vvEPA
United States
Environmental Protection
Agency
Environmental Sciences Research
Laboratory
Research Triangle Park NC 27711
EPA-600/4-79-063
October 1979
Research and Development
Spatial Variability of
Ozone and Other
Pollutants at
St. Louis, Missouri
>
-------
RESEARCH REPORTING SERIES
Research reports of the Office of Research and Development, U S Environmental
Protection Agency, have been grouped into nine series These nine broad cate-
gories were established to facilitate further development and application of en-
vironmental technology Elimination of traditional grouping was consciously
planned to foster technology transfer and a maximum interface in related fields
The nine series arej
1 Environmental Health Effects Research
2 Environmental Protection Technology
3 Ecological Research
4 Environmental Monitoring
5 Socioeconomic Environmental Studies
6 Scientific and Technical Assessment Reports (STAR)
7 Interagency Energy-Environment Research and Development
8 "Special" Reports
9 Miscellaneous Reports
This report has been assigned to the ENVIRONMENTAL MONITORING series.
This series describes research conducted to develop new or improved methods
and instrumentation for the identification and quantification of environmental
pollutants at the lowest conceivably significant concentrations. It also includes
studies to determine the ambient concentrations of pollutants in the environment
and/or the variance of pollutants as a function of time or meteorological factors.
This document is available to the public through the National Technical Informa-
tion Service, Springfield, Virginia 22161
-------
EPA-600/4-79-063
October 1979
SPATIAL VARIABILITY OF OZONE AND OTHER POLLUTANTS
AT ST. LOUIS, MISSOURI
by
Thomas R. Karl
Meteorology and Assessment Division
Environmental Sciences Research Laboratory
Research Triangle Park, North Carolina 27711
ENVIRONMENTAL SCIENCES RESEARCH LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
RESEARCH TRIANGLE PARK, NORTH CAROLINA 27711
-------
DISCLAIMER
This report has been reviewed by the Office of Research and Development,
U.S. Environmental Protection Agency, and approved for publication. Mention
of trade names or commercial products does not constitute endorsement or
recommendation for use.
Mr. Karl is a meteorologist in the Meteorology and Assessment Division,
Environmental Sciences Research Laboratory, Environmental Research Center,
Research Triangle Park, N.C. 27711. He is on assignment from the National
Oceanic and Atmospheric Administration, U.S. Department of Commerce.
-------
ABSTRACT
A network of 25 aerometric stations was part of the U.S. Environmental
Protection Agency's Regional Air Pollution Study (RAPS) in the Greater
St. Louis area. At these stations ozone (Oo) and various other photochemical
pollutants (NO, NCU, and total hydrocarbons (THC)) as well as carbon
monoxide (CO) were analyzed with respect to their spatial variability.
Data were analyzed for the warm months of the year, April through October
of 1975 and 1976--periods during which high 03 concentrations are common.
The results of these analyses indicate that when 03 concentrations are
high (above 100 ppb) the daily 1-h maximum 0, concentration is highly
dependent upon the location of measurement. Measurements made at these
times in both urban and rural locations in and around St. Louis were
frequently found to vary by a factor of two and occasionally by a factor
as large as four. Simultaneous hourly average measurements for each of
the pollutants were correlated across the network of 25 stations.
Rather poor correlations were found for the primary pollutants such as
CO, NO, N02, and THC. Analyses of normalized fields of pollutant concentrations
suggest that the scavenging of 03 by NO dominates the 03-NOX reaction
cycle before 1000 CDT.
-------
CONTENTS
Abstract iii
Figures vi
Tables vii
1. Introduction 1
2. Conclusions 2
3. Data and Station Sites ' 3
4. Procedures and Results 4
Some elementary statistics on pollutant concentrations . 4
Maximum CU concentrations and the air flow 6
Hourly pollutant concentrations and the urban plume. . . 7
Cluster analysis of pollutant concentrations 9
5. Summary 14
References 15
-------
FIGURES
Number Page
1 Stations of the Regional Air Monitoring Systems 17
2 Annual total emissions for the St. Louis area. Diagrams on
the right show emissions for 5-km grid squares in the more
intense emission areas, while diagrams on the left are for
10- and 15-km grids covering the entire region. Emissions
are directly proportional to the area covered by each
asterisk; emissions are specified for the largest asterisk
in each of the six diagrams 18
3 One-H average 0, concentrations and the highest 1-h Oo concen-
tration (ppb) between 1300 and 1400 CDT for the study period. 19
4 Hourly average concentrations of THC (ppm x 10 )
CO (ppm x 10~]) NO (ppb), and N02 (ppb) during 0700-0800 CDT. 20
5 Percentage of the highest 03 concentration in the RAMS for
individual stations on days of downwind and upwind flow ... 21
6 Percentage of the highest 03 concentration in the RAMS for
individual stations on days of high wind speeds (>3.4 m/s)
and on days of low wind speeds (<2.5 m/s) 22
7a.- Normalized concentrations for various pollutants. 03, NO, and
7f. N02 concentrations are in ppb. THC and CO concentrations are
in ppm.x 10" . The zero isopleth is indicated on each map. 23
8a.- Diagram for cluster analysis of hourly pollutant concentra-
8f. tions and maximum 03 concentrations. The first correlation
coefficient inside each block represents the similarity of
the two clusters which are combined into a single cluster.
The second correlation coefficient (or first if the second
is absent) is the minimum correlation coefficient between any
two stations of the previously distinct clusters. The hour
for which each cluster analysis is presented is that for
which the network average concentration was highest 29
VI
-------
TABLES
Number Page
1 Percent of Daily Maximum 1-H 63 Concentrations at Each Station
That are Within Given Percentage Intervals of the Correspond-
ing Daily Highest 1-H 03 Concentration in the Network. N is
the Number of Days of Valid Data for the Stations When the
1-H 03 Concentration at Any RAMS Station Exceeded 100 PPB. . 32
2 Combinations of Stations Whose Concentrations are Most
Frequently Within 80-100% of the Highest Daily Maximum 1-H
03 Concentration in the Network (For Days With Maximum
Concentration Above 100 PPB). N is the Total of Days Above
100 PPB and n is the Number of Days Which Were Within
80-100% of These High Concentrations at the Indicated
Stations 33
vn
-------
SECTION 1
INTRODUCTION
Attainment of the National Ambient Air Quality Standard (NAAQS) for
ozone (Oo) is determined from measurements at fixed monitoring stations.
A number of reports have either directly or indirectly confronted the
problem of attaining the NAAQS for Do with missing or invalid measurements
123
at stationary sites ' . A somewhat similar, but more difficult
problem concerns the representativeness of point measurements. This can
also be regarded as a missing data problem, pertaining to space instead
of time.
The spatial variability of 0., and other pollutants affect the
interpretation of output from photochemical and other air quality models.
A question is raised regarding the representativeness of pollutant
concentrations spatially averaged over some grid scale with respect to
the actual concentrations observed in the atmosphere. Two other questions
then arise. How much do stationary point measurements vary in space in
and around metropolitan areas, and if mobile measurements cannot routinely
be made, where and how many fixed point measurements are necessary to
designate the area of highest 0^ concentrations? Answers to these questions
4
were recently pursued also by others , using more limited data but with
similar conclusions.
This study investigates the spatial variability of pollutant concen-
trations, Oo in particular, using data from the Regional Air Monitoring
System (RAMS) of the U.S. Environmental Protection Agency's Regional Air
Pollution Study (RAPS) in the St. Louis, Missouri area. Using these data
some answers are provided to the questions posed here.
-------
SECTION 2
CONCLUSIONS
Primary pollutants, such as THC, CO, NO, and N02, require a dense
network of stations in order to describe ambient concentrations across a
metropolitan area such as St. Louis. Assuming St. Louis is fairly
typical of many large cities, a much smaller number of carefully selected
sites than the 25 employed in the RAPS would adequately monitor the
maximum 0., concentrations in and around metropolitan areas. The site
selection process must include climatological information on the local
air flow during meteorological regimes that are favorable for high 03
concentrations. This information should be used in conjunction with
emission data to locate the stations. The results demonstrate that to
measure the regional maximum concentration of 0^, it would be misleading
to rely on 03 monitors located only in and around high emission areas.
From this study it is concluded that the grid-point output from
photochemical models should have sufficient resolution to resolve the
differences in 03 concentration between areas of both concentrated and
diffuse emissions or areas within and outside the urban plume. Otherwise,
there is little chance of replicating area-wide maximum 03 concentrations.
-------
SECTION 3
DATA AND STATION SITES
The RAMS was a network of 25 stations in the St. Louis area (Figure 1).
Wind speed, wind direction, 0~, carbon monoxide (CO), nitrogen dioxide
(N02), nitric oxide (NO), and total hydrocarbons (THC) were continuously
measured at each station. Wind speed and direction were measured atop
10-m or 30-m towers, while the effective measurement height for the
5
pollutants was 4 m . During the placement of these sites, domination
of a site by one or more local sources was avoided .
One-minute averages of the data were automatically recorded on
magnetic tape at RAPS headquarters. These averages were checked by
various computer programs for erroneous data . The data were screened
for continuity, relational consistency, and the operational mode of the
instruments. These data for the months of April through October of 1975
and 1976 were used to calculate hourly average quantities. A careful
visual inspection of the hourly averages helped assure against the
inclusion of fictitious data.
An extensive emission inventory for the Greater St. Louis area was
compiled as part of the RAPS-: Figure 2 illustrates the spatial variability
of total (area and point sources) annual emissions of THC, NO , and CO
A
in the St. Louis area. This diagram is based on the annual emissions
for 1975; however, there was no detectable difference when a similar
diagram was constructed using 1976 emissions. Variable grid spacing was
used to conduct the actual RAPS emission inventory. Grid squares were
smallest (1 km) in areas with concentrated emissions and largest (10 km)
in areas with less concentrated emissions. In order to visually display
emissions from these variable grid squares they were apportioned out
over a more uniform grid network. Grid squares of sizes 5, 10, and 15 km
are contained in Figure 2. Apportionments were also made for 2.5 grid
squares, but little additional detail could be found regarding the
spatial variability of the emissions.
3
-------
SECTION 4
PROCEDURES AND RESULTS
SOME ELEMENTARY STATISTICS ON POLLUTANT CONCENTRATIONS
Stations were checked each day for hourly average 03 concentrations in
excess of 100 ppb. When any of the 25 stations had concentrations this
high the maximum 03 concentrations for all the stations for the given day
were set aside for further study. A concentration of 100 ppb was considered
high because at the time of this research a new NAAQS for 03 was proposed
at such a level. A total of 154 days had at least one station exceeding an
03 concentration of 100 ppb. The daily highest 1-h 03 concentration at
each station was converted to a percentage of the network highest 1-h 03
concentration at any station for each of the 154 days, (this value will be
referred to as 03MAXP). For example, if station 107 had a 1-h 03 concen-
tration of 150 ppb on July 14, which was the highest 1-h 03 concentration in
the RAMS on that day, 100 would be the percentage of the highest 1-h 03
concentration for station 107 on July 14. All other stations would have a
smaller percentage associated with them for July 14, unless they had missing
hourly average 03 concentrations close to the time of the peak daily maxi-
mum 03 concentration. Stations were not used with missing hourly averages
within two hours of the estimated time of maximum 03 concentration.
Table 1 summarizes the results of the 154 days of high 03 concentrations
for each station. The variability in the daily maximum 03 concentration
from station to station is large. Ozone monitoring stations that were located
in and near intense emission areas (stations 101 through 107, 111, 112, 113,
119, and 120) were less frequently within 80-100 percent of the highest 1-h
03 in the RAMS than stations outside of these concentrated emission areas. On
the other hand, several stations located outside intense emission areas on a
number of days had very low maximum concentrations when compared to maximum
1-h 03 concentrations at other stations. This is indicated in Table 1 by
-------
the large percentage of time stations 117, 123, 124, and 125 were only
20-40 percent of the maximum 1-h 03 concentration. The fact that stations
outside and north of intense emission areas (station 114, 115, 121, and 122)
rarely (1 percent of the time) measured concentrations only 20-40 percent
of the maximum 1-h Oo concentration can be attributed to the fact that south-
erly winds in St. Louis are more frequently associated with high 0^ concentra-
9
tions than any other wind direction". This implies that stations north of
St. Louis were more frequently in the 'urban plume' during days of high 0-
concentrations than those to the south.
The 154-day data set was used in another way. The combination of
stations was determined, for a given number of stations, that had the greatest
number of days with at least one station's concentration within 80-100 per-
cent of the highest 1-h 0^ concentration in the network. An increase in the
given number of stations did not necessarily increase the number of days
with at least one station's concentration within 80-100 percent of the maximum
1-h 0~ concentration in the network. For instance, the combination of the
three stations 101, 102, and 103 had fewer days within 80-100 percent of
the highest 0^ concentration than did the single station 114. In this analysis.
the number of stations was increased until at least 95 percent of the high
03 days could be represented by a subset of the 25 RAMS sites. The process
was terminated at 95 percent due to the extensive amount of computer time
required to search through all possible combinations of stations.
The results of the search are contained in Table 2. This table illus-
trates the difficulty in measuring the highest 1-h Oo concentration in a
metropolitan area on days when 0~ concentrations are high. Even the most
successful combination of stations required six stations to come within
80-100 percent of the maximum 1-h 0^ concentration on 95 percent of the
days when the concentration in the RAMS exceeded 100 ppb. Notice that these
six stations were all located outside concentrated emission areas. This does
not imply that 0-, concentrations fail to reach high values inside the city.
Figure 3 reveals that the absolute maximum 1-h 0^ concentration in urban areas
is nearly as high as in rural areas. However the frequency of attaining these
high concentrations is considerably less in intense emission areas. This is
apparent by the lower average 1-h 0., concentrations at the urban sites.
-------
Figure 4 contains the 0700-0800 CDT hourly average concentrations of
CO, THC, NO, AND N02 for all stations in the RAMS. For all pollutants
other than 03, the highest average concentration occurred during 0700-0800 CDT.
Unlike the Figure 3 distribution of average 0., concentrations, Figure 4 shows
that for CO, THC, NO and N02, no station outside the intense emission
areas had higher 1-h average concentrations than stations inside these
areas.
MAXIMUM 03 CONCENTRATIONS AND THE AIR FLOW
At each station the 03MAXP value was calculated, and related to daily
network values of wind direction, speed, and steadiness. The daily network
wind speed was the arithmetic average of 24 individual hourly speeds, each
of which was the meJian speed for the 25 stations for .each hour. The daily
network wind direction was the resultant of 24 individual hourly directions,
each of which was the resultant direction for the 25 stations for each hour.
The steadiness of the wind was computed using the following expression:
S =
where V is the 24-h arithmetic average wind speed (i.e., the network speed
as described above), and \TR is the 24-h resultant wind speed (i.e., based on
the network direction described above). For each station (other than 101)
each resultant wind direction was categorized into an upwind or downwind
quadrant that was defined by the relative location of station 101 (the area
of high emissions). For example, in Figure 5 if the resultant wind direction
was between 310° and 40°, station 105 was considered downwind of station 101,
and if the wind was between 130° and 220°, it was classified as an upwind day.
The network wind speed was categorized as high (>3.4 m/s), medium (2.5-3.4 m/s),
or low (<2.5 m/s). These categories were determined so that each had an
equal number of days. At each station the 03MAXP value was calculated for
high and low wind speeds and for days of upwind or downwind flow.
-------
Figures 5 and 6 depict the relationships between the wind (direction and
speed, respectively) and the location of the 03 maximum in the RAMS. As the
distance from the intense emission area increases, there is a large difference
between upwind and downwind 03 concentrations, e.g., at the outer sites
compared to the inner sites (stations 114 through 125 versus stations 102
through 113). When wind speeds were high (Figure 6) a closer approximation
to the network maximum 03 concentration was generally observed at the
stations away from concentrated emission areas than at stations in and
close to these source areas. For example, on days with high wind speed
the percentages of the network highest CU concentrations typically are in
the 60's at stations in the vicinity of the high emissions area, but are
often in the 70's (some 80's) at more distant stations. Therefore, on
days with high CU concentrations (>100 ppb) and relatively strong wind
speeds the highest 03 concentrations are not likely to be found near
intense emission areas. For low wind speeds there is a better chance
that the highest Oo concentration in a metropolitan area is located
closer to the intense emission areas than for high wind speed days. A
specific example of this occurred on October 1 and 2, 1976. The lowest
wind speeds during the entire period of study were reported on these
days and the highest 03 concentrations were all in or near concentrated
emission areas.
The steadiness of the wind was also classified into three categories
and related to the 03MAXP values. The steadiness of the wind did not
appear to have much significance as an indicator of which type of site
would be representative of the maximum 03 concentration in the RAMS.
This was determined by inspection of various contingency tables and t-tests.
HOURLY POLLUTANT CONCENTRATIONS AND THE URBAN PLUME
The network median concentration for each of the five pollutants
THC, CO, NO, N02, and 03 was calculated for the four time periods, 0700-
1000-1100, 1300-1400, and 1600-1700 CDT, based on all available measurements.
These network median concentrations were then subtracted from the corres-
ponding 1-h average concentrations at each station and categorized
according to a network wind direction which was determined in the following
manner. On each day for each of the four 1-h periods of interest the
-------
network wind direction was classified into one of three sectors, 30°-150°
(E), 150°-270° (SSW), and 270-30° (NNW). For each hourly period of
interest, the network wind direction was based on up to 125 hourly wind
measurements (when none were missing), the hour of interest and the four
hours previous at each of the 25 RAMS sites. To be classified into one
of the three specified direction sectors 75 percent of the wind directions
had to fall within a single sector with a speed exceeding 1.5 m/s. For
each pollutant and time period of interest, the concentration deviations
from the network median were then averaged over all the days of identical
wind direction sectors.
The decision to use five hours of data and a minimum wind speed of
1.5 m/s was based en the travel time necessary for the pollution emitted
from the concentrated emission areas to reach the more distant RAMS
sites. The subtraction of the network median concentration for each of
the three sectors normalizes the concentrations of one sector with
respect to the concentrations of the two other sectors. This alleviates
the first major difficulty in understanding some of the more subtle
effects of the city on pollutant patterns. For example, due to the
weather regimes that affect St. Louis, (K concentrations over the entire
network are generally higher with southerly winds than with northerly
g
winds . Because of this it is not uncommon to observe higher 03 concentra-
tions at sites south of the city on days of upwind flow as opposed to
downwind flow.
A second subtraction (or normalization) was required in order to
normalize concentrations within a sector. An example of why this was
necessary can easily be given. Normally CO concentrations are much
higher in and near the city as opposed to stations farther removed.
When the network median CO concentration is subtracted from each station's
CO concentration, the urban stations consistently have concentrations
above the median, regardless of wind direction, and the rural sites
consistently have less than the median CO concentration. In order to
overcome this effect the deviation of a station's pollutant concentration
from the network median concentration, after being averaged for each one
of the three wind direction sectors, was subtracted by the average of
8
-------
the deviations from the network median concentration of the two other wind
sectors. The net result is a normalized field of concentrations repre-
senting the mean departure from the normalized concentration fields of the
two other wind sectors. These concentrations are plotted on Figure 7 for
the four time periods of interest.
Figure 7a depicts higher THC concentrations downwind of the city
for all time periods with winds from the SSE-W sector*. Figure 7b
illustrates that the urban plume concept, at least for mean conditions,
applies even for a pollutant such as CO. This pollutant, which has
considerable spatial variability, shows consistent patterns of higher
concentrations downwind of source areas. Figure 7c depicts higher
concentrations of NO downwind of the city during the early morning and
late afternoon, but during other times around mid-day such a pattern is
not readily apparent. This is at least partially attributable to the
very low NO concentrations, not infrequently near the noise level of the
instrument, that occur during these hours. Figure 7d is very much
similar to Figures 7a and 7b in that higher than normal NOo concentrations
occur downwind of the city's intense emissions areas throughout the day.
In Figure 7e, 03 concentrations were lower than normal downwind of the
city until approximately 1000 CDT, but after 1000 CDT, 03 concentrations
were higher than normal in areas downwind of the city. The lower 0^
concentrations that occur before 1000 CDT are attributed to the scavenging
of Oo by NO since patterns of NO and Oo were similar but opposite in
sign for the period 0700-0800 CDT. Apparently, after 1000 CDT the
energy of the solar radiation and the subsequent reaction time necessary
to provide 0-, are sufficient to overcome this phenomenon. The changeover
time from lower than normal Oo concentrations downwind of the city to
higher than normal concentrations during 1000-1100 CDT is confirmed by
Figure 7f for other wind directions.
* Figures 7a and 7e do not contain the pollutant patterns for the W-NNE
or the NNE-SSE wind sectors since these wind directions had nearly the
same type of patterns as did the SSE-W wind sector.
-------
CLUSTER ANALYSIS OF POLLUTANT CONCENTRATIONS
In order to understand the spatial variability of pollutant concen-
trations, it is essential to investigate correlation between concentrations
at various stations. These correlations can then be reconciled in a
systematic approach through the use of cluster analysis.
Correlation coefficients were calculated for each pollutant at each
pair of stations. The correlations were calculated for one of the four
hours 0700-0800, 1000-1100, 1300-1400, or 1600-1700 CDT, depending upon the
hour which had the highest network average concentration. In order to avoid
inflated correlation coefficients due to the yearly oscillation of pollutant
concentrations, the actual concentrations at each site were subtracted
by their seasonal n.eans before the correlations were run. Three seasons were
formed, consisting of the months April and May, June through August, and
September and October. Since the subtractions effectively removed the
yearly oscillation these deviations from seasonal means were then used to
compute the correlation coefficients.
Cluster analysis was applied to the matrix of correlation coefficients
produced by the interstation correlations. Cluster analysis is achieved by
combining stations into groups or clusters. Stations are combined into
a cluster based on the correlation coefficients between the various stations.
Initially, each station can be thought of as a cluster unto itself (a corre-
lation coefficient of unity). At each step in the clustering technique all
possible combinations of existing clusters are considered. Only the two
clusters which have the highest average correlation coefficients, based on
the station-to-station correlation coefficients, are combined into a new
cluster. Once a cluster is formed it cannot be broken apart. In this
manner clustering continues until all the stations are combined into one
cluster. Since there are 25 stations this process requires 24 steps to
complete.
Cluster analysis does not require any assumption regarding the statis-
tical significance of the correlation coefficients. The use of average
correlation coefficients can be justified because the clustering technique
initially combines stations into a cluster w'lich have "high" correlation
10
-------
coefficients and works down to "low" correlation coefficients by the final
step. In this manner averages of correlation coefficients are considered
for only the pairs of stations which have similar correlation coefficients
(i.e. high, medium, or low). The problem of averaging or combining
correlation coefficients which differ significantly is then sidestepped.
Figure 8 contains the results of cluster analysis for the various
pollutants. For 0^ the cluster analysis is also included for the maximum
daily 1-h 0-, concentration when at least one station reports a concentration
in excess of 100 ppb. Each of the 24 steps for each pollutant is contained
in Figure 8. A new cluster is formed at each step represented by the
stations included within the boundaries of the solid block containing corre-
lation coefficients. The first correlation coefficient is the average
correlation coefficient of all pairs of stations between the previously
distinct clusters. The second correlation coefficient listed is the minimum
correlation coefficient of any pair of stations between the previously dis-
tinct clusters.
Figure 8a is for the hour with the highest network average concentration
of CO. The figure reveals a poor correlation between the RAMS sites with
respect to CO concentrations. By the sixth step the similarities of any
two clusters is so low that the average variance between any two clusters
is less than 50 percent*. An inspection of the stations included in
the clusters indicates that they were formed on the basis of local emissions
around each station, the distance separating the stations, and their orien-
tation with respect to the central urban area of concentrated CO emissions.
Some clusters were created when there was a large disparity in the local
emissions surrounding specific stations in the clusters. This only occurred
when stations were separated by relatively short distances (see step 6,
Figure 8a). In other clusters (steps 2 and 16) the local CO emission around
the station and the orientation of the station with respect to major emission
* This calculation and subsequent calculations in this section regarding
the average variance explained by one cluster for another cluster were
calculated using the average of the square of the correlation coefficients
between stations. The square of the average correlation coefficients
shown in Figure 8 turns out to be a good approximation to the variance.
11
-------
areas was more important than the distance separating the station. The for-
mation of a cluster which is more difficult to explain occurs at step 18. It
is unclear why station 105, in an area of intense CO emission, does not cluster
with other nearby stations in concentrated emission areas before it clusters
with stations 104 and 120. One explanation is that station 120 is affected
more by a local interstate highway approximately 2 km to the west (Figure 1)
than by the large central area of high emission further east. This implies
that stations 104, 105, and 120 were oriented east of important emission
areas.
The emission pattern for THC is remarkably similar to the pattern for
CO, and cluster analyses for these two pollutants also have similarities.
For example, the cluster formed at step 11 in Figure 8b is almost identical
to the cluster formed at step 9 of Figure 8a Clusters formed at steps 7,
8, 12, and 16 of Figure 8b have identical counterparts on Figure 8a The
correlation between stations is slightly higher for THC than CO, but
it is also low by the final step. In fact, if a subset of stations from
the RAMS were desired to explain an average of 50 percent of the variance
in THC concentrations, no less than 16 stations would be required (see step 9,
Figure 8b).
Cluster analysis for NO and N02 (Figures 8d and 8e) also reveals poor
correlation coefficients, particularly for NO. A subset of the RAMS would
require 24 stations for NO and 16 stations for N02 in order to explain an
average of at least 50 percent of the day-to-day variance of the concentra-
tions at each station from the original network.
The correlations for 03 concentrations during 1300-1400 CDT are quite
high as indicated in Figure 8e. The clusters formed in this analysis are
closely tied to the orientation of the stations with repsect to major emis-
sions areas as well as the distance between stations. Clusters formed by
this secondary pollutant are highly related. Unlike the clusters formed
by the four other pollutants, these clusters suggest that stations are
representative of areas rather than points. In Figure 8e step 10 suggests
that a single station could be used to replace most of the stations in and
near the major source areas of urban St. Louis. Even in areas outside the
city there are a number of clusters with sufficiently high correlation
12
-------
coefficients such that a single station could be used in place of two,
three, or even four stations. However, in order to employ one station as
a surrogate for another station the average difference of the 03 concen-
trations between the stations during 1300-1400 CDT must be added to the
surrogate station's concentration during the same hour.
Figure 8f is important in light of the NAAQS emphasis on maximum
1-h 03 concentrations. Maximum 1-h 03 concentrations have higher corre-
lation coefficients between stations than average hourly correlations for
the other pollutants, but these correlations fall short in comparison to
average hourly concentrations of 03 during 1300-1400 CDT, the hour with
the network average highest concentration (see Figure 8e). The formation
of clusters in Figure 8f were closely related to the distance between
stations. This implies that when maximum 0-, concentrations were relatively
high (low) at specific stations an area of high (low) maximum 03 concen-
trations surrounded this site, but the correlation coefficients indicate
that this area was not large in size. A similar result is found in the
analysis of a few days of data during an 0., episode in St. Louis .
The information contained in Figure 8f can be used with Table 1 to
expand the list of stations in Table 2 that have concentrations 80-100
percent of the daily highest 1-h 03 concentration in the RAMS. The six
stations in Table 2; 114, 115, 118, 122, 124, and 125 are in distinct
clusters at step 18 in Figure 8f (initially each station is a cluster unto
itself). At step 17 station 109 is not included in any of the clusters with
these six stations. Table 1 indicates that station 109 has the highest
percentage of days within 80-100 percent of the maximum 1-h Oo concentration
of the 19 stations not listed in Table 2. The seven stations 109, 114, 115,
118, 122, 124, and 125 (i.e., station 109 plus the six stations in Table 2)
were within 80-100 percent of the maximum 03 concentration in the RAMS 97
percent of the time. Figure 8f indicates that at step 16, station 120 is
not included in any of the clusters of these seven stations. Furthermore,
in Table 1, station 120 has the highest percentage of days within 80-100
percent of the highest 1-h 03 concentrations of the remaining stations. With
this addition, at least one of the eight stations, 109, 114, 115, 118, 120,
124, and 125, was within 80-100 percent of the maximum 1-h 03 concentration
in the RAMS on each of the 154 days.
13
-------
SECTION 5
SUMMARY
Stations that monitored 0^ concentrations in and near intense emission
areas were less frequently within 80-100 percent of the highest 1-h 03
concentration in the RAMS on days when the concentration exceeded 100 ppb
than stations away from these areas. On the other hand, several stations
located south and upwind of the city had very low 1-h maximum Oo concentrations
compared with other stations. A total of eight stations, all located outside
of intense emission areas, could have been used in place of 25 to detect
(within 80-100 percentage) the 1-h network maximum 03 concentration when it
exceeded 100 ppb. Mean maximum concentrations of all the other pollutants
studied were located inside areas of intense emissions.
On days when the 1-h Oo concentration exceeded 100 ppb and relatively
strong wind speeds (>3.4 m/s) were observed, the highest Oo concentration
in the network was rarely located near intense emission areas. For low
wind speeds (<2.5 m/s) the highest 0^ concentrations were more often located
closer to the intense emission areas.
The scavenging of 03 by NO resulted in lower than normal 1-h average
0~ concentrations downwind of the city's intense emission areas before 1000
CDT. After 1000 CDT 03 concentrations downwind of the city averaged higher
than normal. The concept of an urban plume was verified for mean concen-
trations for all the primary pollutants, even for the highly variable pollu-
tant CO.
Station-to-station correlations of the primary pollutants was generally
lower than 0.70, except for some closely spaced stations, particularly those
in intense emission areas. Station-to-station correlations of 0^ were
significantly higher than those of the primary pollutants. In fact, a single
station frequently could be used in place of several nearby stations with
little loss of information. This was particularly true for the stations
located in and around the urban area of concentrated emissions.
14
-------
10.
REFERENCES
Federal Register (Thursday, June 22,1978). Photochemical Oxidants;
Measurement of Ozone in the Atmosphere; Requirements for Preparation,
Adoption, and Submittal of Implementation Plans. Vol. 43, No. 121.
Larsen, R. I. A Mathematical Model for Relating Air Quality Measurements
to Air Quality Standards. AP-89, U. S. Environmental Protection
Agency, Research Triangle Park, North Carolina. 1971. 56 pp.
Staff Summary Report. Alternate Forms of the Ambient Ai> Quality
Standard for Photochemical Oxidants. U.S. Environmental Protection
Agency. Office of Air Quality Planning and Standards, Research
Triangle Park, North Carolina. 1978. 20 pp.
McClenny, W. A. and L. W. Chaney. Pollutant Variability in the
Regional Air Pollution Study. J. Air Pollut. Control Assoc. 28,
693-696, 1978.
Myers, R. L. and Reagan, J. A. The Regional Air Monitoring System,
St. Louis, Missouri, U. S. A. Conf. on environmental sensing and
assessment proceedings: CAT, IEEE No. 75-CH 1004-ICESA, Las Vegas
Nevada. 1975. pp 8-6, 1-9.
Pooler, F. J. Network Requirements for the St. Louis RAPS. J. Air
Pollut. Control Assoc. 24, 228-231. 1974.
Jurgens, R. B. and Rhodes, R. C. Quality Assurance and Data Validation
for the Regional Air Monitoring System of the St. Louis Regional
Air Pollution Study. Conf. on environmental modeling and simulation
proceedings: EPA 600/9-76-016, Cincinnati, Ohio, 1976. 730-734.
Littman, F. E. Regional Air Pollution Study Emission Inventory
Summarization. U. S. Environmental Protection Agency, Research
Triangle Park, North Carolina. 1979. 92 pp.
Karl, T. R. and G. A. DeMarrais. Meteorological Conditions Conducive
to High Levels of Ozone. In: Report of International Conference on
Photochemical Oxidant and Its Control, EPA 600/3-77-OOla Sept. 1976
Research Triangle Park, North Carolina. 1977. pp. 75-88.
Brooks, C. P. and N. Carruthers. Handbook of Statistical Methods
in Meteorology. Her Majesty's Stationary Office, London, 1953.
pp. 222-224.
15
-------
11. Ludwig, F. L. and E. Shelar. Site Selection for the Monitoring of
Photochemical Air Pollutants. EPA-450/3-78-013, U. S. Environmental
Protection Agency, Research Triangle Park, North Carolina. 1978.
103 pp.
16
-------
REGIONAL AIR MONITORING STATIONS
5 10 15 20
109
EAST
.LOUIS
• 117
BELLEVILLE
124
Figure 1. Stations of the Regional Air Monitoring System.
17
-------
2.22 X 107 Kg/yr
THC
482 X 10° Kg/yr
THC
Figure 2. Annual total emissions for the St. Louis area. Diagrams on the right
show emissions for 5-km grid squares in the more intense emission areas, while
diagrams on the left are for 10- and 15-km girds covering the entire region.
Emissions are directly proportional to the area covered by each asterisk;
emissions are specified for the largest asterisk in each of the six diagrams.
18
-------
0)
o>
(D
-Q
Q.
Q.
c
O
to O
C 'C
O 0)
3 a
C 3
-------
20
-------
V)
CO
c
o
in
O
'+j
03
03
•
c
CD
.c
c
o
I
o
CJ
OJ
o
0) CO
S-o
in c
0= g
i= O
§,-°
ill's
21
-------
f 1
10 •
DS]
CO
TJ
C
o
en
c
g
'+-1
CO
4-1
V>
!i
o IT:
*+- CN
co V
DC
cu
cu
a
01
c o
If
el
tn C
0) CO
tl
c «
§ Q.
o «
fe-p
a. .E
22
-------
c
01
s»
c E
°.c
t; °
JO CD
£ o>
§ o
C T3
O
CO -C
o
. v>
n—
o o
O'-'
O X
> Q.
<"
c
o
TD c
CD O
"co O
£0
I"
•^ CO
cdCJ
r^I
QJ|-
.
U- Q.
23
-------
T3
c
O)
O)
cu
O)
CO
CD
24
-------
TJ
c
CO
OT
CD
CD
CD
CO
C/5
d
25
-------
c
o>
_OJ
OJ
l_
CD
tn
03
CO
CO
CO
26
-------
CO
o>
01
—
03
3
O)
0)
CO
0)
-------
to
c
cu
D)
(D
OJ
g>
LL
28
-------
071/0 68
0 68/0 66
O.S9/0.39
0.54/0.30
0.61/060
0.58
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
241 ~ 008/-0.08
I 101 | 107 | 113 | 103 | 106 | 111 | 112 I 119 | 102 1 114 | 108 | 109 | 110 | 115 I 116 | 117
052/051
Q.43/0.33
0.37/0.03
CO 0700 - 0800 COT
0 31/.0.06
0 23/-0.06
020/-0.02
0.13/-005
0.45
STEP
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
104 | 120 I 105 | 121 | 123 | 122 I 118 I 124 | 125
Figure 8a. Diagram for cluster analysis of hourly pollutant concentrations
and maximum 03 concentrations. The first correlation coefficient inside
each block represents the similarity of the two clusters which are com-
bined into a single cluster. The second correlation coefficient (or first
if the second is absent) is the minimum correlation coefficient between
any two stations of the previously distinct clusters. The hour for which
each cluster analysis is presented is that for which the network average
concentration was highest. See text for further details.
STEP
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
i STEP
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
0.82
0.73/0 68
0.66/0.50
0 62/0.55
0.63/0.62
057/0.53
0.40/0.08
THC 0700 - 0800 CDT
0,60/0.56
0.49/0.38
0.31/0.04
0.66
0.54/0.41
0.40/0.34
0.17/0.01
1 101 | 106 | 111 | 102 | 107 | 113] 112 | 119 | 103 | 104 | 105 | 120 | 118 | 116 | 117 | 108 | 114 | 121 | 122 | 109 [ 110 | 118 | 124 | 123 | 125 |
Figure 8b. (Same as figure legend 8a.).
29
-------
STATIONS
STEP
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
089 |
0 83/0 82
0 80/0 78
0.80/0 71
069
0 66/0 55
0 77
0 77/0 74
0 67/3 66
0 5//0 30
051/0
0 74
I
080
0 66/0 62
076
061/0 48
0 68
0 54/0 43
15
0 42/0 24
,
..
NO2 0700
069
0 35/0 03
0 19/003
0 19/-0 13
0800
:DT
035
STEP
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
| 101 | 107 [ 112 | 111 f 113 I 119 | 120 J 104 1 106 J 105 | 110 [ 102 | 103 | 108 | 114 | 115 | 116 | 109 | 117 | 123 | 121 | 122 | 125 | 118 | 124 |
Figure 8c. (Same as figure legend 8a.).
STEP
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
[ 077
0.68/0 62
0.58
0.55/0.50
0 64
0.52/0 38
0.60
0 47/0 34
058
0 45/0 34
058
0 53/0 50
i
i
|
0.65
0 53/0 51
0 42/0 30
045
0 36/0 29
0.30/0.01
i
066
NO
0.45/0 36
0700-
0 28/0.26
0.211-003
0800
0.20/0.03
CDT
0.09/-006
STEP
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
101 I 107 | 112 ] 102 [ 113 | 111 | 119 | 103 | 108 1 114 | 121 | 109 I 110 I 120 | 115 | 116 | 117 | 123 | 124 | 104 | 106 1 105 | 118 I 122 | 125
Figure 8d. (Same as figure legend 8a.).
30
-------
STEP
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
i '
| 0.92
0 90/0.90
0.93
0.89/0.88
0.88/037
0.89
0.89
0 85/0.81
083/0.80
1
i
i
0.85/0.84
082/071
I
I
I
i
31
0 70/0 61
300
0.68/0.61
0 68/0.58
111 | 119
112 | 120 | 113 | 109 | 116
123
117 | 102 | 108
114 | 115
124
125
400 C
DT
0.80
121 | 122
STEP
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
Figure 8e. (Same as figure legend 8a.).
STATIONS
STEP I
1
2
3
4
5
6
7
8
|
| 0.91
0.90/0.89
0.92
0 87/035
0.86/0.82
0.86
0.84/0.84
OBO/0 72
0.77
0 67/0.49
078
0 66/0 46
0.86
0.80/0 78
0.71/065
0.62/0 1 3
am
0.75/071
0.57/0.12
O
054/0 12
3 MAX
0.28/-0.01
IMUM
>10(
072
3ppb
026/0.12
0.22/-0 14
101 1 106
107 |
103 I 104
102
105 j 109
110
112
120
113
114
108 I 116
115
123
111
118
119 I 117
124
121
122
125
STEP
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
Figure 8f. (Same as figure legend 8a.).
31
-------
TABLE 1. PERCENT OF DAILY MAXIMUM 1-H 03 CONCENTRA-
TIONS AT EACH STATION THAT ARE WITHIN GIVEN PERCENT-
AGE INTERVALS OF THE CORRESPONDING DAILY HIGHEST
1-H 03 CONCENTRATION IN THE NETWORK. N IS THE
NUMBER OF DAYS OF VALID DATA FOR THE STATIONS WHEN
THE 1-H 0, CONCENTRATION AT ANY RAMS STATION
EXCEEDED 100 PPB.
STATION
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
N
123
132
105
119
123
107
128
130
136
131
110
117
136
128
94
120
122
115
139
88
119
137
66
113
95
PERCENTAGE OF NETWORK HIGHEST CONCENTRATION
0-20
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
20-40
2
3
4
4
6
5
5
4
3
2
4
1
2
1
1
2
7
2
2
0
1
1
5
9
8
40-60
19
25
20
45
22
11
36
24
14
12
27
24
24
11
7
22
30
8
18
11
12
15
35
32
33
60-80
54
51
60
44
41
59
52
45
46
56
57
53
53
30
52
52
45
47
44
54
44
65
31
34
36
80-100
25
21
16
7
11
25
7
27
37
30
12
22
21
58
40
24
18
43
24
35
43
49
29
25
23
32
-------
Table 2. Combinations of stations whose concen-
trations are most frequently within 80-100% of the
highest dailymaximum 1-H 03 concentration in the
network (for days with maximum concentration
above 100 ppb). N is the total number of days above
100 ppb and n is the number of days which were
within 80-100% of these high concentrations at
the indicated stations.
NUMBER OF
STATIONS
1
2
3
4
5
6
STATIONS
114
114,122
114,118,122
114,118,122,125
114, t15, 118, 122, 125
114,115,118,122,124,125
n
74
106
127
137
144
147
N
154
154
154
154
154
154
n/N, percent
48
69
82
89
94
95
33
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
1. REPORT NO.
EPA-600/4-79-063
2.
3. RECIPIENT'S ACCESSIOr*NO.
4. TITLE AND SUBTITLE
SPATIAL VARIABILITY OF OZONE AND OTHER POLLUTANTS
AT ST. LOUIS, MISSOURI
5. REPORT DATE
October 1979
6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
Thomas R. Karl
8. PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
10. PROGRAM ELEMENT NO.
(Same as Block 12.)
1AA603 AE-11 (FY-79)
11. CONTRACT/GRANT NO.
12. SPONSORING AGENCY NAME AND ADDRESS
Environmental Sciences Research Laboratory - RTF, NC
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
13. TYPE OF REPORT AND PERIOD COVERED
11/7R-fi/7q
14. SPONSORING AGENCY CODE
EPA/600/9
15. SUPPLEMENTARY NOTES
16. ABSTRACT
A network of 25 aerometric stations was part of the U.S. Environmental Pro-
tection Aqencv's Reaional Air Pollution Studv (RAPS) in the areater St. Louis area.
At these stations ozone (OJ and various other photochemical pollutants (NO, N0?,
and total hydrocarbons (THC)) as well as carbon monoxide (CO) were analyzed with
respect to their spatial variability. Data were analyzed for the warm months of
the year, April through October of 1975 and 1976—periods during which high 0.
concentrations are common. The results of these analyses indicate that when 0.
concentrations are high (above 100 ppb) the daily 1-h maximum 0- concentration is
highly dependent upon the location of measurement. Measurements made at these
times in both urban and rural locations in and around St. Louis were frequently
found to vary by a factor of two and occasionally by a factor as large as four
Simultaneous hourly average measurements for each of the pollutants were correlated
across the network of 25 stations. Rather poor correlations were found for the
primary pollutants such as CO, NO, N02, and THC. Analyses of normalized fields of
pollutant concentrations suggest that the scavenging of 0^ by NO dominates the
0.,-NO reaction cycle before 1000 CDT. 6
3 X
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.IDENTIFIERS/OPEN ENDED TERMS
COSATI Field/Group
* Air pollution
* Ozone
Nitrogen oxides
Hydrocarbons
* Spatial distribution
St. Louis, MO
13B
07B
07C
13. DISTRIBUTION STATEMENT
RELEASE TO PUBLIC
19. SECURITY CLASS (This Report)
UNCLASSIFIED
21. NO. OF PAGES
42
20 SECURITY CLASS (This page)
UNCLASSIFIED
22. PRICE
EPA Form 2220-1 (9-73)
34
------- |