EPA/600/R-96/134
July 1995
A NEW APPROACH
FOR DEMONSTRATING ATTAINMENT OF THE
AMBIENT OZONE STANDARD
MODELING, ANALYSIS, AND MONITORING CONSIDERATIONS
by
Kenneth L. Demerjian
248 Stow Road
Harvard, MA 01451
Philip M. Roth
Envair
836 Fawn Drive
San Anselmo, CA 94960
and
Charles Blanchard
Envair
526 Cornell Avenue
Albany, CA 94706
for
Basil Dimitriades
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park, NC 27710
Printed on Recycled Paper
-------
DISCLAIMER
The information in this document has been funded wholly by the
United States Environmental Protection Agency under contracts
4D1577NATA and 4D1653NASA, to Dr. Kenneth L. Demerjian and
ENVAIR, respectively. It has been subjected to the Agency's peer
and administrative 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.
-------
TABLE OF CONTENTS
I. PURPOSE
H PERSPECTIVE
A.
B.
C.
D.
The Decision Making and Information
Attainment Demonstration
Shortcomings of the Attainment
Demonstration Process
Improving the Process
1-1
n-
n-
n-
n-
n-
i
i
3
5
9
m. PROPOSED APPROACH TO TRACKING PROGRESS
AND DEMONSTRATING ATTAINMENT m - 1
A. A Schematic Depiction of an Improved
Process HI - 1
B. The Primary Information Flows TTT - 4
C. The Primary Elements: A Synopsis El - 6
IV. ELEMENTS OF THE PROPOSED APPROACH IV - 1
A. Photochemical Air Quality Simulation
Modeling rV - 1
B. Observation-driven Methods IV - 5
C. Analysis of Observations P7 - 23
D. Monitoring IV -27
-------
TABLE OF CONTENTS (cont'd)
V. AN AGENDA FOR RESEARCH AND DEVELOPMENT
A. Data Analysis and Related Activities V-l
B. Modeling V-2
C. Monitoring V-7
D. Longer Term Needs V-9
VI. REFERENCES
VI-1
-------
I. PURPOSE
Over the past 15 years photochemical air quality simulation models (PAQSMs) have
played a major role in elucidating our understanding of the relationship between the
emissions of ozone precursors (volatile organic compounds, or VOCs, and nitrogen
oxides, or NOX) and meteorology with the formation of ozone and other
photochemical oxidants. As a result of provisions of the Clean Air Act Amendments
of 1990, photochemical modeling is now the cornerstone of the air quality
management approach used in the development of emissions control strategies
designed to bring into attainment areas in the United States that exceed the NAAQS
for ozone.
While twenty five years of regulatory efforts have effected substantial improvements
in ozone-related air quality in the United States, these improvements have fallen
short of goals and expectations. The failure to meet the air quality standard for
ozone in a timely manner has raised concern nationally and elicited considerable
attention, one result of which was the commissioning of major scientific studies by
the Office of Technology Assessment (1989) and the National Academy of Sciences
(1991). These studies have identified several basic issues regarding the state of the
science and the operational approaches used in the management of ozone air quality.
Specifically, these studies find that:
• current pollutant monitoring networks are inadequate to measure and track the
performance and progress of the ozone air quality management process,
• representations of the emissions of ozone precursors can have significant errors
and uncertainties and have not been subjected to independent testing and
evaluation procedures,
• uncertainties in biogenic emissions and their contributions to ozone air quality
may mask unanticipated and in some cases unachievable control requirements,
• a single national control strategy for the control of VOC and/or NOX emissions
may not apply uniformly in all areas, and
• regulatory expectations for the use of models in the state implementation planning
(SIP) process exceed their performance capabilities.
The premise of this document is that we cannot and should not continue to rely on
the use of PAQSMs alone (or even substantially) in developing plans for attaining the
NAAQS for ozone. Rather, we should design and adopt a planning approach that (a)
integrates attractive aspects of key elements - modeling (using PAQSMs), data
analysis, and, especially, ambient and source monitoring • in continuing to pursue
attainment of the NAAQS and (b) provides for accountability and feedback, so that
I- 1
-------
corrective actions may be taken as needed. Such an approach is intended to:
• take advantage of the strengths of each of these individual elements,
• minimize the impact on planning of their known limitations, and
• integrate the application of the elements so that planning is a continuing and
efficient process.
In this document we propose a new approach for demonstrating attainment of the
NAAQS for ozone.
1-2
-------
II. PERSPECTIVE
A. The Decision Maker and Information
The Needs of the Decision Maker
A central objective of planning to improve air quality in an urban area or a larger
region is to establish how best to attain the NAAQS for ozone. Specific information
needed by a decision maker includes determining:
• requirements for emissions reductions,
• the extent of transport of pollutants into the region and the impact of these
pollutants on air quality,
• the anticipated response of the air resource to widespread emissions reductions,
both locally and upwind,
• the effectiveness of emissions controls imposed (i.e., the fraction of prescribed
reductions actually achieved),
• the extent of progress toward attainment of the standard through the monitoring
of air quality in a region, and
• the magnitude of export of pollutants out of the region and the potential for
adverse impacts downwind.
Technical Pursuits
To address these informational needs, a variety of technical endeavors are conducted.
These include:
• characterizing air quality through routine ambient monitoring of ozone, and in
some instances, one or more precursor species. Ozone is monitored in all large
urban areas, at varying network density and spatial coverage. VOCs are
monitored in only a fraction of cities; however, with the introduction of PAMS
sites, coverage is improving markedly. NOX monitoring varies widely, from the
dense network of the South Coast Air Basin to the absence of NOX monitoring in
the some metropolitan areas. Here, too, PAMS is improving coverage.
• conducting an intensive measurement campaign to acquire a comprehensive
aerometric data base that will support analysis and modeling aimed at increasing
understanding of the dynamic atmospheric processes associated with adverse air
quality.
II- 1
-------
• planning air quality improvements through the use of PAQSMs. Needs include
assessing if the reduction of VOC or NOX or both limits the formation of ozone,
determining the magnitude of emissions reductions required, establishing the
soundness and reliability of estimation procedures, and determining the
geographical region over which emissions reductions are to be imposed.
• demonstrating progress toward attainment through ambient monitoring. Efforts
include characterizing air quality improvements through the analysis of
concentration trends for both primary and secondary pollutants, adjusting for
meteorological variability,
• tracking emissions change!, Activities may include both ambient and source
monitoring, sampling for categories of lources comprised of a large number of
individual emitters (such ai motor vehicles), undertaking special experiments
(such as tunnel studies), and conducting surveys or audits of surrogates for source
emissions. Source monitoring is now conducted for many but not all categories of
sources.
• evaluating the effectiveness of emissions controls through the analysis of relevant
data. Observational information acquired is used in the analysis of emissions
trends and the assessment of the efficacy of control programs.
• estimating the impacts of pollutants transported into a region, also through
modeling. Often, inadequate attention is given to this need. This information
may be provided as an output of PAQSM simulations.
Note that these technical endeavors, as undertaken now, address only a portion of
the decision maker needs listed above, Spacifically, the first two and last items in the
list merit greater attention than they now receive. An improved approach to
attainment demonstration should address each of the needs listed and incorporate
each pursuit described as effectively as possible.
In addition, there is need to evaluate and weigh the relative costs and benefits of
alternative control strategies in light of uncertainty, preferably through the formalism
of decision analysis. Due to the more limited scope of this effort, these topics are not
now addressed. However, they merit inclusion as components in an overall
framework, and should be considered in any subsequent scoping activities pertaining
to attainment demonstration,
II-2
-------
Air Quality Monitoring
A key element of a new approach must include increased (in quantity) and expanded
(spatially) use of air quality monitoring. The current dependency on modeling as the
central element in air quality management suggests an acceptance of PAQSMs as
accurate estimators of present and future air quality. In fact, as we shall discuss, '
modeling results virtually always possess a degree of uncertainty that limits or
discourages confidence in the result. In contrast, monitoring, being frequently
perceived as "ground truth", engenders a degree of confidence that is often not
warranted. In particular, the lack of specificity of some measurements, such as for
NOz, and the lack of ability to monitor at very low concentrations, such as for NO
and NO2, account for the reservations. Nevertheless, with the use of improved
instrumentation now becoming available, awareness of the pitfalls of monitoring, and
application of a proper level of care in conducting measurements, monitoring
provides a reasonably reliable adjunct to modeling in establishing knowledge of air
quality and in providing feedback to a continuing planning process, one that allows
for the improvement of PAQSMs through incorporation of newly gained knowledge
and the revision or adjustment of their estimates based on "real world" measurement.
In general, we envision several consequences of an expanding role of ambient
monitoring:
• Monitoring information is likely to be incorporated more fully in the planning
process due to the continuing concern about the uncertainties and inaccuracies
that attend modeling.
• Metropolitan areas that choose to expand their routine monitoring network
because of inadequate spatial coverage or the absence or paucity of monitoring of
selected pollutants will wish to take maximum advantage of their new
investments.
• The introduction of PAMS sites in major cities is providing impetus to use the
data in planning and assessment.
• Ambient monitoring of precursors in the vicinity of sources, coupled with in-
source monitoring and special studies, will provide information on the efficacy of
controls that is not now available to the extent needed.
• Tracking of progress and continuing assessment of accountability will be more
feasible.
Overall, there is need for a shift in planning philosophy from a dependency on
modeling results with little feedback to a commitment to a fully integrated planning
effort based on modeling, expanded monitoring, and data analysis, with continuing
II-3
-------
feedback of observational information into the analytical elements of the process.
B. Attainment Demonstration
As discussed, models play a central role in the ozone air quality management process
and are the operational tool used in state implementation planning to quantify the
emission control requirements needed to achieve the ozone air quality standard. The
air quality management approach defines the conceptual framework for linking
pollutant emissions to ambient air quality, and the state implementation plan
provides the operational context for carrying out the management approach.
Schematics of the air quality management approach and the state implementation
planning process are provided in Figures H-l and E-2. Although it is possible to base
the air quality management approach on elements other than models (e. g.,
technology-based emissions controls and/or monitoring), this has not been the case
for ozone. Since ozone is not directly emitted into the atmosphere, but formed
through photochemical reactions of primary emitted precursors (VOC and NOX),
models are thought to be the most effective means for quantifying the complex
chemical and meteorological interactions that relate precursor emissions to ozone
production.
II-4
-------
Determine Air
Quality Status
p
Evaluation
Determine Emission
Reductions
Monitoring
Emission Invento
Modeling
Monitoring
Model!
Implementation
Identify Emission
Control Program
I SIP Rules
NSR Permits
I Enforcement
NSPS
FMVCP
VOC RACT
NOx RACT
Figure II-l. Air Quality Management Approach
The state implementation planning process (Figure II-2) requires data for several
purposes. First, ozone air quality data are needed to determine if the standard has
been violated and to what extent. During the demonstration phase of the SIP
process, air quality, meteorological and emission data must be acquired to support
the operation of a modeling system and the evaluation of its performance in
estimating air quality for the base case year. Models are then used to estimate the
emission reductions (taking into consideration future emission projections) needed to
H-5
-------
attain the ozone air quality standard. Once the control program for achieving the
target year demonstration for ozone attainment is stipulated and all supporting
documentation is prepared, the SIP is submitted for public comment and approval by
EPA. This completes the demonstration phase of the SIP.
As part of the implementation phase of the SIP, a new provision under the Act
requires the tracking of emissions to determine reasonable further progress (RFP).
The RFP requires a 15% reduction in VOC emissions over the first 6 years and 3%
reduction per year thereafter until attainment is achieved in all but "marginal"
classified areas.
Input Data
& Present
AQ Status
Air Quality
BC&IC
O5NOX&VOC
Meteorology
Wind Fields
Mixing Heights
Episodes
fa
Emissions I
Mobile, Area &!
Biogenic/Natura
NOX&VOC
Demonstration Phase Implementation
Phase
Base-year • largel-year EPA
Evaluation" Attainment Approval
Attainment
and
Redesignation
Annual
Inventory
Tracking
Reasonable
Further
Progress (RFP)
Figure II-2. State Implementation Planning Process
n-6
-------
C. Shortcomings of the Attainment Demonstration Process
The Overall Process
As currently mandated under Clean Air Act Amendments of 1990, the air quality
management approach and its use of photochemical air quality simulation models in
the SEP process to develop control strategies for attainment of the ozone standard has,
in effect, failed to meet its objectives. For a given geographical area, this lack of
success stems from one or more of the following potential problems:
• The time and resources availableare inadequate to support SIP modeling
requirements.
• The availability of data are inadequate to support SIP modeling.
• Limitations inherent to PAQSMs strongly suggest that regulatory needs may
exceed performance capabilities.
• Annual inventory tracking of reasonable further progress (RFP) frequently is "a
paper exercise" with little or no substantive accountability.
• Although PAMS has the potential to establish monitoring as an effective means
for tracking progress and introducing accountability in emission control programs,
substantial implementation and operational issues remain unresolved.
Although the air quality management approach, in principle, appeared to be a viable
methodology for addressing the problem of air pollution, it has encountered a
number of failures as a result of flaws in execution and an insufficiency of means and
incentives to track progress and account for the effects of emission control strategies
on air quality. The future success of the ozone control program will depend strongly
on the community's ability and willingness to address these shortcomings. Actions to
be considered include:
• seeking to modify CAAA requirements and/or attendant regulations,
• developing improved guidelines for applying models,
• further developing, refining and testing empirically-based, observation-driven
methods (ODMs),
• improving monitoring capabilities, including PAMS implementation, and refining
PAMS-related methodologies to facilitate addressing existing problems.
H-7
-------
The approach proposed here incorporates each of these responses.
Models
We have expressed concerns about the extent to which modeling systems (i.e.,
comprised of air quality, emissions, and meteorological models) adequately portray
the dynamics of atmospheric processes that govern ozone formation and thus about
the accuracy of simulated air quality representations. These concerns include:
• deficiencies in knowledge of key dynamic processes and thus limitations in the
formulations of these processes embodied in the model. Processes of concern are
emissions, transport and dispersion, gas phase chemistry, and dry deposition. In
the course of model development and application through the 1980s, notable
improvements were made. These include the introduction of prognostic
meteorological modeling in the mid-1980s, the incorporation of mechanistically-
based dry deposition algorithms in place of more empirical relationships, and
with the discovery of underestimation of VOC emissions ,in the late 1980s, efforts
to correct for it, either through improved estimates or adjustments of the incorrect
estimates based on independent evidence, the inclusion of NOX emissions from
soils in 1993, and adjustment of the emissions rates of isoprene in 1994. Each
improvement introduced significant alterations in model inputs or formulation
and thus the potential for changes of similar significance in modeling results.
In recent years, regional scale modeling is receiving increasing attention. As
spatial and temporal scales expand, the simulation of some dynamic processes
becomes more critical and others less critical. For example, the simulation of biogenic
VOC emissions and the deposition of pollutant species requires greater attention and
accuracy. In most cases, the needed shifts in emphasis have yet to be adopted.
Attention should be given to improvements in the representation of individual
processes, insofar as demands for accuracy of depiction increase with increasing
spatial scale.
• additional demands for data and improved performance created by the increasing
complexity of model formulation. As greater attention is given to simulating the
dynamics of individual processes, model complexity increases. Attending
increased model complexity are (a) increased difficulty in simulating each
individual process correctly, (b) the potential for greater cumulative uncertainty
(while imprecision may increase, bias may be reduced) and (c) increased
computational demands. Thus, the potential for uncertainty increases after a
certain point, rendering overall acceptance of the model more difficult. Some feel
that the "bottom of the complexity-uncertainty curve" has already been passed,
and the community is being confronted with evaluation requirements that are
increasingly difficult to meet.
II-8
-------
• limitations in the availability of information needed as inputs to UAM. Typically,
important data deficiencies exist for emissions estimates for VOCs and NOX,
meteorological fields (wind velocities and vertical exchange rates), boundary
conditions, and aerometric data aloft. Such limitations exist even when relatively
rich data bases are available, but they are particularly severe when data are only-
routinely collected in an urban area. Most areas are lacking or deficient in data-
needed to estimate boundary conditions and meteorological and air quality
conditions aloft. Relatively few areas monitor VOC concentrations. Moreover,
. due to limitations in measurement capabilities, monitored concentrations of NO,
NO2, and other nitrogen-containing atmospheric pollutants are often inaccurate.
Where data gaps of consequence exist, modeling accuracy suffers, and the
prospects for identifying and reducing or eliminating the presence of
compensating errors are diminished.
• limitations imposed in estimating the likelihood of attainment based on the results
of simulations for only one or two episodes for which intensive data are acquired
and intensive modeling is pursued. The number of episodes simulated today in
preparing a SIP does not begin to meet the needs of representation of
meteorological regimes of interest. Also, highly adverse episodes often are not
captured in data collection programs. In addition, EPA guidelines permit the
elimination from consideration of "hard to simulate" episodes.
• limitations imposed in estimating the likelihood of attainment that derive from the
deterministic analysis of meteorological processes that are inherently stochastic
and thus the inability to directly relate results to the standard (which is
probabilistic in formulation). The results of PAQSM simulations do not and
cannot tell us the probability of an area's exceeding the NAAQS in the future,
once a stipulated control program is fully in place, given that modeling results
indicate that attainment will be achieved.
• the potential for inaccuracy and imprecision in model-derived concentrations
estimates devolving from the deficiencies and limitations noted above. Model
imprecision for ozone concentrations typically ranges from 25-40%, inaccuracy
(bias) from 5-30%. These uncertainties should be compared with a typical "signal"
of 25-40% - the fraction of reduction in the peak ozone concentration for an area
(150-170 ppb) that is needed to achieve attainment (at 120 ppb).
In most modeling exercises, the modeler adjusts inputs in attempting to improve
performance. A key test of quality of performance is to evaluate the model for other
episodes without adjustments, using the same rules for establishing model inputs for
the second and third episodes as were used for the first. This type of test is rarely
applied. When it is, performance typically is inconsistent.
• the potential for introducing compensating errors into the model as a
H-9
-------
consequence of the inability to evaluate model components independently. It
appears that compensating errors have been present in many past applications.
Their presence introduces the potential for bias into the estimation of the impacts
of emissions control strategies.
• title lack of availability and use of well founded emissions projections techniques.
Complex socioeconomic forces drive population growth, industrial expansion,
residential and commercial development, and emissions changes. Little effort has
been made nationally to develop the capability to project future emissions based
on a quantitative evaluation of alternative futures. Thus, uncertainties in
projections of alternative futures made ten years or more in advance can
significantly exceed inherent uncertainty levels. This deficiency in prediction
capability is not likely to be rectified in the near future.
Implications
These issues, in the aggregate, place into serious question the potential for the
PAQSMs, including UAM-IV and UAM-V and related models, to provide accurate
quantitative representations of the impacts of emissions reductions on peak ozone
concentrations and thus of attainment demonstration. Given the unfavorable noise-
to-signal ratios in model simulations and the frequent presence of compensating
errors in the modeling systems, the high level of dependency being placed on this
type of model seems inappropriate. If model performance is judged to be inadequate
and cannot be improved, no alternatives for developing emissions control strategies
will exist. It thus seems prudent to adopt an approach that either permits "back-up"
or constitutes an alternative, such that a much lower probability of failure
accompanies the use of these models. Unfortunately, today's guiding principle for
the use of a PAQSM might be stated as, "The model is the best tool available," so let's
use it, and not, "Is it good enough to meet the need?"
It is crucial to recognize that the history of air quality projection using PAQSMs is
one of consistently overestimating the reductions in peak ozone concentrations that
are to be realized in a specified period of time. A record of emissions reductions
prescribed in the past and projections of future air quality now exists. SIPs prepared
since 1979 and throughout the 1980s document this information. In the vast
preponderance of situations, projections of air quality made in the past have not been
achieved in the present. While there are many reasons for this, the inaccuracy of
modeling systems (i.e., air quality, emissions, and meteorological models and their
inputs) is a likely contributor to this failure in projection.
As discussed, a second notable deficiency in the SIP process also merits attention: the
lack of (a) accountability for ensuring that plans are met and (b) evaluation and
demonstration requirements. The current SIP process stipulates control programs
H- 10
-------
based on "engineering estimates" that are intended to achieve needed reductions in
precursor emissions, as projected through model simulations. Unfortunately, the
process has lacked requirements or procedures to track progress in reducing
emissions, attaining emissions reduction goals, and achieving planned air quality
improvements. In addition, it has lacked a requirement and a post-SIP effort to
assess the accuracy and reliability of models used in estimating future air quality. '
There is a need to institute a process of retrospective analysis for determining
cause(s) of failures in modeling, such as flawed formulations, erroneous emissions
representations, inadequate data bases, and insufficiently effective implementation of
emissions reductions, and to introduce improvements that will lead to reliable
estimation of control needs.
D. Improving the Process
Needs
Models play a central role in the ozone air quality management process and are the
operational tool used in state implementation planning to quantify the emission
control requirements necessary to achieve the air quality standard for ozone. The air
quality management approach defines the conceptual framework for linking pollutant
emissions to ambient air quality, and the state implementation plan provides the
management approach. The experience gained in SIP preparation and the subsequent
failure to meet air quality standards in the time frame prescribed speaks strongly to
the need for an improved planning process, one which, if conducted thoughtfully,
will not experience failure in its outcome. Needs to be met that are not now being
met include:
• oversight of SIP implementation, including tracking progress and demonstrating
the extent to which the plan being implemented is being met,
• a mechanism for mid-course adjustment or correction, where emissions reductions
fall short of targets as a function of time or where air quality improvements,
adjusted for meteorological variability, fall short of targets as a function of time,
and
• diagnostic analysis of modeling failures and improvement or correction of the
modeling process.
II- 11
-------
A proposed direction
We propose that a new process for attainment demonstration and SEP preparation be
developed and adopted, taking advantage of the strengths of elements (such as
modeling, data analysis, monitoring) now in use, minimizing the impacts of key
limitations of these elements, introducing tracking and feedback, and requiring
diagnostic analysis of the SB?'process during the years following. Section IQ
proposes such an approach, Section IV delineates its elements, Section V offers
recommendations for research and development in support of the new approach, and
Section VI suggests additional improvements that should be considered over the
longer term.
II- 12
-------
III. A PROPOSED APPROACH TO TRACKING PROGRESS AND
DEMONSTRATING ATTAINMENT
A. A Schematic Depiction of an Improved Process
We now describe an alternative approach to attainment demonstration, one that de-
emphasizes but retains the use of PAQSMs and adds other elements which, when
taken together, constitute a process for achieving attainment rather than an act of
attempting to demonstrate attainment. We refer the reader to Figures ffl-la and Ill-lb,
which present a detailed schematic diagram depicting the approach. We now briefly
describe its elements and their purposes.
Conduct initial data acquisition
Supporting modeling and ODM use with a data base adequate to meet the input
requirements of the methods is a requisite for reducing uncertainty, eliminating bias,
and building confidence in the use of these tools in decision making. If formal air
quality planning is contemplated for an area, the planning and implementation of
ambient aerometric monitoring should begin as soon as possible. Sections IVD and
VC of this report address monitoring needs.
Conduct modeling, using both FAOSMs and ODMs
PAQSMs should be used to simulate the temporal and spatial distributions of ozone
concentrations that are anticipated to result from prescribed reductions in the
emissions of VOC and NOX. However, their limitations should be recognized, and
expectations associated with their use should not exceed their demonstrated
performance capabilities. The following principles should be observed in applying
photochemical modeling systems (air quality, emissions, and meteorological models):
• Acquire and use a suitable data base for performance evaluation, control strategy
assessment, and sensitivity analysis.
• Require adequate testing and performance prior to use. Model evaluation
protocols prepared for the Western States Petroleum Association (WSPA)
(Reynolds, et al., 1994) and the San Joaquin Valley ozone study (Roth, et al., 1995 )
provide guidance on the conduct of model evaluations. Among the many issues
meriting attention are the need (a) for "hands off" testing of the model for
episodes after the first studied and (b) to examine the
IE- 1
-------
^J
j^>j
1
•s
S
u
Si
•a o
c« u
3*3
« 00
£«E8
"So
C O co
us
o »s
SI
03 JO
§
Q
O
1
5
A
llll
IMA.
« §
2 g-
e8 g
a, o
e> •«
S
a>
Ml
i
$
T3
S
'B,
W
ts
u
"3
CO
V
S
u
g
fi
«
IS
.-§
CO
S
CO
a
?
«2
4-H
4>
CA
1
3
ca
S
Q
0
1-4
•s
ts
CO
"S
T3
i
t
j£
f -2
tf ^
>> Q
|.|
CO «
°8 I
£ a3
l<
O
§
I
I
|=3
||«
^ 1
b
m-2
-------
from Emission
corroborative
Estimate AO3 for
AEyoc and/or AENOx'
Using PAQSM
Estimate Risk via
Alternate Base
Cases (ABCs)
Revise
Control Strategy
Develop Control
Strategy
Rerun PAQSM;
Update ODM
Project
AO3max = f(Time)
Project
AE = f(Time)
Implement
Strategy
Monitor
AO3 max = f(Time)
Adjust Cone. Trends
for Meteorology
Compare
' Discrepaiitx^xx^ Tracking
Diagnose Cause;
Revise
Model / Inputs
Monitor Precursor
Concentrations
a) ambient, near source
b) insource
Establish E = f(Time)
Improve
Effectiveness
of Contro
Continue
Monitoring
Modify AE
Plan
Figure III.l.B Schematic Depiction of Attainment Demonstration Process (continued)
m- 3
-------
adequacy of individual modules comprising the models in order to eliminate, to the
extent feasible, the possibility of undetected compensating errors.
• Apply and test the model for episodes representing a satisfactory range of
meteorological conditions.
• Give explicit consideration to the need to model an area upwind of the
metropolitan area of interest in order to examine the impacts of upwind sources
on ozone concentration in the metropolitan area and downwind.
• Conduct analyses of alternative base cases to determine the range of simulation
outcomes that are consistent with the uncertainties of the modeling system.
• Provide estimates of uncertainties in outcomes to the extent that these can be
developed.
One or more observation-driven methods (ODMs) should be used to estimate the
spatial and temporal ranges of VOC versus NOX limitation for the atmospheric
system and episodic conditions under study. Several ODMs are under development
and evaluation. Some are available for use; of these, most are available in a
preliminary form. All ODMs are discussed in Section IVB.
Establish corroboration among models and methods
Results should be compared directly with those of the PAQSM, to determine if areas
of VOC and NOX limitation coincide. If significant discrepancies exist, diagnostic
analyses should be conducted to determine possible causes. Corrections should then
be made and the model and methods run again and compared. Proceeding further
along paths on the schematic should be conditioned on the acceptability of
comparisons for several episodes.
Develop alternative candidate emissions control strategies and identify the most
attractive option
Once a successful evaluation of a PAQSM is conducted, following the guidance of the
protocol cited, and the results are corroborated using one or more ODMs, the impacts
of alternative emissions reduction scenarios should be estimated and compared.
Emissions estimates should be developed using two or more independent estimation
methods - source-oriented vs ambient and/or"top-down" vs "bottom-up", whenever
possible to afford a means for reliably evaluating their soundness. Where estimates
are discrepant, diagnostic analyses should be conducted to determine the causes, and
corrections should be made. The attainment demonstration process should proceed
further only when the accuracy of the emissions estimates is reasonably assured.
Ill-4
-------
Emissions projections should be prepared for use for "future year" simulations.
Unfortunately, effective procedures for projecting emissions to future years have not
been developed for use in air quality modeling. See Chapter V for a discussion of
needed research and development in the preparation of emissions projections.
Estimate uncertainties using alternative base case and sensitivity analyses ' -
An alternative base case (ABC) is a simulation (a) having a set of inputs that differs
from the reference base case and (b) displaying performance that does not differ
statistically from that of the reference case. [See Reynolds, et al (1996) for a
discussion of alternative base case (ABC) analysis.] Three or four base cases,
consistent with the uncertainties in the modeling system, should be established and
used as a basis for control strategy evaluation. Sensitivity and control strategy runs
using each base case as a reference provide a basis for estimating a minimum
uncertainty accompanying ozone concentration estimates derived for the preferred
control strategy.
Finalize and implement control strategy
Based on the findings of control strategy, sensitivity, and ABC simulations, develop
estimates of the extent to which one or more selected air quality indicators, such as
peak ozone concentration or exposure, will be reduced. [A new, 8 hour standard, if
adopted, can be accomodated in this structure as well] If the estimates suggest that
attainment will be achieved, then adopt the control strategy for implementation. In
evaluating the merit of the strategy, take into account the fact that past estimates of
air quality improvement in a specified time frame have rarely been realized. If
analysis of the historical record supports it, an adjustment should be made to account
for a potential shortfall in improvement relative to projections.
If it appears that attainment may not be achieved, re-evaluate the merits of
alternative strategies considered earlier and/or increasing the stringency of control
measures in the strategy under consideration. Evaluate the likelihood of success of
this alternative.
Once a decision is made to accept a strategy, implement it. [As noted earlier,
consideration should be given in future work to the use of cost/benefit and decision
analysis in comparing alternative strategies.]
Project changes in emissions and ambient ozone and precursor concentrations with
time
Using emissions estimation and projection procedures in place and the PAQSM,
project changes in emissions and air quality for several years in the future. Create
projected trend plots for use as a reference, to be compared with data collected in the
m-5
-------
field and near or at sources.
Compare projections with monitoring data
Compare projected and observed air quality data, for ozone and precursors.
If unacceptably discrepant, determine the cause and adjust the control strategy
accordingly
Conduct diagnostic analyses to determine the causes of the discrepancies. Conduct
modeling and analysis with inputs modified in accordance with the results of the
diagnostic analyses. Project air quality as a function of time using the modified
modeling system. Re-evaluate the anticipated effectiveness of emissions controls.
Based on the findings, revise the control strategy.
Implement it.
If projections comparable with observations, continue to monitor
In this case, we are on track. No need for changes.
Continue process
Continue process until attainment is achieved. Thereafter, continue process to
maintain or further improve air quality in accordance with regulation.
B. The Primary Information Flows
The Flows
The initial effort centers on acquiring a sound and comprehensive data base,
conducting traditional photochemical modeling, introducing the use of ODMs,
establishing consistency in the results of the PAQSM and the ODM(s), estimating-the
uncertainties attending the calculations, and, based on all the information available,
recommending a control strategy.
Once a strategy is implemented, observational data are used to gauge progress and to
determine if the strategy implemented is likely to be successful. As soon as
information is forthcoming that suggests otherwise, diagnostic, modeling, and data
analysis are conducted to determine what modifications to strategy might be effective
in increasing the rate of air quality improvement to coincide with the target trend
line.
m-6
-------
The cycle of comparing observation with estimates, noting discrepancies, conducting
diagnostic investigation, repeating modeling and data analysis with modifications in
place, and adjusting the control strategy to conform with the revised results
constitutes a pattern that is followed until attainment is achieved and then thereafter
to assure that attainment is maintained.
The Outputs
The main outputs of the various analyses conducted include:
• PAQSM results,
• ODM results,
• estimated reductions in peak ozone concentrations for specified changes in VOC
and NOX emissions,
• estimated uncertainties in control strategy simulations based on the results of ABC
analyses,
• projections of emissions as a function of time,
• projections of air quality as a function of time (for ozone, VOC, and NOX), and
• revisions in all of the above, based on the results of diagnostic analyses and
subsequent modifications in cases where observations and estimates are
discrepant.
Comparisons and Corroboration
Decisions or actions are based on the comparison of:
• PAQSM and ODM results,
• projected and observed air quality as a function of time, and
• projected and observed emissions as a function of time.
Feedback Loops
Feedback loops are a distinguishing feature of the proposed approach. The primary
loops are those instituted by:
• an inability to resolve differences between PAQSM and ODM results, leading to
m-7
-------
acquisition of additional or more suitable data and modification of modeling and
analysis efforts,
• discrepancies between projected and observed air quality, and
• discrepancies between projected and observed emissions.
C. The Primary Elements: A Synopsis
The primary elements of the proposed approach to attainment demonstration
include:
• Using photochemical models to (a) indicate where and if VOC or NOX controls, or
both, are likely to be effective in reducing ozone concentrations; (b) assist in better
representing process dynamics, (c) determine where controls should be applied for
maximum effectiveness, and (d) estimate attainment requirements. However, they
would not be used explicitly for attainment demonstration.
• Introducing observation-driven methods for determining if VOC and/or NOX are
limiting. These methods, which do not depend on knowledge of emissions or
boundary conditions, may be used to determine the choice of pollutant to control
but not the magnitude of the control needed. In this approach, control
requirements (i.e., magnitudes) are estimated using a photochemical modeling
system, but actual determinations of concentration levels achieved will be made
over time through monitoring.
• Monitoring primary and secondary pollutants downwind, including ozone, nitric
oxide, nitrogen dioxide, nitric acid, PAN, and NOy, to determine current air
quality and to provide historical information for estimating trends. Progress
toward attainment is evaluated based on this information and that discussed
under the element following.
• Monitoring VOC and NOX in near source areas and in-stack to determine if annual
emissions reductions requirements prescribed by the CAAA are being met.
This four-element approach takes advantage of the strengths of photochemical air
quality simulation models while avoiding the limitations of most concern. It also
capitalizes on the attractive aspects of monitoring. Note that the approach does not
place undue dependence on the accuracy of models; rather, it relies on monitoring to
provide feedback for those determinations that are difficult to make using modeling.
HI-8
-------
In summary, the proposed approach stresses:
• demonstrating the comparability of the results of PAQSMs and observations-based
analyses before proceeding;
• seeking high quality performance of the PAQSM, as established through testing;
• checking of emissions and air quality projections in the field through monitoring;
• allowing for feedback and correction in a reasonable time frame (1-3 years, as
feasible); and
• assessing the uncertainties associated with the projection of future air quality
through ABC analysis.
The overall approach outlined has the following advantages:
• reduced dependency on modeling results — the accuracy of which cannot be
assured.
• provision for correction of unfavorable variances from planned outcomes.
• increased assurance of meeting prescribed targets.
Its primary disadvantages are that:
• it is "less prognostic" in conception, and
• it provides no immediate answer; rather, it offers a process that is "distributed
over time".
m-9
-------
-------
IV.
ELEMENTS OF THE PROPOSED APPROACH
In this chapter we examine each major element of the proposed procedure:
photochemical air quality simulation modeling, application of observation-driven
methods, expanded monitoring, and analysis of data.
A. Photochemical Air Quality Simulation Modeling
A PAQSM is one of three models that comprise a system of models designed to
simulate the formation of air pollutants and their transport, dispersion, chemical
transformation, and removal from the atmosphere. The other two are a
meteorological model and an emissions estimation capability (in recent times, also a
model). The meteorological model is used to develop a representation of the three
dimensional wind and temperature fields and the dispersion characteristics of the
atmosphere. This information is then used as input to the emissions model, as
emissions are influenced by temperature and winds. Both emissions and
meteorological representations developed using the appropriate models are then
supplied as input to the air quality model. The output of the air quality model is a
three dimensional representation of the concentrations of air pollutants of interest as
a function of time.
Basically, a PAQSM solves a series of equations that describe the conservation of
mass for each pollutant species, taking into account emissions, transport, chemistry,
dispersion, and deposition (and, in some, aerosol physics). The governing equations
are approximated to permit numerical solution on a computer. The numerical
integration method used is one that minimizes numerical error while attempting to
maintain reasonable efficiency (i.e., computational time). Solution is effected on a
three dimensional grid artificially imposed on the geographical area. A typical grid
spacing for an urban area might be 5 km in the horizontal (x and y directions) and
0.02-0.4 km in the vertical (z direction); the number of cells might be (typically) 30 x
20 x 8. For regional application, the horizontal dimension of individual grid cells will
typically be larger, and the number of cells usually will be greater.
To evaluate model performance, estimated pollutant concentration fields are
compared with observations. If performance is flawed, diagnostic analyses are
conducted to determine the cause of the problem, and the problem is corrected if
feasible. Performance evaluation is then repeated. Once a model is accepted for use,
it may be applied for a variety of circumstances. The most common involves
changing emissions in conformance with the anticipated outcomes of a control
strategy, maintaining all other inputs constant (with the possible exception of initial
and boundary conditions), and estimating the impact of the changes on the three
dimensional distribution of pollutant concentrations.
IV- 1
-------
Data required to support model evaluation and application are quite substantial.
Emissions information for all types of sources is needed, including spatial and
temporal distributions. Information should be provided at the spatial resolution of
the modeling grid (typically 5 km x 5 km) and a one hour temporal resolution. For
diagnostic wind models (i.e., analysis procedures based on interpolation of wind
data), the availability of observed wind velocities at a number of sites is essential.
For prognostic wind models (e.g., models based on the solution of fundamental
governing equations), few input data are needed in principle. However, when used
in the "data assimilation" mode, a common practice in attempting to achieve
accuracy, sounding data are required to bring estimates in line with observations at
regular time intervals. Pollutant concentration data are needed to establish initial and
boundary conditions for air quality models and to compare with model estimates.
In general, the following key categories of aerometric data are often sparse or lacking
(see also Section C):
• VOC concentrations,
• NOX concentrations (usually available, but often at too few sites, with too low a
sensitivity and with ill-defined measurements interferences or artifacts),
• air quality and wind data aloft,
• mixing heights (for models that require them, such as UAM-IV), and
• upwind boundary conditions at the surface and aloft (for the air quality and
meteorological models).
Reliance is usually placed on the availability and accuracy of routinely collected data.
Often, this data base is inadequate to support model evaluation and application.
When an area determines that it wishes to seriously pursue photochemical modeling,
it augments routine monitoring for a short period (3 to 8 weeks) during the "smog
season" with a suite of additional measurements selected to significantly enhance the
range and quantity of data available. Generally, data for one to three episodes of
interest are acquired through the monitoring effort, depending on the "favorability" of
the meteorology at the time. Note, however, that these episodes may not include:
• the highest ozone concentrations observed,
• one or more categories of meteorological conditions of interest, or
• particular attributes of interest, such as high boundary concentrations or high
concentrations of ozone aloft.
IV-2
-------
Thus, an adequate data base may not be acquired even when a special monitoring
study is conducted.
The most advanced and attractive PAQSMs available for use today include the two
latest versions of the Urban Airshed Model, UAM-TV and UAM-V, the CalTech
model and variants, and the San Joaquin Valley Air Quality Simulation Model
(SAQM, an advanced version, of RADM).
While we have provided this brief background for those unfamiliar with PAQSMs,
we recognize that most readers will be knowledgeable about the formulation, data
requirements, performance characteristics, and other attributes of this class of models.
Further, applications practice is well known and will not be discussed here.
Limitations of the PAQSM are summarized in Chapter H See Tesche et al (1992) for
a detailed critical commentary on the UAM-IV; many of the topics discussed in this
report pertain to other PAQSMs as well. See also Roth (1993).
In general, conducting a sound and well conceived photochemical modeling effort
having a high probability of simulating atmospheric dynamics with reasonable
fidelity requires giving attention to a broad list of needs, several of which may
require significant funding, staff time, and calendar time. Historically, the extent to
which these needs have been satisfied ^varies enormously - from the conduct of
extensive multi-year research programs involving the acquisition of a comprehensive
data base to several month efforts using what data are available. While
comprehensive efforts have a greater chance of success, decision makers are
sometimes confronted with budgets and/or schedules that are sufficiently tight that
they view their only option as being to forge ahead, respecting existing constraints.
The proposed procedure is intended to provide a general framework, one that will
apply for the full range of situations encountered.
In the ideal situation, with adequate funding, time, and staff, it is desirable to:
• acquire a comprehensive aerometric (air quality and meteorological) data base,
• undertake a careful and detailed construction of an emissions representation, with
corroboration of individual elements through at least two independent means of
estimation,
• conduct detailed prognostic meteorological modeling with four dimensional data
assimilation (FDDA), supported by adequate data acquisition,
• evaluate the estimation capability of the modeling system for a range of
meteorological episodes,
• conduct "hands off" performance evaluation (i.e., conducting simulations without
IV-3
-------
adjustment of inputs in an attempt to improve performance) for at least half of the
episodes,
• require high standards of performance for the full range of episodes studied
(representing a variety of meteorological regimes),
• evaluate the performance of individual models (meteorological and emissions) and
process modules (e.g., chemistry, boundary conditions, deposition) in an effort to
minimize the potential for compensating errors remaining undetected and to
correct errors as feasible,
• conduct ABC analyses to assess uncertainty, and
• use uncertainty information in decision analysis.
However, circumstances rarely permit this full range of efforts to be undertaken and
completed. The press of time, the availability of data, and/or the limitations of
budget inevitably restrict the extent of effort. To the extent that one must truncate or
omit one or more of the activities listed, the risk increases that modeling results may
not be reliable. The advantage of the proposed procedure is that it provides a means
for detecting flaws in the control strategy, as implemented, and correcting for them,
regardless of the accuracy of the modeling. In effect, a "safety net" is provided to
offset the risks associated with conducting an incomplete or limited modeling effort.
One might consider this aspect of the attainment demonstration procedure ''modeling
without guilt".
Consider the extremes. If a comprehensive modeling effort is conducted, the
probability is reduced of developing an inaccurate air quality management plan and
falling short of air quality targets. Perhaps a "one or two pass" emissions control
effort might be required, minimizing the uncertainties attending the amounts of
control needed and minimizing the costs associated with frequent revisitation of the
control issue on the part of industry. If a limited modeling effort is conducted
and/or if the supporting data base is limited or insufficient, the converse situation
might be expected. In this case the probability of having to revisit the specification of
an area's control strategy is increased in some relation to the shortfall in the modeling
effort. Consequently, the overall cost of control to industryand regulators is likely to
remain uncertain for a longer time and, in the long run, to be higher. In effect, then,
the decision to conduct or not conduct a data acquisition program and to undertake a
modeling program at a given level of effort is tantamount to weighing the trade off
between the level of investment that might be made in planning and the uncertainty
in control needs and, possibly, the overall costs of control and reassessment over
time.
IV-4
-------
Finally, a particular issue in the use of PAQSMs is the design and conduct of a study
such that proper attention is given to the emissions from sources in upwind areas
that have a significant influence on air quality in the region of interest. Specifically, it
is important to (a) characterize emissions of upwind sources (b) characterize their
transport, transformation, and deposition downwind and (c) estimate the effect on
downwind air quality of controls that might be placed on these sources. This need-
may be addressed in one of three ways:
• to model the entire region of interest, placing a grid of finer resolution on
metropolitan areas that are not in attainment,
• to incorporate knowledge of upwind conditions in the boundary condition for the
metropolitan area, omitting modeling of the upwind area, or
• to incorporate knowledge of upwind conditions in the boundary condition for the
metropolitan area, and to derive their conditions through modeling the upwind
area independently or properly coupled.
Data bases needed to support comprehensive regional modeling or modeling of the
upwind area alone have been inadequate. Further, focusing on boundary conditions
without analysis of the impacts of controls in upwind areas falls short of meeting
needs. How best to address the regional scale impacts of upwind sources is a topic
of considerable interest and a focus of the newly initiated NARSTO program. We
thus will not pursue this topic further here, except to say that consideration of
upwind sources can be accommodated in the proposed framework.
IV- 5
-------
B. Observation-Driven Methods
Definition and Capabilities
Observation-driven methods (ODMs) use measurements from one or more
monitoring locations to assess the relative sensitivities of ozone concentrations to •
reductions in VOC or NOX emissions. Four such methods are:
• the correlation between ozone and NOy or NOZ (e.g, Trainer et alv 1993; Jacob et
al., 1995),
• the Integrated Empirical Rate (IER) model of Johnson (1984), as revised by
Blanchard et al. (1994) and Chang et al. (1995). [Blanchard et al. (1994) name
their revised version the Smog Production (SP) algorithms],
* the use of indicator species and ratios (Sillman et al., 1990; Milford et al., 1994;
Sillman, 1995; Watkins et al, 1995; Jacob et al., 1995), and
• the observation-based model (OEM) of Cardelino and Chameides (1995).
The ODMs vary in their degree of empiricism.1 Correlation of ambient ozone and
NOy is the most empirical of the methods; however, key chemical processes are
responsible for the correlation and, thus, the method is linked to current
understanding of ozone formation. The IER and SP algorithms are primarily but not
exclusively empirical: they are sets of algebraic relationships that describe the
formation of ozone in environmental chambers. Blanchard et al. (1994) show that
the SP algorithms also describe the formation of ozone as calculated using a box
model with an initial charge of pollutants and two different chemical mechanisms.
The use of NOy concentrations (Milford et al., 1994) and indicator ratios (Sillman,
1995) was proposed based upon consideration of key chemical reactions; the values
suggested by Milford et al. (1994) and Sillman (1995) as threshold criteria for
distinguishing VOC from NOX control preferences were derived from one or more
applications of one or more grid-based PAQSMs. However, this method could have
a more empirical character as well, because the concepts are potentially verifiable
using field data. The OBM, as its name suggests, is a model that utilizes ambient
measurements rather than emissions estimates to drive a photochemical simulation.
Because the ODMs have in common the use of ambient measurements as inputs, we
group them together under the title "Observation-Driven Methods." The ODMs are
all diagnostic, rather than prognostic, tools. They are capable, in principle, of
1As will be described below, some ODM's are themselves dependent on some of the
same assumptions as photochemical models.
IV-6
-------
qualitatively estimating the sensitivity of ozone concentrations to VOC or NOX
reductions.
The ODMs share some other characteristics as well. Because they are all driven by
ambient measurements, the accuracy of their findings depends upon the precision
and accuracy of the ambient measurements (as well as upon the details of the
formulation of the methods). Moreover, conclusions are also dependent upon the
representativeness of the locations of the ambient monitoring sites, since they hold
only for monitored locations (see Section C for further discussion).
Because most of the ODMs have been developed only recently, and because no
comprehensive review of ODMs has been published, this section is somewhat more
detailed than other portions of the report. We summarize and evaluate each of the
four methods listed above according to the following categories:
• Description
• The key assumptions
• Nature and levels of uncertainty associated with the method
• The type of output of the method and the degree of consistency between this
output and the types of information needed for regulatory applications
• Availability of data needed as input to the method
• Availability of equipment, computational algorithms, or software needed for
implementing the method
• Potential value of the method as part of the proposed new attainment
demonstration procedure (including for use in comparisons among ODMs)
Following the discussion of each method, this section discusses similarities and
discrepancies among methods, where identifiable, and characterizes the potential for
developing confirmatory cross-comparisons through applications of the methods to
specific data bases. In Section V, recommendations are presented for the conditions
of use of each ODM and for further research and development of the ODMs.
Correlation Between Ozone and NOy or NOZ
Description. Unlike the other ODMs, this method was not originally proposed for
use as a discriminator between VOC and NOX sensitivity of ozone. Rather, the
method, which was applied primarily at sites where ozone concentrations were
thought to be limited by the availability of NOX, was intended to help elucidate
IV-7
-------
aspects of ozone formation at low concentrations of NOX (e.g., confirmation that ozone
levels appeared to be related to NOX concentrations at locations thought to be NOX
limited; estimation of rates of ozone formation with respect to NOX consumption).
However, the procedure appears to have value as a discriminator between VOC and
NOX sensitivity of ozone; moreover, it can be linked to the methods proposed by
Milford et al. (1994) and by Sillman (1995).
The method consists of computing correlations and regression relationships between
ozone and either NOX (NO + NO2), NOy (the sum of NOX and its reaction products),
or NOZ (NOy - NOX). For example, Trainer et al. (1993) used measurements obtained
from six rural sites in the eastern United States and Canada to show that ozone
concentrations in photochemically aged air increased from about 30 to 40 ppbv at
NOy concentrations below 1 ppbv to values of 70 to 100 ppbv at NOy concentrations
of abcat 10 ppbv. The data were selected to reflect photochemically aged air through
the choice of remote or rural monitoring locations (four of which were along the crest
of the Appalachians), restriction to the hours from 1:00 to 5:00 p.m., and restriction to
measurements where NOx/NOy was less than 0.40 (these two restrictions were not
applied to the four mountain-top sites). Ozone was found to correlate more closely
with NO,, the products of the oxidation of NOX (i.e., NOZ = NOy - NOX), than with
'z'
NOy Earlier studies have shown correlations between ozone and NOX (e.g.,
Fehsenfeld et al., 1983; Kelly et al., 1984; Parrish et al., 1986), NOZ (Spicer, 1977; Hanst
et al., 1982; Tuazon et al., 1980, 1981) or NOy (Fahey et al., 1986). In particular,
Fehsenfeld et al. (1983) showed that the rate of daytime increase in ozone
concentration at Niwot Ridge, Colorado, was proportional to the concentration of
NOX when NOX was in the range of 0.4 to 3 ppbv.
The key assumptions. Ozone production in the troposphere requires the photolysis
of NO2. In turn, NO2 is produced by reaction of NO with either ozone or peroxy
radicals. Ozone can accumulate only when sufficient peroxy radicals are present.
NOX is removed from the NO-to-NO2-to-NO cycle by reaction of NO2 with OH
radical to yield HNO3/ which, in turn, may form aerosol nitrate; another reaction
product of NOX is PAN, formed by reaction of NO2 with peroxyacetyl radicals. Since
radical reactions result in both ozone formation and conversion of NOX to its reaction
products, a correlation of ozone with NOZ should occur whether or not NOX is the
limiting precursor. However, because the number of ozone molecules produced per
NOX consumed decreases as NOX concentrations increase (e.g., Liu et al., 1987), the
relationship between ozone and NOZ appears to deviate from linearity when NOZ
reaches about 5 to 10 ppbv (Trainer et al., 1993). The use of the correlation between
ozone and NOZ as a discriminator between VOC and NOX limited conditions thus
requires that monitoring data exhibit statistically distinguishable slopes, AO3/ANO2,
under conditions of VOC and NOX limitation.
The correlation of ozone with NOy should be observed when NOx/NOy is low;
otherwise, concentrations of NOX in excess of a few ppbv might obscure the
IV-8
-------
correlation. Ozone levels have been shown to correlate with NOX concentrations
when NOX is less than a few ppbv (e.g., Fehsenfeld et al., 1983; Kelly et al., 1984;
Parrish et al., 1986), but at higher NOX concentrations the correlation can break down
since NOX need not be limiting the rate of ozone formation. Where ozone is
scavenged by NO, the correlations described above would be expected to break
down.
Nature and levels of uncertainty associated with the method. The existing
measurement studies do not provide firm guidance for delineating the range of
slopes, AO3/ANOZ, that would be indicative of NOX limitation. Trainer et al. (1993)
reported two sets of regression results bearing on ozone production efficiency, i.e., on
the ratio of ozone produced to NOX consumed. For the site at Scotia, Pennsylvania,
the data were well described by a line having an intercept of 35 ppbv ozone and a
slope of 8.5 ppbv ozone per ppbv NOX oxidized for NOZ concentrations up to 10
ppbv (Trainer et al., 1993). Data from the Los Angeles area, in which NOZ
concentrations ranged from about 20 to 85 ppbv, showed a slope of about 5 ppbv
ozone per ppbv NOZ (Trainer et al, 1993); because the NOZ concentrations excluded
aerosol nitrate, the true slope was thought to be lower (Trainer et al., 1993). These
two data sets represent extremes (i.e., high and low NOX concentrations) on a
continuum of precursor limitation, thus suggesting that application of the method to
areas in which precursor limitations are less clearly defined would require
discrimination between slopes somewhat exceeding 4:1 and those that were
somewhat less than 8:1. Jacob et al. (1995) used the slope, AO3/ANOZ, in conjunction
with two other indicators described by Sillman (1995) to argue that data from a site
in Shenandoah National Park exhibited a transition from NOX- to VOC-limited
conditions during September 1990. The slopes were 18:1 and 7:1. While no error
bars were given for the slopes, the correlation coefficients were modest (r2 values
were 0.49 and 0.23 for the two periods), thus suggesting that the slopes may not have
been statistically different. Moreover, the value of the slope occurring under
conditions that were asserted to be VOC-limited, 7:1, was almost certainly statistically
indistinguishable from the 8.5:1 value reported by Trainer et al. (1993) for a NOX-
limited situation. It is possible that additional guidance could be derived from
modeling studies. Sillman (1995) (see discussion below) presents ranges of O3/NOZ
for which modeling results showed sensitivity of ozone to reductions of VOC and
NOX.
While the existence of discernible correlations between ozone and NOX (or NO ) at
rural sites that are thought to be NOx-limited does help to confirm the expected NOX
limitation, little or no guidance is available regarding the opposite situation. That is,
while poor correlations could be indicative of VOC limitation, "poor" is not a well-
defined criterion. Presumably, firm indication of VOC limitation might also require
demonstration of a good correlation between ozone and either VOC concentration or
radiation. Alternatively, as for the case of the correlation between ozone and NOZ, it
is possible that additional guidance could be derived from modeling studies. Sillman
IV-9
-------
(1995) (see discussion below) presents ranges of O3/NOZ and O3/NOy for which
modeling results showed sensitivity of ozone to reductions of VOC and NOX
The regression relationships described by Trainer et al. (1993) use binned data, i.e.,
each data point shown is the average of ten percent of the data set. Time averages of
the measurements range from 5 minutes to one hour. The results for the unbinned.
data show more scatter and would thus be more indicative of the uncertainties
associated with application of the method to individual ozone episodes.
The type of output of the method and the degree of consistency between this output
and the types of information needed for regulatory applications. Output of the
method consists of regression relationships indicating linearity (r2 values) and
changes in ozone with respect to changes in NOX, NOy/ or NOZ. To be of most use
for regulatory applications, these outputs would need to be expressed in qualitative
terms, i.e., as indicating VOC sensitivity, NOX sensitivity, sensitivity to both
precursors, or as being indeterminate. Criteria for making these qualitative
distinctions would also need to be provided.
Availability of data needed as input to the method. Application of the method
requires measurements of ozone, NOX, and NOy. Both NOX and NOy would probably
need to be measured with accuracies of about one ppbv down to concentrations of
about one ppbv; however, this need has not been explicitly demonstrated. Virtually
no measurements of true NOX or NOy are available from routine monitoring
programs at present. However, the technology exists for measuring both NOX and
NOy at the stated concentration and levels of accuracy.
Availability of equipment, computational algorithms, or software needed for
implementing the method. Application of the method requires only monitoring
equipment. No specialized computational algorithms or software are needed, other
than commerically available database or statistical software.
Potential value of the method as part of the proposed new attainment demonstration
procedure (including for use in comparisons among ODMs). Use of the method
appears promising. Measurements of both NOX or NOy at the stated levels of
accuracy are technically feasible. Coupled with the indicator levels proposed by
Milford et al. (1994) and the ratios suggested by Sillman (1995), the method appears
capable of qualitatively identifying VOC or NOX sensitivity. Additional research and
development should be carried out, as described in Section V.
Tlie IER model and SP algorithms
Description. The IER model was derived from environmental chamber experiments
carried out at the Commonwealth Scientific and Industrial Research Organization
IV- 10
-------
(CSIRO) in Sydney, Australia Qohnson, 1984; Johnson and Quigley, 1989; Johnson et
al., 1990; Johnson and Azzi, 1992). Johnson (1984) defined smog produced (SP) as:
SP(t) = O3(t) - O3(0) + NO(0) - NO(t)
where all species are expressed in units of mixing ratios (not mass). [In situations in
which NOX (NO + NO2) enters a system over a period of time, NO(0) would be
replaced by NO(i), which denotes the concentration corresponding to the sum of NOX
inputs.] SP is a useful variable because: (1) the time derivative of SP is an indicator
of the rate of NO oxidation by peroxy radicals and a robust measure of the time
required to reach the NO-NO2 crossover and the NO2 and O3 maxima in
environmental chambers (Carter and Lurmann, 1991); (2) the environmental chamber
studies carried out at the CSIRO in Sydney indicate that SP displays a linear
relationship to cumulative light flux, provided sufficient NOX is present to sustain
ozone^ production; and (3) since ozone and NO react rapidly and reversibly, their
concentrations are not independent (the definition of SP accounts for this
dependence).
The original CSIRO environmental-chamber studies exhibited an empirical
relationship between the maximum potential SP (SPmax) and initial NOX concentration
Qohnson, 1984):
SPmax=P[NOx(0)]
where the parameter P was assumed to be constant with a value estimated from the
CSIRO experiments of 4.1 ± 0.4. P is an estimate of the maximum amount of SP
potentially produced (SPmax) per unit NOX input.
The extent of reaction (E) was defined by Johnson (1984) as:
E(t) = SP(t) / SPmax
When the extent of reaction reaches one in Johnson's model, smog production ceases
because virtually all of the NOX has reacted. Subsequent photochemical formation of
ozone is negligible because the system has consumed all of the NOX and, therefore,
can no longer produce ozone. Situations in which extent reaches one correspond to
cases in which peak ozone concentrations could be lowered by reducing NOX inputs
to the system. • Regions where the calculated extent of reaction is predominantly less
than one during the periods of high ozone concentrations are classified as VOC (or
light) limited, indicating that peak ozone concentrations could be lowered by
reducing VOC inputs to the system. When extent approaches but does not reach one
(e.g., E > 0.8), a system is usually in a transitional regime in which peak ozone might
be responsive to reductions of either VOCs or NOX.
IV- 11
-------
Chang and Suzio (1995) modified the IER model using the relationship
(OsLax = Y[NOX(0)]1/2
based upon the chamber data reported by Akimoto et al. (1979). Chang et al. (1995)
further modified the EER through substitution of a parameterization of deposition for
one of the parameters of Johnson's model.
Blanchard et al. (1994) re-defined SP as:
SP(t) = 03(t) + D03(t) - 03(0) + N0(0) - N0(t)
where they added the term DO3(t) to the definition of SP to account for the
cumulative ozone lost to deposition from time zero to t (i.e., ozone that has been
produced but deposited). Further, Blanchard et al. (1994) proposed the following
equation:
SPmax=p[NOx(0)]«
Data from environmental chamber experiments carried out at the University of North
Carolina (UNC), the Statewide Air Pollution Research Center (SAPRC) at the
University of California at Riverside, and the CSIRO were reviewed to identify those
experiments that used a mixture of VOCs and that went to completion. Values of
0.680 to 0.721 were obtained for a, depending upon the treatment of wall effects
(Blanchard et al., 1994). Simulations carried out using a photochemical box model
(OZIPR) (Gery and Grouse, 1992) with the Carbon Bond 4 Mechanism (CBM-4) (Gery
et al., 1989) yielded a=0.701; with the Carter-Atkinson-Lurmann-Lloyd (CALL)
mechanism (Lurmann et al., 1986),a=0.658 (Blanchard et al., 1994). Correlation
coefficients were about 0.9 for the chamber data and greater than 0.99 for the box
model, with no relationship to VOC concentration (Blanchard et al., 1994).
Blanchard et al. (1994) described further revisions of the model, which included the
incorporation of a deposition algorithm to account for deposition losses of NOy and
ozone, and provided equations appropriate for use in applications to ambient data.
A computationally tractable version of the equations was described, which used the
value a=2/3.
The key assumptions. The SP algorithms (Blanchard et al., 1994) are based on the
analysis of environmental-chamber data and simulations carried out using a box
model. These systems all have fixed volumes and an initial charge of NOX and
VOCs. Because the algorithms are not yet fully developed for systems subject to
dilution and continuous injection of emissions, the application of the algorithms to
ambient data is potentially subject to biases that have not yet been determined.
IV- 12
-------
Additional work, which is now in progress, should help clarify the potential
importance of these biases.
The value of the exponent proposed by Chang and Suzio (1995) (i.e., 1/2) differs
significantly from that proposed by Blanchard et al. (1994) (i.e., 2/3). Since the
former value (1/2) was derived from single-species (propylene) experiments, whereas
the latter (2/3) was based upon both mixture experiments and box-model simulations
using urban-mixture speciation, the value a=2/3 appears to be more appropriate for
application to ambient data. Although other details of the revisions proposed by
Chang et al. (1995) and Blanchard et al. (1994) vary, the two versions appear to have
converged in most respects.
Nature and levels of uncertainty associated with the method. The following sources
of uncertainty could affect the accuracy and replicability of the calculated extent of
reaction (E =
1) inaccuracy of the functional form (i.e., SP^ = f[NOx(0)]) that is used to derive the
expression for E, including incomplete specification of the arguments of the function;
2) failure of the equations to adequately treat physical processes that are insignificant
in smog chambers but that are of importance in the ambient atmosphere (e.g.,
dilution - which is potentially testable in a chamber with dilution);
3) inaccuracy (i.e., bias) of measurements (O* NO, NO2, NOy), parameters (a, p),
and estimated or modeled quantities (deposition of O3);
4) imprecision of measurements (O3, NO, NO2, NOy), parameters a, P), and
estimated or modeled quantities (deposition of NOy and O3);
The first three factors would introduce biases into the calculation of the extent of
reaction, whereas the last factor would affect the replicability, or precision, of the
calculated extent.
Although biases can be difficult to identify, the agreement between chamber
experiments and chemical mechanisms described by Blanchard et al. (1994) suggests
the absence of serious biases of the first type when applying the algorithms to fixed-
volume, initial-charge systems.
The principal uncertainty associated with the method has to do with its applicability
to ambient measurements, as described above. The potential for biases of the second
type requires further investigation.
Because the SP algorithms are algebraic in nature, they lend themselves to direct
analytical calculation of uncertainties deriving from the third and fourth items listed
IV- 13
-------
above. Differentiation of the function defining E with respect to each of its
arguments indicates the sensitivity of the calculated extent to variation in each
measurement or parameter. These sensitivities, denoted AE/AM, where M is a
measurement or parameter, are usually known as sensitivity coefficients. Sensitivity
coefficients may be used to evaluate potential biases in E that would result from
biases in the measurements or parameters and can also be used in determining sE2, •
the overall variance of the calculated extent.
At present, some but not all uncertainties have been estimated,2 (see Blanchard et al.,
1994). For example, a change of 10 ppbv in ozone concentration yields about a 0.01
to 0.02 change in extent. Measurements of NOX that are accurate to approximately 2
ppbv at NOX concentrations of about 2 to 5 ppbv permit the calculation of extent to
be accurate to within about 0.02 to 0.1 units. If NOy is accurate to within 10 percent,
biases in E can be limited to about 0.05 units. Changes of 10 ppbv in the estimated
background ozone concentration yield changes of about 0.01 to 0.02 when extent is
calculated using the NOX version of the algorithms; the changes are about 0.05 to 0.1
units for the NOy version of the algorithms.
The type of output of the method and the degree of consistency between this output
and the types of information needed for regulatory applications. The output of the
method consists of calculations of the extent of reaction. Extent can be related to the
sensitivity of ozone concentrations to precursor reductions, thus permitting a
qualitative determination of control preferences.
Availability of data needed as input to the method. Application of the method
requires measurements of ozone, NOX, or NOy. Both NOX and NOy should be
measured with accuracies of about one ppbv down to concentrations of about one
ppbv. Virtually no measurements of true NOX or NOy are available from routine
monitoring programs at present. However, the technology exists for measuring both
NOX and NOy at the stated concentrations and levels of accuracy. It is also possible
to utilize present routine measurement of "NOX", which are biased by the inclusion of
unquantified levels of interference from PAN and other species, in the algorithms in a
way that bounds the calculated extent of reaction (e.g., Blanchard et al., 1995).
Availability of equipment, computational algorithms, or software needed for
implementing the method. Measurements of ozone, NOX, or NOy at the stated
accuracy levels are technically feasible using existing measurement instruments. The
computational algorithms are explicitly delineated in Blanchard et al. (1994). A
software program is available that permits easy implementation of the method.
2It should be noted that these estimates all presume the applicability of the algorithms
to ambient measurements.
IV- 14
-------
Potential value of the method as part of the proposed new attainment demonstration
procedure (including for use in comparisons among ODMs). Use of the method
appears promising. Additional research and development should be carried out, as
described in Section V.
The Use of NOy and Indicator Ratios "''
Description. Milford et al. (1994) concluded that ozone concentrations would be
responsive to reductions of NOX where afternoon concentrations of NOy fall below a
threshold ranging from about 10 to 25 ppbv NOy The value of the threshold, which
was derived from analysis of modeling results, was given as a range to reflect the
ranges exhibited by the different modeling studies. Sillman (1995) proposed using
several indicator species or ratios to distinguish between NOX and VOC limited
situations: NOy, NOZ, O3/NOy/ O3/NOZ, O3/HNO3, HCHO/NOy, H2O2/HNO3,
H2O2/NOy, and H2O2/NOZ. Ranges of threshold values were given by Sillman (1995)
based upon variation of a number of model inputs"
The key assumptions. Milford et al. (1994) drew upon results from three PAQSMs,
which used two different chemical mechanisms, applied to four geographical regions.
The findings were based upon a range of models, modeling assumptions, and model
applications to help reduce the possibility that the results would be overly dependent
upon the accuracy of a single modeling application. Sillman (1995) used a single
model [which was denoted as 'UMICH' by Milford et al. (1994)], applied to two
regions with four cases per region (the cases were obtained byVarying factors such
as the base-case anthropogenic VOC emissions).
Nature and levels of uncertainty associated with the method. The uncertainty range
of the NOy threshold was given by Milford et al. (1994) as 10 to 25 ppbv NOy. This
range does not, however, encompass the full range of uncertainty that would have to
be ascribed to the method. No applications of the UAM were included. Two of the
models (ROM and UMICH) used relatively coarse grids (18.5 km horizontal and
three-layer vertical grid for ROM; 20 km horizontal and two-layer vertical grid for
UMICH), while the CIT model application to the Los Angeles basin used a finer grid
resolution (5 km horizontal and 5-layer vertical resolution). The sharpness of the
delineation of the NOy threshold varied among the models. The sharpest thresholds
were obtained from the UMICH applications. The ROM application exhibited more
scatter, while, for the CIT application, the sensitivities of ozone to its precursors
exhibited so much scatter that no well-defined threshold appeared. The range of the
threshold, 10 to 25 ppbv NOy, thus represents the range exhibited among the
applications, considering for each application the average over all locations within the
modeling domain. Were additional model applications considered, the range might
be larger. Consideration of individual locations would also expand the uncertainty
range, and would necessitate consideration of the large scatter shown by the CIT
model application.
IV- 15
-------
Although Sillman (1995) considered four base cases for each of two regions, the
results were nonetheless obtained through application of a single model. The
threshold values of the indicators varied by about a factor of two among the
alternative cases. According to Sillman's (1995) results, the transition into NOX
sensitivity was observed for values of (a) NOy less than 7 to 31 ppbv (i.e., the
threshold depended upon the simulation), (b) [O3- 40]/NOZ greater than 3.6 to 6.4,
-------
DOAS, and instruments described by Watkins et al. (1995) and Munger et al. (1995).
None of these methods may be considered tested and proven for routine application.
Availability of equipment, computational algorithms, or software needed for
implementing the method. Application of the method requires only monitoring
equipment. No specialized computational algorithms or software are needed.
Although instruments for measuring NOX and NOy are available and could be used
routinely, instruments for measuring HCHO, H2O2, and HNO3 appear to require
further development before they may be considered for use in routine monitoring
networks.
Potential value of the method as part of the proposed new attainment demonstration
procedure (including for use in comparisons among ODMs). Use of the method
appears promising in part, although the stated ranges in the threshold values are
large. While measurements of both NOX or NOy at the stated levels of accuracy are
technically feasible, measurements of HCHO, H2O2, and HNO3 may not be possible at
present in routine monitoring networks. Additional research and development
should be carried out, as described in Section V.
The Observation-Based Model (OEM)
Description. Cardelino and Chameides (1995) describe the use of a box model for
calculating the sensitivity of ozone to VOC or NOX reductions, which they name the
"Observation-Based Model" (OBM). The model is an ODM because it uses ambient
concentrations rather than emissions estimates to drive the calculations. The
calculation is carried out separately for each monitoring location. Unlike a trajectory
model, each box is fixed at the location of its monitor. The OBM utilizes some
features of the OZIPM4 model to account for dilution and employs a modification of
the CBM-4 (Cardelino and Chameides, 1995).
Cardelino and Chameides (1995) define a quantity, Ps03-No/ which is the net ozone
formed plus the net NO consumed over a 12-hour period (this quantity is similar to
SP of the IER model but is computed as an integral over time rather than an
instantaneous concentration). The fractional change in Ps03-No/ divided by the
fractional change in the "source strengths" of precursors, are used to define relative
incremental reactivities (RIRs). For each measured species, instantaneous source
strengths are calculated from the measurements and from production and loss terms
using a set of coupled nonlinear equations resembling the continuity equations but
which are expressed for a volume rather than an infinitesimal volume element. The
RIRs for each site are (or can be) averaged to generate area-averaged RIRs. Cardelino
and Chameides (1995) sum RIR terms so as to yield RIRs for NO, anthropogenic
hydrocarbons (AHC), and natural hydrocarbons (NHC). The split between RIR-AHC
and RIR-NHC is accomplished by summing RIRs for species arising from
anthropogenic and biogenic emissions, respectively.
IV- 17
-------
The key assumptions. Numerous assumptions are made in the procedure. Some of
the assumptions that appear to have a potentially substantial effect on the
calculations are listed here. As with any method, there is a concern that
assumptions could lead to biases. Magnitudes of biases cannot be identified without
carrying out specific analyses, nor can the likely direction of most biases be
ascertained a priori. However, the following factors appear to introduce a potential
bias toward overestimating the benefits of anthropogenic NOX reductions:
• The OEM calculates RDRs for ARC and NHC, but does not separate NOX into
anthropogenic and natural components. In many cities, the natural NOX source
might be small; however, in some situations, e.g. where anthropogenic fertilizer
use is high, soil emissions of NOX can be a significant fraction of the total NOX
source. The OBM is able to compute RIRs for AHCs and NHCs because some
chemical species can be identified as anthropogenic in origin, while others (e.g.,
isoprene) are known to be primarily of biogenic origin. However, no such
subdivision of NOX is possible.
• The RIRs are computed from the fractional changes in Ps03_No, divided by the
fractional changes in the "source strengths" of precursors. The RIR for NO may,
under some circumstances, not properly reflect the increases that can occur in
ozone concentrations subsequent to NOX reductions. For example, consider the
following simplified case (assume a fixed volume). Let the initial ozone and NO
concentrations be zero and 10 ppbv, respectively. After 12 hours, suppose the
ozone and NO concentrations are 90 and zero ppbv, respectively. Now compare
initial ozone and NO concentrations of zero and 5 ppbv, and suppose the final
concentrations are again 90 and zero ppbv (i.e., the ozone concentrations at the
end of 12 hours are identical for the two cases). The PS03-NO term is actually
computed by integrating over the 12-hour period. However, for purposes of
illustration, consider the 12-hour period as a single time step. Then, the
numerator of the RIR-NO would be (90 + 10) - (90 + 5) divided by 100, which is a
positive quantity (i.e., 0.05); hence, the RIR-NO will be positive. However, ozone
was actually unchanged. Thus, an apparent sensitivity is obtained that is entirely
due to the reduction in the amount of NO consumed. While this example is
highly artificial, it does appear to indicate the potential for a generic bias arising
from the definition of the RIRs in terms of the Ps03-No term- &1 more general
terms, since PS03-NO integrates the areas under both the ozone and NO curves, any
reduction in NO concentrations must translate into a positive contribution to the
RIR-NO, thus potentially overstating the sensitivity of ozone to NOX.
The effects of other assumptions and choices made in the development of the method
appear difficult to predict. Some factors to consider include the following points; any
consequent biases might range from inconsequential to substantial and cannot be
determined without further study.
IV- 18
-------
Depending upon the choice of chemical mechanism, the reactivities of natural
VOCs may be too high, thus affecting the computed RIRs for NHCs as well as the
apparent relative benefits of controlling VOCs versus NOX.
In application to data from Atlanta, the RIR-NO and RIR-AHC were sensitive to
the afternoon NO concentrations (Cardelino and Chameides, 1995); the crossover
point occurred at 0.75 ppbv, a concentration below the detection limit of most
currently-operating NO instruments. Thus, the value of the method may be
limited until NO measurements can be made routinely with an accuracy of about
0.1 ppbv down to about 0.1 ppbv.
Many assumptions appear to underlie the box modeling. The conditions under
which the assumptions hold or fail are not obvious, nor is it apparent which
conditions would occur for typical applications.
Uncertainties appear to arise due to temporal and spatial averaging in the
calculation of the source functions.
The horizontal transport of calculated species, which includes NO2 and secondary
paraffins, are treated as negligible (Cardelino and Chameides, 1995); the potential
uncertainties cannot be identified at present.
Only surface data are used to drive the calculations. Aloft concentrations are
assumed to be those of the free troposphere. The effect may or may not be to
limit the applicability of the procedure.
Cardelino and Chameides (1995) do not explicitly describe the treatment of
deposition, although the process may be incorporated into the loss terms.
Area-averaged RIRs would appear to lump together potentially dissimilar sites.
This effect, however, should be apparent through computation of the standard
error function described by Cardelino and Chameides (1995) and through
examination of the site-specific RIRs.
Nature and levels of uncertainty associated with the method. Analyses of uncertainty
have not been carried out for the OEM. Taking into account the uncertainties
deriving from factors such as those listed in the preceding section, as well as the
inaccuracies and unavailability of data for typical nonattainment cities, there is a clear
need to quantitatively characterize the resulting overall uncertainties in the model
output. At present, some of the assumptions appear to generate biases that would
enhance the apparent benefits of controlling anthropogenic NOX.
IV- 19
-------
The type of output of the method and the degree of consistency between this output
and the types of information needed for regulatory applications. The output consists
of RIRs for NO, AHC, and NHC. RIRs can also be output for anthropogenic area
and point sources, as well as for other disaggregations that would be of use. The
form of the output, as RIRs, can be directly translated into qualitative control
preferences. However, RIRs derived from 12-hour PS03-NO terms may hot be
appropriate given the present form of the ozone standard, which requires compliance
with a one-hour ozone average. Presumably, the time averaging could be modified
should this appear necessary, though Cardelino and Chameides (1995) do not
explicitly indicate that such modification is possible. Moreover, because the RIRs are
expressed in terms of P^NO terms, it is unclear that the sensitivities of ozone are
properly characterized, as described above.
Availability of data needed as input to the method. The method requires
measurements of NO that are accurate at sub-ppbv concentrations (Cardelino and
Chameides, 1995), which appears to exceed the capabilities of most instrumentation
that has been, or will be, deployed in routine monitoring networks. The OEM also
appears to require continuous gas chromatograph (GC) measurements of
hydrocarbon species (multi-hour averages appear to be inadequate given the need to
interpolate concentrations to relatively fine time resolution). Such data are largely
nonexistent at present, although the technology is available.
Availability of equipment, computational algorithms, or software needed for
implementing the method. As indicated above, equipment for measuring NO at sub-
ppbv concentrations with sub-ppbv accuracies generally has not been deployed,
although it is available. Presumably, the OEM consists of one or more computer
programs that could be made available to interested parties. However, the
availability and ease of use of these programs is unknown.
Potential value of the method as part of the proposed new attainment demonstration
procedure (including for use in comparisons among ODMs). The OEM is a much
more complex procedure than the other ODMs described here and involves a level of
effort that appears to be on a par with the use of emission-based dispersion models.
Several aspects of the OEM require further study and evaluation, as indicated above,
before the method may be considered available, useful, and accurate in the guidance
that it provides. The full range of assumptions underlying the OEM, and the
attendant uncertainties, cannot be characterized as readily as for other ODMs; both
assumptions and uncertainties should be clearly delineated. There is a need to
determine if the uncertainties associated with the OEM are greater or less, than those
characterizing emissions-driven models; if the uncertainties are as large as for
emissions-driven models, the level of effort involved in operating the OEM may not
be warranted. Finally, the necessary measurements may be largely unavailable, both
at present and in the near future, thus limiting the utility of the method even if it
IV-20
-------
otherwise appears attractive. Nonetheless, further research efforts are both needed
and recommended, as discussed in Section V.
Cross-comparisons Among the ODMs
Both the SP algorithms (Blanchard et al., 1994) and the use of NOy as an indicator
species appear to be broadly consistent with the ozone-NOy (or NOZ) correlations
described by Trainer et al. (1993). However, the comparisons described here are
limited and should be carried out in greater depth.
Trainer et al. (1993) selected measurements that indicated photochemically aged air
(in which ozone levels would be limited by the availability of NOX) . With the
exception of a single day at one site, the NOy and NO2 concentrations reported by
Trainer et al. (1993) were less than 14 and 10 ppbv, respectively. In comparison,
Milford et al. (1994) concluded that ozone concentrations would be responsive to
reductions of NOX where afternoon concentrations of NOy fall below a threshold
ranging from about 10 to 25 ppbv NOy. A more exact comparison of consistency
would require consideration of the sensitivities of the modeling results to the
calculated deposition rates and of the comparability of model-calculated grid-cell
average concentrations to ambient measurements (Milford et al., 1994).
As noted earlier, the data from Scotia were well described by a line having an
intercept of 35 ppbv ozone and a slope of 8.5 ppbv ozone per ppbv NOX oxidized for
NOZ concentrations up to 10 ppbv; data from the Los Angeles area, in which NOZ
concentrations ranged from about 20 to 85 ppbv, showed a slope of about 5 ppbv
ozone per ppbv NOZ (Trainer et al, 1993). For the indicator [O3- 40]/NOZ, Sillman
(1995) reported that values of 1.8 to 7.4 denoted VOC sensitivity, whereas values of
3.6 to 11 indicated NOX sensitivity (see earlier discussion). Thus, the Scotia data fall
within the NOx-sensitive regime according to the Sillman indicator; the Los Angeles
data are indeterminate.
The SP algorithms use the simple relationship SPmax = 0 [NOx(0)]a, with p = 19 and
a = 2/3. For a fixed-volume system, the relationship implies a ratio of ozone [where
O3 = SP - NO(0)] to NOy of 7.8:1 when NO(0) = 10 ppbv and NO2(0) = 0 ppbv,
excluding consideration of depositional losses of ozone and NOy. For systems in
which NOX concentrations have decayed to low (sub-ppbv) values, the ratio of ozone
to NOy would be approximately equal to that of ozone to NOZ. The corresponding
ozone-to-NOz ratio at NO(0) of 55 ppbv is 4.0:1. These ratios would be somewhat
higher if the deposition algorithm and the initial NO/NOX fractions used in the SP
algorithms were included in the calculation; as calculated, the ratios nonetheless
appear broadly consistent with the data presented by Trainer et al. (1993). On the
other hand, the SP algorithms use a nonlinear relationship even below 10 ppbv NO
whereas the data of Trainer et al. (1993) appear approximately linear below 10 ppbv
NOZ.
x/
IV-21
-------
No co-applications of the ODMs have been conducted (see recommendations in
Section V).
Summary
Four ODMs were reviewed. Each is under active development, and some are
available for application in preliminary forms. Further development and evaluation
of each of the ODMs is warranted (see Section V). The ODMs should be co-applied
to several study areas to permit assessment of the consistency of the guidance
provided (see Section V). Interim applications of the ODMs should be encouraged so
as to facilitate familiarity with their use, provide incentives for developing the
necessary monitoring information, and develop experience with their performance
capabilities. The developers of the ODMs should provide users with clear guidance
as to present capabilities and limitations so that the findings from interim
applications may be interpreted appropriately.
Finally, one other caveat should be mentioned. With the exception of correlation
analyses of the observation measurements, all other observations-driven methods can
be traced back to empirical chamber data. This derivation is direct in the case of the
IER method or indirect, through the use of chemical mechanisms, in the other two
approaches, which use smog chamber data in their development. Therefore, should
smog chambers have an undetected systematic bias which affects the quantitative
relationship between ozone and its precursors, these methods couls also be biased.
C. Analysis of Observations
Analysis of air quality data has predominantly considered studies associated with
ozone concentrations, and with the exception of the past few years, very little
consideration given to ozone precursor concentrations ( volatile organic compounds
(VOC) and oxides of nitrogen (NOx)). This lack of analysis activity by the scientific
and air quality communities was mainly the result of severe limitations in the
availability of high quality concentration data of the requisite precursor species and
to a lesser extent limited funds in support of this area. As a result of recent
criticisms regarding this data and analysis void (NRC,1991), the U.S. Environmental
Protection Agency has initiated an extensive monitoring effort to be implemented by
state agencies (U.S. EPA, 1993) within specified ozone Nonattainment areas. The
Photochemical Assessment Monitoring Stations (PAMS) are to measure ozone, NO,
NO2/ NOX, formaldehyde, acetaldehyde and acetone and fifty-five (55) hydrocarbon
species in and around the specified nonattainment areas.
With the onset of this data, come a variety of data analysis opportunities which
historically were not possible or were so data limited as to be impractical for credible
scientific inquiry. These new analysis opportunities, open new options for
IV -22
-------
consideration in the air quality management process as well as a rethinking of the
role monitoring and analysis can play in the SIP processes. Using the PAMS network
as a prototype, this section identifies some of the analysis tools and applications that
will be possible with a "new generation" of air quality data and discusses the role
such analyses can play in implementing the "Attainment Demonstration Process
(ADP)" proposed in this report.
In a later section on monitoring, a discussion is presented regarding specific
improvements in the PAMS program and its related operational infrastructure as well
as additional monitoring activities which must be considered to augment the ADP.
Basic Data Analysis Tools and Objectives
Historically most air quality data analysis studies have been confined to
characterizing ozone air quality and establishing its overall trend with time (U.S.
EPA, 1974; Chock et al,, 1982). Although these analyses have provided information
on ozone air quality, exposure assessments and basic trend information within the
monitored regions, they have shed little light on the reasons for the unexpected and
limited progress made in mitigating ozone pollution in the region. In recent years,
research measurement programs have collected data sets over limited time periods,
which have provided detailed characterizations of the chemical composition of
regional atmospheres. These data sets have resulted in the consideration of more
ambitious data analyses studies that present new opportunities for understanding the
photochemical processes associated with ozone formation and its quantitative
relationship with VOC and NOX precursor emissions. These analysis techniques
should become routinely possible with the availability of measurement data sets such
as those being collected as part of the PAMS, SOS and NARSTO-NE programs.
The purpose is not to provide a critical review of all possible data analysis techniques
available for air quality assessment, but to identify a select series of analysis
approaches to address specific objectives identified within the ADP. As new analysis
techniques are identified and accepted as better serving the analysis objectives
outlined in the ADP, they should be incorporated into the process.
A summary of ADP needs and supporting data analysis techniques is presented in
Table rV.C-1. Selected examples of the application of these techniques in addressing
air quality related objectives identified as part of the ADP will now be discussed.
IV-23
-------
Table IV.C-1. Summary of ADP Needs and Supporting Data Analysis
Techniques
Needs
Techniques
Chemical Species
^characterizing ozone exposures and
NAAQS attainment
Air Quality Characterization:
1) extreme value and exceedance
statistics;
2) meteorological data adjustment
a) ozone
b) ozone precursors
c) target indicator
species
«• assessing effectiveness of emission
control
tsr bracking air quality
«• tracking (reasonable further progress)
RFP
Ambient Trend Analysis:
1) time series of hourly, daily,
monthly, and annual statistics
2) means and extreme values
3) detrended data
a) ozone
b) ozone precursors
c) target indicator
species
rar assessing effectiveness of emission
control
«• tracking RFP
«• corroborating NOX and VOC emission
budgets
Emissions Trend Analysis:
1) time series of hourly, daily,
monthly, and annual statistics
2) means and extreme values
a) ozone precursors
b) target indicator
species
tr assessing effectiveness of emission control
«• corroborating NOX and VOC emission
budgets
Source Attribution:
1) principal component analysis
(PCA)
2) chemical mass balance (CBM)
3) source apportionment by factors
with explicit restrictions (SAFER)
a) ozone precursors
b) target indicator
species
«• corroborating of NOX and VOC emission
budgets
«• assessing effectiveness of emission control
«r tracking RFP
Correlation Analysis:
1) NMHC and NOx>y, CO and NOXiy,
NMHC and CO
a) ozone and its
precursors
b) target indicator
species
ADP Emission Related Analysis Components
The emissions related data analysis components identified within the ADP scheme
must be concerned with the tracking of emissions changes, corroborating inventories,
establishing the effectiveness of control programs, attributing source contributions to
observed precursor levels and providing feedback and accountability to the emission
control/projection process.
Example analyses include:
• Using ambient concentration ratios (i.e., NMHC/NOX , NMHC/NOy , CO/NOX,
CO/NOy , SO2/NOX, SO2/NOy , NMHC/CO) to determine if the relative
proportions of the precursor components of the emissions inventory are consistent
(Parrish et al., 1991; Buhr et al., 1992; Fujita et al., 1992);
IV-24
-------
• Using ambient concentration ratios from early to midmorning measurements of
NMHC/NOX and NMHC/NOy to provide insights with regard to expected ozone
responses to hydrocarbon vs NOX emission control strategies within monitored
regions (Wolff and Korsog, 1992);
• Using ambient concentration ratios from early to midmorning measurements of CO/NOX
, CO/NOy , and NMHC/NOX , NMHC/NOy to identify sources of error in
emission inventories (Fujita et al., 1992);
• Using ambient concentrations to detect trends in total NMHC, NOX, NOy/ CO,
NMHC/NOX, NMHC/NOy , CO/NOX/ CO/NOy and target indicator
hydrocarbon species for the purpose of tracking reasonable further progress (RFP)
towards ozone attainment; the choice of the most appropriate analysis parameters
will depend on the control strategies selected to meet the RFP for the monitored
region, (e.g. if Reformulated Gasoline (RFG) were part of a region's RFP, ambient
trends in NMHC, CO, NMHC/NOX , NMHC/NOy , CO/NOX, CO/NOy , benzene
and benzene/NMHC would all prove to be relevant analyses for tracking the
progress of emission reductions and the effectiveness of the RFG control strategy);
« Using ambient measurements of speciated hydrocarbons and ratios of indicator
species to reconcile source contributions to ambient concentrations (Stephens and
Burleson, 1967,1969; Altsuller et al.,1971; Mayrsohn et al.,1976; Lonneman et
al.,1968; Pilar and Graydon,1973; loffe et al.,1979; Whitby and Alrwicker,4978; and
McKeen et al.,1990; Jobson et al., 1994) and to derive specific source contributions
source apportionment techniques (Henry et al., 1994)
ADP Air Quality Trends and Control Strategy Evaluation Analysis Components
Air quality monitoring to support feedback and accountability in the air quality
management approach is a central feature of the ADP. Data analysis procedures to
track and verify air quality responses to implemented control strategies and related
emissions projections are used in the ADP: to assess the effectiveness of control
programs; to provide the opportunity to evaluate the progress of the control program
and consider mid-course corrections; to establish that the ozone response function to
measured ambient precursor concentration changes is consistent with that predicted
by the PAQSM. Example analyses include:
• Using linear regression, analysis of variance, and nonparametric statistical
analyses of the means of early to midmorning measurements of NMHC, NOX/
NOy/ CO and target indicator hydrocarbon species concentrations within urban
monitoring sites to evaluate trends in precursor air quality; using commensurate
analyses of maximum O3 concentrations at the respective downwind sites evaluate
trends in ozone air quality (Chock et al., 1982, Kumar and Chock, 1984; Rao et al.,
IV-25
-------
1992);
Using statistical filtering methodologies remove the influence of meteorological
fluctuations on the concentrations of O3, NMHC, NOx, NOy, CO, and target
indicator hydrocarbon species so as to detect changes in ozone air quality due to
changes in precursor emissions and/or selective control strategies (Korsog and
Wolff, 1991; Rao and Zurbenko, 1994);
apply extreme value theory - chi-square distribution and Poisson process to
evaluate and track trends in ozone summary statistics based on daily maximum
ozone concentrations and the attainment of the NAAQS (Roberts, 1979; Smith,
1989; Shively, 1991)
apply correlation analyses of the concentrations of O3 and NOy; O3 and NOy -NOX
; O3 and CO; O3 and NMHC to discriminate if region is NOX or VOC limited with
regard to ozone formation (Trainer et al, 1993; Parrish et al.,1993; Chin et al.,
1994; Olszyna el al., 1994) These methods are discussed in further detail in Section
IV.B.
D. Monitoring
Siting and Analysis Objectives
The successful application of measurement data to support analysis approaches,
whether they be related to the evaluation of theoretical models or the development
and application of diagnostic observational data analysis tools, are contingent upon
the measurement/monitoring stations meeting specified data quality objectives and
siting attributes. The PAMS program has identified four distinct site types,
summarized in Table IV.C-2 and identified their intended monitoring objectives for
each site type (U.S.EPA, 1991).
IV-26
-------
Table IV.C-2. PAMS Siting Criteria i
Type (1)
Type (2)
Type (3)
Type (4)
Attributes
located in the pred
morning upwind d
the local area of m;
precursor emission
located immediate]
maximum precursc
located 10 to 30 mi
from the edge of tr
located in the pred
afternoon downwii
from the local area
precursor emission:
The scope of the PAMS network, alt!
contributions to ADP.
md Contributions to ADP
ominant
irection from
iximum
5
y downwind of
r emissions
es downwind
e urban area
Dminant
d direction
of maximum
Intended Contributions to ADP *
specifying pollutant inflow and trends;
specifying model boundary conditions,
present and future.
corroborating NOX and VOC emissions;
tracking emissions and progress in
implementation strategy; evaluating pollutant
exposures, models, and NAAQS attainment.
evaluating pollutant exposures, models, and
NAAQS attainment; tracking emissions and
progress in implementation strategy
evaluating ozone control strategies;
specifying pollutant outflow and trends;
specifying model boundary conditions,
present and future.
lough ambitious, is inadequate for meeting all the intended
In the discussion which follow,
the PAMS network is taken as a starting point from
which modifications or additions will be considered to meet the goals and objectives
of the proposed ADP. The esta
under PAMS is critical to the si
be applied for tracking, feedbai
note that the network has significant limitations with regard to its coverage and
spatial representativeness. In addition, there are currently no operational examples of
State PAMS networks which considered all four site types nor are there demonstrated
data analyses protocols to assess the representativeness of the PAMS sites which are
operational. To some extent it has tried to serve too many masters and as a result has
over stated its intended object!
inadequate and potentially mis
is its ability to specifying trend
lishment of the four designated site types identified
iccessful implementation of key analysis procedures to
k and verification of the ADP, but it is important to
es. For example, a single upwind monitor is
eading for specifying model boundary conditions, as
in pollutant inflow fluxes.
IV- 27
-------
Measurement Parameters, Methods and Data Quality Objectives
Chemical parameters measured in PAMS sites include: NO, NO2/ NOX/ O3
formaldehyde, acetaldehyde, acetone and the fifty-five non-methane hydrocarbon
species listed in Table IV.C- 3.
Table IV.C-3. PAMS Speciated Non-Methane Hydrocarbons
Acetylene
Ethylene
Ethane
Propylene
Propane
Isobutane
1-Butene
n-Butane
trans-2-Butene
cis-2-Butene
3-Methyl-l-Butene
Isopentane
1-Pentene
n-Pentane
Isoprene
trans-2-Pentene
2-Methyl-2-Butene
2,2-Dimethylbutane
Cydopentane
2,3-Dirriethylbutane
2-Methylpentane
3-Methylpentane
2-Methyl-l-Pentane
n-Hexane
trans-2-Hexene
cis-2-Hexene
Methylcyclopentane
2,4-Dimethyl pentane
Benzene
Cyclohexane
2-Methylhexane
2,3-Dimethyl pentane
3-Methylhexane
2,3-Dimethylpentane
3-Methylhexane
2,2,4-Trimethylpentane
n-Heptane
Methylcyclohexane
2,3,4-Trimethylpentane
Toluene
2-Methylheptane
3-Methylheptane
n-Octane
Ethylbenzene
p-Xylene
Styrene
o-Xylene
n-Nonane
Isopropylbenzene
n-Propylbenzene
a-Pinene *
1,3,5,-Trimethylbenzene
1,2,4-Trimethylbenzene
p-Pinene *
Total Non-Methane HC
* Recent testing and evaluation of the automated Gas Chromatographic hydrocarbon analysis metnoa
deployed in PAMS have reported that these species cannot be measured quantitatively and have been
removed from the required species list.
IV-28
-------
Meteorological parameters to be measured at a height of 10 meters from the surface
include: wind speed and direction, temperature, relative humidity, barometric
pressure and solar radiation. The PAMS network must also have available regional
representative upper air meteorological measurement data including: winds,
temperature, dew point, and pressure which can be used to help characterize the
synoptic flow and mixed layer dynamics of the region.
Measurement methods and techniques for PAMS are outlined in EPA reference
methods and technical assistance documents (U.S. EPA, 1993; U.S. EPA,1991), those
associated with criteria pollutants (i.e., O3, NO, NO2, and NOX) are reasonably well
established, though, as we will discuss later, issues still remain regarding detection
limits and accuracy. The measurement methods required for speciated hydrocarbons
and carbonyls in PAMS are not well established and quality assurance
procedures/standards for these methods are still under development.
A scoping study has been initiated in the northeast portion of the United States in
1995 as part of the North American Research Strategy for Tropospheric Ozone
(NARSTO). This multi year program is designed to address a series of policy relevant
scientific issues associated with oxidant control and ozone nonattaiment in the
northeast region. This summer a research measurement network was established, that
is designed to augment PAMS monitoring through characterization of chemical
backgrounds in the northeast region. In addition, several PAMS sites in the region
were augmented with NOy instrumentation. Chemical measurements and methods
deployed at these research measurement sites are summarized in Table rV.C-4. The
table also provides performance criteria and operating schedules for these
measurements and is indicative of what should routinely be possible in air quality
monitoring networks.
In addition, a standard complement of meteorological measurements (wind speed
and direction, relative humidity, temperture, solar radiation and pressure) will be
collected at site each from 10 meter towers. The program also has supplemented
upper air meteorological measurements, through a network of accoustical and radar
sites in the region.
Other chemical measurements of interest in support of model evaluations and
diagnostic testing of chemical process relevant to the ozone precursor relationships
include: radical termination products (e.g. HNO3, H2O2/ ROOR, and RONO2); VOC
oxidation products (e.g. PAN, MEK and MA); and oxygenated VOC sources (CH3OH,
C2HSOH, and MTBE)
Aircraft platforms are capable of providing equivalent sets of measurements as
identified in Table P/.C-4, but quality measurements from these platforms requires
significantly more effort as a result of additional factors affecting instrumentation
performance (e.g. air sampling inlet designs, pressure and temperature variations, in-
IV-29
-------
flight quality assurance procedures)
TABLE IV.C-4. Typical Chemical Measurements Performed at Regional
Enhanced Monitoring Sites in NARSTO-NE, 1995
Measurement/ Method
NO/NO2/NOX
Oiemiluininescence
NOy- Mo/Au
Conversion1
Qiemiluminescence
03
UV absorption
CO1
Gas Filter Correl. Enh.
Speciated Hydrocarbons
GC-FID2; dual column auto
GC3
Carbonyl Compounds
DNPH Silica Gel & C-18
Cartridges4 - HPLC
SCV
Pulsed Fluorescence
Range/MDL
0-50 ppb
100 ppt
0-50 ppb
200 ppt
0-200 ppb
2 ppb
0 - 1000 ppb
20 ppb
0 - 500 ppbC
0.2 ppbC
0 - 10 ppbC
0.5 ppb
0-10 ppb
100 ppt
Averaging Times
1 minute/hourly
1 minute/hourly
1 minute/hourly
5 minute/hourly
grab or integrate (l-6hr)
cans, 4 to 6 samples/day; in
line cryotrap continuous ~ 1
hr
integrated (1-6 hr)
cartridges, 4 to 6
samples/day
1 minute/hourly
Operation Schedule •
Photochemical Season
or
Year Around
Photochemical Season
or
Year Around
Photochemical Season
or
Year Around
Photochemical Season
or
Year Around
Photochemical Season
or
Year Around
(reduced sampling
schedule)
Photochemical Season
or
Year Around
(reduced sampling
schedule)
Photochemical Season
or
Year Around
'Chemical species currently not measured in PAMS network
2 Hydrocarbon speciation C2 - C10
3Automated dual column Gas Chromatographs are currently under evaluation
4DNPH cartridge sampling systems are currently under evaluation
The depolyment of instrumented aircraft to address the very important
characterization of the vertical dimension of the air quality is very much dependent
of the range, performance, and payloads of the aircraft. Even the most ambitious of
aircraft measurement programs will characterizes only a few episodes well enough to
be useful for selective model evaluation studies and provide an indication as to the
importance boundary conditions may have on a designated modeling domain. There
is critical need for routine chemical concentration data in the vertical, but the costs
associated with aircraft measurement programs to provide the data are prohibitive.
IV-30
-------
Therefore the development of innovative techniques, ( remote sensing, instrumented
tether ballons, instrumented unmanned aircraft) must be considered to support the
future adavacement of the field.
Ambient Monitoring in Support of PAQSM
Ideally, the information needed to support the operation and evaluation of the entire
photochemical air quality simulation modeling system, including all modules and
preprocessor models requires a data base which completely characterizes the spatial
and temporal features of requisite input variables and the final estimated
concentration fields. The meteorological and air quality models provide estimates
that represent spatial averages over volumes the size of a grid cell, which may have
dimensions of several kilometers in the horizontal and several tens to a few hundred
meters in the vertical. But in reality, current measurement technology typically
provides observations at a "point" (i.e., the sampling volume is much smaller than
that of a grid cell). Although the time averaging of fixed surface observations may
be comparable to that of the model estimates (i.e., one-hour averages), measurements
from instrumented aircraft typically represent averages over an interval of less than
one minute. Thus, both the spatial and temporal scales of aircraft data are much
smaller than those of the model estimates. This disparity of scales problem pervades
all current observations. Using current measurement technologies, it is important to
collect information in a way that allows the "point" observations to provide
information that is as representative as possible of values averaged over larger spatial
and time scales. Therefore, site selection criteria are a critical part of assuring data
representativeness of the measurement site.
Ambient monitoring data in support of model evaluation activities may include: area-
specific emissions measurements; combined chamber and air parcel tracking studies
to assess the chemical mechanism; inert tracer releases and associated downwind
monitoring to test treatment of pollutant transport; dry deposition experiments; and
grid-averaged, time varying information on primary and secondary pollutant
concentrations along upwind and downwind boundaries; and hourly- and grid-
averaged concentrations of O3, NOX, NO2/ VOCs (speciated), PAN, and other
secondary reaction products collected over all significant portions, of the modeling
region.
Initial.pollutant concentrations for use in the air quality models are derived using
interpolation techniques and available air quality data collected at the surface and
aloft. Surface monitoring sites situated in the immediate vicinity of large sources
may not be useful in establishing the subregional estimates of initial concentrations
needed by the model. If significant amounts of pollutants reside aloft at the start of a
simulation, their concentrations will be difficult to characterize in the absence of
aircraft or other suitable measurements. Speciated VOC data are needed to specify
the concentrations of organic species required by the kinetic mechanism and must
IV-31
-------
usually be collected as part of a supplemental field study. Alternatively, values must
be derived based on an analysis of data coUected in previous studies in other urban
settings. In general, an attempt is made to minimize problems associated with
specifying initial concentrations through the performance of multi-day simulations.
Boundary concentrations are derived using available local data at the sectors upwind
of the modeling domain. These routine measurements of pollutant levels upwind of
an urban area must characterize horizontal spatial distributions of the incoming
concentration fields. The specifications of pollutant levels aloft are much more
difficult to characterize. Instrumented aircraft flying along the upwind boundaries
can be used to help establish boundary concentrations, but is important to recognize
their limitation as well, since at best these data rarely provided more than a few
periods during a day. The entrainment of pollutants from aloft, representative of the
previous day's mixed layer, may also be estimated from upwind surface monitors/if
they are not influenced by local source emissions. For example by reviewing the
time rate of change in the diurnal profiles of species concentrations of O3, NOx and
CO during the morning periods, reasonable assumptions can be made of
concentrations aloft.
As with the lateral boundary concentrations, measurement data for the top boundary
of the modeling domain, are very limited. Though the exchange of background
concentrations of ozone and its precursors in the free troposphere (or stratosphere)
into the modeling domain can occur through significant larger scale connective
activity, it is generally of secondary importance in photochemical ozone episodes.
Data for this top boundary can be collected via aircraft or from elevated mountain-
top sites in the region.
Routine Monitoring Networks
Federally mandated air quality monitor networks [State or Local Air Monitoring
Stations (SLAMS) and National Air Monitoring Stations (NAMS) ] were not designed
to support model evaluation studies, but to measure criteria pollutants in terms of
characterizing 1) maximum concentrations, 2) population exposure, 3) source impacts,
4) background concentrations and 5) in the case of NAMS to track air quality. These
monitoring sites are typically located in an urban, multi-source impacted areas with
high population density. Their greatest value in model performance evaluations is in
their ability to characterize the spatial distribution of ozone and its maximum
concentration within an urban area.
These surface air quality measurements will generally be impacted by local sources,
thus limiting their spatial representativeness for use in model performance
evaluations. Also the networks have provided only limited data on VOC and NOX
concentrations, which are essential for model performance evaluations, moreover,
sensitivity and specificity of the NO/NOX instruments used in these networks is
IV- 32
-------
insufficient to accurately determine low ambient concentration levels needed to
effectively test the O3-VOC-NOX relationship.
As mentioned above, PAMS, the newest federally mandated measurements network,
is intended to provide information to assist in control strategy development and
evaluation, emission tracking, and trend analysis, and exposure. The ozone precursor
measurements to be performed will be a vast improvement in the air quality data
base and should significantly benefit the model performance evaluation processes. It
is unlikely that the spatial coverage provided by PAMS will be sufficient to satisfy
current requirements specified for model performance evaluations (Roth et al., 1992).
Criteria for network design, including station density, areal coverage, chemical and
physical parameters, and precision and accuracy of the measurements in support of
the model performance evaluation process, have not been established. The choice of
chemical and physical parameters and the required precision and accuracy of their
measurement have been driven by what was available and not by the rational
development of data quality objectives designed to meet specific model performance
evaluation standards. This result is not surprising, since the establishment of
unambiguous model performance standards, remains as elusive now as it did ten years
ago. In a similar manner sample frequency and averaging time are generally driven
by instrumentation capabilities or cost constraints. Current generation O3, NO/NOX ,
SO2 and CO instrumentation can deliver virtually continuous data and should be
archived for averaging times of one minute, five minute and one hour. Speciated
VOC and carbonyl compounds currently being measured under the PAMS program
have minimum sampling times of one hour and three hours respectively.
We can clearly state that the attributes of the four types of sites identified under the
PAMS program must be represented in any model domain being considered for
evaluation. What is less clear is the number of each type of site required to provide
acceptable spatial coverage of the modeling domain for the performance evaluation.
The Role of Intensive Field Measurement and Special Study Programs
In addition to routine monitoring data, intensive field study programs have been'
carried out to support regulatory applications of photochemical models. Programs
such as the Southern Oxidant Study (SOS; Chameides and Cowling, 1995), Lake
Michigan Ozone Study (LMOS; ENSR, 1991), Southern California Air Quality Study
(SCAQS; Blumenthal, Watson, and Roberts, 1987), South Central Coast Cooperative
Aerometric Monitoring Program (SCCCAMP; Dabberdt and Viezee, 1987), and
SJVAQS/AUSPEX (Blumenthal and Watson, 1991) have been carried out in part to
provide data needed to support the application and evaluation of photochemical
models in these study areas because existing routine monitoring activities were
inadequate.
IV-33
-------
The design for an intensive monitoring program begins with consideration of the
overall modeling program objectives. Considering the existing routine monitoring
network, various measurements are added to fulfill one or more modeling needs.
The goal is to fully characterize important transport and chemical transformation
phenomena in the study area of concern. There are three major categories of data
which must be enhanced; meteorological, emissions inventory and air quality. The
discussion here is limited to air quality data and the measurements/monitoring
required.
Enhancements in surface air quality monitoring undertaken as part of these field
intensives and typically operated continuously throughout one or more
photochemical seasons include:
• supplementing existing networks with instrumentation to provide a complete
complement of ambient air quality measurements, including (NO, NOX, O3, VOC,
PAN, etc.);
• establishing new monitoring sites or even networks to provide regional coverage
or a better characterization of the upwind and downwind areas;
• testing and evaluation of new advanced instrumentation capable of monitoring
low level concentrations of species important to assessing the quantitative
relationship between ozone and its precursors;
• establishing enhanced high sensitivity monitoring sites in rural areas, near the
boundaries of the model domain to provide both boundary condition data for
model inputs and also data to evaluate model performance, and perform
diagnostic and corroborative analyses.
Intensive measurement programs, as part of these field intensives, are carried out on
an intermittent basis, typically during episode periods. These programs typically
measure air quality aloft through the use of aircraft and operate research
instrumentation at selected surface monitoring site(s) to measure unique chemical
species of importance to understanding photochemical oxidation processes and the
ozone-VOC- NOX relationship. Aircraft data provide initial and boundary
concentration inputs for model applications and a three-dimensional characterization
of air quality for use in model performance evaluation. Unfortunately the temporal
and spatial limitations of aircraft data are significant. As a result, a very limited
number of realizations are available to test/evaluate the model which leaves model
performance testing in this region highly uncertain. Surface and airborne lidar have
become increasingly useful for characterizing air quality conditions aloft and may
help in establishing inputs as well as in evaluating model performance. One or more
inert tracers may be released from key source areas to document actual transport
phenomena; associated air sampling may be conducted at downwind surface
locations and by instrumented aircraft.
IV-34
-------
The quality of a model performance evaluation is directly related to the availability of
suitable meteorological, air quality, and emissions data. Routine measurement
programs do not adequately characterize the spatial and temporal variability of key
meteorological and air quality variables and thus cannot be relied upon to support
the model evaluation effort. Thoughtfully designed intensive field measurement
programs can provide the needed supplemental information. At a minimum, such'
programs should include additional meteorological and air quality monitoring (both
at the surface and aloft), collection of day-specific emissions data for large industrial
sources, VOC speciation measurements, and upwind monitoring of O3 and precursors
at the surface and aloft.
Ambient Monitoring in Support of Observation-Driven Methods
The PAMS network should provide an excellent foundation for the support of
observation- driven methods and related data analysis and diagnostic techniques.
Several operational issues regarding PAMS and its supporting infrastructure remain
unresolved and present some cause for concern. The deployment of PAMS is
significantly behind schedule, though many sites are identified as operational,
delivery of quality assured data is not forthcoming. This seems to be a systemic
problem that can be traced to 1) a significant underestimation of the manpower
resources and technical expertise required to maintain and operate the PAMS sites; 2)
the inability of state agencies to commit sufficient manpower resources to the PAMS
operation; 3) premature deployment of unproven instrumentation/measurement
systems for routine operation; 4) limited infrastructures to support routine data
assessment, analysis and distribution of the PAMS data to facilitate quality assurance
and scientific interactions.
A fully deployed PAMS as prescribed in (U.S. EPA, 1993) must be maintained and
operated for a minimum of ten years to meet the most important policy relevant and
operational analysis objectives it was designed to address. The significant
commitment required to maintain a high quality monitoring network like PAMS is a
new and unprecedented challenge to the state agencies involved and suggests the
need for external technical assistance in the deployment and operation of these sites
as well as scienific assistence in the assessment and analysis of the data collected. The'
scientific community recognizes the potential value of the PAMS and has
recommended augmentation of the network through the addition of complementary
enhanced regional measurement sites and suggested instrumentation upgrades and
measurement parameter additions to the existing PAMS networks (NARSTO, 1994).
The community also recognizes the scientific and technical assistance needs of the
state agencies and is in the process of building partnerships, mainly through
university and federal research groups, to facilitate PAMS related operation and
analysis activities with state agencies.
-------
The measurements being performed under PAMS and the enhanced PAMS sites
under the NARSTO program, should very likely meet the needs of the most
ambitious observation-driven methods. The routine delivery of high quality assured
data from these networks is currently not a reality, but should be possible with the
proper infrastructure development and support.
Specific measurement requirements for ODMs were addressed in Section IV.B for
each of the four methods reviewed. These requirements are summarized in Table
IV.D-1, along with suggested siting and spatial representation requirements. It
should be noted that the measurement requirements for emission-based models, even
at the most rudimentary level of model application and performance evaluation, are
far more demanding than ODM requirements. Although the full extent of the
measurements requirements for the ODMs are not established at this time, since
these methods are still under development, it is not likely that they will approach the
needs of the PAQSM.
Finally, there are several measurement/mstrumentation issues associated with critical
parameters identified in the measurement requirements for both the ODMs and
PAQSM as well as in the analysis of observations which will be addressed in Section
V.
Ambient Monitoring in Support of Analysis of Observations
There are many similarities in the data needs in support of the analysis of
observations approaches outlined in Section IV.C and those identified with regard to
the ODMs. What is a somewhat unique objective identified in the analysis of
observations approaches, but not emphasized in the other techniques are the emission
analysis components which require near source and in-source measurements.
Type(2) monitors identified under the PAMS network, can provide useful data for
addressing these near source measurement needs (see Table IV.C-1 and Table TV.C-ty.
But intensive measurement programs designed to specifically target source categories
will likely be needed to augment these sites.
In-source measurements have been mainly associated with special studies and
research programs, and with the exception of the recent regulatory requirement
(Clean Air Act, 1990) to provide continuous NOX measurements for point sources
emitting more than 100 tons per year, the direct measurement of source emissions has
not been considered in the monitoring programs.
IV-36
-------
Table IV.D-1 Measurement Requirements in Support of ODMs
Observation-Driven Method
(ODM)
Correlation; O3-NOy or O3-NOZ
Correlation; O3-CO, O3-NMHC,
C0-N0x/C0-N0y/
Integrated Empirical Rate (IER)
Indicator Species and Ratios
Observation-based Model (OBM)
Parameters
03
NO-NO2/ NOy
CO
NMHC
(speciated)
03
NO-NO2/ NOy
03
NO-NO2, N0y
HCHO
H202
HN03
03
NO-NO2, NOy
NMHC
(speciated)
Range/MDL1
0 - 200 ppb
2ppb
0 - 200 ppb
500 ppt
0 - 1000 ppb
20 ppb
0 - 1000 ppbC
500 pptC
0-200 ppb
2 ppb
0 - 200 ppb
500 ppt
0 - 200 ppb
2 ppb
0 - 200 ppb
100 ppt
0 - 10 ppb
200 ppt
0-200 ppb
2 ppb
0 - 200 ppb
100 ppt
0 - 1000 ppbC
500 pptC
Site Location and Spatial
Representation2
1) one or more monitors (20-40
km) upwind and downwind of
major urban emission regions;
2) several monitors in rural
areas which characterize
regional pollution events; 3)
one or more monitors
immediately downwind of
major urban emission regions
(<20km)
1) one or more monitors (20-40
km) upwind and downwind of
major urban emission regions;
2) one or more monitors
immediately downwind of
major urban emission regions
(<20km); 3) several monitors in
rural areas which characterize
regional pollution events
1) one or more monitors (20-40
km) upwind and downwind of
major urban emission regions;
2) one or more monitors
immediately downwind of major
urban emission regions (<20ktn);
3) several monitors in rural areas
which characterize regional
pollution events
1) one or more monitors (20-40
km) upwind and downwind of
major urban emission regions;
2) one or more monitors
immediately downwind of
major urban emission regions
(<20km)
Jxaiigeb will vary with site location, values identified are typical for downwind urban sites
2 Location requirements are in priority order
3 Analyses identified in SectionlV.C
IV-37
-------
The majority of emissions related characterization studies have been directed towards
mobile source emissions, with many of these studies specifically design to evaluate
the contribution of mobile source emissions of VOC and NOX to ambient
concentrations in the region and assess the accuracy of mobile source emission
models used in the ADP process. It is important to distinguish these studies from
source emission factor studies which are associated with measurements made on
various generic source types with the objective of establishing a quantitative
relationship between the rate of pollutant emitted from the source and a set of
appropriate operational and process parameters.
Special studies to characterize mobile source emissions have included:
Tunnel Studies
In these studies ambient measurement of hydrocarbons, carbon monoxide and
NOX have been carried out in parking garages and roadway tunnels (Lonneman et al.,
1974; Pierson, 1978; Gorse and Norbeck, 1981; Hampton, et al., 1983; Gorse, 1984; and
Lonneman et al., 1986; Ingalls, 1989; Ingalls et al, 1989) to determine the emissions
factors for the measured species for comparison with model estimated emissions for
the vehicles mix observed in these somewhat controlled environments. The contained
tunnel sample, has reduced uncertainties associated with chemical transformations
and transport, and as a result of high concentration levels found in these
environments reduced uncertainties in measurement. Tunnel studies have proven
useful in identifying errors in the emission estimates derived from mobile source
emission models (Pierson et al., 1990).
Roadside Measurements
In these studies ambient measurements immediately downwind or within the
roadway environment are performed to determine emissions from the in-use vehicle
mix sampled. Experimental studies have considered:
• measurements designed on the principals of conservation of mass, measuring the
vertical concentration profile of chemical constituents and appropriate
meteorological parameters upwind and downwind of the source region
(roadway). Measurements determining the net mass flux due to vehicle emissions
on the roadway have been used, (Bullin et al, 1980, and Hlavinka and Bullin,
1988) to determine CO emission factors along an interstate highway for
comparison to emissions models.
• measurements using inert tracers releases within roadway traffic to determine
dilution effects (Zweidinger et al., 1988, Cadle et al., 1976) have been applied to
study on-road vehicle emissions and estimate the emission rate of the composite
IV- 38
-------
vehicle fleet on the roadway.
• real-time in situ long path CO and CO2 on-road measurements via remote sensing
(Bishop et al., 1989; Bishop and Stedman, 1990; Lawson et al., 1990; Stephens and
Cadle, 1991; Bishop et alv 1993; Bishop et alv 1994); have reported the CO
emission rates for large populations of vehicles, indicating that a small percentage
of the vehicle population (< 10%) is responsible for ~ 50% of the total CO
emissions, averaged on a gm CO per gallon of fuel burned basis.
Finally, analysis applications using the continous emission monitoring (CEM) data
that is now being collected as required under the Clean Air Act have not been
addressed by the scientific community to date. This is likely the result of the data
being very new and not readily accessible. But, once the data are in the main stream,
analysis opportunties include:
• performing quality assurance checks on the major points source emission
inventory estimates;
• evaluating projection and disaggregation techniques for estimating daily emission
rates;
• providing feedback and coupling between point source emissions and receptor
measurements for the purpose of monitoring emissions projections, tracking
projected air quality improvements, diagnosing modeling and/or analysis failures,
and monitoring transport impacts of point sources with changing meteorological
conditions
Data Management
Cuurent practice is that all data collected under federally mandated air quality
networks be submitted to the EPA AIRS data base. AIRS is neither user friendly nor
state of the art. It represents one of the major road blocks relating to the distribution
and scientific application of air quality monitoring data compiled by EPA. It is highly
likely that the PAMS data sets will be seriously underutilized, if AIRS remains the
primary source for data compilation, access, and retrievial. A discussion in Section
V.A provides suggestions as to how data assessment, analysis and distribution
systems can be developed to resolve this problem.
IV-39
-------
-------
V. AN AGENDA FOR RESEARCH AND DEVELOPMENT
A.
Data Assessment, Archive, Distribution and Analysis System
The utility of measurement data and its successful application in emission-based and
observation-driven methods is directly related to its demonstrated quality and its
accessibility. The PAMS network has major deficiencies with regard to providing the
necessary infrastructure needed to process the extensive data sets generated by the
continuous automated gas chromatographic systems and other continuous gas phase.
These processing needs include:
• performing routine visual quality assurance checks of gas chromatogram and
concentrations time series;
• performing matrix scatter plots of paired hydrocarbon or chemical species, paired
chemical species and meteorological parameters for the purpose of checking data
quality, representativeness of the site and influence of local sources, and sample
contamination
The generation of high quality data must be supported with an appropriate data
archiving and distribution system. The data system must be user friendly, allowing
the measurement community both easy access with respect to data entry as well as
retrieval, be accessible on the internet, and have available to the data user a standard
set of analysis tools, demonstrated to be useful in assessing data quality and
supporting needs identified in the attainment demonstration process.
One approach for developing the required infrastructure to support data assessment,
archiving, distribution and analysis is to establish regional data analysis center.
These centers will act as the central repository for the PAMS data for the state
agencies within the designated regions and will develop and maintain a client -
server network which will provide the necessary training, access, and analysis tools
to support the various analysis activities outlined in Section IV.B and IV.C. The
regional centers would preferably be associated with Universities having active
atmospheric-air quality research programs and computational expertise and facilities
to support the development and operation of the data system.
V- 1
-------
B. Modeling
Development and Evaluation
No co-applications of the ODMs have been conducted. The general levels of
agreement and disagreement among the ODMs could be much better evaluated were
one or more co-applications to be carried out. At least one such co-application
should be for an area for which results are available from one or more recent PAQSM
applications. Several possible data bases exist. The 1995 Nashville study, which is
not yet ready for analysis, may be one of the best data bases for comparing ODMs.
Among the presently available data, the Southern California Air Quality Study
(SCAQS) provides a unique data base in two respects: many different types of
measurements were carried out, including for species such as HNO3, PAN, H2O2,
HCHO, and others, and the measurements were taken in highly urbanized locations
[thus contrasting with the data presented by Trainer et al. (1993)]. Sillman (1995)
uses one hour of data from the SCAQS by way of illustration; however, many days of
data are available. A major shortcoming of the SCAQS data is that, to the best of our
knowledge, true NOy was not measured. Thus, the applicability of the correlational
approach would need to be examined. Calculations using the SP algorithms have
been, or could readily be, carried out for all SCAQS days (including both intensive
and nonintensive sampling days). The SCAQS data base might lend itself to
application of the OBM, although, as with -the Atlanta study (Cardelino and
Chameides, 1995), the detection limits for NO may not have been adequate to
support the method fully (also, VOC samples were collected as one-hour averages 3
to 6 times a day using canisters).
Three other major data bases should be considered: the San Joaquin Valley Air
Quality Study/AUSPEX (SJVAQS/AUSPEX), the Lake Michigan Ozone Study
(LMOS), and the Coastal Oxidant Assessment for Southeast Texas (COAST) study
data bases. The SP algorithms have been applied to the COAST data (Blanchard et
al., 1995). Because species such as HNO3, H2O2, and NOy were not measured in these
three studies, the applicability of each of the other ODMs would need to be
considered further.
Because the OBM has, to date, been applied only to the 1990 Atlanta study, the
Atlanta data should also be evaluated using one or more of the other ODMs. The
diagnostic value of these comparisons may be limited, however, because some of the
data were of inadequate sensitivity or accuracy to drive the OBM (Cardelino and
Chameides, 1995). Nonetheless, there appears to be value in carrying out the
comparisons.
Finally, it may be possible to utilize one or more of the data sets included in Trainer
et al. (1993) to drive ODMs other than the correlational method.
V-2
-------
In the next section, we consider development and evaluation needs that are specific
to each ODM.
Correlation Between Ozone and NOy or NO, . Because the procedure consists of
reasonably straightforward regression analyses of ambient data, the statistical
methods do not themselves require much further development. The two statistical
topics that would benefit from some further development are, first, the treatment of
data points as individual time periods (e.g., hourly) versus binned averages, and,
second, the use of robust regression methods [e.g., Jacob et al. (1995) allude to the use
of a different regression procedure without explanation; also, various types of robust
procedures are described in the statistical literature].
Further attention should be devoted to delineating specific criterion values for
differentiating NOX and VOC-limited situations. For example, the existing
measurement studies do not provide firm guidance for delineating the range of
slopes, AO3/ANOZ, that would be indicative of NOX limitation. Moreover, the
existence of good correlations between ozone and NOX (or NOy) at rural sites does
help to indicate NOX limitation, no unambiguous definition of a "poor correlation",
which would be indicative of VOC limitation, exists. Thus, there is a need to apply
the method to a broad sampling of field measurements, representing a range of
conditions from urban to remote, to help delineate empirically a range of slopes and
correlation coefficients that could be used as guideline values.
Because this method is the only ODM that has been developed using exclusively
ambient measurements, it is important to carry out the developmental procedures
described above (they permit the method to remain independent of findings derived
from modeling and chamber experiments). However, there is also merit in drawing
upon modeling studies, such as Sillman (1995), that present ranges of O3/NOZ and
O3/NOy for which modeling results showed sensitivity of ozone to reductions of VOC
and NOX (see discussion below).
Guidance will also be needed for accounting for variability and uncertainty (see also
the discussion below). A particular need specific to the correlational approach is to
develop guidance for the indeterminate cases, i.e., those situations that cannot be
clearly delineated as VOC- or NOx-limited because the slopes or correlations fall into
some intermediate range.
SP algorithms. The SP algorithms were derived from analyses of chamber
experiments and box modeling using two existing chemical mechanisms. The
accuracy of the algorithms, when applied to the lower concentrations that occur in
the ambient atmosphere, is unknown. For example, the lowest initial NOX
concentrations in the UNC and UC Riverside chamber experiments were 90 to 100
ppbv; a limited number of the CSIRO experiments used 25 to 50 ppbv initial NOX.
V-3
-------
Some, though not all, of the same chamber experiments were used to develop the
chemical mechanisms; thus, the accuracy of the mechanisms at low NOX
concentrations is essentially untested. In contrast, ambient NOX concentrations are
often in the range of 5 to 100 ppb. During the past few years, attempts have been
made to carry out chamber experiments at increasingly lower NOX concentrations.
Results from such experiments should be obtained and used to further evaluate the
SP algorithms.
The SP algorithms were derived from systems that all have fixed volumes and an
initial charge of NOX and VOCs. Because the algorithms are not yet fully developed
for systems subject to dilution and continuous injection of emissions, the application
of the algorithms to ambient data is potentially subject to biases that have not yet
been determined. Additional work, which is now in progress, is needed to clarify the
importance of these possible biases.
Additional applications of the SP algorithms should be carried out using data sets
that would permit further evaluation of their performance under ambient conditions.
For example, studies having a high density of monitors and sufficient meteorological
information with which to construct trajectories could be used to evaluate the
consistency of the calculations along trajectories (in the absence of fresh emissions,
extent of reaction should increase with time and values of SPmax at earlier times
should not be less than SP at later times, etc.).
The Use of NO., and Indicator Ratios. To help reduce the possibility that results
would be overly dependent upon the accuracy of a single modeling application,
Milford et al. (1994) drew upon a range of models, modeling assumptions, and model
applications, while, for the same reason, Sillman (1995) used a single model applied
to two regions with varying assumptions. Nonetheless, the range of studies could be
expanded, for example, by including applications of the UAM. The consistency of
the findings should then be re-examined.
The utility of the indicator species and ratios should be examined using field data.
Sillman (1995) provides three limited examples for illustrative, rather than evaluative,
purposes. Similarly, Jacob et al. (1995) and Watkins et al. (1995) apply the concepts
to two data sets, but, in each case, the application is largely illustrative rather than
critical. A larger number of field studies should be selected, spanning a broad range
of conditions, and indicator species and ratios computed for each. The consistencies
among the different species should be examined for each field study. Where
possible. Cross-comparisons should be made using the method of correlations, the SP
algorithms, or independent modeling results.
At present, all indicator species and ratios are expressed as a range of threshold
values, rather than tight crossover points. Further work is needed to determine if the
widths of the ranges are artifacts of modeling assumptions or if such ranges also can
V-4
-------
be demonstrated to characterize ambient field data. If the ranges cannot be
tightened, additional guidance will be needed for interpreting cases that are
indeterminate, i.e., those situations that cannot be clearly delineated as VOC- or NOX-
limited because the concentrations or ratios fall into some intermediate range.
The OEM. Several developmental and evaluational needs are apparent. First, there
is a need to determine if the fundamental basis for calculating relative incremental
reactivities (RIRs) provides information that is appropriate and relevant. RIRs
derived from 12-hour PS03-NO terms may not be appropriate given the present form of
the ozone standard, which requires compliance with a one-hour ozone average.
Moreover, because the RIRs are expressed in terms of PS03-NO terms, it is unclear that
the sensitivities of ozone are properly characterized, as described in Section IV.
Many assumptions appear to underlie the box modeling approach used in the OEM.
Sensitivity analyses should be undertaken to clarify the consequences of the various
assumptions. Once a sufficiently large set of sensitivity analyses becomes available,
there will be a need to determine if the uncertainties associated with the OEM are
greater or less than those characterizing emissions-driven models. If the uncertainties
are as large as for emissions-driven models, consideration should be given to whether
or not the level of effort involved in operating the OEM is warranted.
Data Support
For each of the ODMs, improvements are needed in sampling methods. In some
cases, improvements in accuracy and detection limits are required for species that are
currently sampled. In other cases, the methods require measurements of species that
are not now routinely monitored.
Present measurements of NO are generally unbiased and many are sufficiently
accurate for use in the correlational methods and the SP algorithms. The indicator
species and ratios do not make explicit use of NO. However, the OEM requires
measurements of NO that are accurate to about 0.1 ppbv at concentrations as low as
0.1 ppbv. While measurement technologies exist that are capable of providing NO
measurements at concentrations as low as about 20 pptv, the performance of such
high-sensitivity instrumentation when used in routine monitoring networks may
degrade considerably.
Measurements of true NOy and NO2 are needed for use in the correlational methods,
the SP algorithms, and the indicator species and ratios. At present, measurement of
NOy appears more feasible in a routine mode than does the measurement of NO2.
While each of these ODMs can be used with NOy data only, better results could be
obtained if NO2 were also measured (thus providing measurements of true NOX and
of NOZ), The development of accurate routine measurements of both NOy and NO2 is
of considerable importance. Such measurements are expected to be of value
V-5
-------
regardless of the state of development of the ODMs reviewed here.
The use of indicator species and ratios other than NOy and O3/NOy requires
measurements of HNO3/ fiA, or HCHO, none of which are measured routinely at
present. As noted in Section IV, measurement of nitric acid has been problematic in
at least one routine monitoring network. Because it appears that substantial effort '
would be required to develop routine measurement capabilities for these species, the
focus of effort at present should be on the evaluation of the utility of the indicator
ratios using special data sets (see earlier discussion).
The OEM may require continuous GC instrumentation to provide speciated
hydrocarbon measurements with appropriate time resolution (the adequacy of, e.g.,
one-hour average concentrations from canisters is not known to us). At present,
efforts to improve VOC measurement methods are under way. With respect to the
use of the OBM, the need is to evaluate the accuracy of the method given current
VOC measurement capabilities.
For each of the ODMs, additional effort should be devoted to providing guidance
regarding the necessary numbers and locations of monitoring stations needed for
reliable results.
Analysis of Uncertainty
Analyses of uncertainties are needed for each of the ODMs. In each case,
uncertainties fall into two broad categories. The first category consists of
methodological limitations and uncertainties that should be addressed as previously
described under titie heading "Development and Evaluation." Addressing this type of
uncertainty is necessary for eliminating methodological biases. The second category
consists of the irreducible uncertainties that exist even for fully developed methods
that may be considered fundamentally sound and accurate. Such uncertainties can
generally be quantified through statistical methods or sensitivity analyses; a necessary
assumption is that the ODMs are accurate and appropriate for the intended
applications.
The development of the ODMs should proceed so that outputs always consist of both
findings and associated uncertainty estimates. In some cases, the statistical form of
the procedures lends itself to the calculation of a confidence limit. For example, the
regression analyses of the correlational procedures fall into this category. Similarly,
the computation from ambient measurements of mean values for indicator species or
ratios generates a straightforward confidence band. In other cases, such as for the SP
algorithms, the simplicity of the method permits a direct calculation of an uncertainty
in the result as a function of the uncertainties in measurements and parameters.
V-6
-------
Uncertainty analyses should, however, be translated into a format that would be of
more utility when the ODMs are used to help evaluate emission control strategies.
That is, the output of the ODM analyses needs to be expressed so as to qualitatively
indicate the precursor (VOC or NOX) whose control would be most effective in
reducing peak ozone concentrations; similarly, the uncertainty analyses should be
expressed in a manner that indicates the probability that the stated conclusions are'
erroneous. Additional efforts are needed to translate the confidence intervals
mentioned above into the type of probability statements that are actually needed.
Finally, further attention should be devoted to the treatment of variability. Because
differing conclusions may result for different days or locations, guidance is needed
for both presenting and interpreting results that exhibit substantial day-to-day or site-
to-site variability. This need applies equally to all ODMs and should be addressed
generically.
C.
Monitoring
Significant measurement issues remain with regard to supporting the modeling,
analysis , and monitoring considerations in support of the attainment demonstration
process. These measurement issues fall into three general classes.
• Measurements involving standard commercial monitoring instrumentation, where
questions remain regarding their performance, sensitivity and specificity in
routine operational monitoring.
• Measurements involving new commercially available instrumentation or
methodologies which have not been fully demonstrated or evaluated for routine
monitoring
• Measurements which require instrumentation development and or advancement
in new technologies
Monitoring of the Oxides of Nitrogen
Historically routine monitoring of the oxides of nitrogen has been quite limited. The
method of choice for NO measurements has been chemiluminescence detection, based
on the chemiluminescent reaction of O3 with NO. Coupling this method with a
catalytic converter to reduce NO2 to NO, it is possible to infer NO2 concentrations by
difference between the NOX and NO detection channels of the instrument. These
monitors, when used in routine networks, have had two major shortcomings which
affect their use in analysis related activities outline in this report
V-7
-------
• The catalytic converter is not NO2 specific and measures some unquantifiable
fraction of organic nitrates and nitric acid resulting in measured NO2
concentrations which are biased systematically high
• The detection limit and sensitivity for NO is insufficient to measure the critical
NO2/NO ratio during peak ozone production periods.
High sensitivity commercial NOX instruments are now available which can provide
NO detection limits of 100 ppt, but it remains to be demonstrated that such
measurements can be routine performed in operational monitoring networks.
Programs like SOS and NARSTO-NE are currently evaluating these systems and
developing the necessary methodologies to quality assure data at these low levels.
The development and evaluation of NOy research instrumentation has proceeded for
almost a decade/ but it was not until this year that a commercial instrument become
available in the U.S. This technology, which is an extension of the NOX technology
mentioned above, takes care to deliver all the oxides of nitrogen immediately to the
catalytic converter and assure that they are reproducibly and efficiently reduced prior
to entering the chemiluminescence detector. Programs like SOS and NARSTO-NE are
currently evaluating these systems and developing the necessary methodologies to
quality assure data at low concentration levels (i.e. 100 ppt).
The current state of monitoring science for the oxides of nitrogen indicates that
further work is needed in the following areas:
• develop NO2 specific instrumentation technologies, promising areas include
photolytic reduction, spectroscopic methods using tunable diode laser absorption
and differential optical absorption spectroscopy and luminol chemiluminescence
detection
• test and evaluate low level, high sensitivity NO chemiluminescence instruments to
quenching effects of water vapor and other trace gases found in polluted
environments
• test and evaluate the sensitivity of NOy converter efficiencies to water vapor and
other trace gases found in polluted environments
• develop instrumentation technologies for research measurements of HNO3 and
organic nitrates (including PAN) and which provide opportunities for the
development of routine continuous monitors
V-8
-------
Monitoring of Spedated Hydrocarbons and Volatile Organic Compounds
The measurement of hydrocarbons in the atmosphere has historically been
performed as part of research and special study programs. Routine, continuous
monitoring of speciated hydrocarbons has never been performed mainly due to the
lack of acceptable instrumentation. As a result of the PAMS initiative and its
requirements for speciated hydrocarbons, instrument vendors developed and
introduced several automated gas chromatographic hydrocarbon analysis systems.
These systems have been deployed in PAMS somewhat prematurely and have not
undergone rigorous field testing and evaluation, nor have acceptable SOPs or quality
assurance programs been established for these systems. The measurement of
carbonyl compounds has followed a similar track and is likely even less established
than the hydrocarbon methods. The SOS and NARSTO-NE programs have performed
some testing and evaluation of these systems in the field during the summer of 1995,
but much remains to be done.
Further work is needed in the following areas:
• establish the detection and speciation limits for operational automated continuous
gas chromatographic hydrocarbon analysis system
• develop low cost continuous speciated NMHC and VOC instrumentation
• evaluate current technologies and develop new techniques for the measurement of
carbonyls and other oxygenated hydrocarbons (e.g. alcohols, ethers, organic acids)
• develop SOPs and quality assurance programs for the automated continuous gas
chromatographic hydrocarbon analysis system
One or more commercial instruments have recently become available which use
luminescence reactions in the liquid phase to measure concentrations of
formaldehyde and hydrogen peroxide on a continuous basis. These techniques
should be evaluated both for their application for research as well as routine
measurement programs.
D. Longer Term Needs
Recognizing the stochastic nature of the system in the attainment demonstration
process
Simulating stochastic atmospheric processes using deterministic models is clearly a
simplifying approximation. The primary output of a deterministic approach, for a
simulation of an air quality episode, is one realization of the dynamic processes of
V-9
-------
interest. If two or three episodes are studied, the overall output will be simulated
results for two or three determinations. The ambient standard, however, is
probabilistic. Comparing the results of PAQSM simulations with the standard is at
best artificial and, in some circumstances, is an inadequate exercise that can be
misleading. Were a probabilistic model available for use, results could be compared
in terms of estimated and observed frequencies of occurrence of concentrations at
monitoring sites in an area. Relative to deterministic modeling little research and
development effort has been devoted to stochastic modeling. We recommend that
NARSTO consider pursing the development of this type of modeling system.
Assessing risk
Once modeling studies have been conducted, decision makers have to interpret their
results in light of the uncertainties that accompany them. Adopting a conservative
approach to control will lead to greater protection of the public and higher costs of
control, which eventually are absorbed by the larger community. Adopting a more
lenient approach will give more weight to economic concerns. Stated otherwise, the
decision maker must weigh the risks of underprotection of health and
overexpenditure for control. As now conducted, balancing risk is a wholly subjective
pursuit. While human judgment is irreplaceable, the scientific community can
develop and provide tools for use in clarifying decision options, quantifying and
assessing risks, and presenting information in ways that are readily communicated.
[See, for example, Clemen (1991).] We recommend that research and development be
conducted to provide analytical tools that will aid the decision maker in evaluating
risk.
V- 10
-------
VI. REFERENCES
Akimoto, H., F. Sakamaki, and M. Hoshino. "Photochemical ozone formation in
propylene-nitrogen oxide-dry air system." Environ. Sci. Technol., 13, 53-58,1979.
Altshuller, A.P., W.A. Lonneman, F.D. Sutterfield, and S.L. Kopczynski, Hydrocarbon
composition of the atmosphere of the Los Angeles basin-1967. Environ. Sci. Technol.,
5, 1009-1016,1971.
Anastasi, C, R.V. Gladstone and M.G. Sanderson, Chemical amplifiers for detection
of peroxy radicals in the atmosphere, Environ. Sci. Technol., 27, 474-482, 1993.
Andronache, C., W.L. Chameides, M.O. Rodgers, J. Martinez, P. Zimmerman and J.
Greenberg, Vertical distribution of isoprene in the lower boundary layer of the rural
and urban southern United States, J. Geophys. Res., 99, 16,989-16,999,1994.
Apel, E.G., J.G. Calvert and F.C. Fehsenfeld, The nonmethane hydrocarbon
intercomparison experiment (NOMHICE): tasks 1 and 2, J. Geophys. Res., 99, 16,651-
16,664,1994.
Aschmann, S.M. and R. Atkinson, Formation yields of methyl vinyl ketone and
methacrolein from the gas-phase reaction of O3 with isoprene, Env. Sci. Technol., 28,
1539-1542, 1994.
Beaton, S.P., G.A. Bishop, and D.H. Stedman, Emission charcteristics of Mexico City
vehicles , J.air Waste Manage. Assoc., 42, 1424-1429,1992.
Bishop, G.A., J.R. Starkey, A. Ihlenfeldt, W.J. Williams, and D.H. Stedman, IR long-
path photometry, a remote sensing tool for automotive emissions, Anal. Chem., 61,
671A, 1989.
Bishop, G.A., and D.H. Stedman, On-road carbon monoxide measurement
comparisons for the 1988-89 Colorado oxy-fuels program, Environ. Sci. Technol., 24
843,1990.
Bishop, G.A., D.H. Stedman, J.E. Peterson, T.J. Hosick and P.L. Guenther, A cost-
effectiveness study of carbon monoxide emissions reduction utilizing remote sensing,
technical paper, J. Air Waste .Manage. Assoc., 43, 978-988, 1993.
Bishop, G.A. et al., Enhancements of Remote Sensing for Vehicle Emissions in
Tunnels, J. Air Waste Manage.Assoc., 44,169-175, 1994.
Black, F.M., Motor vehicles as sources of compounds important to tropospheric and
stratospheric ozone, in Atmospheric Ozone esearch and its Policy Implications,
VI- 1
-------
Studies in Environmental Science 35, T. Schneider, S.D. Lee, G.J.R. Wolters and L.D.
Grant, eds. Elsevier, Amsterdam, pp. 85-109,1989.
Blanchard, C.L., P.M. Roth and H.E. Jeffries, Spatial mapping of preferred strategies
for reducing ambient ozone concentrations nationwide, 93-TA-37A.04, proceedings for
presentation at the 86th Annual Meeting & Exhibition Air & Waste Management
Assoc., Denver, CO, June 13-18,1993.
Blanchard, C. L. and H. M. Michaels. Regional Estimates of Acid Deposition Fluxes
in California. Sacramento, CA: California Air Resources Board, 1994. Contract
number A132-149.
Blanchard, C.L., F. W. Lurmann, M. Korc, and P. M. Roth. The Use of Ambient Data
to Corroborate Analyses of Ozone Control Strategies. Research Triangle Park, NC:
U.S. EPA, 1994. Contract number 68D30020.
Blanchard, C.L., F. W. Lurmann, M. Korc, and P. M. Roth. "Application of Smog
Production (SP) algorithms to the TNRCC COAST data". Air & Waste Management
Association 88th Annual Meeting & Exhibition, San Antonio, TX, June 18-23,1995.
Paper number 95-TP15P.04.
Buhr, M.P., M. Trainer, D.D. Parrish, R.E. Sievers and F.C. Fehsenfeld, Assessment of
pollutant emission inventories by principal component analysis of ambient air
measurements, Geophys. Res. Letters, 19,1009-1012,1992.
Bullin, J.A., N.J. Grenn, and J.C. Polasek, Determination of vehicle emission rates
from roadways by mass balance techniques, Environ. Sci. Technol., 14, 6,1980.
Cadle, S.H., D.P. Chock, J.M. Heuss, and P.R. Monson, 1976: Results of the General
Motors sulfate dispersion experiments, General Motors Research Laboratories, GMR-
2107, Warren, ML
Calvert, J.G., J.B. Heywood, R.F. Sawyer and J.H. Seinfeld, Achieving acceptable air
quality: some reflections on controlling vehicle emissions, Science, 261, 37-45,1993.
Cardelino, C. A. and W. L. Chameides. "An observation-based model for analyzing
ozone precursor relationships in the urban atmosphere." J. Air Waste Manage. Assoc.,
45,161-180,1995.
Carter, W. P. L. and F. W. Lurmann. "Evaluation of a detailed gas-phase atmospheric
reaction mechanism using environmental chamber data." Atmos. Environ., 25A, 2771-
2806,1991.
VI-2
-------
Chameides, W.L., F. Fehsenfeld, M.O. Rodgers, C. Cardelino, J. Martinez, D. Parrish,
W. Lonneman, D.R. Lawson, R.A. Rasmussen, P. Zimmerman, J. Greenberg, P.
Middleton and T. Wang, Ozone precursor relationships in the ambient atmosphere, J.
Geophys. Res., 97, 6037-6055,1992.
Chang, T. Y and M. J. Suzio, Assessing ozone-precursor relationships based on a '
smog production model and ambient data, J. Air & Wast Manage. Assoc., 45, 20-28,
1995.
Chang, T. Y., D. P. Chock, B. I. Nance, and S. L. Winkler, A photochemical extent
parameter to aid ozone air quality management, Air & Waste Management
Association 88th Annual Meeting & Exhibition. San Antonio, TX, June 18-23,1995.
Paper number 95-TA5.03.
Chin, M., D.J. Jacob, J.W. Munger, D.D. Parrish, E.G. Doddridge, Relationship of
ozone and carbon monoxide over North America, J. Geophys. Res., 99(D7), 14,565-14-
573,1994.
Chock, D.P., S. Kumar, and R.W. Herrmann, An analysis of trends in oxidant air
quality in the South Coast Air Basin of California, Atmos. Environ., 16(11), 2615-2624
1982.
Clean Air Act Amendments, 1990, P.L. 101-549,101st Congress, 104 Stat. 2399-2712,
November 15,1990.
Clemen, R.T., Making hard decisions: An introduction to decision analysis, PWS Kent
Publishing Co., Boston, MA, 1991.
Croes, B.E., J.R. Holmes and A.C. Lloyd, Reactivity-based hydrocarbon controls:
scientific issues and potential regulatory applications, J. Air Waste Manage. Assoc.,
42, 657-661, 1992.
Fan, S.-M., D.J. Jacob, D.L. Mauzerall, J.D. Bradshaw, S.T. Sandholm, D.R. Blake, H.B.
Singh, R.W. Talbot, G.L. Gregory and G.W. Sachse, Origin of tropospheric NOX over
subarctic eastern Canada in summer, J. Geophys. Res., 99,16,867-16,877,1994.
Fahey, D. W., G. Hiibler, D. D. Parrish, E. J. Williams, R. B. Norton, B. A. Ridley, H.
B. Singh, S. C. Liu, and F. C. Fehsenfeld. "Reactive nitrogen species in the
troposphere: measurements of NO, NO2, HNO3, particulate nitrate,
peroxyacetylnitrate (PAN), O3, and total reactive odd nitrogen (NOy) at Niwot Ridge,
Colorado." T. Geophys. Res.. 91, 9781-9793, 1986.
VI-3
-------
Fehsenfeld, F. C, M. J. Bellinger, S. C. Liu, D. D. Parrish, M. McFarland, M. Trainer,
D. Kley, P. C. Murphy, and D. L. Albritton. "A study of ozone in the Colorado
mountains." Journal of Atmospheric Chemistry, 1, 87-105,1983.
Finlayson-Pitts, BJ. and J.N. Pitts, Jr., Atmospheric chemistry of tropospheric ozone
formation: scientific and regulatory implications, Air & Waste, 43,1091-1100,1993. '
Gery, M.W., G. Z. Whitten, J. P. Killus, and M. C. Dodge. "A photochemical kinetics
mechanism for urban and regional scale computer modeling." J. Geophys. Res., 94,
12,925-12,956,1989.
Gery, M. W. and R. R. Crouse. User's Guide for Executing OZIPR. Research Triangle
Park, NC: U.S. Environmental Protection Agency, 1992. 166 pp.
Gorse, R.A.Jr., 1984: On-road emission rates of carbon monoxide, nitrogen oxides,and
gaseous hydrocarbons, Environ. Sci. Technol., 18, 500-507.
Gorse, R.A.,Jr. and L.M. Norbeck, CO emission rates for in-use gasoline and diesel
vehicles, JAPCA, 31,1094-1096,1981.
Hampton, C.V., W.R. Pierson, D. Schuetzle and T.M. Harvey, Hydrocarbon gases
emitted from vehicles on the road. 2. Determination of emission rates from diesel
and spark-ignition vehicles, Environ. Sci. Technol., 17, 699-708,1983.
Hanst, P. L., N. W. Wong, and J. Bragin. "A long-path infrared study of Los Angeles
smog." Atmos. Environ., 16, 969-981,1982.
Harley, R.A., A.G. Russell and G.R. Cass, Mathematical modeling of the
concentrations of volatile organic compounds: model performance using a lumped
chemical mechanism, Environ. Sci. Technol., 27,1638-1649,1993.
Harriss, R.C., G.W. Sachse, J.E. Collins, Jr., L. Wade, K.B. Bartlett, R.W. Talbot, E.V.
Browell, L.A. Barrie, G.F. Hill and L.G. Burney, Carbon monoxide and methane over
Canada: July-August 1990, J. Geophys. Res., 99,1659-1669,1994.
Heikes, B.C., Formaldehyde and hydroperoxides at Mauna Loa Observatory, J.
Geophys. Res., 97,18,001-18,013,1992.
Hess, G. D., F. Carnovale, M. E. Cope, and G. M. Johnson. "The evaluation of some
photochemical smog reaction mechanisms - n. Initial addition of alkanes and
alkenes." Atmos. Environ., 26A, 642-650,1992.
Hlavlinka, M.W. and J.A. Bullin, Validation of Mobile Source Emission Estimates
using Mass Balance Techniques. J. Air Pollut. Control Assoc., 38,1035,1988.
VI-4
-------
Ingalls, M.N., L.R. Smith, and R.E. Kirksey, 1989: Measurement of on-road vehicle
emission factors in the California south coast air basin. Vol. I. Regulated Emissions.
Final Report: SCAQS-1, Southwest Research Institute, San Antonio, TX, June 1989.
Ingalls, M.N., On-road vehicle emission factors from measurements in a Los Angeles
area tunnel. Presented at Air & Waste Management Association 82nd Annual
Meeting & Exhibition, Anaheim, CA, June 25-30,1989.
loffe, B.V., V.A. Isidorov, and I.G. Zenkevich, Certain regularities in the composition
of volatile organic pollutants in the urban atmosphere. Environ. Sci. Technol., 13,
864-868,1979.
Jacob, D.J., L.W. Horowitz, J. W. Munger, B. G. Heikes, R. R. Dickerson, R. S. Artz
and W. C. Keene, Seasonal transition from NOx- to hydrocarbon- limited conditions
for ozone production over the eastern United States in September, J. Geophys. Res.,
100, 9315-9324,1995.
Jaffe, D.A., R.E. Honrath and J.A. Herring, Measurements of nitrogen oxides at
Barrow, Alaska during spring: evidence for regional and northern hemispheric
sources of pollution, J. Geophys. Res., 96, 7395-7405,1991.
Jeffries, H. E., K. G. Sexton, and J. R. Arnold. Validation Testing for New
Mechanisms With Outdoor Chamber Data. Volume 2: Analysis of VOC Data for the
CB4 and CAL Photochemical Mechanisms. Cooperative Agreement CR-813107.
Research Triangle Park, NC: U.S. Environmental Protection Agency, 1989.
Jeffries, H. E., J. R. Arnold, and K. G. Sexton. Development of a Standard Chamber
Data Base and Protocol for Evaluating Oxidant Mechanisms for Urban and Regional
Models. Volume U: UNC SegFile Plots. Cooperative Agreement 815779. Research
Triangle Park, NC: U.S. Environmental Protection Agency, 1993.
Jobson, B.T., Z. Wu and H. Niki, Seasonal trends of isoprene, C2 -C5 alkanes, and
acetylene at a remote boreal site in Canada, J. Geophys. Res., 99,1589-1599,1994.
Johnson, G. M., "A simple model for predicting the ozone concentration of ambient
air." Proc. Eighth Inter. Clean Air Confer.. Melbourne, Australia, May 2, 1984, pp.
715-731.
Johnson, J. M., and S. M. Quigley. "A universal monitor for photochemical smog."
Air & Waste Management Association 82nd Annual Meeting and Exhibition.
Anaheim, California, June 25-30,1989. Paper number 89-29.8.
Johnson, G. M., S. M. Quigley, and J. G. Smith. "Management of photochemical smog
using the AIRTRAK approach." 10th International Conference of the Clean Air
VI-5
-------
Society of Australia and New Zealand. Aukland, New Zealand, March 1990, pp. 209-
214.
Johnson, G. M. and M. Azzi. Notes on the Derivation: the Integrated Empirical Rate
Model (V2.2X North Ryde, NSW, Australia: CSIRO Division of Coal and Energy
Technology, 1992.
Johnson, G. M., T. Trieu, M. Azzi, and S. Chamberlain. Results From AIRTRAK 2000
Systems: Lake Michigan Ozone Study 1991 Field Program. North Ryde, NSW,
Australia: CSIRO, 1993. Report no. CET/IR145R.
Kelly, N. A., G. T. Wolff, and M. A. Ferman. "Sources and sinks of ozone in rural
areas." Atmos. Environ., 18,1251-1266,1984.
Kelly, N.A. and PJ. Groblicki, Real-world emissions from a modern production
vehicle driven in los angeles, technical paper, J. Air Waste Manage. Assoc., 43,1351-
1357,1993.
Kleinman, L., Y.-N. Lee, S.R. Springston, L. Nunnermacker, X. Zhou, R. Brown, K.
Hallock, P. Klotz, D. Leahy, J.H. Lee and L. Newman, Ozone formation at a rural site
in the southeastern United States, J. Geophys. Res. 99, 3469-3482,1994.
Korsog, P.E. and G.T. Wolff, An examination of urban ozone trends in the
northeastern U.S. (1973-1983) using a robust method, Atmos. Environ., 25B, 47,1991.
Kumar, S., and D.P. Chock, An up-date on oxidant trends in the South Coast Air
Basin of California, Atmos. Environ., 18(10), 2131-2134,1984.
Lawson, D.R., P.J. Groblicki, D.H. Stedman, G.A. Bishop, and P.L. Guenter, Emissions
from in-use motor vehicles in Los Angeles: A pilot study of remote sensing and the
inspection and maintenance program, J. Air Waste Manage. Assoc., 40, 1096,1990.
Lee, J.H., D.F. Leahy, I.N. Tang and L. Newman, Measurement and speciation of -gas.
phase peroxides in the atmosphere, J. Geophys. Res. 98, 2911-2915,1993.
Liu, S. C, M. Trainer, F. C. Fehsenfeld, D. D. Parrish, E. J. Williams, D. W. Fahey, G.
Hiibler, and P. C. Murphy, "Ozone production in the rural troposphere and the
implications for regional and global ozone distributions." J. Geophys. Res., 92, 4191-
4207,1987.
Lonneman, W.A., T.A. Bellar, and A. P. Altshuller, Aromatic hydrocarbons in the
atmosphere of the Los Angeles basin. Environ. Sci. Technol., 2,1017-1020,1968.
VI-6
-------
Lonneman, W.A., S.L. Kopczynski, P.E. Daley and F.D. Sutterfield, Hydrocarbon
composition of urban air pollution, Environ. Sci. Technol. 8, 229-236,1974.
Lonneman, W.A., R.L. Seila and S.A. Meeks, Non-methane organic composition in the
Lincoln Tunnel, Environ. Sci. Technol. 20, 790-796,1986.
Lurmann, F. W., A. C. Lloyd, and R. Atkinson. "A chemical mechanism for use in
long-range transport/acid deposition computer modeling." J. Geophys. Res., 91(D10),
10905-10936,1986.
Lurmann, F. W., W. P. L. Carter, and L. A. Coyner. A Surrogate Species Chemical
Mechanism for Urban-Scale Air Quality Modeling. EPA-600/3-87-014. Research
Triangle Park, NC: U.S. Environmental Protection Agency, 1987.
Mackay, G.I., Measurement of Atmospheric Nitric Acid at Azusa by Tunable Diode
Laser Absorption Spectrometry. Sacramento, CA: California Air Resources Board,
1994. Contract number 93-300.
Mathur, R., K.L. Schere, A. Nathan, Dependencies and sensitivity of tropospheric
oxidants to precursor concentration over the Northeast United States: a model study,
J. Geophy&. Res., 99,10,535-10,552,1994.
Mayrsohn, H., and J.H. Crabtree. 1976. Source reconciliation of atmospheric
hydrocarbons. Atmos. Environ., 10,137-143.
Meyers, E. L. 1987. A review of 1986 Philadelphia NMOC, NOX and NMOC/NOX
ratio data, an interim report. U.S. Environmental Protection Agency, Research
Triangle Park, N.C.
Milford, J.B., D. Gao, S. Sillman, P. Blossey and A.G. Russell, Total reactive nitrogen
(NOy ) as an indicator of the sensitivity of ozone to reductions in hydrocarbon and
NOX emissions, J. Geophys. Res., 99, 3533-3542,1994.
Miyoshi, A., S. Hatakeyama, N. Washida, OH radical-initiated photooxidation of
isoprene: an estimate of global CO production, J. Geophys. Res., 99,18,779-18,787,
1994.
National Academy of Sciences, 1991 - "Rethinking the Ozone Problem in Urban and
Regional Air Pollution", National Academy Press, Washington, DC.
Nelson, P.P., and S.M. Quigley, The m,p-xylenes : ethylbenzene ratio. A technique
for estimating hydrocarbon age in ambient atmospheres. Atmos. Environ., 17, 659-
662,1983.
VI-7
-------
Office of Technology Assessment, U.S. Congress, 1989 - "Catching Our Breath-Next
Steps for Reducing Urban Ozone", OTA-O-412, Washington, DC.
Olszyna, K.J., E.M. Bailey, R. Simonaitis and J.F. Meagher, O3 and NOy relationships
at a rural site, J. Geophys. Res., 99,14,557-14,563,1994.
Parrish, D. D., D. W. Fahey, E. J. Williams,, S. C. Liu, M. Trainer, P. C. Murphy, D.
L. Albritton, and F. C. Fehsenfeld. "Background ozone and anthropogenic ozone
enhancement at Niwot Ridge, Colorado." Journal of Atmospheric Chemistry, 4, 63-80,
1986.
Parrish, D.D., M. Trainer, M.P. Buhr, B.A. Watkins and F.C. Fehsenfeld, Carbon
monoxide concentrations and their relation to concentrations of total reactive oxidized
nitrogen at two rural U.S. sites, J. Geophys. Res., 96, 9309-9320,1991.
Parrish, D.D., M.P. Buhr, M. Trainer, R.B. Norton, J.P. Shimshock, F.C. Fehsenfeld,
K.G. Anlauf, J.W. Bottenheim, Y.Z. Tang, H.A. Wiebe, J.M. Roberts, R.L. Tanner, L.
Newman, V.C. Bowersox, K.J. Olszyna, E.M. Bailey, M.O. Rodgers, T. Wang, H.
Berresheim, U.K. Roychowdhury and K.L. Demerjian, The total reactive oxidized
nitrogen levels and the partitioning between the individual species at six rural sites in
eastern North America, J. Geophys. Res., 98, 2927-2939,1993.
Parrish, D.D., J.S. Holloway, M. Trainer, P.C. Murphy, G. L. Forbes and F.C.
Fehsenfeld, Export of North American ozone pollution to the North Atlantic Ocean,
Science, 259,1436-1439,1993.
Parrish, D.D., J.S. Holloway and F.C. Fehsenfeld, Routine, continuous measurement of
carbon monoxide with parts per biUion precision, Environ. Sci. Techno!., 28,1615-
1618,1994.
Pilar, S., and W.F. Graydon, Benzene and toluene in Toronto atmosphere. Environ.
Sci. Technol. 7, 628-631,1973.
Pierson, W.R, A.W. Gertler, and R.L. Bradow, 1990: Comparison of the SCAQS tunnel
study with other on-road vehicle emission data, J. Air Waste Manage. Asssoc. 40,
1495,1990.
Poulida, O., K.L. Civerolo and R.R. Dickerson, Observations and tropospheric
photochemistry in central North Carolina, J. Geophys. Res., 99,10,553-10,563,1994.
Rao, S.T., G. Sistla and R. Henry, Statistical analysis of trends in urban ozone air
quality, J. Air-Waste Manage. Assoc., 42,1204-1211,1992.
VI-8
-------
Rao, S.T. and I.G. Zurbenko, Detecting and tracking changes in ozone air quality,
technical paper, J. Air & Waste Manage. Assoc., 44,1089-1092,1994.
Reynolds, S.D., P.M. Roth and T.W. Tesche, A Process for the stressful evaluation of
photochemical model performance. Final Report, Western States Petroleum
Association,
Glendale, CA January, 1994.
Reynolds, S.D., H.M. Michaels, P.M. Roth ,T.W. Tesche, D. McNally, L. Gardner, and
G. Yarwood, Alternative base cases in photochemical modeling: Their construction,
role and value, Atmos. Environ., (accepted for publication 1996).
Roberts, E.M., Review of statistics of extreme values with applications to air quality
data, Part H. Applications, J. Air Pollut. Control Assoc., 29(7), 733-740,1979.
Roberts, J.M., F.C. Fehsenfeld, S.C. Liu, M.J. Bellinger, C. Hahn, D.L. Albritton, and
R.E. Sievers, Measurements of aromatic hydrocarbon ratios and NOX concentrations in
the rural troposphere: Observation of air mass photochemical aging and NOX removal,
Atmos. Environ., 18, No. 11, 2421-2432,1984.
Roth, P.M., Uses and limitations of modeling in analyzing and designing ozone stratgeies,
Air, Water and Soil Pollution, 67, Nos. 1-2, 195-203, 1993.
Roth, P.M., T.W. Tesche, and S.D. Reynolds, Protocol for evaluating the performance of the
SARMAP model. SARMAP Technical Committee, April, 1995.
Shively, T.S., An analysis of the trend in ground-level ozone using nonhomogeneous Poisson
processes, Atmos. Environ., 25B(3), 387-395, 1991.
Sillman, S., J. A. Logan, and S. C. Wofsy, The sensitivity of ozone to nitrogen oxides
and hydrocarbons in regional ozone episodes, J. Geophys. Res. 95,1837-1851,1990.
Sillman, S., The use of NOy, H2O2, and HNO3 as indicators for ozone-NOx-
hydrocarbon sensitivity in urban locations, J. Geophys. Res., 100,14175-14188,1993.
Slemr, P., W. Seiler, Field measurements of NO and NO2 emissions from fertilized
and unfertilized soils, J. Atmos. Chem., 2, 1-24, 1984.
Smith, R.L., Extreme value analysis of environmental time series: An application to
trend detection in ground-level ozone, Statist. Sciences, 4, 367-393,1989.
Spicer, C. W., D. W. Joseph, P. R. Sticksel, and G. R. Ward, Ozone sources and
transport into the northeastern United State, Environ. Sci. Technol., 13, 975-985,1979.
VI-9
-------
Stephens, E.R., and F.R. Burleson, Analysis of the atmosphere for light hydrocarbons.
J. Air Pollut. Control Assn., 17,147-153,1967.
Stephens, E.R., and F.R. Burleson, Distribution of light hydrocarbons in ambient air.
J. Air Pollut. Control Assn., 19, 929-936,1969.
Stephens, R.D. and S.H. Cadle, Remote sensing of carbon monoxide emissions from
on-road vehicles, J. Air Waste Manage. Assoc., 41, 39,1991.
Tesche, T.W., P.M. Roth, S.D. Reynolds, and F.W. Lurmann, Scientific assessment of
the urban airshed model dJAM-IV), American Petroleum Institute. Report No. AG-
90/TS16, Crested Butte, CO, 1992.
Trainer, M., D.D. Parrish, M.P. Buhr, R.B. Norton, F.D. Fehsenfeld, K.G. Anlauf, J.W.
Bottenheim, Y.Z. Tang, H.A. Wiebe, J.M. Roberts, R.L. Tanner, L. Newman, V.C.
Bowersox, J.F. Meagher, K.J. Olszyna, M.O. Rodgers, T. Wang, H. Berresheim, K.L.
Demerjian and U.K. Roychowdhury, Correlation of Ozone with NOy in
photochemically aged air, J. Geophys. Res., 98, 2917-2925,1993.
Tuazon, E. C, A. M. Winer, R. A. Graham, and J. N. Pitts, Jr. "Atmospheric
measurements of trace pollutants by kilometer-pathlength FT-IR spectroscopy." Adv.
Environ. Sci. Technol., 10, 259-300,1980.
Tuazon, E. C., A. M. Winer, and J. N. Pitts, Jr., Trace pollutant concentrations in a
multiday smog episode in the California south coast air basin by long path length
Fourier Transform Infrared Spectroscopy, Environ. Sci. Technol., 15,1232-1237, 1981.
Tuazon, E. C., C. L. Blanchard, S. V. Hering, D. Lucas, and G. I. Mackay. A Review
of Nitric Acid Measurements by Tunable Diode Laser Absorption Spectroscopy
(TOLAS). Sacramento, CA: California Air Resources Board, 1995. Contract number
93-340.
U.S. Environmental Protection Agency, Ambient Air Quality Surveillance; Final Rule,
40CFR Part 58, Federal Register Vol. 58 (28), 8452-8475, February 12,1993.
U.S. Environmental Protection Agency, Implementation Strategy for the Clean Air Act
Amendments of 1990. Office of Air and Radiation, U.S. Government Printing Office,
Washington, DC, 1991.
U.S. Environmental Protection Agency, Guideline for the evaluation of air quality
trends. Office of Air Quality Planning and Standards, Research Triangle Park, NC,
1974.
VI- 10
-------
Watkins, B. A., D. D. Parrish, M. P. Buhr, R. B. Norton, M. Trainer, and F. D.
Fehsenfeld. "Factors influencing the concentration of gas-phase hydrogen peroxide
during the summer at Kinterbish, AL." Submitted to T. Geophys. Res.. 1995.
Watson, J. G., J. C. Chow, R. T. Egami, J. L. Bowen, C. L. Frazier, A. W. Gertler, D.
H. Lowenthal, and K. K. Fung. Measurements of Dry Deposition Parameters for the
California Acid Deposition Monitoring Program. Sacramento, CA: California Air
Resources Board, 1991. Contract number A6-076-32.
Went, F.W., Organic matter in the atmosphere and its possible relation to petroleum
formation, Proc. Natl. Acad. Sci., 46, 212-221,1960.
Whitby, R.A., and E.R. Altwicker, Acetylene in the atmosphere: Sources,
representative ambient concentrations and ratios to other hydrocarbons. Atmos.
Environ., 12,1289-1296,1978.
Williams, E. J., D. D. Parrish, and F. C. Fehsenfeld, Determination of nitric oxide
emissions from a grassland site in Colorado, United States, J. Geophy. Res., 92, 2173-
2179,1987.
Williams, E. J., D. D. Parrish, M. P. Buhr, F. C. Fehsenfeld and R. Fall, Measurement
of soil NOX emissions in Central Pennsylvania, J. Geophy. Res., 93, 9593-9546,1988.
Williams, E.J., and F.C. Fehsenfeld, Emissions of nitrogen oxides from soil in three
North American ecosystems, EOS, 43,1008,1989.
Wolff, George T. and P.E. Korsog, Ozone control strategies based on the ratio of
volatile organic compounds to nitrogen oxides, J. Air Waste Manage. Assoc., 42,
1173-1177,1992.
Wolff, G. T., On a NOX -focused control strategy to reduce O3, Notebook, Air &
Waste , 43,1593-1596,1993.
/
Zhang, D.H. Stedman, G.A. Bishop, P.L. Guenther, S.P. Beaton, and J.E. Peterson, On-
road hydrocarbon remote sensing in the Denver area, Environ. Sci. Technol., 27,1885-
1891,1993.
Zweidlinger, R.B., J.E. Sigsby, Jr., S.B. Tejada, F.D. Stump, D.L. Dropkin, W.D. Ray,
and J. W. Duncan, Detailed Hydrocarbon and Aldehyde Mobile Source Emissions
from Roadway Studies. Environ. Sci. Technol., 22, 956,1988.
VI- 11
•&U.S. GOVERNMENT PRINTING OFFICE; 1997 - 549-001/60159
-------
-------
-------
rn
1
o
9>
8g s
3=82.
s-F
^ 03
CD
c
o o m
-j' CD 5.
Q.a£
3 ® §
-
SI
o) 3
oo CD
a
co
3 £2
CD to
"O .
<3
I
i
30 !^
CD (O
CO CD
CD 3
83
O
m
33
I
fcg
•mS
------- |