Telling the Ozone Story with Data
Final Report, Vol. I: Executive Summary
OTAG Air Quality Analysis Workgroup
Dave Guinnup and Bob Collom, Co-chair
Draft Submitted to the Policy Group, June 2, 1997
Introduction
The Air Quality Analysis (AQA) Workgroup of OTAG has been formed to provide assessments of
air quality and meteorological data relevant to the mission of OTAG. The Workgroup placed special
emphasis on ozone transport since some nonattainment areas are experiencing considerable influx of
ozone across their boundaries and they cannot demonstrate attainment by local measures only.
Specifically, the workgroup's purpose statement reads:
The Air Quality Analysis Workgroup shall identify, characterize, compare, and assess
observational data and studies including but not limited to air quality trends analysis,
overflight data, and meteorological studies for the purpose of evaluating the effects of the
transport of ozone and its precursors on ozone nonattainment in the eastern United States.
The Workgroup process has entailed the development and presentation of individual work products
in Workgroup meetings, followed by open review of work by the group. This process has been
greatly facilitated by the early development of an interactive Internet website which has been used to
communicate data sets, analytical tools, results, interpretations, and critical feedback. In addition,
the Workgroup developed policy-relevant summaries and interpretations and subjected these
summaries to group review.
The Workgroup members were affiliated with the EPA, state agencies, industry (power,
transportation) consultants and academia. The activities undertaken by this group have included
reviewing existing air quality studies and analyses; developing analyses and visualizations of air
quality and meteorological data to help in the understanding of ozone formation and transport;
comparing modeling results against available air quality data; and integrating air quality analyses and
modeled results into conceptual interpretations of ozone transport for use in policy development.
The multiple types of analyses included: spatial and temporal pattern and trend analysis, trajectory
and residence time analysis, correlation and cluster analysis, detailed evaluations of intensive (SOS
and NARSTO-NE) field studies, and comparison of model results with the data. The model
comparison complemented the comprehensive model evaluation performed by the OTAG Modeling
Workgroup.
In general, these results provide the background assessment of the current ozone problem necessary
to place the modeling results in the appropriate context. That is to say, they "set the stage" for the
useful interpretation of the modeling of future-year control strategies. The air quality analysis is
based on measurements, and in many cases they can provide a good indication of the inherent
strengths and weaknesses of the photochemical grid modeling. They also provide more of a
"climatological" view of the ozone problem, which extends beyond the modeled episode days and
provides a broader perspective of the ozone problem and its characteristics. Once one has developed
an understanding of this broad perspective, the modeling results can be used to subsequently provide
a more focused view of individual episodes and potential control strategy impacts during these
episodes.
In the paragraphs below, we present the major policy-relevant results, conclusions and
recommendations from the AQA Workgroup. The summary falls into the following major
categories:
Origins and patterns of ozone in the OTAG domain.
2
Ozone transport in the OTAG domain.
Description and climatological representativeness of OTAG episodes.
Air quality management implications of the data analysis results.
In each major category a set of policy-relevant technical questions are posed, followed by and brief
statement of pertinent results. The supporting materials for this Volume I - Executive Summary, can
be found in Volume II - Summary and Integration of Results and in Volume III - Summaries of
Individual Analyses, all accessible through the AQAWG website, http:capita/wustl.edu/OTAG.
Origins and Patterns of Ozone in the OTAG Domain
What is the pattern of ozone precursor emissions in the OTAG domain?
Anthropogenic emissions of volatile organic compounds (VOC) and nitrogen oxides (NOx) within
the OTAG domain from area and point sources contribute to the formation of excess ozone which
adds to tropospheric background levels. Point sources are generally rich in NOx emissions, area
source emissions are rich in VOC as well as NOx. Area sources of NOx occur primarily in large
urban metropolitan areas (e.g., the Washington-New York corridor, Chicago, Atlanta, Dallas-Ft.
Worth, Houston, St. Louis), while elevated point sources of NOx are prevalent in non-urban,
industrial regions such as the Ohio River Valley. (Figure 1) Temporally, area sources tend to have a
strong diurnal and weekly cycle, while point sources typically vary less in time.
What is the pattern of ozone on the edges of and within the OTAG domain?
The ozone concentrations have been evaluated using a comprehensive set from 600 monitoring
stations in the OTAG region of routine EPA and research databases. The entire network includes 102
urban sites, 259 suburban sites, and 238 rural or remote sites (13 sites in the network are not
classified).
The OTAG region is ventilated generally by air coming from outside the domain whose ozone
concentrations average about 30-40 ppb, corresponding to typical tropospheric background levels
measured in the northern hemisphere. (Figure 2) Thus, with the notable exception of the Canadian
border along the Windsor-Quebec corridor, there are no significant external source impacts on the
OTAG domain, at least on a regional scale. It is reasonable to assume that in the absence of
anthropogenic emissions, the average summertime ozone concentration would be about 30-40 ppb
throughout the OTAG domain. The average ozone concentration within the OTAG domain is highest
near major metropolitan areas. A large pool of ozone is also generated in the central subregion of
OTAG domain along the Ohio River as shown in Figure 2.
3
What is the range of ozone concentration?
Ozone exhibits strong day-to-day variation that can be quantified by examining "cleaner" and
"dirtier" days across the domain. (Figure 3) The urban impact is virtually undetectable during
cleaner, low-ozone days (i.e., the lowest 10th percentile of all measured ozone concentrations).
However, even these clean days exhibit a broad area of higher average ozone (>40 ppb) in non-urban
portions of the OTAG states from Illinois to Pennsylvania relative to the rest of the OTAG domain.
During the dirtier, high-ozone days (the 90th percentile of all measured ozone concentrations) the
urban influence is very pronounced but confined to about one or two hundred miles from major
metropolitan areas. It is therefore observed that urban areas are making substantial local
contributions to the highest 1-hour daily maximum ozone concentrations as well as exhibiting the
highest variation in ozone concentration (difference between 90th and 10th percentile).
4
Figure 1. Area and point source emissions for NOx. The high emission densities for area sources are located in the vicinity of urbanmetropolitan areas. The major point source emission regions are located mostly in the industrial Midwestern states.
Figure 2. Average 1-hour daily maximum ozone concentration in the OTAG region. All four corners have about 30-40 ppb ozone, which
corresponds to tropospheric background entering the region. The crossection indicates a general increase from west to east.
5
Figure 3. A) Spatial pattern of the 10th percentile of daily maximum ozone. Elevated ozone
concentrations are evident throughout the central OTAG region. B) Spatial pattern of the 90th
percentile of daily maximum ozone. The highest ozone occurs near metropolitan regions.
6
How much does ozone vary in time?
Ozone exhibits temporal variability over hourly, diurnal, synoptic (3-5 days), weekly, seasonal, and
long-term (5-20 years) scales. At hourly, diurnal, and seasonal time scales, the urban areas exhibit
much more variability than rural sites. The ozone changes on weekly and long-term scales are
caused primarily by anthropogenic emission changes, while changes at the hourly, diurnal, synoptic
and seasonal scales are also influenced by meteorology.
The long-term trends show that the number of 120 ppb, 1-hour exceedances has been steadily
declining over the past decade, indicating that historical control strategies appear to have been
effective at shaving peak concentrations. (Figure 3a) However, the downward trend of 80 ppb, 8hour exceedances is less pronounced, indicating that new types of control approaches may be needed
to attain future standards.
Where are the areas of exceedances for the current and proposed ozone
standards?
The spatial pattern of the current (1991-1995) exceedances (120 ppb, one-hour) shows that the
Washington-New York corridor, Chicago, Atlanta, Dallas-Ft. Worth, Houston, St. Louis, are the
major metropolitan areas that exceed the current standard. (Figure 4) However, there are other
metropolitan areas throughout the OTAG region which remain in nonattainment with the current
standard. Virtually all areas where exceedances of the current standard occur are confined to the near
vicinity (<150 miles) of metropolitan areas. In contrast, areas exceeding the proposed 8-hour 80 ppb
ozone standard are more numerous, extend further from metropolitan areas and include a large
portion of the central part of the OTAG domain. Distances between nonattainment areas projected
under the proposed standard are significantly shorter than those under the current standard.
7
Figure 4. Counties not meeting the 0.12 and 0.08 ppm standard according to EPA.
8
Ozone transport in the OTAG Domain
Ozone transport: beneficial or harmful?
In general, the atmosphere acts as a dispersing agent for air pollutants. Dispersion through increased
wind speed can cause both beneficial pollutant reduction near a source and harmful pollutant
increases downwind of a source. (Figure 5) Dispersion takes place through horizontal transport and
vertical mixing in the lower atmosphere. Horizontal transport depends on location as well as on
elevation above ground, and to some extent on time of day. The horizontal transport for elevated
emission sources (greater than 100m above ground) is substantially higher than the transport of lowlevel sources, due to higher wind speeds aloft, particularly at night during nocturnal jet conditions.
In general, ozone transport takes place in synoptic (>800m), channeled (200-800m) and near surface
(<200m) flow regimes. (Figure 7)
Examination of dispersion conditions (Figure 6) during high-ozone periods across the OTAG domain
shows that dispersion in the Southeast is typically poor due to stagnating air masses, whereas the
western and northern sections of the domain are typically influenced by strong southerly and westerly
winds, respectively. This supports the notion that ozone exceedances in the Southeast are more
predominantly “homegrown,” than exceedances in other sections of the OTAG domain. In contrast,
on low-ozone days the air flow is from outside the domain (e.g., Canada and the Gulf of Mexico).
Many of the most widespread episodes of high ozone across the OTAG domain result from a
stagnation event somewhere in the domain (particularly central and south) which is followed by
strong unidirectional flow to some other part of the domain (particularly in the upper Midwest and
Northeast). While this pattern is not always seen for every ozone episode, it points out what appears
to be the necessary condition for a region-wide ozone event; namely stagnation followed by
transport. Such a condition is seen in three (88, 91, 95) out of four episodes chosen for OTAG
domain modeling.
In summary, the "good news" about transport is that it can disperse, or clean up, the ozone formed in
an area during a stagnation event. The south-central and southeastern portions of the OTAG domain,
which experience relatively more stagnation, can benefit from this aspect of transport. The "bad
news" about transport is that it can carry significant concentrations of ozone from one portion of the
domain to another, and this aspect of transport tends to cause more problems in the midwestern and
northeastern portions of the domain.
What are the ranges of ozone transport implied from various studies?
AQA researchers have assessed the transport issue based on examination of meteorological data,
trajectory analyses and field observations of ozone and ozone precursors. It is important to note that
AQA analyses have not estimated future ozone transport, per se.
The distances of ozone impact deduced from multiple types of analysis range from 150 to 500 miles.
The direct influence of specific urban areas can be directly traced to some 150-200 miles before
merging indistinguishably into the regional ozone pattern. Ozone transported at night can have a
significant impact hundreds of miles downwind the next day. Statistical correlation analyses of the
9
regional ozone pattern suggest ozone transport distances of up to 300-500 miles, but it is not clear to
what extent this actually represents transport of ozone and/or precursors, or is a meteorological
correlation. Time-lagged correlation analyses also suggest linkages between ozone concentrations
separated in time by one to two days within the domain. Based on observed mean wind speeds, the
interpretation of such a time correlation as an atmospheric lifetime suggests potential downwind
impact areas of about 400 miles. While an estimate of impact distance based on any single method is
rather uncertain, the coinciding range of the various methods increases confidence in the accuracy of
these estimated ozone transport distances.
The perceived distance of transport can vary considerably depending on the statistic used to describe
the ozone concentrations or exposures. In general, the spatial scale of perceived ozone transport
decreases rapidly with an increasing concentration threshold or increasing percentiles (percentage of
a cumulative concentration distribution). For example, the average daily maximum ozone may have
a transport scale of hundreds of miles, while ozone exceedances of 1-hour, 120 ppb thresholds may
extend to only tens of miles. This finding underscores the notion that an 8-hour, 80 ppb ozone
standard will implicate larger areas and longer transport scales, than the 120 ppb 1-hour standard.
10
Figure 5. Average ozone concentration during low (< 3 m/s) and high (>6 m/s) wind speed conditions. Low wind speeds cause ozone to
accumulate near the source areas. Higher wind speeds cause more regional ozone transport.
Figure 6. Transport winds during low (10%-ile) and high (90%-ile) local ozone days. Low ozone occurs on days with transport from outside
the region. The regions of influence are higher. High ozone in the Southwest, Southeast, and the Midwest occurs when the transport winds
are slow and variable. The regions of influence are shorter. In the Northeast, high ozone is associated with strong westerly winds.
12
Figure 7 . Schematics of transport in the N.E. Ozone transport occurs in synoptic, channeled, and
near surface flow regimes.
Figure 8 . Daily maximum ozone concentration averaged over all monitors in the OTAG domain.
Regional episodes above 70 ppb have a duration of up to nine days.
14
Description and climatological representativeness of OTAG
episodes
What are the characteristics of four modeling episodes?
At a specific location, high ozone concentration, above 80 or 120 ppb, can occur during regional,
sub-regional, or local episodes. The OTAG domain-wide regional episodes are caused by slowmoving or recirculating airmasses over the center of OTAG domain. Coincidentally, the central
Ohio River Region is a high NOx emissions area. A regional episode is defined here (Figure 8) as
the condition when the daily maximum ozone concentration averaged over all the OTAG domain
monitoring sites, exceeds 70 ppb. All four periods selected for OTAG domain modeling are
regional. During the 1988 episode, the OTAG domain average ozone exceeded 70 ppb for 9
consecutive days with a peak of 103 ppb. The 1991 and 1995, episodes lasted for about 6 days with
peaks between 80-90 ppb. The 1993 are best characterized as a subregional episode over the
Southeast, since the OTAG domain-average ozone barely exceeds 70 ppb only on one day. The
pattern of average ozone concentration (Figure 9A) shows that during the four modeling episodes
ozone is high over the industrial Midwest, as well as over Pennsylvania and New Jersey and low
along the Gulf Coast.
The OTAG domain-scale regional episodes occur about 3 (1993) to 8 (1988) times, covering about
10% of the April-September ozone season. Multi-day ozone accumulation causes about half of the
OTAG-wide recorded exceedances over 120 ppb occur during these regional episodes. However,
these episodes account for only 30% of the 80 ppb exceedances. Hence, their role for the proposed
new standard is diminished, compared to the old standard.
15
What is the representativeness of the selected modeling periods?
The four modeling periods are typical for regional ozone episodes that occur 3-8 times during every
ozone season. The average ozone concentration during the four OTAG domain episodes is about 5
ppb less over the Illinois-Ohio and Michigan corridor, as well as over Texas than the typical episode.
On the other hand, the southeastern OTAG domain, including Alabama, Georgia, Carolinas, and
Virginia is about 5 ppb higher than during typical episodes. Overall, however, the average
concentration pattern closely resembles the typical regional episode condition.
The high concentrations during the OTAG domain episodes are not representative of high
concentrations that occur locally. A measure of modeling period representatives is a geographic
comparison with the highest ozone (90th percentile) concentrations. In the Pennsylvania-New Jersey
region, the modeling periods roughly correspond to the 90th percentile values. However, over the
Gulf Coast and Texas, the modeling episodes represent about the 60th percentile of ozone. (Figure
9B) Thus, the high ozone concentrations in the Gulf Coast belt and Texas are under-represented in
the OTAG domain modeling selection days, which should be kept in mind when using the episode
simulations for policy.
How do the transport conditions during the OTAG episodes compare to the
climatological average?
The transport conditions during the selected OTAG domain modeling episodes are similar to the
transport during OTAG domain-wide regional ozone episodes. (Figure 10A) Such episodes are
characterized by slow meandering transport over Kentucky, Tennessee, and West Virginia, with is
strong clockwise transport around this region of stagnation. However, the transport during OTAG
domain modeling periods differs significantly from the conditions of highest local ozone
concentrations. (Figure 10B) The main difference is that the net transport speeds are mostly higher
during the OTAG domain episodes, and the direction of net transport over Illinois-Pennsylvania
corridor is from the west, rather than the typical southwesterly flow.
16
Figure 9. A) Average daily maximum ozone concentrations during the four OTAG domain episodes. B) Difference between the OTAG
domain episodes and 90th percentile ozone concentration. The OTAG domain episodes under represent the ozone in the deep South.
Figure 10. A) Comparison of transport winds during the 1991, '93, and '95 OTAG domain episodes and the regionally high ozone
conditions. B) Comparison of transport winds during the three OTAG domain episodes and the locally high ozone conditions.
17
Air quality management implications of the data analysis results
Is ozone controllable by measures within the OTAG domain?
Ozone is indeed controllable by measures within the OTAG domain. The high ozone concentrations
in excess of the tropospheric background originate from within the OTAG region (Figure 11);
tropospheric background levels of ozone are found at nearly all of the borders of the OTAG domain
(Figure 12), with the exception of the Windsor-Quebec corridor. This means that the high ozone
concentrations are not due to external influences but contributed by sources internal to the OTAG
region. Furthermore, the high ozone concentration regions roughly coincide with the pattern of
anthropogenic precursor emissions as modified by atmospheric dispersion. Consequently, it can be
inferred that most of the excess ozone concentrations observed within the domain result from
anthropogenic emissions from within the domain.
Is there evidence that precursor emission changes cause ozone concentration
changes?
There is empirical evidence that anthropogenic emission changes do cause changes in ambient ozone
concentrations. The data show that there is a pronounced weekly cycle of ozone exceedances, with
the 1-hour 120 ppb exceedances on Sundays reduced by a factor of three compared to Friday
exceedances. This reduction is most pronounced in urban areas, while in the central portion of the
OTAG domain, the weekly ozone cycle is virtually nonexistent (Figure 12). Hence, any control
scenario that simulates the weekday-weekend emission changes would be effective in reducing the 1hour 120 ppb exceedances. It should be noted that the 8-hour 80 ppb exceedances show less weekly
fluctuation, indicating that such a control scenario would be less effective in reducing exceedances of
the new standard. While this natural "emission control" experiment provides interesting results, we
do not have good information on the actual emission changes that occur between weekdays and
weekends. Examination of long-term trends of ozone as well as NOx and VOC showed that the
largest ozone declines occurred over approximately the same geographic area where simultaneous
VOC and NOx reduction occurred. This provides further evidence, but not proof, that emission
reductions cause declines on ozone concentrations.
Do air quality data suggest high-leverage means for controlling ozone?
Based on spatial pattern, temporal pattern, and transport considerations, some general control
approaches appear to have higher leverage. Clearly, urban nonattainment areas contribute
significantly to their own ozone problems as well as to downwind areas within their area of
influence. Spatial pattern analysis indicates that the central area of the OTAG domain along the Ohio
River is chronically characterized by moderately high ozone levels nearly all the time (Figure 3).
Forward and backward trajectory analyses (Figure 11) and surface wind analyses implicate the
central portion of the OTAG domain for being involved in transport-related ozone events more than
any other portion of the domain, although the degree of impact cannot be quantified from this type of
analysis. Nonetheless, controls implemented in this area may be effective at reducing ozone transport
more often than controls anywhere else. Further, given the density of NOx-rich point sources in this
18
portion of the domain and the observation that ozone formation appears to be NOx-limited in nonurban areas, it follows that NOx controls may be more effective in this regard. It should be noted that
this suggestion is consistent with regional modeling results to date.
19
A
B
Figure 11a) Back trajectories for 22 receptor sites for low ozone days. These back trajectories point
to outside the OTAG region as the source of low ozone. b)Back trajectories for 22 receptor sites for
high ozone days (upper 50th percentiles). These back trajectories point to the OTAG region as the
source of high ozone.
Figure 12. Map of exceedances on Fridays and on Sundays. Throughout the OTAG domain the
Sunday 120 ppb exceedances are one third of the Friday values.
20
Do air quality data support the use of UAM-V model use to evaluate control
options?
From the point of view of AQA, the utility of modeling results as a basis of policy making can be
evaluated based on two criteria: model performance during the selected (36 day) episodes, and
secondly the representatives of these episode conditions for “typical” or climatological conditions.
The evaluation of representativeness indicates that these episode days well represent the conditions
for regional-scale episodes but not representative of high local ozone occurrences (Figures 9 and 10)
With regard to model performance, visual comparison of maps of measured ozone during modeled
episodes with model-predicted ozone values shows that the simulations capture the large-scale
features of each episode. Although not seen in every episode, there appear to be some tendencies for
the model to underpredict daily maximum ozone levels in the northern portion of the domain (Figure
14) where high ozone is frequently associated with strong, synoptic-scale flows, and overpredict the
same for the southern portion of the domain where high ozone is typically associated with local
stagnation. Further, aircraft measurements taken at selected locations for a few simulation days
indicate that ozone levels above the surface layer of the atmosphere may be underpredicted by the
model. One possible interpretation of these observations (although not the only one) is that transport
impacts may be understated by the model. This, and the generally longer scales of transport indicated
by the climatological air quality analyses should be kept in mind when interpreting model results
regarding transport distances. Also, transport impacts derived from any specific historical episodic
periods should not be taken as representative of the full range of meteorological flow conditions
which may be anticipated in the future.
Comparison of model predictions to ozone precursor data, while limited by the availability of
measurements, shows marginal to poor agreement, especially for isoprene from biogenic VOC
emissions. In addition, comparison of modeled and observed CO (a tracer of automotive emissions)
shows poor agreement for the 1995 episode. Comparison of ozone to total reactive nitrogen ratios in
non-urban locations tends to agree with model predictions.
Based on the ability of the model to capture the large-scale features of the ozone concentration
patterns and non-urban ozone to total reactive nitrogen ratios, it is suggested that the UAM-V model
is a useful tool for evaluating the general features of future large-scale control options. However, the
limitations of model applications, such as those suggested by these analyses, reinforce the notion that
modeling results should be interpreted not as definitive or absolute, but rather as one part of a full
assessment. The full assessment, relative to the mission of OTAG, would thus include modeling, air
quality analysis, and information relating to control strategy feasibility and cost.
21
Figure 14. A) Average daily maximum UAM-V model ozone concentrations during the four OTAG
episodes. B) Difference between the average UAM-V and measured ozone concentrations.
22
Recommendations to Foster Future Analyses
The infrastructure consisting of the community of analysts brought together by the OTAG process
and associated communications capabilities (regular meetings, conference calls, web sites, mailing
lists) has proven to be a valuable resource for integrating scientific research and policy-making;
efforts should clearly be made to maintain this infrastructure in the future.
Now that ozone transport has been identified as a significant problem by atmospheric scientists,
efforts are needed to address the lack of routine air quality and meteorological data suitable for
analysis of ozone formation and transport over a broad range of episode conditions. As the EPA
reviews its current monitoring networks and plans monitoring programs for the future, this
workgroup urges the Agency to consider enhancing the coverage and implementation of such
measurement programs. In particular, the current monitoring network is primarily focused on the
larger metropolitan areas whereas spatial coverage in less-populated areas or along boundaries of
political jurisdictions is quite limited. The current monitoring network is also primarily geared
toward monitoring ozone attainment status and trends. As well, measurements of ozone aloft or of
ozone precursors are quite limited. All these measurements are needed both for transport
assessments and to better evaluate the performance of photochemical models used to estimate control
strategy impacts. In particular, the PAMS network should be enhanced to provide the information
necessary for the evaluation of control strategy effectiveness.
Issues related to spatial representativeness and measurement methods (notably NOy) should be
addressed. All sites need not measure the identical suite of species, but all must collect data of
sufficient quality to support needs of future research efforts. Measurements at high elevation sites ridgetops and tall buildings - are particularly valuable and should be maintained and enhanced where
possible. Co-located measurements of ozone, precursors and fine particle composition are needed to
better understand pollutant interactions and source contributions.
Finally, because of the inherent scientific uncertainties in the air quality analysis and modeling, it is
clear that the most prudent course of action toward solving the ozone problem entails the
development of and commitment to a stakeholder-based air quality management process, such as that
being undertaken through the OTAG process, which promotes a continuous and measurable progress
toward attainment.
23