM.
Meteorology Generally, data are available which can be used to
develop the wind fields and mixing heights for the UAM. The
PLANR approach relies on the best use of available data from
local monitoring networks, National Weather Service (NWS),
Federal Aviation Administration (FAA) and other sources such as
existing special field studies. For clarification throughout
-------
this report, routinely available data refer to data sources
separate from specially designed field programs.
Diagnostic techniques A more complete explanation of the PLANR
approach as it is imbedded in the modeling protocol described
above is provided in Report 4. However, diagnostic analyses can
play an integral role in modeling and should be discussed
briefly. Diagnostic analyses are an evaluation of the adequacy
of model inputs in characterizing the system (i.e., prevailing
meteorological, emissions and chemical characteristics). The
need for diagnostic analyses often arises from attempting to
explain poor model performance; although, diagnostic analyses
should be conducted on model inputs independent of simulation
results to check for obvious errors of omission (e.g., major
emissions sources) and mischaracterization. The most
controversial diagnostic analyses are model simulations in which
input variables are adjusted to determine possible reasons for
poor model performance. Any adjustments are based on a
reasonable understanding of the system and within constraints
related to inherent uncertainty in data and best technical
judgement as to why certain available data are not
representative. At the time of this project, clear guidelines on
model performance criteria or acceptable use of diagnostic
analysis procedures were not available. In the context of the 5-
City project, the PLANR diagnostic analysis procedures consisted
of:
1. checking the meteorological, emission and air quality
fields for obvious omissions and mischaracterized
events,
2. performing a minimal number of simulations (1 to 3
beyond the unadjusted base case simulation).
Specifics regarding diagnostic analyses for the individual cities
are provided in the relevant reports, and discussed briefly in
appropriate sections in this summary report.
RATIONALE BEHIND THE PLANR APPROACH
The PLANR approach is best suited to urban areas relatively
isolated from long distance transport of ozone and its
precursors, and in regions without complex meteorology (e.g., sea
breeze effects) or terrain. Therefore, Atlanta, Dallas-Fort
Worth (DFW) and St. Louis are considered good candidates for the
PLANR techniques.
-------
Uncertainty due to data base design and resolution exists in all
three types of data. Historically, emission inventories have not
included major sources such as automotive evaporative running
losses and biogenic emissions, and they lack adequate spatial,
temporal, and chemical resolution. Typically, air quality
monitoring data reflect a small number of receptor/impact sites
located in and downwind of central urban cores. While these
sites are useful for evaluating model performance, upwind
locations are required to specify boundary concentrations. Air
quality data generally are inadequate in their temporal and
chemical resolution, and there is often a total lack of
measurements to characterize pollutant concentrations in grid
cells aloft. Also disturbing is the lack of meteorological data
to characterize surface and upper air flows and mixing heights.
For example, FAA surface wind monitors located in urban areas
report wind speed and direction only averaged over a few minutes
each hour; the UAM requires true hourly averages of speed and
direction. Upper air soundings which are used to define hourly
elevated wind flows and mixing heights often are located outside
the domain of interest and are reported only two times a day.
Obviously, assumptions must be invoked with available data to
attain satisfactory model performance.
The task becomes tractable when one assumes that the available
emissions inventory is fixed and not subject to alteration (other
than episodic adjustments such as temperature sensitivity of
evaporative emissions not accounted for in "typical" summer day
inventory). This assumption allows us to focus on procedures for
overcoming uncertainties in the available air quality and
meteorological data base, and may be appropriate in that
emissions are the ultimate target of control and analysis. The
starting point for emission inputs is the available emission
inventory (e.g., the 1985 NAPAP inventory was used in this
project). Emission inventories which are being compiled as part
of the post-87 SIP requirements are essentially similar to the
NAPAP inventory with respect to -spatial (county level), temporal
(summer day average), and chemical resolution (not speciated).
The UAM Emission Processing System translates the NAPAP level
inventory into resolved UAM files by cross referencing spatial,
temporal and speciation surrogates with different source
category codes. Example surrogates include population
distribution, historical diurnal traffic profiles. An analogous,
intensive approach might require developing an inventory for each
grid cell by every hour. When guidance and further documentation
on the use of the EPS are completed, the system will be a useful
tool for model users in bridging the gap between SIP inventories
-------
and UAM emission files.
The influence of initial conditions on hourly concentrations of
interest is reduced as the simulation period is lengthened
Simulations often start at noon the day before a high ozone
event. In effect, the concentration field is then influenced to
a greater extent by local emissions relative to initial
concentrations. To reduce dependence on boundary conditions, the
upwind modeling domain is specified sufficiently large so that
the flux of species entering the inner area of greatest interest
has been developed primarily by upwind emission fields.
Sensitivity tests on the necessary simulation period and extent
of upwind domain needed to minimize the effects of initial and
boundary conditions can be performed to optimize computer storage
and speed requirements. Recommended "clean" default initial and
boundary air quality concentrations, representative of background
conditions, are presented in the reports.
Meteorological data present very difficult problems when
constrained to routinely available data, particularly when
limited to NWS and FAA data. Mixing heights and wind speeds set
dilution rates for ozone, primary precursors and intermediate
species involved in ozone formation. Misalignment of urban
plumes due to errors in wind direction could have significant
impact on assessing the relative contributions of different
source groups. The PLANR approach uses available morning and
afternoon National Weather Service (NWS) soundings (or other
available sources) to develop diurnal profiles of mixing heights
and upper wind velocities. Soundings can be located in rural
areas outside the modeling domain. Preferably, the most
representative sounding is used. Available surface wind data
combined with upper wind data are used to drive the diagnostic
wind model (DWM) which generates hourly, gridded wind velocities
for all vertical levels in the UAM domain. The DWM interpolates
wind velocities between stations and accounts for terrain induced
effects.
Once the input files are developed and the first simulation
completed, results are analyzed to determine how adequately
resultant meteorological and concentration fields represent
actual conditions. At this point advantage of diagnostic
simulations is taken, if necessary, to establish a better
representation of episode meteorology.
-------
3 DEMONSTRATION OF LOW-COST APPLICATION OF THE MODEL TO THE
CITY OF ATLANTA AND THE DALLAS-FORT WORTH METROPLEX REGION
(REPORT 1)
The Atlanta and Dallas-Fort Worth base case applications have
been separated from the other cities because historical modeling
data bases are not available. Hence, they best reflect PLANR
approaches using routinely available data and perhaps offer a
glimpse into similar future applications for other cities.
Current year model inputs were based on 1985 NAPAP emissions
inventory and air quality and meteorological data from 1984
(Atlanta) and 1985 (Dallas-Fort Worth).
•
ATLANTA RESULTS
The Atlanta simulation for June 4, 1984 (starting at 1200 June,
3) produced reasonable temporal and spatial tracking of observed
ozone, but tended to underpredict peak values. The model
produced a region wide daily peak ozone value of 132 ppb,
relative to observed value of 147 ppb. A diagnostic run which
accounted for additional upwind biogenic VOC and eliminated a
poorly sited surface wind station was considered the final base
case simulation. Note, that the Atlanta study included biogenic
VOC emissions which constituted a significant fraction (59% -
mass basis) of total VOC emissions. Figure 1 presents the
station plots of predicted versus observed hourly ozone for three
monitors in the Atlanta domain.
Given the level of peak ozone underprediction, a sensitivity
analysis was performed which assumed a 50% reduction in the
reported FAA wind speeds to determine the importance of surface
level wind measurements. FAA wind stations report wind speed and
direction averaged over a 1-3 minute time span; the UAM requires
hourly average representations. Consequently, the FAA reported
measurements would tend to overstate hourly average speeds.
Further justification for this adjustment was based on quantified
differences between collocated hourly average and instantaneous
monitors in California. The model simulation with reduced wind
speeds resulted in improved peak ozone performance (predicted =
observed=147 ppb) and better diurnal tracking (Figure 2).
DALLAS-FORT WORTH RESULTS
The DFW simulation spanned two and a half days (August 29-31,
1985). The last day was a Saturday which required weekend
emissions adjustments (provided as part of the NAPAP inventory).
-------
™nrt T1SSlons were also included. Model performance was
considered very good as predicted hourly ozone closely tracked
observed values for most surface stations (Figure 3) The
predicted regional peak ozone value (164 ppb) over the 2-1/2 day
simulation compared favorably with an observed value of 170 ppb
Also, model predictions accurately tracked (spatially and "
temporally) observed NOx concentrations.
A preliminary diagnostic analysis revealed that the available
Cnn^^Uf? S
-------
4 EVALUATION OF BASE CASE MODEL PERFORMANCE FOR THE CITIES OF
ST. LOUIS AND PHILADELPHIA USING RICH AND SPARSE DATA BASES
(REPORT 2)
Previous UAM studies using intensive data bases obtained in
special field studies had been conducted for St. Louis and
Philadelphia as part of the Regional Air Pollution Study (RAPS)
and Philadelphia Oxidant Study (POS), respectively. The
intention of this part of the project was to compare both base
case model performance and ozone response to VOC control by
conducting side by side UAM simulations based on detailed (i.e.,
the RAPS and POS data bases) and sparse (i.e., data routinely
available) observations. Such a comparison might provide insight
on the limitations of the PLANR approach as well as isolate key
input parameters which could optimize data gathering in future
field studies.
BASE CASE RESULTS
St.Louis
The Regional Air Pollution Study provided intensive field data
for an ozone episode on July 13, 1976. This day was modeled
previously with an earlier version of the UAM with carbon bond-II
(CB-II) chemistry (the current UAM utilizes CB-IV, a more
recently developed mechanism). Model runs based on the RAPS data
are referred to as RAPS simulations. A third approach referred
to as SIMPLE used no diagnostic simulations and very limited
meteorological data. Base case model performance statistics for
four approaches (RAPS-CBII, RAPS-CBIV, PLANR, and SIMPLE) are
presented in Table 1. This discussion is restricted to
comparisons between RAPS-CBIV and PLANR; that is , the same model
with different inputs. The PLANR approach was based on routine
air quality and meteorological data. The 1976 emission files
from the RAPS modeling were used in all approaches (with
translation of CB-II to CB-IV categories), and biogenic emissions
were not included. The PLANR results presented in Table 1
reflect an assumed 50% wind speed reduction in NWS/FAA surface
monitors (refer to the related discussion above on Atlanta). The
PLANR peak predicted ozone of 195 ppb is compared with observed
(222 ppb), RAPS-CB-IV (242), RAPS-CBII (174) and SIMPLE (118).
Good overall performance resulted from the PLANR application,
although not as impressive (or expected) as the RAPS-CBIV.
The analysis is complicated by constraining the domain size and
simulation period for the PLANR approach to that used by the
11
-------
RAPS-CBIV. Recall that the PLANR approach recommended expanding
upwind domain and simulation time in the absence of adequate air
quality data. However, to be consistent with the existing RAPS
simulations, a short one day simulation period and limited domain
was retained for the PLANR application. Moreover, the PLANR
initial and boundary values were based on relatively clean
defaults, the impact of which would have been mitigated by
imposing a larger domain and simulation period. A recommended
improvement for these analyses would utilize the original RAPS
initial and boundary conditions with the PLANR meteorological
inputs. This would isolate differences between the PLANR and
RAPS approaches to a set of meteorological variables, rather than
sets of both meteorological and air quality variables - as we
have seen, these simulations are very sensitive to meteorology.
Philadelphia
The Philadelphia Oxidant Study (POS) provided detailed data for
an ozone episode on July 13, 1979. Base case model performance
statistics (Table 2) exhibit poor peak ozone performance for the
PLANR application. Although the predicted PLANR peak of 187 ppb
compares favorably with the observed peak (205 ppb) and the POS
peak (236 ppb), the PLANR peak is located about 45 km east of
the observed peak, and at the location of the observed peak the
PLANR simulation underpredicts by 46%.
Two factors complicate the Philadelphia analysis: 1) the area is
not isolated from intercity transport of ozone and its precursors
and 2) July 13 was an extreme stagnation day which limits
thorough description of the wind fields and impairs the models
ability to spatially track concentration gradients. These
results suggest that the PLANR approach is not recommended for
complicated ozone episodes.
Differences in the simulation results may result from different
diagnostic approaches as well as the quality of observed data
The original POS modeling with CB-II chemistry has been
criticized for seemingly excessive iterations of diagnostic
simulations. In contrast, only two diagnostic simulations were
performed for the Philadelphia PLANR application.
12
-------
CONTROL STRATEGY ANALYSIS
Perhaps of greater interest are concentration prediction
comparisons from across-the-board VOC control strategies between
the PLANR and RAPS approaches. Ultimately, the UAM will have to
demonstrate attainment of the ozone NAAQS. An important
interpretive problem exists when peak model predictions under or
overpredict observed peak values. Three approaches (for a single
episode) can be used to define attainment for a particular
episode1:
1. unbiased acceptance of model predictions, i.e.,
disregard difference between peak and observed values
and accept the control required to attain predicted
value of 120 ppb,
2. decrement approach (in terms of ppb) which requires a
predicted peak ozone reduction equivalent to the
observed peak minus 120 ppb, and
3. the percentage approach which requires a percentage
decrease in predicted peak ozone equivalent to
percentage decrease required to lower observed peak
ozone to 120 ppb.
These approaches have separate advantages/disadvantages. For
example, the unbiased approach might be desirable in cases of
overprediction to provide a conservative buffer. On the other
hand, biased techniques might .be needed to prevent understated
control targets when model predictions underpredict observed
values.
Results of the three methods for St. Louis are presented in
Figure 4. In terms of estimating the level of VOC control to
demonstrate attainment, excellent agreement exists between PLANR
(73% control) and RAPS (78%) with the decrement procedure. The
Philadelphia results (Figure 5) exhibit large differences (at
various VOC control levels) with all three approaches, and
suggest that the PLANR approach must be considered carefully
before being applied in complex areas.
issue of how many episodes are required to demonstrate
attainment based on the statistical representativeness of the
ozone NAAQS is beyond the scope of this project
13
-------
CONCLUDING REMARKS ON PLANR APPLICATIONS
Some additional comments based on the PLANR applications
R^nn^6? ^ J^^L1 and 2 are Provided before proceeding to
Reports 3-5 which address control strategy issues. The base case
performance results for Atlanta, Dallas-Fort Worth and St Luis
generally track observed hourly ozone data in space and time but
typically underpredicted peak ozone. In these analyses
diagnostic simulations emphasized meteorological parameters-
*?ooVe^™he emission data bases in these studies - the 1976
s^nifielV08' anV985 NAPAP " are susP*cted of containing
significant gaps. For example, the RAPS and POS data do not
include biogenic or running loss emissions and the NAPAP data
p?£J»designeduto be accurate on a national scale. Since the
PLANR approach underpredicted and the detailed approach
overpredicted in both cities, it is likely that inclusion of
these missing sources would upgrade the PLANR performance
relative to the RAPS and POS applications.
Model users must be aware of the dangers of using limited data
bases When diagnostic analysis work is phrased with seemingly
logical expressions like "adjustments within the limits of data
uncertainty", the likelihood that good results are achieved for
the wrong reasons is increased. Accordingly, very explicit and
careful guidance on diagnostic analysis procedures must
eventually be provided to model users. Equally important is the
recommendation that routine meteorological monitoring networks
which rely heavily on NWS/FAA stations be enhanced and redesigned
to reduce uncertainty in key model inputs.
One finding of significance is that estimated control
requirements depend on the model's ability to predict base case
ozone accurately. This raises important questions which
supercede the rich/sparse data base question. in concept, the
model provides a tool for the decision maker to judge which of
several different technically feasible strategies will produce
the greatest benefit. Several metrics (e.g., grid cells above
certain values, different averaging times, total risk, population
exposure) can be produced which take advantage of the spatial and
temporal resolution provided by the model. When applied in a
relative context for evaluating the merits of different control
approaches, the requirement for superior model performance
probably is less important, and lower cost applications become
more viable.
14
-------
5 LOW-COST APPLICATION OF THE MODEL TO ATLANTA AND EVALUATION
OF THE EFFECTS OF BIOGENIC EMISSIONS ON EMISSION CONTROL
STRATEGIES (REPORT 3)
Interest in the role of VOC emissions from vegetation (biogenic
emissions) in ozone formation and precursor control escalated
when Chameides et al. of Georgia Institute of Technology
published a paper in the journal Science in the fall of 1988.
The Chameides study modeled an Atlanta ozone episode (June 4,
1984; peak observed ozone = 147 ppb) with EPA's OZIPM4 model, a
Lagrangian approach lacking the spatial resolution of the UAM.
The analysis looked at the amount of anthropogenic VOC control
required with and without biogenic emissions in the total VOC
inventory. The OZIPM4 model indicated that significantly greater
VOC control would be required to reduce peak ozone to 120 ppb
when biogenic emissions are included relative to anthropogenic
emissions only.
A major concern with the OZIPM4 application was the model's
inability to properly account for relative spatial distributions
of anthropogenic and biogenic emissions. The 5-City project
provided a timely vehicle to reproduce the Georgia Tech work with
more detailed grid modeling. Base case performance results for
Atlanta have been discussed above. Results of across-the-board
anthropogenic VOC controls and constant NOx emissions for both
cases (with and without biogenics) in Figure 6 are based on the
original wind fields which produced a predicted peak of 132 ppb
with biogenic emissions. The results clearly show the same
general effect observed in the Georgia Tech. study - inclusion of
biogenic emissions will increase the predicted level of
anthropogenic VOC control required to reduce peak ozone for
Atlanta on June 4, 1984. To reach predicted attainment based on
normalized percentage reductions of peak ozone, the biogenics and
nonbiogenics cases require 100% and 62% anthropogenic VOC
control, respectively.
Estimated biogenic emissions constitute roughly 59% of all VOC
emissions (mass basis) for the modeling day. Isoprene which is a
very reactive VOC category comprises about 19% of the biogenic
estimates. An uncertainty factor of +/- 200% typically is
associated with biogenic emission estimates. It must be
emphasized that these results are not generalized for other urban
areas and, perhaps, for different episodes in Atlanta. Atlanta
may be distinctive for three reasons. First, Atlanta is
predominantly a residential/commercial city with a limited urban
core relative to other major urban areas. Hence, the ratio of
15
-------
biogenic to anthropogenic emissions might also be greater
throughout the domain, relative to other cities. Second,
biogenic emission models exhibit extreme temperature sensitivity
and southern cities are characterized by high summer
temperatures. Lastly, the distribution of vegetation types in
the Atlanta area might be classified as high emitters.
The original base case performance results were improved when
reported FAA wind speeds were reduced 50% (Figure 2) Results
from identical VOC control strategy simulations with the new
windfield (Figure 7) suggest a reduced impact, relative to the
original windfield, on control calculations due to biogenic
emissions. Simulations based on the new windfield produced VOC
control levels of 63% and 52% to reach attainment for biogenic
and nonbiogenic cases, respectively (compared to 100% and 62%
with the original windfield). Rather than deciding which
windfield is more representative, the new windfield might be
viewed as an additional modeling day with a slower mean
trajectory. in this scenario anthropogenic emissions now
comprise a greater fraction of total VOC emissions because the
slower wind speed increases the residence time ratios of
anthropogenic to biogenic emissions, relative to the original
windfield with higher wind speeds.
16
-------
-------
1. base case with mobile emissions at current RVP (9.0
psi ) .
2. base case with low RVP (7.8 psi).
5-
10W RVP (7'8 psi) and 100% Penetration
7. base case with mobile emissions at current RVP (9 psi)
and a near doubling of evaporative running losses.
'SIP RUN X. base case with low RVP and enhanced I/M and
other stationary source measures to achieve
an overall VOC reduction of 40%.
SIP RUN Y. base case with a 40% VOC reduction targeting
most reactive VOC classes.
REID VAPOR PRESSURE
produces voc emission reductions of 24% and 15% for
"
indicate 11 3) ' M°del r6SUltS for
inaicate negligible impact on ozone for New York and
approximately a 3% decrease for St. Louis (Table 4). The New
York results are confounded by regional transport of ozone and
ozone precursors (i.e., high boundary concentrations) which
severely limit sensitivity of local emission changes in these
simulations, boundary conditions were held constant A
commensurate reduction in boundary conditions might 'have produced
loull L^°ne sen^itivitv- The apparent small effect on ^?
Louis peak ozone (i.e., 3% ozone decrease for 15% VOC emissions
4 Th SIP RUN Y refer to SIP A and SIP B in
****
convention in
- ^ — — -^f m
18
-------
decrease) is attributed to the nonlinear ozone response to
changes in VOC emissions, which is especially flat during initial
incremental emission changes. RVP reductions were brought about
by removing the necessary C-4 fraction (mostly butane). No
reformulation assumptions to boost octane content following the
removal of high octane butane were assumed (ethanol is an octane
enhancer).
FUEL BLENDS
Two important factors which potentially impact ozone should be
recognized when analyzing modeling studies on oxygenated fuel
blends:
1. Splash blending ethanol with gasoline produces RVP
increases which result in increased evaporative
emission losses. However, evaporative VOCs typically
are less reactive relative to total composition of all
VOC emissions.
2. The increased oxygen content enhances combustion which
should reduce CO emissions. However, CO emission
benefits from alcohol blends are a strong function of
mobile source distribution. Since-modern automobiles
with advanced combustion control systems emit less CO,
reductions in CO due to use of ethanol blends would be
expected to diminish in future years.
[Reformulation assumptions can also enter into these analyses as
reformulation can affect the exhaust VOC composition and thus
change emissions reactivity.]
The results for New York and St. Louis show that use of ethanol
blends have negligible impact (either increase or decrease) on
peak ozone beyond that already achieved with RVP reduction. Note
that the proper comparisons in Table 4 are between scenarios 2
and 3 (New York) and 2 and 5 (St. Louis). Thus, the effects of
increased VOC and decreased CO emissions attendant with use of
ethanol blends approximately counterbalance one another.
The ETBE scenario for St. Louis shows a negligible benefits (only
a 1 ppb decrease in ozone). ETBE is attractive in that splash
blending does not result in RVP increase, yet the combustion
benefits due to oxygen increase are retained.
19
-------
RUNNING LOSS AND SIP SCENARIOS
Running losses are evaporative emissions of VOC occurring while a
motor vehicle is in use. The New York scenario 4 and St. Louis
scenario 7 were sensitivity analyses motivated by the large
uncertainty in these evaporative mobile source emissions. All
other scenarios used best estimates of running losses. Reasons
for the apparent small changes in peak ozone are identical to
those presented above in the RVP discussion. However, at the
start of this study EPA used very rough running loss estimates
based on limited data. These estimates were added to the
evaporative and exhaust emission factors provided by Mobile 3 9
(Mobile 3.9 did not account for running losses). More recent'
data incorporated in EPA's more recent mobile source model
(Mobile 4) suggest that running losses were overstated in the
1995 scenarios in Report 4. [The Atlanta and Dallas-Fort Worth
scenarios as well as all 1995 strategies in Report 5 were
developed with Mobile 4 - running losses were not considered in
the RAPS, POS and OMNYMAP data bases.]
The St. Louis SIP scenarios illustrate the potential differences
due to consideration of VOC emissions composition. Scenario X
(10% peak ozone reduction) reflects a 40% reduction of all VOC
categories. Scenario Y (19% peak ozone reduction) reduced the
most reactive carbon bond classes by 80% to obtain a "total VOC
emissions decrease of 40%.
20
-------
7 LOW-COST APPLICATION OF THE MODEL TO FUTURE YEAR SIP CONTROL
AND ALTERNATIVE FUEL STRATEGIES FOR DALLAS-FORT WORTH,
ATLANTA, PHILADELPHIA, AND ST. LOUIS (REPORT 5)
The last part of the 5-City study attempted to illustrate
development of future year SIP strategies representing both
across-the-board (referred to as Type A) and source category
specific (Type B) emission reductions. As the project also
coincided with preparation of numerous Clean Air Act proposals,
simulations reflecting proposed new gasoline regulations and
alternative fuels were conducted for St. Louis, Philadelphia and
Dallas-Fort Worth. The alternative fuel strategies are based on
vehicles fueled by Ml00 and CNG and offer more radical approaches
than alcohol blends (Report 4) which are technologically
compatible with the existing automobile fleet. Simulations based
on proposed gasoline regulations go beyond the simple RVP
analyses of Report 4 by incorporating specific tailpipe emission
constraints.
Due to time and budget restraints similar strategies were not
performed in every city. In the strategy descriptions listed
below, note that two "1995 base cases" exist for the alternative
fuel cities (St. Louis, Dallas and Philadelphia). The first is
referred to as "base case" and reflects growth projections of the
1985 NAPAP emissions to 1995 with existing controls. A second
base case referred to as "GAS" reflects new proposed gasoline
regulations was developed after the original future year base
case. Finally, all the 1995 strategies included biogenic VOC
emissions; a comprehensive tabulation of all emission totals is
provided in Report 5 (as Table 2). The 1995 strategies are
listed below:
1. Base case; reflecting source growth projections to 1995
and existing controls with enhanced mobile source I/M
and reduced RVP (Atlanta, St. Louis, Dallas and
Philadelphia).
2. GAS: base case reflecting most recent CAA proposals
including enhanced I/M, reduced RVP, and tighter
tailpipe emission limits (St. Louis, Dallas and
Philadelphia).
3. M100: base case with assumed 100% penetration (total
mobile source fleet) of pure methanol fueled vehicles
(St. Louis, Dallas and Philadelphia).
21
-------
4. CMC: base case with 100% penetration of vehicles fueled
with compressed natural gas (St. Louis, Dallas and
Philadelphia).
5. SIP Type A; base case with multiple percentage across-
the-board VOC emission reductions (Atlanta-30%,60%,90%-
Dallas 30%,60%). ' '
6- SIP Type B Scenario 1r base case with VOC reductions
from stationary solvent sources and NOx reductions from
tighter tailpipe limits and control of major boilers
(Atlanta, Dallas and St. Louis).
7.
SIP Type B Scenario ?; Identical to SIP Type B Scenario
1 with further control of major boilers to achieve
additional NOx reductions.
Please note that the three fuel strategies (GAS, M100, CNG) were
based on a 20 mph average speed assumption, which is somewhat
^We?aa£aK SPSedS ln the 1985 NAPAP inventory used for developing
the 1995 base case emissions estimates. Care should be exercised
when comparing results among the 1995 base case and fuel
strategies, since exhaust emissions factors increase with
decreasing vehicle speed. Potentially larger differences in peak
ozone between the 1995 base case and fuel strategies than
reported in Tables 5a,b might result if all 1995 scenarios
utilized identical speed assumptions.
RESULTS OF FUEL RELATED STRATEGIES
A listing of daily peak ozone values for fuel strategies are
provided in Tables 5a,b - the results suggest:
1. proposed gasoline regulations, M100, and CNG all impart
small beneficial impacts on peak ozone,
2. M100 and CNG produce very small peak ozone benefits
relative to implementation of new gasoline regulations;
CNG produces the greatest benefit.
In addiiton to the above caveat concerning dissimilar speeds, the
apparent small impacts from the fuel scenarios might be due to
the projected future decrease of the mobile source fraction of
total VOC emissions. For example, the 1995 base case emissions
for Atlanta_and Dallas (Table 2, Report 5) indicate considerable
reductions in mobile source VOC relative to 1984 and 1985 base
years, despite expected increases in vehicle miles. The
22
-------
reductions between 1984/1985 and 1995 base cases are
substantially larger than the differences between the 1995 base
cases (i.e., with and without new gas regulations). The major
factor causing these projected differences is automobile fleet
turnover. Evidently, the 1995 fleet is very efficient relative
to 1985. Interpretations of these analyses should recognize the
substantial emission reductions from this "built in" control
associated with fleet turnover. Additional substantial VOC
reductions (relative to the 1995 base case) are projected with
implementation of new gasoline regulations which require tighter
tailpipe emission limits.
Emission factors for the M100, CNG and GAS strategies (Table 6 -
light duty vehicles) were provided by the Office of Mobile
Sources and, to the extent possible, were based on available test
data. Relative differences in VOC emission totals between GAS
and M100 strategies are due primarily to reduced evaporative loss
in Ml00 vehicles. CNG strategies assume no evaporative losses
and lower exhaust emission factors relative to the GAS and Ml00
strategies. Identical CO and NOx emissions were assumed for the
three strategies (Ml00, CNG and GAS).
Speciation of exhaust gas VOCs for the M100 and CNG scenarios
were based on work performed by the California Air Resources
Board (CARB). The order of scenarios by exhaust gas reactivity
determined-by weighted k-OH values (min"1) at 298 K is: GAS -
3107, M100 - 1683, CNG - 130.
In summary, predicted region-wide peak ozone values indicate very
small decreases attributed to either M100 or CNG, relative to
proposed new gasoline regulations. Given the significant
differences among scenarios in both quantity and reactivity
within the mobile source fraction, the most plausible explanation
for the minimal impact on peak ozone is due to substantial
emission reductions predicted to be achieved by both fleet
turnover and the new gasoline regulations. The mobile source
fraction of total VOC decreases as one proceeds to future years.
Imbedded in those emission projections are assumptions on
deterioration rates and effectiveness of control devices and
efficiencies of I/M programs. Although best estimates of these
factors are used in the emissions models, it is possible that
they will overstate achievable control.
RESULTS OF SIP STRATEGIES
A listing of the hydrocarbon and combined hydrocarbon and NOx
23
-------
strategies and associated region wide daily peak ozone values is
given in Table 7. From a city specific perspective the
simulations suggest:
Dallas-Fort Worth;
VOC control appears effective
NOx control appears to offset benefits from VOC
controls
St. Louis;
combined VOC/NOx controls show benefits; but additional
NOx control to maximum levels might exacerbate ozone
Atlanta;
combined VOC/NOx controls are more effective than VOC
only
combined VOC/NOx with maximum NOx control yields
additional benefits
NOx control by itself is effective.
All SIP strategy results are expressed relative to the 1995 base
case. It should be noted that the Type A across-the-board VOC
strategies are hypothetical reductions which can not be
translated directly to applied control measures. Nevertheless,
Type A strategies are simple to construct and useful for
developing a general picture of ozone response to VOC controls -
they can indicate broad control levels that a package of
technology based, source specific control measures should meet.
On the other hand, Type B strategies reflect reductions derived
from source specific control measures and explicitly account for
source distribution and composition (i.e., reactivity). The Type
B strategies in this project included control measures resulting
in combined reductions in VOC and NOx emissions. Reductions were
based on a set of prospective control measures prepared for the
South Coast Air Quality Management District (SQAMD) The
principal source for VOC reductions came from solvent control;
NOx reductions were based on tailpipe restrictions and control
from major boilers.
The blend of VOC only and combined VOC and NOx strategies in
three cities provides a small opportunity to speculate on the
widespread effects of VOC and NOx control measures. For example
24
-------
the peak ozone results in Table 7 suggest a strong relation
between city and effectiveness of NOx control. Aggressive NOx
control appears to reduce peak ozone in Atlanta (scenarios 1 and
2), while additional NOx control (Scenario 2) exacerbates peak
ozone in Dallas and St.Louis. Optimum control approaches appear
to be case specific and influenced by the emissions
(anthropogenic and biogenic) profiles of individual cities. One
is tempted to generalize that areas with strong biogenic emission
rates are good candidates for NOx control. For example, consider
Atlanta and Dallas with 66% and 41% biogenic contributions to the
projected 1995 total VOC emissions (Table 2, Report 5),
respectively. Such a simple approach should acknowledge the
uncertainty factors introduced with the inclusion of natural
emissions. What if biogenics actually contribute only 30% in
Atlanta, or 60% in Dallas? While natural emissions represent a
necessary component in determining control strategy
effectiveness, they have added complexity to the control decision
process. Clearly, increasing the precision and accuracy of
natural and anthropogenic emission rates would increase the
confidence in modeling results and provide regulators with clear
information for basing decisions on control approaches.
25
-------
-------
8 CONCLUDING REMARKS
The 5-City study addressed many issues broad in scope ranging
from model application procedures to impacts on ozone resulting
from specific control strategies. As a result, key findings were
spread throughout the five reports and the various sections in
this summary report. The following remarks provide a synopsis of
important findings and major issues related to project objectives
presented in the Introduction.
PLANR APPROACH AND COMPARISON WITH RICH/SPARSE DATA BASES
The PLANR application has been demonstrated in Atlanta and
Dallas-Ft. Worth, and to a lesser extent in St. Louis and
Philadelphia (i.e., recognizing previous modeling data bases in
St. Louis and Philadelphia). The base case performance results
ranged from very good (Dallas-Fort Worth) to poor (Philadelphia).
Overall, the approach is a useful and relatively inexpensive
means of designing and evaluating relative benefits of ozone
precursor control strategies with certain provisions:
the application is limited to areas not complicated by .
long range transport or difficult meteorology,
adequate emissions inventories are supplied, and
uncertainty due to gaps in air quality data is
mitigated by expanding the upwind modeling domain and
simulation period.
In addition, important fundamental modeling issues including
minimum data requirements and appropriate diagnostic analysis
procedures were identified (e.g., the adequacy of NWS/FAA wind
data in Atlanta). Resolution of those issues must be addressed
in the development of guidance on use of PLANR, as well as more
detailed applications of this model.
Results from Report 2 on comparing simulation results with rich
and sparse data inputs were generally inconclusive. Comparisons
were complicated by constraining the domain size and simulation
time for the PLANR approach to that used in the original modeling
analyses, and suspected omissions in the "so-called" intensive
emissions data bases. Three methods for demonstrating attainment
when model results do not match observed ozone were presented: 1)
unbiased, 2) decrement and 3) percentage. Each method provided a
different VOC control target depending on the approach used
26
-------
(i.e., PLANR or intensive). A consensus approach for
demonstrating attainment when model predictions are not closely
matched with observed data should be developed.
OVERVIEW OF STRATEGY RESULTS
Although related to one another through photochemical
considerations, three areas of interest have been addressed: 1)
alternative fuels (including gasoline measures -eg RVP
reductions), 2) biogenic VOC emissions, and 3) VOC and NOx
control measures. Most of the fuel-related strategies showed
Tn% JSn 1?/impa?t onwpeak ozone concentrations. Substitution of
10% ethanol/gasoline blends produced no significant ozone impact
- the increase in evaporative VOCs from a 1 psi exemption
apparently is balanced by reduction in CO emissions The
proposed gasoline regulations, conversion to M100 or CNG all
impart small, beneficial effects on peak ozone. The level of
impact may be viewed as small, but can be explained by the
projected decreasing mobile source fraction of total VOCs in
combination with the nonlinear response of ozone to reductions in
Based on our current ability to quantify natural VOC emissions,
the Atlanta biogenic sensitivity simulations indicate that the
level of anthropogenic VOC control may be understated for some
urban areas when these emissions are not considered. This
finding generally agrees with earlier EKMA modeling performed by
the Georgia Institute of Technology. Results of these analyses
have implications concerning the usefulness of NOx controls in
certain areas. In regions strongly influenced by biogenic
emissions the ambient daytime VOC/NOx ratios might be elevated to
levels where the system is saturated with respect to
hydrocarbons. This emission pattern partly explains the minimal
effect on peak ozone due to substantial VOC emission changes
(Figure 6). In a crude sense areas like Atlanta may be in a NOx
limited condition where ozone is more sensitive to changes in
ISTDv =
NOx.
An illustration of the role of biogenic emissions on the relative
effectiveness of VOC or combined VOC and NOx control is provided
by the SIP strategies for Atlanta, Dallas-Fort Worth and St
Louis. Implementation of aggressive NOx controls (combined with
VOC control) showed peak ozone benefits only in Atlanta
(considered a "high" biogenics area). In contrast, peak ozone
was exacerbated by vigorous NOx control in both St. Louis and
Dallas-Fort Worth - areas not as strongly influenced by biogenic
27
-------
emissions. All three cities benefitted from anthropogenic VOC
control (in the absence of NOx control), although Atlanta
exhibited a relatively insensitive peak ozone response. In fact,
the vast response differences among cities indicate that case
specific control measure approaches probably yield optimum
benefits. Also, implementation of nationwide NOx control
measures might exacerbate peak ozone in certain areas.
TECHNOLOGY DEVELOPMENT AND TRANSFER TO PARTICIPATING STATES
Emissions preprocessing software
Significant resources are applied in processing basic emissions
inventory data into UAM emissions input files. The task includes
translating county level, daily VOC, NOx and CO emissions into
gridded, transient (i.e., hourly) and speciated (VOCs and NOx)
UAM inputs. As part of the 5-City project, SAI developed (to
near completion for public use) a package of FORTRAN utility
programs referred to as the Emissions Preprocessing System (EPS)
which automates the translation process and provides a means for
invoking various levels of control strategy measures (e.g., from
across-the-board to source specific) and other tasks such as
merging anthropogenic area source emissions with biogenic
emissions. The EPS was applied throughout the 5-City study and
is expected to be widely used by State and local air pollution
agencies. SAI is finalizing the EPS and will include the
software in a public domain UAM system package including UAM main
source code with CB4 chemistry, all preprocessors and complete
documentation in the Spring of 1990.
Transfer of DAM to participating States
As part of the project the air pollution control agencies of New
York, Georgia, Texas and Maryland have or will receive the UAM
with CB4 chemistry plus one or two days of installation and set-
up assistance. New York and Georgia have the UAM installed;
Texas and Maryland requested that installation be delayed until
appropriate hardware was inplace. Maryland was selected because
of their location and interest in working with EPA to develop a
regional modeling center and conduct several Region III analyses
including Philadelphia, Washington D.C and Baltimore.
28
-------
-------
TABLE 1. Comparison of performance statistics for four modes of application
of the UAM to St. Louis for July 13, 1976.
UAM RAPS PLANR SIMPLE
Performance Measure (CB-II) UAM UAM UAM
Hourly Ozone Concentrations (matched by time and location)
Numbers of pairs 184* 265 265 265
Average observed (pphm) 8.3 6.8 6.8 6.8
Average predicted (pphm) 7.4 7.5 5.4 4.3
Bias (pphm) 0.9 -0.7 1.4 2.5
Average percent difference 11% 10% 21% 37%
Average absolute (gross) error N/A 1.7 2.02 2.9
Gross error percent difference N/A 25% 30% 43%
Correlation coefficient 0.95 0.91 0.90 0.79
Daily Maximum Ozone Concentration (matched by location but not time)
Number of pairs N/A 14 14 14
Average observed (pphm) N/A 15.0 15.0 15.0
Average predicted (pphm) N/A 16.0 12.3 8.2
Bias (pphm) N/A -1.0 2.7 6.8
Average percent difference N/A 7% 18% 45%
Average absolute (gross) error N/A 2.4 3.9 6.8
Gross error percent difference N/A 16% 26% 45%
Correlation coefficient N/A 0.68 0.55 0.20
Peak Ozone Concentration
Peak observed (pphm) 22.2 22.2 22.2 22.2
Unmatched by time or location:
Predicted region-wide maximum (pphm) 17.4 24.2 19.5 11.8
Ratio of prediction to observation 0.78 1.09 0.88 0.53
Matched by location but not time:
Predicted maximum (pphm) 16.8 21.9 11.7 10.0
Ratio of prediction to observation 0.76 0.99 0.75 0.45
Hours difference in prediction N/A +2 0 -2
* Due to differences in sample sizes between the historical UAM (CB-II) and
the current UAM performance statistics, statistics for non-peak ozone results
cannot be directly compared.
29
-------
In$Lm AMD S?^1^" °t Performance statistics for the UAM (CB-II), POS UAM
and PLANR UAM applications to Philadelphia for July 13, 1979.
Performance Measure (CJMI) JAM
Hourly Ozone Concentrations (matched by time and location)
Number of pairs N/A 330
Average observed (pphm) N/A 6.23
Average predicted (pphm) N/A 7 >6J
(pphm) N/A - 41 -080
Average percent difference N/A 23 13
Average absolute (gross) error (pphm) N/A 2 54 28?
Gross error percent difference N/A 41 4*
Correlation coefficient N/A 0.80 0 68
Daily Maximum Ozone Concentration (matched by location but not time)
Number of pairs 19 ig lg
Average observed (pphm) 13.90 13.90 13 90
Average predicted (pphm) 14.90 13.21 11 75
Jias (PPhm) -1.00 0.69 2 15
Average percent difference 7 5 J
Average absolute (gross) error (pphm) 2.02 2 21 3 60
Gross error percent difference 15 15 26*
Correlation coefficient 0.73 0.72 -0.10
Peak Ozone Concentration
Peak observed (pphm)
Unmatched by time or location:
Predicted region-wide maximum (pphm)
Ratio of prediction to observation
Matched by location but not time:
Predicted maximum (pphm)
Ratio of prediction to observation
Hours difference in prediction
to observation
20.5
26.6
1.30
18.5
0.90
+2
20.5
23.6
1.15
17.7
0.86
0
20 5
£w • +J
18 7
J> W • /
0 Ql
v • y i
11 i
* J. • JL
0 54
W * w~
o
V
30
-------
TABLE 3
SUMMARY OF REGION-WIDE TOTAL EMISSION RATES, VOC-TO NOX RATIOS, AND PERCENT
DIFFERENCES WITH THE 1995 SCENARIOS 1 AND 2 FOR THE 1995 EMISSION SCENARIOS
Scenario*
New York
Scenario 1
Scenario 2
Scenario 3
Scenario 4
St. Louis
Scenario 1
Scenario 2
Scenario 5
Scenario 6
Scenario 7
SIP Scenario X
SIP Scenario Y
Total
(tons/day)
4429
3344
34%
2960
676
575
586
568
849
404
405
VOC
Percent
Change from
Scenario 1
_
-24
-21
-33
_
-15
-13
-16
+26
-40
-40
CO
Percent
Change from
Scenario 2
+32
-
+5
-11
+18
-
+2
-1
+48
-30
-30
Total
(tons/day)
3547
3298
2763
3547
592
555
523
511
592
592
592
Percent
Change from
Scenario 1
_
-7
-22
0
_
-6
-12
-14
0
0
0
Percent
Change from
Scenario 2
+8
-16
+8
+7
_
-6
-8
+7
+7
+7
Total
(tons/day)
1181
1183
1209
1182
350
350
352
352
350
350
350
NOx
Percent
Change from
Scenario 1
0
+2
0
0
+1
+1
0
0
0
I*/
V(
Percent
Change from
Scenario 2
0
+2
0
0
+1
+1
0
0
0
K-to-NOx
Ratio
10.8
8.1
8.3
7.2
5.6
4.7
4.8
4.6
7.0
3.3
3.3
*See text for elaboration of scenarios (pages 49-50).
-------
TABLE 4
COMPARISON OF PREDICTED PEAK OZONE, PERCENT CHANGE IN PEAK OZONE, AND VOC EMISSIONS BETOEEN
SCENARIOS 1 AND 2 AND ALL OTHER 1995 EMISSION SCENARIOS.
Scenario*
New York*
Scenario 1 (current RVP)
Scenario 2 (low RVP)
Scenario 3 (low RVP with ethanol)
Scenario 4 (current RVP with no
running losses)
St. Louis
Highest Ozone
Concentration
(PPM
17.4
17.4
17.5
17.4
Percent Change
of Highest
Ozone with
Scenario 1
0
0
+ 0.5
0
Percent
Change of VOC
Emissions with
Scenario 1
0
-24
-21
-33
Percent Change
of Highest
Ozone with
Scenario 2
0
0
+ 0.5
0
Percent
Change of VOC
Emissions with
Scenario 2
+32
0
+ 5
-11
Scenario 1 (current RVP)
Scenario 2 (low RVP)
Scenario 5 (low RVP with ethanol)
Scenario 6 (low RVP with ETBE)
Scenario 7 (current RVP with high
running losses)
SIP Scenario X (40% VOC reduction)
SIP Scenario y (40% VOC reduction
based on reactivity)
15.0
14.5
14.5
14.4
15.4
13.5
12.3
0
- 3.1
- 3.1
- 3.8
+ 2.7
-10.0
-18.7
o
-15
-13
-16
+26
-40
-40
+ 7 n
T I . U
_ n •)
U.I
+ 6.2
- f. a
D. 7
-15.7
. 1 O
+lo
+ f\
2
1
- 1
+4ft
T4O
Oft
-JU
_on
•jU
*See text for elaboration of scenarios (pages 49-50).
+The analysis did not consider the effects of changes in emissions of the different emission scenarios on the long-range transport of pollutants
(i.e boundary conditions). This contributes heavily to the aaarent insensitivity of the ozone peak to large changes in ttoSSc i vet or or the
New york applications. The reader is referred to the text for further explanation pages 132-133) *
-------
Table 5a. Region-wide maximum ozone concentrations (pphm) for
observed, current and 1995 base cases.
Strateav Dallas-Fort
Observed
Current base
21995 base
17.
16.
13.
Worth
0
4
7
Philadelphia
20.
23.
18.
5
6
6
St. L
22
24
14
ouis
.3
.4
.5
1 - 1985, 1979, and 1976 are current base years for Dallas-
Fort Worth, Philadelphia, and St. Louis, respectively.
2 - The 1995 base emissions projections utilized vehicle
speeds based on the NAPAP inventory; these speeds are
generally higher than the 20 mph used in the fuel
strategies in Table 3-3b, below. Care should be taken
when comparing results from this strategy with those
below since a 20 mph assumption would increase the 1995
base emissions.
Table 5b. Region-wide maximum ozone concentrations (pphm) for
1995 fuel strategies.
Strateav Dallas-Fort Worth Philadelphia St. Louis
New Gas Regs 13.0 18.2 14.3
M1°0 12.4 18.2 14.1
CNG 12.2 18.0 13.9
33
-------
NMHC
0.45
0.184
7.8 psi RVP
_CJL NQx
5.56 0.71
NMHC
0.50
0.26
9.0 psi
_£0_
7.36
RVP
NQx
0.73
Table 6a
Projected In-Use Emissions For
Light-Duty Gasoline Vehicles (grams per mile)
Dallas (T min. 77°, T max. 102°)
Type of
Emission
Exhaust
Evap
Running
Losses 0.154 0.39
Uncontrolled
Refueling 0.17 0.20
Gasoline NMHC speciation guidance provided in the August 23, August 30, and
September 2, 1988 memos from Phil Lorang to Ralph Morris, and a September 7,
1988 memo from Phil Lorang to Gene Durman.
Exhaust emission factors calculated at an average speed of 20 mph.
66% control of these refueling emissions should be assumed.
34
-------
TABLE 6b
PROJECTED IN-USE EMISSIONS FOR
LIGHT-DUTY ALTERNATIVE-FUELED VEHICLES
(grains per mile)
co
en
Type of
Emission
Exhaust
Hot Soak/
Diurnal
Running Loss
Refueling
Vehicles Optimized for 100% Methanol
NMHC
0.049
0
0
0
HeOH
0.490
0.020
0.007
See Note f 3
HCHO
0.015
0
0
0
CO
Same as LDGV
0
0
0
NOx
Same as LDGV
0
0
0
NMHC
0.186
0
0
0
MepH
0
0
0
0
Dedicated
HCHO
0.00
0
0
0
CNG Vehicles
CO
Same as LDGV
0
0
0
NOx
Same as LDGV
0
0
0
1. Speciation for H100 and CNG exhaust NMHC should be taken from the May 19, 1989 CARB technical report titled, "Definition of a Low-Emission Motor
Vehicle in Compliance with the Mandates of Health and Safety Code Section 39037.05 (Assembly Bill 234, Leonard, 1987)."
2. Exhaust emission factors calculated at an average speed of 20 mph.
3. For the M100 case, assume 24% as much methanol as there is NMHC in the gasoline case. This is 91% control of the mass.
-------
Table 7. Peak Ozone (ppb) Results from 1995 SIP Control
Strategies.
St. Louis Observed Peak
1976 Base 244 223
11995 Base 145
224% VOC/25% NOx 133
324% VOC/38% NOx 134
Atlanta
1984 Base 132 147
41995 Base 125
530% VOC 119
60% VOC in
90% VOC 106
18% VOC/32% NOx 113
318% VOC/50% NOx 108
Dallas-Fort: Wnrl-h
1985 Base 164 170
61995 Base 137
30% VOC H6
60% VOC 108
24% VOC/40% NOx 133
424% VOC/60% NOx 134
1all 1995 reductions relative to 1995 base"
2 - combined VOC/NOx controls are source specific from LA Tier 1
control measures plus NOx control on major boilers; these
are referred to as SIP type B strategies in text
3 - additional NOx control beyond previous strategy from major
boilers .
4 - Atlanta 1995 base represents 9% increase and 35% decrease in
NOx and VOC, respectively, over 1984 levels
5 - all VOC only controls are across-the-board and are referred
to as SIP type A strategies in text
6 - Dallas 1995 base represents 32% increase and 17% decrease in
NOx and VOC, respectively, over 1985 levels (weekday
emissions)
36
-------
160
I I I I I | I 1 I I I | I I I I I | I I I | |
i i i I i i i i i I
- DKLB
—
U PREDICTED — -
12
TIME (HOURS)
18
24°
12
TIME (HOURS)
18
24
160
120
CD
8: so
40
12
18
24
i i i i | i I i
- DLLS
OBSERVED
PREDICTED
Q
160
120
80
40
6 12 18
TIME (HOURS)
24
0™, lm observed and nearest-neighbor predicted (two-cell
search) ozone concentrations (ppb) for the Atlanta evaluation run
37
-------
160
1SO
- 120
12
TIME (HOURS)
2*
160
12
18
24
i I I I 1 | I 1 II I | I I I 1 I | I I 1 I
• DLLS •
OBSCRVCO Q
PREDICTED — -
120 -
8 .
160
120
Figure 2. Observed and nearest-neighbor predicted (two-cell
search) ozone concentrations (ppb) for the Atlanta reduced wind
speed sensitivity test.
38
-------
200
I I I I I I | I I I I I | I I I | I I 200
MOCK
I OBSERVED 03
6 12 18
August 30. 1985
2* 30 36 42
Time (hours) August 31. 1985
NSTR
_ OBSERVED O
150
m >
& 100
50
16
**
' '
3°
36
111 I ' ' ' ' • I ' i i i i | I I I I I | | | | | | |
/ .
i i i I i i i i i I
12 18
I
I I ]
' ' ' ' ' I i I i I I l i i i
36
6 1Z « 24 30 36 42
August 30. 1985 Time (houra) August 31. 1985
200
150
100
50
48
Figure 3. Observed and nearest-neighbor predicted (two-cell
search) ozone concentrations (ppb) for the Dallas-Ft. Worth 30-
31, August 1985 base case simulation.
39
-------
20O
12
I I I I I I I I I I
ILLI
_ OBSERVED D
150
100
50
18
30
36
42
48
200
150
100
50
6 12 18
August 30. 1985
24
Time (hours)
30 36 42
August 31. 1985
48°
20O
6
i I i 1 T
BNVW
_ OBSERVED
150
OQ
fc 100
O
O
12
18
24
' ' I ' ' ' i i I i i i i i I i i i
30
1 ' ' I ' ' I ' I | I I I I I 1 I I I I I | | |
36
42
48200
150
100
50
36 42
August 30. 1985 Time (hours) August 31 1985
48
Figure 3. continued
40
-------
6 12 is
August 30, 1985
30 36 42
Time (hours) August 31. 1985
200
ROSS
_ OBSERVED Q
m
150
100
SO -
6 12 18
August 30. 1985
24
Time (hours)
30 36 42
August 31. 1985
150
100
50
Q-
Figure 3. cdntinued
41
-------
200
I I I I I I l . . i , I . . , ,
6 12 18 24
AUgUSt 30. 1985 Time (hours)
30 36 42
August 31. 1985
Figure 3. continued
42
-------
UNBIASED
DECREMENT
APPROACH
PERCENTAGE
APPROACH
40 60 go
% VOC REDUCTION
Figure 4. Peak ozone response for St. Louis to across-the-board VOC
controls using three different approaches when predicted
peak ozone does not match observed value.
43
100
-------
250
240
£N 230 -
gp 220 -
%< 210 -
&*, 200 -
r^ 19° -
S 180-
Z no -
O 160 -
N 150 -
O 140 -I
UNBIASED
APPROACH
DECREMENT
APPROACH
PERCENTAGE
APPROACH
% VOC CONTROL
Figure 5. Peak ozone response for Philadelphia to across-the-board VOC
controls using three different approaches when predicted
peak ozone does not match observed value.
44
-------
biogenics included
20
40
% VOC REDUCTION
Figure 6. Peak ozone response in Atlanta to across-the-board VOC
controls with original (higher speed) wind speeds.
biogenics included
20
40
% VOC REDUCTION
60
Figure 7. Peak ozone response in Atlanta to across-the-board VOC
controls with new (lower speed) wind speeds.
45
-------
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
EPA 450/4-90-006A
3. RECIPIENT'S ACCESSION NO.
>'URBAENAADIRSHEDLMODEL STUDY OF FIVE CITIES
(Summary Report)
5. REPORT DATE
April 1990
6. PERFORMING ORGANIZATION CODE
8. PERFORMING ORGANIZATION REPORT NO.
Richard D. Scheffe
FORMING ORGANIZATION NAME AND ADDRESS
U. S. Environmental Protection Agency
Office of Air Quality Planning and Standards
Research Triangle Park, NC 27711
10. PROGRAM ELEMENT NO.
11. CONTRACT/GRANT NO.
ONSORING AGENCY NAME AND ADDRESS
U.S. Environmental Protection Agency
Office of Air Quality Planning and Standards
Research Triangle Park, NC 27711
13. TYPE OF REPORT AND PERIOD COVERED
14. SPONSORING AGENCY CODE
NOTES
16. ABSTRACT
This document provides an overview and summary of the 5-City UAM Study.
17.
KEY WORDS AND DOCUMENT ANALYSIS
'
DESCRIPTORS
Urban Airshed Model
Ozone
Control strategies
18. DISTRIBUTION STATEMENT
Unlimited
EPA Form 2220-1 (R«v. 4-77) PREVIOUS EDITION is OBSOLETE
b.lDENTIFIERS/OPEN ENDED TERMS
19. SECURITY CLASS !TMs Report)
Unclassified
20. SECURITY CLASS 'Tins page)
Unclassified
c. COSATI Field/Group
21. NO. OF PAGES
47
22. PRICE
-------
INSTRUCTIONS
1. REPORT NUMBER
Insert the EPA report number as it appears on the cover of the publication.
2. LEAVE BLANK
3. RECIPIENTS ACCESSION NUMBER
Reserved for use by each report recipient.
4. TITLE AND SUBTITLE
KSa^bSSrt SSn^e"^ IT"""1"*- So, suN,,., if used. ,n snw,,r
number and include subtitle for the specific title. prepared in more than one volume, repeat the primary title, add volume
5. REPORT DATE
6. PERFORMING ORGANIZATION CODE
Leave blank.
7. AUTHOR IS)
Giveoname(S, ,n conventional order (John R. Doc. J. Rotor, Doc. „,.,. List author's .mu.no. if i, differs from ,hc performing (,rpan,.
8. PERFORMING pRGANIZATION REPORT NUMBER
insert if performing organization wishes to assign this number.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Give name, street, city, state, and ZIP code. List no more than two levels of an organizational hirearchy.
10. PROGRAM ELEMENT NUMBER
Use the program element number under which the report was prepared. Subordinate numbers msiy be included in parentheses
11. CONTRACT/GRANT NUMBER
Insert contract or grant number under which report was prepared.
12. SPONSORING AGENCY NAME AND ADDRESS
Include ZIP code.
13< niPE OF REPORT AND PERIOD COVERED
Indicate interim final, etc., and if applicable, dates covered.
14. SPONSORING AGENCY CODE
Insert appropriate code.
15. SUPPLEMENTARY NOTES
• such as:
16. ABSTRACT
SStaw
"• ^fXJSiS AND DOCUMENT ANALYSIS
SgnCant n°"na'ion «»«-«' '» "- —,. ,r .he report «,nuin, a
.
18. DISTRIBUTION STATEMENT
i^bffSSSta'S^" limiUti°n f°r feaSOnS °thW than ^""^ f" «""'" "^'— UnlimiM." Cite any avail.hih.y „,
19. & 20. SECURITY CLASSIFICATION
DO NOT submit classified reports to the National Technical Information service.
21. NUMBER OF PAGES
Insert the total number of pages, including this one and unnumbered pages, but exclude distribution list, if uny
22. PRICE
Insert the price set by the National Technical Information Service or the Government Printing OtT.c-. if known.
EPA Form 2220-1 (Rev. 4-77) (R.v.r,,)
------- |