United States
Environmental Protection
Agency
Office of Air Quality
Planning and Standards
Research Triangle Park NC 27711
EPA-450/4-79-013
March 1979
Air
Weekend/Weekday
Differences in
Oxidants and Their
Precursors
-------
EPA-450/4-79-013
Weekend/Weekday Differences
in Oxidants and Their Precursors
by
Yuji Horie, Joseph Cassmassi,
Larry Lai, and Louis Gurtowski
Technology Service Corporation
2811 Wilshire Boulevard
Santa Monica, California 90403
Contract No. 68-02-2595
EPA Project Officer: Gerald L. Gipson
Prepared for
U.S. ENVIRONMENTAL PROTECTION AGENCY
Office of Air, Noise, and Radiation
Office of Air Quality Planning and Standards
Research Triangle Park, North Carolina 27711
March 1979
-------
This report is issued by the U.S. Environmental Protection Agency
to report technical data of interest to a limited number of readers.
Copies are available free of charge to Federal employees, current
contractors and grantees, and nonprofit organizations in limited
quantities from the Library Services Office (MD-35), Research Triangle
Park, North Carolina 27711; or, for a fee, from the National Technical
Information Service, 5285 Port Royal Road, Springfield, Virginia 22161
This report was furnished to the Environmental Protection Agency by
Technology Service Corporation, 2811 Wilshire Boulevard, Santa Monica,
California 90403, in fulfillment of a contract. The contents of
this report are reproduced herein as received from Technology Service
Corporation. The opinions, findings and conclusions expressed are
those of the author and not necessarily those of the Environmental
Protection Agency. Mention of company or product names is not to
be considered an endorsement by the Environmental Protection Agency.
Publication No. EPA-450/4-79-013
-------
CONTENTS
FIGURES v
TABLES vii
Section Page
1. INTRODUCTION AND SUMMARY 1
1.1 OBJECTIVES OF THIS STUDY 2
1.2 SUMMARY OF CONCLUSIONS 3
2. SELECTION OF STUDY LOCATIONS AND DATA BASE PREPARATION . . 7
2.1 AIR QUALITY DATA 9
2.2 METEOROLOGICAL DATA 12
3. DIFFERENCES IN DAILY MAXIMUM OXIDANT LEVELS 16
3.1 SPATIAL VARIATION OF DAILY MAX OX LEVELS 16
3.2 WE/WD DIFFERENCES IN DAILY MAX OX LEVELS 21
4. METEOROLOGICAL ADJUSTMENT OF WE/WD OXIDANT LEVELS 33
4.1 WE/WD DIFFERENCES IN METEOROLOGY 34
4.2 WE/WD DIFFERENCE IN OX WITHIN METEOROLOGICAL
CLASSES 38
4.2.1 AID Decision Tree Analysis 39
4.2.2 Meteorological Classification 48
4.2.3 WE/WD Differences Within High-OX-Potential
Class 58
4.3 METEOROLOGICAL ADJUSTMENT OF WE/WD OXIDANT LEVELS . . 60
4.4 CLASSIFICATION OF SITES 67
5. DIFFERENCES IN OXIDANT PRECURSOR LEVELS 70
5.1 DIFFERENCES IN 6-9 A.M. AVERAGE NO AND NO,
CONCENTRATIONS * 70
5.1.1 WE/WD Differences in NO Concentrations .... 72
5.1.2 WE/WD Differences in N02 Concentrations ... 74
5.2 WE/WD DIFFERENCES IN N0?/N0v RATIOS 73
C, y\
m
-------
CONTENTS
Section Page
5.3 WE/WD DIFFERENCES IN 6-9 A.M. AVERAGE NMHC AND
THC CONCENTRATIONS 83
5.3.1 WE/WD Differences in 6-9 A.M. Average
NMHC Levels 85
5.3.2 WE/WD Differences in 6-9 A.M. Average
THC Levels 87
5.4 WE/WD DIFFERENCES IN CO CONCENTRATIONS 90
6. RESULTS AND CONCLUSIONS 96
6.1 OXIDANT-PRECURSOR RELATIONSHIP IN URBAN CORE AREAS ... 96
6.2 OXIDANT-PRECURSOR RELATIONSHIP IN DOWNWIND
AREAS 103
6.3 WE/WD OXIDANT CHANGE VS. LONG-TERM OXIDANT CHANGE ... 108
7. REFERENCES 110
APPENDICES
A. STUDENT T-TEST AND WELSH T-TEST 112
B. WILCOXON RANK SUM TEST 114
C. CONFIDENCE INTERVAL FOR THE MEDIAN 115
D. THE AID DECISION-TREE PROGRAM 116
E. BINARY DECISION TREES DEVELOPED FOR CLASSIFYING DAILY
MAXIMUM OX LEVEL BY METEOROLOGY 118
-------
FIGURES
Number Page
1 Map of the northeastern United States, showing the five large
metropolitan areas selected for the WE/WD air quality
analysis 8
2 Location of 22 monitoring stations used for the WE/WD oxidant
air quality analysis 11
3 Location of 22 air monitoring stations, 12 surface meteoro-
logical stations, 3 upper-air meteorological stations
used for the analysis 15
4 Mean weekday levels of daily maximum OX concentrations 17
5 Mean weekend levels of daily maximum OX concentrations 19
6 Percentage changes of daily maximum OX concentrations,
WE-WD (SS versus MTWTF) 20
7 Percentage changes of daily maximum OX concentrations,
WE-WD (Sun. versus TWTF) 22
8 Plot of P-values for testing the WE/WD difference in daily
maximum oxidant levels 24
9 Box-display of WE/WD daily maximum oxidant concentrations—
under normal and Sunday definitions—for the Newark,
New Jersey, site 26
10 Box-display of WE/WD daily maximum oxidant concentrations —
under normal and Sunday definitions—for the Medford,
Massachusetts, site. 28
11 Box-display of WE/WD daily maximum concentrations--
under normal and Sunday definitions—for the Greenwich,
Connecticut, site 30
'.'(. '•;-.->:-- -.nci;-iv- tree for the Newark, New JGrs~y, :"?t~' ^ 1-
13 Binary decision tree for the Mamaroneck, New York, station .... 44
-------
FIGURES
Number
14
15
16
17
18
19
20
Binary decision tree for the Salem, Massachusetts,
station
Preferred wind direction of high-oxidant air pollution
at each of the 22 sites
Example of Type I station, which exhibits a consistently
lower weekend OX level in every WE/WD comparison ....
Example of Type II station, which exhibits a consistently
higher weekend OX level in every WE/WD comparison ....
Example of Type III station, which does not exhibit a
consistent relationship among meteorological classes . .
P-values for testing WE/WD differences in 6-9 A.M.
average NO by t-test
Graphical presentation of t-test results for WE/WD dif-
Page
45
47
64
65
66
75
ference in 6-9 A.M. average NO concentrations (Sun.
vs. TWTF) 76
21 P-values for testing WE/WD difference in 6-9 A.M. average
N02 by t-test 79
22 Graphical presentation of t-test results for WE/WD difference
in 6-9 A.M. average NO concentrations (Sun. vs. TWTF) . . 80
23 P-values for testing the WE/WD difference in N0?/N0 ratio by
the Wilcoxon rank sum test (Sun. vs. TWTF) . 7 ...... 82
24 P-values for testing WE/WD differences in CO concentrations
during the 6-9 A.M., daytime, and 24-hr periods (SS vs.
MTWTF) 94
VI
-------
TABLES
Number Pa9e
1 Monitoring Stations Used for Analyzing Oxidant Air
Quality Data 10
2 Meteorologically Grouped Air Monitoring Stations 14
3 Significance of the WE/WD Difference in Daily Max OX
Level at Each Station by T-Test and Overall Judgment 31
4 Meteorological Station and Corresponding Air Monitoring
Station Used in WE/WD Meteorology Analysis 35
5 T-Test Results for Testing the WE/WD Differences in
Each of Seven Meteorological Variables 36
6 List of Meteorological Variables Used for Site-Specific
Meteorological Classification of Daily Maximum Oxidant
Concentration by AID 41
7 Qualitative Description of Meteorological Characteristics
for High, Middle, and Low Classes at Each Oxidant
Monitoring Site 49
8 Meteorological Characteristics of High-Oxidant Class
For Each Site 59
9 WE/WD Differences in Daily Max OX Levels for All Cases
and for High Met Class Only (SS vs. MTWTF) 61
10 Classification of WE/WD Differences 68
11 Monitoring Stations Used for Analyzing NO and N0? Air
Quality Data 71
12 WE/WD Differences in 6-9 A.M. Average NO Concentrations 73
13 WE/WD Differences in 6-9 A.M. Average N02 Concentrations . . . .' 77
14 WE/WD Differences in 6-9 A.M. Average (N02/NOX) Ratios 84
VII
-------
TABLES
Number Page
15 WE/WD Differences in 6-9 A.M. Average NMHC
Concentrations 86
16 Statistical Significance of WE/WD Differences in 6-9 A.M.
Average NMHC Concentrations 88
17 WE/WD Differences in 6-9 A.M. Average THC
Concentrations 89
18 Monitoring Stations Used for Analyzing Carbon Monoxide
Air Quality Data 91
19 WE/WD Differences in CO Concentrations During the 3-Hr
(6-9 A.M.), Daytime (6 A.M.-9 P.M.), and 24-Hr
Periods 92
20 WE/WD Means of Daily Maximum OX Concentrations and 6-9
A.M. Average Precursor and CO Concentrations at Urban
Core Sites (SS vs. MTWTF) 97
21 Percentage Changes (WE-WD) in Daily Maximum OX Concen-
trations and 6-9 A.M. Average Precursor and CO
Concentrations at Urban Core Sites (SS vs. MTWTF) ... 99
22 WE/WD Means of Daily Max OX and of 6-9 A.M. Average
Levels of NO, N02, and NMHC: SUN. Versus TWTF 101
23 Percentage Difference (WE - WD) in Daily Max OX and
in 6-9 A.M. Average Levels of NO, NO?, and NMHC:
Sun. Versus TWTF 102
24 Weekend Reductions in 6-9 A.M. Urban Precursor Levels and
Downwind Daily Max OX Levels (SS vs. MTWTF) 105
25 WE/WD Changes in Daily Maximum Oxidant Levels in Regions
Upwind of, Downwind of, and Within the Large
Metropolitan Areas (SS Vs. MTWTF) 107
viii
-------
1. INTRODUCTION AND SUMMARY
The National Ambient Air Quality Standard (NAAQS) for photochemical
oxidants (ozone) is not attained in many geographical areas, including Los Angeles,
Houston, and the northeast megalopolis from Maryland to Maine- These wide-
spread high-oxidant levels indicate the need for more comprehensive control
measures for oxidant precursors than the currently implemented emission
controls. To formulate a comprehensive and cost-effective control policy
requires a better understanding of how sensitive ambient levels of oxidants
are to changes in spatial patterns of precursor emissions, as well as to
changes in amounts of precursor emissions. A study of the weekend/weekday
(WE/WD) differences in oxidant and precursor concentrations would be useful
for gaining a better understanding of the precursor emissions/ambient oxidant
sensitivities.
The WE/WD differences in daily maximum oxidant levels have been examined
by a number of researchers. In 1974, two studies conducted for different
geographical regions (one for the New York-New Jersey region and the other for
the Los Angeles region) reached, coincidentally, the same conclusion: That
although primary-pollutant concentrations are measurably lower on weekends
than on weekdays, daily maximum oxidant levels are not necessarily lower on
weekends [California Air Resources Board (CARB) 1974; Cleveland et al. 1974a
and 1974b], These earlier findings were later revised when the WE/WD
differences were examined for a wider geographical area and for a larger
-------
data base. Recent studies all indicate that although the oxidant level may
be higher on weekends at some very-source-intensive sites, the weekend
oxidant levels are measurably lower in most areas [Levitt and Chock
1975; Cleveland and McRae 1977; Elkus and Wilson 1977; Horie et al. 1977].
The present study is performed for the northeast megalopolis, extending
from Washington, D.C., to Boston, Massachusetts, and uses large data bases
of air quality and meteorology. The spatial pattern of the WE/WD differences,
as well as that of weekday levels, is investigated for both oxidant and
precursor concentrations in order to gain some insight into control strate-
gies that are likely to be the most effective for reducing ambient levels
of oxidants.
1.1 OBJECTIVES OF THIS STUDY
Despite observed reductions of ambient precursor levels on
weekends, earlier studies [Cleveland et al. 1974a; CARB 1974; Elkus and
Wilson 1977; and Levitt and Chock 1975] have indicated that the WE/WD
differences in ambient oxidant levels are small and inconclusive. The
weekend levels are higher than the weekday levels at some monitoring sites,
while at other sites the weekend levels are lower or about the same. In
view of these findings, the so-called weekend effect raises intriguing
questions about the effectiveness of simultaneous reductions in hydrocarbons
and nitrogen oxides as a policy for controlling photochemical air pollution.
The objectives of this study are as follows:
1. To determine if there are significant differences between WE and
WD concentrations of precursors and oxidants;
-------
2. To determine the extent to which these differences may be
functions of meteorological variations;
3. To examine differences between WE/WD spatial and temporal patterns;
4. To assess whether further study is indicated to quantify the effec-
tiveness of reducing oxidant levels by reducing precursor levels; and
5. To gain greater insight into control strategies that are likely to
be the most effective for reducing oxidant levels.
To accomplish the above objectives, we compiled a large data base of air
quality and meteorological data, including ambient pollutant concentrations
at 22 sites, surface meteorological observations at 12 sites, and upper-air
meteorological measurements at 3 sites. The data analysis is focused on the
five summer months, May through September, during the years 1973 through 1976.
1.2 SUMMARY OF CONCLUSIONS
The paragraphs below summarize the findings and conclusions reached in
this study. For convenient reference, the summary is organized according
to the order of the sections.
Selection of Study Locations and Data Base Preparation (Section 2)
• Twenty-two air quality monitoring sites in and around each of the five
large metropolitan areas in the northeast United States were selected
for study. The five metropolitan areas are: Washington, D.C.,
Baltimore, Philadelphia, New York-Newark, and Boston.
• The 22 study locations represent a wide variety of site characteristics.
The study sites were selected from stations upwind of, downwind of,
and within the metropolitan area. However, no true rural sites are
included in the analysis because of inadequate oxidant data at those
s i tes.
• The 22 air quality monitoring sites are grouped so that sufficient
weather data for the analysis can be provided by 12 surface weather
stations and 3 upper-air weather stations.
-------
• All the oxidant and precursor-pollutant data, as well as the meteor-
ological data, were labeled as to the day of week as well as the day
of year. These labels were used to compile the data for weekends
and weekdays, separately.
• Two definitions of weekend/weekday are used: normal WE/WD defines
weekends as Saturdays and Sundays, and weekdays as Mondays through
Fridays; Sunday WE/WD defines weekends as Sundays only, and weekdays
as Tuesdays through Fridays.
Difference in Daily Maximum Oxidant Levels (Section 3)
• The WE/WD differences in daily maximum oxidant levels are examined
by applying two different statistical tests to the oxidant data:
the t-test, and a visual test of the box-displayed WE/WD oxidant
data. The final judgment on the WE/WD oxidant difference at each
site is made by considering the consistency of the two test
results.
• Of the 22 stations examined, only three stations—Lancaster,
Pennsylvania; Greenwich, Connecticut; and Danbury, Connecticut-
exhibit a consistently significant WE/WD difference by the two
different statistical tests. These stations all have a significantly
lower oxidant level on weekends.
t Only the Newark station exhibits a higher weekend oxidant level with mar-
ginal significance in the t-test. However, the box-display of the WE/WD
oxidant data has revealed that the t-test result is probably biased
by one extremely high oxidant value in the weekend data set.
t The percentage decrease in daily maximum oxidant levels on weekends
is generally higher at sites downwind of the large metropolitan areas
and is lower at sites within or upwind of the metropolitan areas.
0 The percentage change (WE-WD) tends to be larger when Sundays are
compared with Tuesday through Friday (Sunday WE/WD) than when Saturdays
and Sundays are compared with Monday through Friday (normal WE/WD).
Meteorological Adjustment of WE/WD Oxidant Levels (Section 4)
0 The WE/WD differences in meteorology are examined for seven meteor-
ological variables at 12 surface meteorological stations. Of the
seven variables, at least a marginally significant difference between
the WE/WD levels is found in five variables at two or more meteorological
stations.
-------
• The five variables having at least a marginally significant difference
are visibility, at six sites; surface pressure, at eight sites; max
temperature, at two sites; wind speed, at two sites; and mixing height,
at two sites.
• A pattern recognition computer program called AID was
applied to find the oxidant-meteorology relationship at each
air monitoring site, using weekday data only. The results
are expressed in a binary classification tree, in which the
relationship between daily max oxidant levels and meteorological
conditions can be easily seen.
• The most important meteorological variable to explain the daily max
oxidant levels is daily max temperature, which appears on 20 out of
the 22 binary classification trees developed. Other important
variables are visibility, on 16 trees; wind direction, on 14 trees;
mixing height and wind through the mixing layer, on 10 trees; wind
speed, on 8 trees; and surface pressure, on 5 trees. Average
temperature never appeared, probably because of the high col-
linearity with max temperature.
• Three site-specific meteorological classes, high, middle, and low,
were developed for each site by using the AID decision tree.
WE/WD differences in daily max oxidant levels were computed for
five cases: high, middle, and low meteorological classes, all
days, and all days with meteorological normalization. Then, each
site was categorized into one of three station types: stations
showing a consistently lower weekend level among 4 out of the 5 cases;
stations showing a consistently higher weekend level among 4 out
of the 5 cases; and stations showing np_ consistent WE/WD difference
among the 5 cases.
• The stations with a consistently lower weekend level among 4 out of
the 5 cases are found in the periphery of or outside the large
metropolitan areas; the stations with a consistently higher weekend
level are found within the metropolitan areas; and the stations
without a consistent WE/WD difference among the 5 cases are found
in the transitional zones.
Differences in Oxidant Precursor Levels (Section 5)
• WE/WD differences in 6-9 A.M. average concentrations of NO, N02, and
(N02/NOX) are examined for eight stations reporting an adequate amount
of NO and N02 data for the analysis. The eight stations are Bethesda,
Silver Spring, Suit!and, and Baltimore in Maryland; Camden and Newark
in New Jersey; Welfare Island and Mamaroneck in New York.
-------
• Five stations, Silver Spring, Baltimore, Camden, Newark, and
Mamaroneck, exhibit a significantly lower 6-9 A.M. average NO
level on weekends. One station, Bethesda, shows a higher weekend
NO level with marginal significance. The remaining two stations,
Suitland and Welfare Island, do not show a significant WE/WD
difference in 6-9 A.M. NO concentrations.
t For 6-9 A.M. average NOg concentrations, all eight stations examined
exhibit a significantly lower level on weekends than on weekdays.
• The WE/WD difference in 6-9 A.M. nonmethane hydrocarbons (NMHC) and
total hydrocarbons (THC) is examined for six stations. Among the
four stations examined for NMHC, the Baltimore and Suitland stations
exhibit a significantly lower weekend level under the Sunday WE/WD
definition. Under the normal WE/WD definition, all four stations
showed a significantly lower weekend level. The two stations,
Bethesda and Camden, examined for THC under the normal WE/WD
definition showed a significantly lower weekend level.
• The WE/WD difference in carbon monoxide (CO) was examined for 17
stations under the normal WE/WD definition. All but two stations
exhibit significantly lower 6-9 A.M. concentrations on weekends.
• The results of the analyses indicate that 6-9 A.M. primary-pollutant
concentrations of NO , NMHC, THC, and CO are indeed lower on weekends
than on weekdays.
Results and Conclusions(Section 6)
• The average oxidant improvement on weekends is 9% when Sundays are
compared with Tuesday through Friday (Sunday WE/WD), and 5% when
Saturdays and Sundays are compared with Monday through Friday
(normal WE/WD).
• A greater oxidant improvement (11% to 12%) is observed in areas
downwind of the large metropolitan areas than in the source-
intensive urban core areas.
• The oxidant improvement (5% to 12%) on weekends is small compared
with the 40% to 50% reductions in 6-9 A.M. urban NOX and NMHC levels
on weekends. However, the weekend reductions in daytime and 24-hr
precursor levels appear to be considerably smaller than those in
6-9 A.M. precursor levels.
• The small oxidant improvement on weekends, brought about by simul-
taneous reductions in NOX and NMHC, may be an indication that the
simultaneous control of NOX and NMHC is less effective for con-
trolling urban oxidants than are hydrocarbon controls.
-------
2. SELECTION OF STUDY LOCATIONS AND DATA BASE PREPARATION
The areas in this study encompass the entire northeast megalopolis from
Washington, D.C., to Boston, Massachusetts. Although cities and towns are
scattered throughout the 400-mile-long study region along the Atlantic Coast,
a great concentration of the population is found in the five large metro-
politan areas (Figure 1): Washington, D.C., Baltimore, Philadelphia, New
York-Newark, and Boston. According to the 1970 SMSA population summary
[U.S. Commerce Dept. 1973], the study region contains a population of well
over 26 million people (New York = 11.6, Philadelphia =4.8, Washington,
D.C. = 2.9, Boston = 2.8, Baltimore = 2.1, and Newark = 1.9 million people).
Ambient air monitoring data [U.S. EPA 1977] and special studies
[Ludwig and Skelar 1977; Wight et al. 1977; Wolff et al. 1977; Cleveland
1976] indicate high-oxidant air pollution over the study region. To study
the response characteristics of photochemical oxidants to changes in
precursors-pollutant levels between weekends and weekdays, we compiled a
large data base of air quality and meteorological data. The air quality data
include not only oxidant (OX) concentration data (measured as ozone), but
also concentration data of nitric oxide (NO), nitrogen dioxide (N02), oxides
of nitrogen (NOX), total hydrocarbons (THC), nonmethane hydrocarbons (NMHC),
and carbon monoxide (CO). The meteorological data include both surface
meteorological observations and upper-air meteorological observations.
Because high-oxidant air pollution occurs primarily in the summer season,
the air quality and meteorological data were compiled for the five warmer
months (May through September) of the years 1973 through 1976.
7
-------
NEWARK-
NEW YORK CITY
- State Boundary
Metropolitan Area
0
Miles
•4o
Figure 1. Map of the northeastern United States,
showing the five large metropolitan areas
selected for the WE/WD air quality analysis,
8
-------
2.1 AIR QUALITY DATA
After considering the data completeness of hourly oxidant and pre-
cursor concentration data during the study period, 22 oxidant monitoring
stations were selected from the areas in and around the five large metro-
politan areas (Figure 1), Table 1 provides the station name, the EPA's
SAROAD I.D. code, the site type, and the number of valid observation days
for each of the 22 stations used for the oxidant air quality analysis.
These stations were chosen such that the WE/WD differences could be
examined in areas upwind of, within, and downwind of each of the five
metropolitan areas. The geographical locations of these monitoring sta-
tions are shown in Figure 2 by the abbreviated station names; full station
names can be found in Table 1.
Precursor concentration data are less complete than the oxidant data.
As a result, fewer monitoring stations are used for the analysis of
precursor-pollutant data. Adequate NO and N02 data are found at eight
stations: Bethesda (BE), Silver Spring (SS), Suitland (SU), Baltimore (BA),
Camden (CA), Newark (NE), Welfare Island (WI), and Mamaroneck (MA) (see
Figure 2 for their location). NO data at these sites are computed by
X
summing NO and N02 concentration values. Adequate NMHC data are found at
four stations only: Silver Spring, Suitland, Baltimore, and Norristown.
Only two stations, Bethesda and Camden, are used for the analysis of THC
concentration data.
CO was also analyzed because in many cases it can be used as an
indicator of automotive-related pollutant levels.
-------
TABLE 1. MONITORING STATIONS USED FOR ANALYZING
OXIDANT AIR QUALITY DATA
Metropolitan
Area
Washington,
D.C.
Bal timore
Philadelphia
Newark -
New York
Boston
Station
Name
Bethesda (BE)
Gaithersburg (GA)
Silver Spring (SS)
Suitland (SU)
Baltimore (BA)
Lancaster (LA)
Camden (CA)
Camden Co. (CC)
Norristown (NO)
Philadelphia (PH)
Asbury Park (AP)
Babylon (BB)
Danbury (DA)
Greenwich (GR)
Mamaroneck (MA)
Middletown (MI)
Newark (NE)
Welfare Is. (WI)
Medford (ME)
Quincy (QU)
Salem (SA)
Worchester (WO)
SAROAD I.D.
210200005 F01
210780004 G01
211480006 F01
211560001 F01
210120018 F01
394660007 F01
310720003 F01
310740001 F01
396540013 F01
397140026 HOI
310060001 F01
330280002 F01
070175123 F01
070330004 F01
334100002 F01
070570003 F01
313480002 F01
334680050 F01
221220003 F01
221880002 F01
221980001 F01
222640012 F01
EPA's
Site Code*
S-R
S-R
S
S
CC-C
S-R
S-R
Ru-A
S-R
CC-C
CC-C
S-I
CC-R
Ru-NU
CC-C
CC-I
CC-C
CC
S-I
S-I
S-R
CC-C
Number of Valid
Observation Dayst
286
140
248
351
253
232
371
224
230
251
197
238
250
340
377
284
278
362
415
458
184
350
*Code key:
A - Agricultural
C - Commercial
CC - Center City
I - Industrial
R - Residential
Ru - Rural
S - Suburban
NU - Near-Urban
t
A "valid observation day" is defined as a day having at least 10 observations
for the period 6 A.M.4^ P.M., with the maximum hourly concentration not
exceeding 50 ppb after 9 P.M. and before 6 A.M.
10
-------
N
^Metropolitan Areas
0
25 50
75
Miles
NOTE: See Table 1 for station names.
Figure 2. Location of 22 monitoring stations used for the WE/WD
oxidant air quality analysis.
-------
Adequate CO data are found at as many as 17 monitoring stations among the
22 stations listed 1n Table 1. These 17 stations supplement the limited
area coverage of the precursor-pollutant data.
2.2 METEOROLOGICAL DATA
Two types of meteorological data were obtained from the National
Climatic Center: surface meteorological data and upper-air meteorological
data. From the surface meteorological data, daily values of the following
six surface meteorological variables were computed:
T - Daily Maximum Temperature
T - Daily Average Temperature
P - Dally Average Surface Pressure
V1s - Daytime (6 A.M.-9P.M.) Average Visibility
WD - Dally Vector Average Wind Direction
WS - Dally Arithmetic Average Wind Speed
From the upper-air meteorological data, two pertinent meteorological
variable values were extracted:
MH - Afternoon Mixing Height
V - Wind Speed Averaged Through the Mixing Layer
Most of these eight meteorological variables have been correlated
with high oxidant air pollution levels. For example, high oxidant levels
are usually observed on days with high temperature. The vector average
wind direction and arithmetic average wind speed were selected to indicate
the direction of pollutant air mass transport and the atmospheric dis-
persion conditions, respectively.
12
-------
Table 2 lists the three upper-air and 12 surface stations at which
the meteorological data used in this analysis were collected. The
geographical locations of the three upper-air meteorological stations, 12
surface meteorological stations, and 22 air monitoring stations are depicted
in Figure 3 with three different symbols. The name and I.D. number of
each station are listed in Table 2. The circles and dashed lines in the
figure indicate how each air monitoring site is grouped to a nearby surface
meteorological station and to one of the three upper-air meteorological
stations. The groupings of these stations are given in Table 2.
13
-------
TABLE 2. METEOROLOGICALLY GROUPED AIR
MONITORING STATIONS
Air Monitoring
Station
Medford
Qu i ncy
Salem
Worchester
Middletown
Danbury
Greenwich
Mamaroneck
Welfare Is.
Babylon
Newark
Asbury Park
Camden
Camden Co.
Norristown
Philadelphia
Baltimore
Bethesda
Gaithersburg
Silver Spring
Suitland
Lancaster
SAROAD I.D.
221220003 F01
221880002 F01
221980001 F01
222640012 F01
070570003 F01
070175123 F01
070330004 F01
334100002 F01
334680050 F01
330280002 F01
313480002 F01
310060001 F01
310720003 F01
310740001 F01
396540013 F01
397140026 HOI
210120018 F01
210200005 F01
210780004 G01
211480006 F01
211560001 F01
394660007 F01
Surface
Met. Station
Boston, Mass.
Worchester, Mass.
Windsor Locks,
Conn.
Bridgeport, Conn.
La Guardia, N.Y.C.
J. F. Kennedy,
N.Y.C.
Newark, N.J.
Atlantic City, N.J
Philadelphia, Pa.
Baltimore, Md.
Washington, D.C.
(National )
Harrisburg, Pa.
(Station
No.)
(14739)
(94746)
(14740)
(94702)
(14732)
(94789)
(14734)
.(93730)
(13739)
(93721)
(13743)
(14751)
Upper-Air
Met Station
(Station No.)
J.F. Kennedy
(94789)
Wash. , D.C.
(13743)
Albany, N.Y.
(14735)
14
-------
CJI
Air Monitoring Stations
Surface Meteorological
Stations
Upper-Air and Surface
Meteorological Stations
Figure 3, Location of 22 air monitoring stations, 12 surface meteorological
stations, and 3 upper-air meteorological stations used for the
analysis. [Several air monitS'Hng stations are grouped to a nearby^
surface meteorological station (in circle) which is, in turn, grouped
to one of the three upper-air meteorological stations (dashed line).]
-------
3. DIFFERENCES IN DAILY MAXIMUM OXIDANT LEVELS
This section discusses the spatial variation of daily maximum oxidant
levels and the WE/WD difference at each of the 22 monitoring sites selected
for the analysis. The oxidant data used for the analysis are daily maximum
1-hr oxidant concentrations during the five warmer months, May through
September, in the years of 1973 through 1976. In developing the data base,
both a data quality check for valid observation days and a data screening
were performed. A "valid observation day" is defined as a day having at
least 10 observations for the period 6 A.M.-9 P.M. (EOT) with the maximum
hourly concentration not exceeding 50 ppb before 6 A.M. and after 9 P.M.
(The 50 ppb threshold is used to eliminate potentially erroneous periods
of data since nighttime oxidant concentrations are usually less than the
threshold value. The station name, site code, and sample size of all the
22 stations are listed in Table 1. The geographical location of each
station can be found in Figure 2.
3.1 SPATIAL VARIATION OF DAILY MAX OX LEVELS
The average weekday level of daily maximum oxidant concentrations at
each station is plotted in Figure 4. The average daily maximum oxidant
levels do not vary much from one metropolitan area to another, except for
those of metropolitan Boston, which are somewhat lower than the others. The
regionwide oxidant level, as computed by the 22-station average, is 68 ppb.
16
-------
Regional Avg. 68 ppb
Metropolitan Areas
0 25 50 75
'' Miles' '
Figure 4, Mean weekday levels of daily maximum OX concentrations (weekdays as Monday
through Friday).
-------
The average weekend level of daily maximum oxidant concentrations at
each station is plotted in Figure 5. The regionwide average of weekend
oxidant levels (63 ppb) is about 7% lower than that of weekday oxidant
levels (68 ppb). Both figures show that the daily maximum oxidant levels
are higher at stations distant from the large metropolitan areas than at
those near or within the metropolitan areas—a phenomenon which indicates
the importance of considering the area downwind of a city when
formulating oxidant-control strategies.
A comparison of Figures 4 and 5 Indicates that, while the weekend
oxidant levels at stations away from the metropolitan areas are measure-
ably lower than the weekday levels, the same is not necessarily true for
stations within the metropolitan areas. Figure 6 shows the percentage
change in daily maximum OX concentrations at each station between weekends
and weekdays. The regionwide average improvement of oxidant air quality
on weekends 1s about 7%. The three stations in Connecticut exhibit a
greater percentage improvement (= 15%) than the regional average. The
greatest improvement (18%) of weekend oxidant air quality occurred at
Danbury, Connecticut. The five sites in Maryland exhibit a moderate im-
provement of about 9%.
In contrast to these improvements, the weekend oxidant air quality
at several urban core sites is slightly deteriorated. These urban core
sites are Newark, New Jersey; Welfare Island, New York; Philadelphia,
Pennsylvania; and Medford, Massachusetts. However, the weekend deteriora-
tion of oxidant air quality at the four sites is small and may be due to
mere sample biases of the weekend and the v/eekday oxidant data.
18
-------
Regional Avg. 63 ppb
Metropolitan Areas
0 25 50 75
'" Miles' '
Figure 5, Mean weekend levels of daily maximum OX concentrations (weekends as
Saturday and Sunday).
-------
PO
o
Regional Avg.
63 68
x 100- -
Metropolitan Areas
0 25 50 75
1 ' Miles' '
Figure 6. Percentage changes of daily maximum OX concentrations, WE-WD
(SS versus MTWTF).
-------
Figure 7 shows the percentage changes in daily maximum oxidant
levels when the oxidant levels on Sunday are compared with those on
Tuesday through Friday (TWTF). The percentage decrease on Sundays (Fig-
ure 7) appears to be generally greater than that on Saturdays and Sundays
combined (Figure 6). Indeed, the regional average percentage reduction
on weekends is increased slightly from 1% to 10% when Sundays are compared
with Tuesdays through Fridays.
The next subsection examines the significance of the WE/WD differences
depicted in Figures 6 and 7.
3.2 ME/KD DIFFERENCES IN DAILY MAX QX LEVELS
This subsection examines the significance of the WE/WD differences
observed in daily maximum oxidant data by applying several different
statistical methods to the original data. A weekend is normally defined
as Saturday and Sunday (SS) and a weekday as any day from Monday through
Friday (MTWTF). However, it has been noted that the oxidant behavior on
Saturdays and Mondays may be somewhat influenced by the carryover effect
from the preceding day. Therefore, statistical testing methods are also
applied to the data set of the other WE/WD definition that treats a week-
end as Sunday only (Sun.) and a weekday as any day from Tuesday through
Friday (TWTF). To distinguish between the two definitions, the SS week-
end and MTWTF weekday will be referred to as the normal WE/WD, and the
Sunday-only weekend and the TWTF weekday will be designated the Sunday
WE/WD.
21
-------
ro
ro
Regional Avg.
10%
•10%)
\
Metropolitan Areas
0 25 50 75
' ' Miles' '
Figure 7. Percentage changes of daily maximum OX concentrations, WE-WD (Sun. versus TWTF).
-------
The statistical t-test is employed to determine 1f the WE/WD dif-
ference observed in the mean daily maximum oxidant data is due to mere
chance (see Appendix A). The t-test is applied to the two data sets
derived from the normal and Sunday WE/WD definitions. The results of the
t-test are plotted in Figure 8 by using the P-value corresponding to
the value T of the t-test statistics. A small P-value, say, P _< 0.05,
indicates that the difference in the two means (WE and WD oxidant
levels) is real and significant. A moderate P-value, 0.05 < P <_ 0.10,
implies that the difference between the WE/WD levels is likely to be
real but yet may be due to mere chance. The larger the P-value, the
more likely that the observed WE/WD difference is caused by mere chance
and thus is not significant.
Under the normal WE/WD definition, four stations—Lancaster, Pennsylvania;
Camden County, New Jersey; Danbury, Connecticut; and Greenwich, Connecticut--
exhibit a statistically significant difference between the WE and WD daily
maximum oxidant levels (Figure 8). Under the Sunday definition, three
stations, Danbury, Greenwich, and Quincy, show a significant difference. All
these stations have considerably lower oxidant levels on weekends than on week-
days. Several other stations show a marginally significant WE/WD difference in
daily maximum oxidant under either the normal or the Sunday definition. Among
them, only the Newark station exhibits a higher weekend oxidant level.
However, the weekend levels at the Newark station are only significantly
higher than the weekday levels under the normal WE/WD definition.
23
-------
P-Value
1 * U
0.5
0.1
0.05
0.01
0.005
0.001
—
*
•
•
if if
•
*
•
V
• • •
*
A
^— —
*
V*'«AV^AAJ
*
» •
• •
*
*
•
*
*
. • •
•
** '
(Insic
RE
(Marginally
Re<
(Sign
Reg
«* . W *A>>*A, W. '^
•
•
r * *
nificant
gion)
Significant
ion)
ficant
on)
1 1
•&,.**. ^
\
• --Under the Sunday WE/WD Definition
* --Under the normal WE/WD Definition
Figure 8. Plot of P-values for testing the WE/WD difference
in daily maximum oxidant levels.
24
-------
Note that in most cases, the P-values are larger for the Sunday
WE/WD definition than for the normal definition, implying that the Sunday
WE/WD differences are less significant than the normal WE/WD differences.
However, these larger P-values are primarily due to the smaller sample size
associated with the Sunday WE/WD data set. As can be seen from Figures 6
and 7, the percentage difference between weekends and weekdays is greater
under the Sunday WE/WD definition for most of the stations.
To confirm the t-test results, we employed a visual test comparing
box-displayed data, seeking consistency of the two results. The box-display
presents, graphically, several pertinent statistical parameters of the data:
the sample size, median value, 25 and 75 percentiles, first- and second-
highest values, and lowest value. The box-display of data is more descrip-
tive than the usual presentation of mean, standard deviation, and t-statis-
tics. Also, the median, which is less affected by outlying data points
than the mean, provides a more robust means of comparing wE/WD differences.
First, the box-display method is applied to the stations that
exhibit a higher oxidant level on weekends under at least one of the
two WE/WD definitions. These stations are Newark, Welfare Island,
Philadelphia, Medford, and Asbury Park (see Figures 6 and 7). Figure 9
demonstrates the box-display for the WE/WD daily maximum oxidant
concentrations at Newark, New Jersey. Although the Newark station
25
-------
O Highest
(243 ppb)
160
150
130
110
90
70
50
30
10
1
-
-
-
-
-
-
-
•
_
-
-
O
o
A
SUN
(150)
O
o
v 95%
VConfl-
' dence
Interval
• \
/ 95%
- \ Confi-
i dence
' Inter-
val
A
TWTF
(43)
O
0
A
ss
(186)
75 p-
tlle
HD
^-Mean
median
?•; n-
" H
tile
o 2nd Highest
p*— HE Mean
A
MTWTF
(88)
Figure 9. Box-display of WE/WD daily maximum oxidant
concentrations—under normal and Sunday defi-
nitions—for the Newark, New Jersey, site.
26
-------
exhibits a higher weekend oxidant level with marginal significance
under the normal WE/WD definition, the box-display of the same data does
not show any large difference in the medians and the 25 and 75 per-
centiles. In Figure 9, the confidence interval of the median is also
shown (see Appendix C). The confidence interval of the weekday data
is completely overlapped by that of the weekend data, indicating that no
significant difference exists between the weekend and weekday oxidant
levels.
One reason why the t-test result has indicated a marginally
significant difference between the WE/WD oxidant levels under the normal
WE/WD definition can be found in Figure 9. As seen from the box-display
for the norma1 WE/WD data, one extreme value has significantly affected
the mean of the weekend oxidant levels. The mean of the weekend oxidant
levels was re-computed without this high value (243 ppb). As a result,
the P-value increases to 0.15, indicating no significant difference between
the WE/WD oxidant levels.
In addition to the Newark station, four other stations, Philadelphia,
Welfare Island, Medford, and Asbury Park, also exhibit a slightly higher
oxidant level on weekends than on weekdays (see Figures 6 and 7). Comparison
of the medians indicates that almost no difference exists between the WE/WD
oxidant levels. For example, Figure 10 presents the box-display of the
WE/WD oxidant data at Medford, Massachusetts. The WE/WD median values
are almost identical, and the WE confidence interval overlaps that of the
WD.
27
-------
ro
CO
.a
a.
a.
100
90
70
50
30
10
0
OHighest O
92nd Hiohest 9
[-H75 p-tile
<- WE Mean
Median
25 p-tile
O
° Lowest
WD Mean
Median
SUN.
(70)
TWTF
(261)
WE Mean->
95*, Con-
fidence
Interval
I
SS
(149)
Box-Display of UE/WD Daily Max 0.. Concentrations at Mndford Site.
O Highest
9 2nd Hiqhest
75 n-tile
•WD Mean
95% Confi-
dence> Intent]
25 p-tile
OLowest
MTWTF
(316)
Figure 10. Box-display of WE/WD daily maximum oxidant concentrations—under
normal and Sunday definitions—for the Medford, Massachusetts, site.
-------
Next, the box-display method was applied to the stations that exhibit
a lower oxidant level on weekends in both Figures 6 and 7. Figure 11 shows
the box-display of the WE/WD oxidant data at Greenwich, Connecticut, where
the t-test result exhibits significantly lower weekend oxidant levels under
the normal and the Sunday WE/WD definitions (see Figure 7). The box-
display shows that the median of the weekend oxidant levels is significantly
lower than that of the weekday levels under both the normal and Sunday WE/WD
definitions. Furthermore, the confidence intervals of the two medians are
completely separated from each other, indicating that a significant difference
exists between the weekend and weekday oxidant levels.
Similarly, a clear separation of the confidence intervals of the two
medians is observed in the box-displayed data for the WE/WD daily maximum
oxidant levels at the Danbury and Lancaster stations. At the Camden
County station, such clear separation occurs only in the SS vs. MTWTF
comparison. The box-displayed data for Bethesda, Silver Spring, Suit!and,
Norristown, Middletown, and Quincy all failed to show a significant difference,
although the t-test results for those stations showed at least a marginally
significant difference under one of the two WE/WD definitions. The rest of
the stations do not exhibit a significant difference in either the t-test
results or the box-displayed data.
The results of the significance test for the WE/WD difference in daily
maximum oxidant levels are summarized in Table 3. The t-test results are
categorized by three levels: significant, marginally significant, and
29
-------
o
>.
p Highest
i
i
A 2nd Highest
" -WE Mean
median
s
o
A A
TWTF SUN.
(230) (93) (189)
(44)
Figure 11. Box-display of WE/Wu daily maximum concentrations-
under normal and Sunday definitions—for the
Greenwich, Connecticut, site.
30
-------
TABLE 3. SIGNIFICANCE OF THE WE/WD DIFFERENCE IN DAILY
MAX OX LEVEL AT EACH STATION BY T-TEST AND
OVERALL JUDGMENT
Metropolitan
Area
Washington, D.C.
Baltimore
Philadelphia
New York -
Newark
Boston
Station
Name
Bethesda (BE)
Gaithersburg (GA)
Silver Spring (SS)
Suitland (SU)
Baltimore (BA)
Lancaster (LA)
Camden (CA)
Camden Co. (CC)
Norn's town (NO)
Philadelphia (PH)
Asbury Park (AP)
Babylon (BB)
Danbury (DA)
Greenwich (GR)
Middletown (MI)
Mamaroneck (MA)
Newark (NE)
Welfare Is. (WI)
Medford (ME)
Quincy (QU)
Salem (SA)
Worchester (WO)
SS vs. MTWTF
t-Test
0
0
0
e
0
0
0
0
0
0
0
9
e
0
0
0
0
0
Overal }
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
Sun. vs. TWTF
t-Test
0
0
0
0
0
e
0
0
0
0
0
0
0
0
0
0
0
0
0
Overall
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
Note
U 9
WE level is significantly higher than WD level
WE level is significantly lower than WD level
WE and WD levels have no significant difference
WE/WD differences are marginally significant
31
-------
insignificant. The overall judgment on the WE/WD difference is also
made by considering the consistency of test results by the two different
tests employed. Because the overall judgment involves a subjective judg-
ment (particularly in the case of the box-display results), the overall
test results are categorized by two levels only: significant and
insignificant.
The overall test results are as follows:
0 Under the normal WE/WD definition (SS versus MTWTF), four
stations exhibit a significant WE/WD difference in daily
maximum oxidant levels; the other 18 stations do not
exhibit a significant WE/WD difference. All four stations,
Lancaster, Camden County, Danbury, and Greenwich, have a
significantly lower oxidant level on weekends.
t Under the Sunday WE/WD definition (Sun. versus TWTF), only
three stations exhibit a significant WE/WD difference in
daily maximum oxidant levels. All three stations, Lancaster,
Danbury, and Greenwich, exhibit a significantly lower
oxidant level on weekends.
In the next section, WE/WD differences in oxidant levels are examined
further, taking into account the effect of meteorology.
32
-------
4. METEOROLOGICAL ADJUSTMENT OF WE/WD OXIDANT LEVELS
In the preceding sections, we have analyzed weekend/weekday (WE/WD)
differences in oxidant concentrations, using all hourly data during the
May 1 through September 30 period in four years, 1973 through 1976. Given
the long period of air quality data, the results of such an analysis should
have only a minimal meteorological bias occurring between weekdays and
weekends [Peterson and Flowers 1977; Cleveland et al. 1974a]. Nevertheless,
there exists a possibility that a meteorological bias occurring between
weekdays and weekends may have affected the WE/WD oxidant differences
examined in the preceding sections.
In Section 4.1, seven meteorological variables usually correlated
with high oxidant levels are examined for WE/WD differences in order to
detect a potential bias on either weekends or weekdays. The impact that
any meteorological differences might have on oxidant WE/WD differences are
assessed. In Section 4.2, WE/WD oxfdant concentrations occurring under
similar meteorological conditions are compared. Here, stratifying the oxi-
dant data by meteorological class allows examination of the WE/WD dif-
ferences on only those days with high oxidant potential. It is under these
conditions that reductions in HC and NO levels, are likely to be most im-
A
portant. This procedure also removes any masking effect that might occur
by including days with meteorological conditions that are not conducive to
ozone formation. In Section 4.3, the meteorological classes are used to
normalize oxidant levels for meteorology, and these results are compared
to WE/WD differences using the other schemes. Finally, each site is cate-
gorized according to the oxidant WE/WD differences observed under the
vario'js analysis
33
-------
4.1 WE/WD DIFFERENCES IN METEOROLOGY
The WE/WD difference in meteorology is examined for each of the 12
surface weather stations whose names and site codes are listed in Table 2.
The WE/WD differences in each of seven meteorological variables are com-
puted for each surface weather station. Because it is more pertinent to
find the WE/WD difference on sampling days with oxidant data rather than
on all days, each of the 12 meteorological stations is paired to one of
the nearby air monitoring stations. The paired air monitoring station and
the number of sampling days at that station are listed in Table 4. The
number of sampling days is gfven separately for weekends and for weekdays.
Table 5 lists the WE/WD levels and the significance level of the WE/
WD difference in each of seven meteorological variables. Surprisingly,
WE/WD differences are found in nearly all the meteorological variables
examined. Some variables show a significant WE/WD difference at several
sites, while other variables show a consistent WE/WD difference (in
values) among all the sites examined.
The daytime average visibility is consistently better on weekends
at all 12 sites examined. At LaGuardia, JFK, Newark, and Philadelphia,
the WE/WD differences in visibility are statistically significant. Two
other sites, Washington, D.C., and Baltimore, Maryland, exhibit a
marginally significant difference. This consistently better visibility
on weekends may be due to a decrease in visibility-reducing pollutants
on weekends [Peterson and Flowers 1977].
The surface pressure 1s consistently lower on weekends than on week-
days at all 12 sites. Four sites—Windsor Locks, Connecticut; Bridgeport,
34
-------
TABLE 4. METEOROLOGICAL STATION AND CORRESPONDING AIR MONITORING
STATION USED IN WE/WD METEOROLOGY ANALYSIS
Meteorological Station
Air Monitoring Station
#Valid Cases WE/WD
Boston, Mass.
Worchester, Mass.
Windsor Locks, Conn.
Bridgeport, Conn.
LaGuardia, N.Y.
JFK, New York
Newark, N.J.
Atlantic City, N.J.
Philadelphia, Pa.
Washington, D.C.
Baltimore, Md.
Harrisburg, Pa.
Medford, Mass.
Worchester, Mass.
Middletown, Conn.
Greenwich, Conn.
Welfare Is., N.Y.
Babylon, N.Y.
Newark, N.J.
Asbury Park, N.J.
Camden, N.J.
Suit!and, Md.
Baltimore, Md.
Lancaster, Pa.
148/368
129/314
99/244
114/283
126/305
88/211
88/214
92/234
108/278
113/301
80/219
84/195
The number of normal weekend days and weekdays with adequate oxidant air
quality data.
35
-------
TABLE 5. T-TEST RESULTS FOR TESTING THE WE/WD DIFFERENCES IN EACH
OF SEVEN METEOROLOGICAL VARIABLES
Surface Weather
Station
Boston, Mass.
Uorchester.Mass.
Windsor Locks,
Com.
Bridgeport,
Conn.
La Guardla.
N.Y.
JFK, N.Y.
Newark, N.J.
Atlantic City,
N.J.
Philadelphia.
Pa.
Washington,
D.C.
Baltimore, Md.
Harrlsburg, Pa.
Max Twp, »F
WE/WD Olff.
73.7/74.2 0
73.7/74.0 0
75.2/75.8 0
75.2/75.5 0
75.6/76.4 0
74.2/75.2 0
78.1/78.7 0
77.3/78.4 0
78.6/79.9 0
81.6/82.0 0
78.0/79.7 0
77.7/79.0 0
Avg. Teap, °F
WE/WO Dlff.
65.7/66.5 0
65.7/66.2 0
68.0/68.9 0
68.3/68.5 0
68.7/69.5 0
67.4/68.0 0
70.7/71.4 0
68.6/69.2 0
70.4/71.1 0
73.3/73.7 0
69.6/70.6 0
69.0/69.7 0
Avg.V1s1b1l1ty.Ml
WE/WD Dlff.
11.2/10.7 0
11.4/10.9 0
10.5/10.2 0
10.7/10.6 0
9.9/9.0 +
11.1/9.6 +
11.3/9.6 +
7.7/7.4 0
10.5/9.0 +
11.2/10.4 •
10.0/9.0 6
9.0/8.3 0
Avg. Wind Speed, kn
WE/WD Dlff.
8.3/8.5 0
8.7/9.1
8.5/8.2 0
8.7/8.4 0
8.0/8.0 0
9.1/9.1 0
7.4/7.2 0
7.6/7.3 0
7.5/7.3 0
6.4/6.6 0
6.2/6.2 0
5.2/4.8 «
Nixing Height, m
WE/WD Dlff.
1044.3/998.6 0
1088.5/1059.6 0
1112.3/1067.7 0
1069.8/1069.8 0
1056.4/1018.5 0
1040.1/1099.2 0
1123.2/1038.7 0
1064.0/1029.0 0
1303.6/1363.4 0
1295.6/1389.3 0
1228.2/1382.2 -
1489.5/1381.5 0
Wind Thru M.L.,n/s
WE/WD Dlff.
6.1/6.1 0
6.1/6.2 0
5.0/5.9 0
6.1/6.0 0
6.1/6.0 0
6.1/6.2 0
6.2/6.1 0
6.0/6.0 0
5.0/5.1 0
5.1/5.1 0
5.1/5.0 0
6.5/6.0 •
Avg. Pressure, mb
HE/WD Dlff.
1015.6/1016.2 0
1015.4/1016.2 9
1015.6/1016.8 -
1015.9/1017.2 -
1015.8/1016.7 0
1016.1/1017.0 0
1015.3/1016.1 0
1016.3/1017.3 0
1015.7/1016.7 -
1015.9/1017.1 -
1016.3/1017.1 0
1016.5/1017.3 0
00
WE level Is significantly lower than WD level
WE level Is significantly higher than WD level
WE/WD difference Is not significant
WE/WD difference Is Marginally significant
-------
Connecticut; Philadelphia, Pennsylvania; and Washington, D.C.—exhibit
a significantly lower pressure level on weekends. Marginally significant
weekend drops in surface pressure are found at four additional sites:
Worchester, Massachusetts; LaGuardia, New York; JFK, New York; and Newark,
New Jersey. These consistent pressure decreases on weekends are rather
difficult to explain. A more detailed study of the WE/WD meteorological
difference appears to be needed.
A consistently lower weekend level is found in both the daily
maximum temperature and the daily average temperature at all the 12 sites.
A marginally significant difference in daily max temperature is found at
the Philadelphia and Baltimore sites. The WE/WD differences in daily max
temperature are not significant at the rest of the 12 sites. The WE/WD
differences in daily average temperature are found to be not statistically
significant at all the sites. However, because of the consistent bias
and because of the known high correlation with daily maximum oxidant levels
[Breiman and Meisel 1976], the temperature bias between weekends and
weekdays may have affected the WF/WD differences in daily maximum oxidant
levels, discussed in the preceding section (Section 3).
The daily average wind speed, afternoon mixing height, and mean wind
through the mixing layer also exhibit some difference between weekends and
weekdays. However, the WE/WD differences are not consistent and change
randomly among the 12 sites. Therefore, we do not expect any significant
underlying cause for the WE/WD differences in these meteorological variables
37
-------
Although several WE/WD differences are found in the meteorological
variables examined, we cannot make any quantitative estimate of the impact
of such meteorological biases on the WE/WD oxidant levels. Instead of
making a quantitative estimate, we discuss, in the following subsections,
a method of removing meteorological biases by comparing the WE/WD oxidant
levels under similar meteorological conditions.
4.2 WE/WD DIFFERENCE IN OX WITHIN METEOROLOGICAL CLASSES
In this subsection, meteorological classes for daily oxidant potential
are developed for each of the 22 study sites. Then, WE/WD oxidant levels
are compared within the same meteorological class. The purpose of
meteorological classification is to determine a common meteorological
condition under which WE/WD oxidant levels can be compared without a
confounding effect of different meteorological conditions. A difference
in WE/WD oxidant levels in the same meteorological class is then relatable
to differences in WE/WD precursor emissions.
While the effect of meteorology upon air pollution, in particular,
oxidant air pollution, has long been recognized, very little has been
written concerning actual methods of removing that effect from observed
oxidant levels. One method is to develop meteorological classes for a
similar daily oxidant potential and to compare WE/WD oxidant levels within
the same meteorological class. By noting that meteorological classificatior
is a kind of meteorological pattern recognition, we applied a pattern
recognition algorithm called AID (Automatic Interaction Detector, developed
38
-------
at Survey Research Center, University of Michigan, some years ago
[Sonquist et al. 1973]) to develop meteorological classes for daily
oxidant potential at each study site.
Because meteorological classification should be independent of
emissions, we used only weekday data of oxidant and meteorology as
inputs to AID. Three meteorological classes, high, middle, and low,
are then formed by grouping several original classes produced by the
AID program into the high, middle, and low ranges of oxidant concentrations,
4.2.1 AID Decision Tree Analysis
After some unsuccessful attempts at using scatterplot approaches, we
have chosen to use the AID computer program to develop meteorological
classes for daily oxidant potential. The AID program [Sonquist et al.
1973] simulates the procedures of a good researcher in searching for
predictors (meteorological variables) that best account for the variations
of a dependent variable (daily maximum oxidant concentration). AID
searches over all independent variables for the best split that separates
the original data set into the two most uniform subsets. AID keeps
dividing each subset until no statistically significant split can be made
(see Appendix D). The result is a binary decision (classification) tree
that shows graphically the dependence of daily maximum oxidant (OX) levels
on the meteorological variables and their ranges.
The AID program is applied to only the weekday data of oxidant and
meteorology collected for each of the 22 locations in the northeast
United States. The data for the analysis covers the warmer seasons in
39
-------
four years (May through September, 1973 through 1976). The reader is
referred to Section 2 (e.g., Figure 2 and Table 2) for a description of
the 22 OX monitoring sites and their pairings with surface and upper-air
weather sites. The data base for each of the 22 locations includes daily
maximum oxidant concentrations and the corresponding values of the eight
meteorological variables listed in Table 6. The meteorological data were
extracted or computed from weather data obtained from the National Climatic
Center (NCC) for 12 surface monitoring sites and 3 upper-air monitoring
sites. Of the eight meteorological variables used, the first six
variables (Tmav, TA, P, Vis, WD, and WS) were obtained from surface
ma X
X,
weather tapes (NCC-TDF 14); the last two variables (MH and V) were
obtained from mixing-height tapes.
Because AID requires the independent variables to be in discrete
form, the actual values of the meteorological variables are converted into
a discrete class by.setting a proper interval of values for each variable.
Table 6 lists the intervals used for the meteorological variables. The
intervals are determined such that an approximately equal number of data
points fall into each interval.
Because we would like to obtain meteorological classes that are
independent of emissions, AID is applied to the weekday data only. By
excluding all the weekend days from the data base, we can isolate the
relationship between daily maximum oxidant level and meteorology from
the WE/WD differences in precursor emissions. Given meteorological
40
-------
TABLE 6. LIST OF METEOROLOGICAL VARIABLES USED FOR
SITE-SPECIFIC METEOROLOGICAL CLASSIFICATION
OF DAILY MAXIMUM OXIDANT CONCENTRATION BY AID
Variable Name
Daily Maximum Temperature
Daily Average Temperature
Daily Average Sea-Level Pressure
Daytime (6A.H.-9 P.M.) Average Visibility
Vector Average Hind Direction
Daily Average Wind Speed
Afternoon Mixing Height
Mean Wind Speed through Mixing Layer
Symbol
max
TA
F
V1s
WD
MS
MH
7
Unit
°F
°F
mb
mi
deg
kn
m
m/sec
1
[-.55)
[—,50)
[0.1000)
[0.2)
[-45.0)
[0.2)
[0,250)
[0.2)
2
[55,65)
[50,55)
[1000.1004)
[2.4)
[0.45)
[2.4)
[250,500)
[2.3)
Ranges Used to Place the
3
[65.70)
[55,60)
[1004.1008)
[4.6)
[45.90)
[4.6)
[500,700)
[3.4)
4
[70.75)
[60,65)
[1008,1012]
[6,8)
[90.135]
[6,7)
[700.1000]
[4,5)
Variable into a Discrete Class
5
[75.60)
[65.70]
[1012,1016]
[8.10)
[135,180]
[7.8]
[1000.1250)
[5.6)
6
[80,85]
[70,75)
[1016,1020)
[10.12]
[180,210]
[8.10)
[1250,1500)
[6.7)
7
[85.90]
[75.80)
[1020.1024)
[12.16)
[210,240)
[10.12)
[1500.2000)
[7.8]
8
[90.95]
[80.85]
[1024.1028)
[16.20]
[240,270]
[12.14)
[2000,2500]
[8.12)
9
[95.-]
[85.-]
[1028.-]
[20..]
[270.315)
[14.-]
[2500.-]
[12.-]
[ — Implies endpoint is included in the range.
( — implies endpoint is excluded.
-------
classes that are independent of emissions, we will be able to compare
the WE/WD oxidant levels under similar meteorological conditions.
Examples of AID decision trees are given in Figures 12, 13, and 14.
Figure 12 shows a binary decision tree for the Newark, New Jersey, station,
which Is the most source-intensive station among the 22 stations analyzed.
The decision tree Indicates that the highest oxidant air pollution occurs
on days with daily maximum temperature (Tmaw) 185°F, daytime average
max
visibility (Vis) < 8 mi, and mean wind speed through mixing layer
2.0 ±V<6.0 m/sec. In other words, the highest oxidant pollution days
are characterized by high temperature, low visibility, and low wind speeds
throughout the mixing layer. Such high-pollution days with mean daily OX
of Y = 98.7 ppb occurred 21 days (N=21) out of the 199 days sampled during
the summer-month periods (May through September) of four years, 1973 through
1976. On the other hand, the lowest oxidant air pollution occurred on cool
days (T_av < 75°F). On those days, the mean level of daily maximum OX con-
ma x
centrations over the 54 sample days was as low as Y = 24.4 ppb.
A similar interpretation can be derived easily for other classes from
the binary decision trees of Newark and the other two stations:
Mamaroneck, New York, and Salem, Massachusetts. A complete set of binary
decision trees for all the 22 air monitoring sites 1s given 1n Appendix E.
By scanning these decision trees, one may find that the dally maximum
temperature, T , 1s, by far, the most Important of the eight meteoro-
logical variables considered for explaining the dally maximum oxidant level.
Other meteorological variables that have appeared frequently on the
42
-------
0)
f4 - 199
Y = 49.1
Tmax >- 85
W
N -
Y = 36.
<85 8 < Vis Vis < 8
Class VIII
1000
-------
T«ax < 75
(4)
N = 97
Y = 30.7
75±Tn,a*<85
-45 < WD < 135135 1 WD<-45
02)
N --56
Y = 25.6
10 _< US 2 <. HS < 10
8 < Vis V1s < 8
Figure 13, Binary decision tree for the Mamaroneck, New York, station.
-------
T <85
max
85
-------
binary decision trees are: vector average wind direction (WD), daily
average wind speed (WS), daytime average visibility (Vis), and mean wind
speed through mixing layer (V). Meteorological variables that have
appeared only seldom are: daily average temperature (TA), daily average
sea-level pressure (P), and afternoon mixing height (MH).
The output of the AID computer program contains much more information
than that expressed by the decision tree. One other useful kind of
information derived from the AID output is an approximate pollution rose
for daily maximum oxidant levels. The sample mean of daily maximum OX
concentration and the number of cases are given for each 45-degree sector
of daily resultant wind direction. From this information, we have con-
structed a map of an approximate pollution rose for each site (Figure 15),
indicating the wind direction associated with high-oxidant air pollution.
From Figure 15 we can see that the preferred wind direction for high-
oxidant air pollution is from the southwest quadrant. It should be noted
that even those sites located to the south or the west of the major
metropolitan areas have a higher oxidant air pollution when the wind
blows from the southwest quadrant instead of from the metropolitan area.
Only the Asbury Park, New Jersey, station has a preferred wind direction
from New York City as well as from the southwest quadrant. It is
interesting to note that the Welfare Island, New York, station does
not have any preferred wind direction for high-oxidant air pollution.
46
-------
Metropolitan Area
0 25 50 75
I _.____._ i ... - - L f
Miles
Figure 15. Preferred wind direction for high-oxidant air pollution at each of the 22 sites.
-------
4.2.2 Meteorological Classification
The site-specific meteorological classes Identified by a binary
decision tree provide us with easily interpretable Information about the
dependence of oxldant concentration on meteorological factors. However,
the detailed meteorological classes for each monitoring site are rather
inconvenient for 1nterstat1on comparison when we want to quantify
oxidant-concentratlon dependency on meteorology from one site to
another. To make such interstation comparisons easily, individual
meteorological classes exemplified by nodes of a decision tree were
combined to form three standard meteorological classes for each station:
a high, middle, and low oxidant-forming-potential class.
Table 7 provides a summary description of site location and meteor-
ological characteristics for each of the three standardized classes.
In addition, Table 7 lists, for each site, the WE/WD oxidant levels in
each meteorological class and the terminal nodes of the decision tree that
form each of the three standardized classes. For example, the table
provides us with the following Information on the Bethesda station:
t The Bethesda station 1s located to the northwest of
Washington, D.C., and 1s close to a major traffic artery.
t The high meteorological class 1s defined by dally max temp
> 80° F, with wind speed < 6 kt or with wind direction from
FForth-northeast to southwest (clockwise).
• The high class has mean oxldant levels of 105 ppb on
weekends and 114 ppb on weekdays.
t The low meteorological class 1s defined by dally max temp
< 80°F and has mean oxldant levels of 40 ppb on weekends and
50 ppb on weekdays.
48
-------
TABLE 7. QUALITATIVE DESCRIPTION OF METEOROLOGICAL CHARACTERISTICS FOR
HIGH, MIDDLE, AND LOW CLASSES AT EACH OXIDANT MONITORING SITE
U3
SITE/SAROAD No.
Bethesda, Md. (BE)
210200005 F01
Gaithersburg, Md.(GA
210780004 601
Silver Spring, Md.
(SS)
211480006 F01
Site Description
The Bethesda monitoring site is located to the
northwest of Washington, D.C.»in a trailer and 1n
an open field. Located at the National Institute
of Health close to Wisconsin Ave., a major traffic
artery leading into Metro-Wash., D.C. The area is
suburban and subject to rush-hour emissions.
The monitoring site is located at the environment
lab on Fredrick St. The building is recessed from
the street with the monitor placed 35 ft above
ground. Major emissions come from traffic.
The Silver Spring monitoring station 1s located
due north of downtown Washington, O.C. The site Is
located 50 ft north of interstate 495, in a
trailer 12 ft above ground. Emissions are heavily
traffic oriented. The immediate area surrounding
the site is dominated by a golf course and
residential neighborhood with little heavy industry.
Class/
Terminal
Node #
High
7,13
Middle
6,8,12
Low
10,11
High
5
Middle
4,7
Low
6
High
6,8,9
Hiddle
4,11
Low
12.13
Avg.
Cone.
JpPb)
WE=105 '
WD=114
WE=69
WD=76
WE=IO
WD=50
WE=81
WD=81
WE=53
W0=62
WE=60
WD=51
WE=71
W0=86
WE=52
WD=54
WE=33
WD=36
Qualitative Description
of Meteorology
Max temps >80°F, avg visibility
<8 miles and avg surface wind
speed <6 knots, or max temp >80°F,
avg visibility >8 miles and avg
wind direction from north-northeast
to southwest (clockwise)
Max temp >80°F and avg visibility
<8 and avg surface wind speed
>6 knots, or max temp >80"F
and avg visibility >8 miles and
wind from the southwest to the
northeast (clockwise)
Max temp <80°F
Max temp >85eF and avg
visibility <8 miles
Max temp >85°F and avg visibility
>8 miles, or max temp <85°F
Fut >80°F
Hax temp <80°F
Max temp >85°F and avg surface
wind speed" <8 knots.
Max temp >85°F ancTavg surface
wind speed" >8 knots, or max temp
<85°F and avg visibility >.16 miles
Max temp <85£F and avg visibility
<16 miles
-------
TABLE 7. CONT'D
SITE/SAROAO No.
Sultland, Hd.(SU)
211560001 F01
Baltimore. Nd. (BA)
210120018 F01
Site Description
Site 1s located at the Sultland Federal Center.
Major arteries 1n the vicinity are Sultland Parkway
and Silver H111 Road. The station will probably
reflect rush-hour traffic emissions and some
Industrial emissions. The monitor 1s elevated 10
ft above ground.
The Baltimore air monitoring site 1s located on
22nd St. and Calvert St. The station 1s located
1n a parking lot and 1s subjected to mainly traffic
emissions. There is some industry; however, the site
is mainly a center-city type.
Class/
Terminal
Node 1
High
10,11,
14,15
Middle
8,17,19
Low
6,12,18
High
4.8,9
Middle
10,11
Low
6
Avg.
Cone.
(DDb)
WE=107
W0=107
WE=73
WD=74
WE=47
HD=49
WE=75
MD=88
WE=59
WD=54
WE=35
WD=32
Qualitative Description
of Meteorology
Max temp >85°F and mean velocity
through tTie mixed layer <4.0 m/s
or max temp >85°F and the mean
velocity > 470 m/s and avg. visibil-
ity <12 miles
Max temp >85°F and mean velocity
through tFe mixed layer >4.0 m/s
and avg visibility >12 mTles, or
max temp <85°F and mixing height
>750 m and avg surface wind speed
<8 mph and either max temp >80°F
or max temp <80°F but >55°F and
avg pressure >.1020 mb
Max temp <85°F and mixing height
<750 m, or max temp <85°F and
mixing height >750 m and avg
surface wind speed >8 knots, or
max temp <85°F with mixing height
>750 m, surface wind speed <8 mph
max temp <80°F and avg pressure
<1020 mb.
Max temp >_ 85°F
Max temp <85°F but >75°F
lax temp <75°F
01
o
-------
TABLE 7. CONT'D
SITE/SAROAD No.
Lancaster, Pa. (LA)
394660007 F01
Camden, N.J. (CA)
310720003 F01
Site Description
Located In a residential area to the northeast of
downtown Lancaster. Emissions are mainly traffic
oriented and are generally light. Main arteries are
Broad St. and Lehigh Street. The monitor is located
on Abraham Lincoln Jr. High School on Gofftown Rd.
The site 1s located in a trailer in a residential
area. Davis & Copewood Sts. are not subject to
heavy traffic emissions. Industrial emissions are
located to the northeast and west at about a
distance of two miles. There is a significant
number of parks and cemeteries in the general
vicinity making the site representative.
Class/
Terminal
Node S
High
5,7
Middle
6,13
Low
8,10,12
•RT^JT"
5,12,13
Middle
9,17,19
Low
10,14,
16,18
Avg.
Cone,
Jppbl
WE=104
WD=114
WE=90
WD=94
WE=57
WD=61
WE=95
WD=103
WE=72
WD=70
WE=49
WD=41
Qualitative Description
of Meteorology
Max temp ^ 80°F with an avg. visibil-
ity < 6 miles or with an average
visibility ^ 6 miles and a mean
velocity through the mixed layer
^ 6.0 m/s.
•lax temp > 80°F and an avg. visibil-
ity >^ 6 mTles and a mean velocity
through the mixed layer < 6.0 m/s,
or max temp > 70° F but < 80°F and a
mixing height >_ 1000 m and an avg.
visibility < 10 miles.
1ax temp < 70°F, or max temp > 70°F
but<80°F and a mixing height < 1000 m
or max temp > 70°F but < 80°F with a
mixing height > 1000 m and an avg.
visibility >. 10 miles.
Max temp > 90°F or max temp >_ 80°F
)ut < 90°F and an avg. surface wind
speed < 6 knots.
Max temp > 80°F but < 90°F and avg.
wind speeds > 6 knots and either
winds from tFe southwest to northwest
or winds from northwest to southwest
(clockwise) with mixing heights
j_ 1250 m, or max temp < 80°F but
>_ 70°F with winds from east to north-
west (clockwise) and mixing height
,1 1000 m.
lax temp < 70°F, or max temp _> 70°F
but < 80°F and avg. wind from north-
west to east, or max temp 21 70°F
but < 80°F,winds from east to north-
west and mixing height < 1000 m, or
nax temp > 80°F but < 90°F with avg.
wind speed" > 6 knots, winds from the
southwest to northwest and a mixing
height < 1250 m.
-------
TABLE 7, CONT'D
SITE/SAROAD No.
CMden Co., N.J.(CC)
310740001 F01
en
ro
Norrlstown, Pa. (NO
396540013 F01
Site Description
The site 1s located on the grounds of the N.J. State
Hospital at Ancorla. The surrounding area 1s malnlj
agricultural. The oxldant monitor Is 45 ft above
ground on the main building of the hospital. The
site 1s subject to local traffic emissions from
Rt. 30 and Is southeast of Camden.
The site Is located 10 miles to the northwest of
Philadelphia. Placed on Belvolr Rd.,the site 1s
subjected to moderate traffic emissions. Minor
Industrial emissions come from the southwest and
Norrlstown proper while the station Is within 1 mile
of the Penn. Turnpike. The monitor Is placed 12 ft.
above ground at the St. Armory 1n a residential area
Class/
Terminal
Node I
High
5
Middle
7
Low
2,6
High
7,12,13
Middle
4,8,15
Low
10,14
Avg. I Qualitative Description
of Meteorology
HE=112
HD=102
WD=82
WD=57
WE=108
HD=120
WE=84
WD=89
NE-65
WD=55
Max temp >_ 75°F and avg. surface
wind from the southwest to northwest
(clockwise).
Max temp > 1
lax temp >_75°F and avg. winds from
northwest to southwest (clockwise)
and an afternoon mixing height
> 1250 m.
Max temp < 75°F or max temp > 75°F
and avg. winds from northwest" to
southwest (clockwise) and afternoon
mixing height < 1250 m.
Max temp ^ 90°F and avg. wind speed
< 8 knots or a max temp >_ 80° F but
< 90°F and a velocity through the
mixed layer >_ 3.0 m/s.
Max temp >_ 80°F and avg. wind speed
8 knots, or max temp > 80°F, avg.
wind speed < 8 knots and" a max temp.
<90°Fw1th a mean velocity through the
mixed layer < 3.0 m/s, or a max temp.
< 80°F but > 75°F and winds from
east to northwest.
Max temp < BOTwlth winds from
northwest to east (clockwise), or max
temp < 75 and winds from east to
northwest (clockwise).
-------
TABLE 7. CONT'D
SITE/SAROAD No.
Philadelphia, Pa.
(PH)
397140026 HOI
Asbury Park, N.J.
(AP)
310060001 F01
Babylon, N.Y. (BB)
330280002 F01
Site Description
The Broad and Spruce St. monitoring site is located
at the corner of a parking site in downtown Phila-
delphia. Emissions are heavy and traffic oriented.
Broad St. is a main artery running north-south. The
site is subject to both industrial and vehicular
emissions and is considered characteristic of a down-
town metropolitan area.
Located on Mattison Ave. between Bond and Main Sts— •
the commercial center of Asbury Park. The site is
elevated (second floor) and is subject to mainly
traffic emissions and minor local industry. The
location of the monitor is within one mile of the
Atlantic Ocean.
The site is located in the center of an industrial
park. The general area is residential (parks and
cemetery) with major traffic emissions from Rt. 110
which is heavily traveled. In addition Interstate
495 is within 1 mile to the north with continuous
heavy traffic. The site is downwind of NYC with the
Atlantic Ocean 5 miles to the south.
Class/
Terminal
Node #
High
3,8,9
Middle
10,11,15
Low
12,14
High
5
Hi d'dle"
4,7
Low
6
High
4,5
Middle
9
Low
6,8
Avg.
Cone.
(PPbj
WE=93
WD=81
WE=61
WD=51
WE=38
WD=31
WE=82
W0=89
WE=51
WD=56
WE=52
WD=46
WE=107
WD=101
WE=60
WD=72
WE=50
WD=42
Qualitative Description
of Meteorology
T max >_ 90°F, or T max > 80°F but
< 90°F and with a mixing height
>. 1500 m.
T max > 80°F but < 90°F and mixing
height < 1500 m, or T max > 75°F but
< 80°F and a mixing height > 1250 m.
T max < 75°F, or T max > 75°F but
< 80°F with a mixing height < 1500 m.
Mixing height > 1250 m with T max
>_ 80°F.
Mixing height > 1250 m with T max
< 80° F, or mixing height < 1250 m
and winds from the southwest to
northwest (clockwise).
Mixing height < 1250 m and winds from
northwest to southwest (clockwise).
Max temp. >_ 80°F.
Max temp < 80°F, wind from the south-
east to northwest (clockwise) with
max temp >_ 70°F.
Max temp < 80°F and wind direction
from southeast to northwest (clock-
wise) but 'max temp < 70°F, or max
temp < 80°F and wind direction from
northwest to southeast (clockwise).
en
CO
-------
TABLE 7. CONT'D
SITE/SAROAO No.
Danbury, Conn. (DA)
070175123 F01
in
Greenwich, Conn.(GR)|
070330004 F01
Mamaroneck, N.Y.
(MA)
334100002 F01
Site Description
Located on a college campus 1n a residential portion
of Danbury. Local emissions are light and come fron
vehicular traffic and local boilers. The site 1s
downwind (40 miles) of central NYC and 1s considered
very representative of the general area.
rhe site 1s located on Bruce Golf Course near the NY-I
Conn, state line. The site 1s considered represents-]
tlve of the general area with local emissions being
light, coming from King St., Rt. 684 and Hestchester
County Airport. NYC 1s approximately 25 ml to the
southwest.
.ocated at the exit to the New England Thruway—35 ft
above ground and 50 ft from the curb on 5th St. and
•lad 1son. The site Is subjected to heavy traffic
emissions and not necessarily representative of the
general area, which 1s commercial and situated 1n a
residential community.
Class/
Terminal
Node I
High
12,13
TTOfle
4,6,11
Low
High
6.7
8,9
Low
10,11
High
3,11
Middle
6,8,10
Tow
12.13
Avg.
Cone.
(ppb)
WE=88
W0=127
WD=94
HE=54
WD=51
MD=92
WE=63
HD=64
WE=79
HD=87
WD=53
WD°31
Qualitative Description
of Meteorology
Avg. visibility < 8 miles and avg.
wind direction from the southeast to
southwest (clockwise).
Svg. visibility _ 8 miles with winds from
southeast to northwest and max temps
75°F.
Avg. visibility ^ 8 miles with winds
from the northwest to southeast, or
avg. visibility ^ 8 miles and winds
from the southeast to northwest with
max temps < 75°F.
Max temp > 75°F and avg. visibility
s n ml lac
< 8 miles. _
Hax temp >_ 75°F and avg. visibility
> 8 miles.
Max temp < 75°F,
T max >_ 85°F, or T max >. 75°F but
< 85°F and avg. wind speed < 10 knot'
with visibility < 8 miles and max
temp < 80°F.
T max >_ 75°F but < 85°F "an3~avg.
wind speed >^ 10 knots, or T max
>_ 75°F but < 85°F and avg. wind
speed < 10 knots and avg. visibil-
ity > 8 miles or T max >, 75°F but
< 80*F with avg. wind speed < lOmph
tvg. visibility < 8 miles.
enip. < 75°F. ~~"
-------
TABLE 7. CONT'D
SITE/SAROAD No.
Middletown, Conn.
(MI)
070570003 F01
tn
en
Newark, N.J. (NE)
313480002 F01
Site Description
The site 1s located at the city hall, approximately
20 ft. above ground. The general area emissions are
mixed,Including both Industry and traffic sources.
The site is approximately 70 miles to the northeast
of NYC, south of Hartford and north of New Haven.
Located In downtown Newark on Washington St. and
Branford St. The area has Industry but Is mainly
commercial. The station Is subjected to moderate-
to-heavy traffic from urban congestion. The monitor
is placed 12 ft above the ground.
Class/
Terminal
Node #
High
6,7
Middle
4
Low
2
High
12,13,15
Middle
6,10,14
low
4,8
Avg.
Cone.
fnnhl
WE=100
WD=124
WE=77
WD=78
WE=46
WD=47
WE=87
WD=77
WE=63
W0=49
WE=33
WD=26
Qualitative Description
of Meteorology
Avg. wind direction from southeast to
northwest with avg. visibility
< 8 miles.
Avg. wind direction from southeast to
northwest and an avg. visibility
> 8 miles.
Avg. wind direction from northwest to
southeast (clockwise).
Max temp ^ 85°F and avg. visibility
< 8 miles, or max temp >_ 80°F but
< 85°F and avg. pressure >_ 1012 mb,
and mean velocity through the mixed
layer < 6.0 m/s.
Max temp ^ 85°F and visibilitity
^ 8 miles, or max temp >^ 75°F but
< 85°F and pressure > 1012 mb and
either max temp < 80TF or if max temp
> 80°F then having a mean velocity
Through the mixed layer >_ 6 m/s.
Max temp < 75°F, or max temp >_ 75°F
but < 85°F and an avg. pressure
< 1012 mb.
-------
TABLE 7. CONT'D
SITE/SAROAD No.
Welfare Is., N.Y.
(HI)
334680050 F01
en
Nedford, Mass. (ME)
221220003 F01
Site Description
Hie site Is located at ground level on the north
snd of the Island. The Island lies between mldtown
Manhattan and Queens with the East River Drive along
:he Manhattan shore. Traffic on the Island was
limited for the study period;however, the site Is
lowmdnd of heavy-Industrial, power-plant and traf-
fic emissions. The site Is well ventilated and
probably gives a good representation of East River
Dxldant.
Located to the northeast of downtown Boston.The
station Is situated on a grass-covered median strip
at the Intersection of Rt. 16 and Rt. 128 and Is
subjected to moderate-to-heavy traffic volumes. The
area 1s basically commercial and Industrial. The
station 1s well ventilated and 1s considered to be a
representative neighborhood station.
Class/
Terminal
Node I
High
11,14,15
Middle
8,10,17
,16
HTgh
7,10,11
Middle
14,15.
16,17
low
8,12
Avg.
IDDb}
WE=93
WD=94
HE=64
W0=63
ME=39
WD=36
WE=62
WD-69
WE-53
WD-47
WD=31
Qualitative Description
of Meteorology
Max temp ^ 85°F, or max temp >. 75°F
but <85°Fand winds from the north to
southwest with a mixing height ^
>.750 mand avg. visibility < 10 miles.
k > i *• ~if~ o co c
Max tempyT5°F but < 85"F with winds
from the northwest to southwest with
either a mixing height < 750 m or
with a mixing height ^ 750 m and an
avg. visibility >_ 10 miles, or a max
temp < 75°F with wind speed < 8 knots
and max temps %_ 70°F.
Max temp >. 75°F and < 85°F with winds
from the southwest to northwest, or
max temp < 75°F and wind speed
< 8 knots and a max temp < 70°F, or
a max temp < 75°F and an avg. wind
speed >_ 8 knots.
Max temp > 85°, or max temp >_ 75"F
and < 85°Fw1th mean wind 1n the
mixed layer >_ 5.0 m/s.
Max temps >. 75" but < 85° and mean
wind speeds 1n the mixed layer
< 5.0 m/s, or max temp >. 70°F but
<-75°F and surface winds from
the east to northwest (clockwise)
Max temp < 70°F, or max temp < 7
and wind from northwest to east
(clockwise).
-------
TABLE 7, CONT'D
SITE/SAROAD No.
Quincy, Mass. (QU)
221880002 F01
Salem, Mass. (SA)
221980001 F01
Worchester, Mass.
222640012 F01
Site Description
.ocated to the southeast of Boston on Rt. 3A and
Washington St. The site is a curbside location sub-
jected to periodically heavy traffic emissions. The
site is well ventilated with numerous industrial
sources in the area (i.e., General Dynamic Shipyard,
Proctor & Gamble and Edison Power). Site experiences
noderate-to-heavy summer daytime traffic volume.
-ocated approximately 10 miles northeast of Boston.
The area surrounding the site is residential, located
on Highland Ave., and it is subject to moderate
traffic. Industrial sources consist of local boiler:
and a power-generating station approximately 2 miles
away.
Located in a commercial district of Worchester, the
station is primarily subjected to traffic emissions.
The monitor is situated at the corner of new Salem &
Washington Sts. in a monitoring trailer.
Class/
Terminal
Node #
High
8,9,14,15
Middle
11,13
Low
10,12
High
3,7
Middle
6,11
Low
8,10
High
4.5
Middle
8,9
Low
6
Avg.
Cone.
_(DPb)
WE=87
WD=89
WE=52
WD=54
WE=43
WD=42
WE=69
WD=86
WE=42
WD=46
WE=43
WD=36
WE=74
WD=79
WE=52
WD=58
WE=46
WD=40
Qualitative Description
of Meteorology
Max temps >^ 80°F.
Max temps < 80°F but >_ 75°F and avg.
visibility < 10 miles, or max temps
< 75°F and the avg. surface winds
from the east to northwest (clockwise)
Max temp < 80°F but >_ 75°F and avg.
visibility >^ 10 miles, or max temp
< 75°F and avg. surface winds from
the northwest to east (clockwise).
Max temp >_ 85 °F, or max temp > 75°F
and avg. surface winds from northwest
to southwest (clockwise).
Max temp >_ 75*F but < 85°F and avg.
surface, wTnd from southwest to north-
west (clockwise), or max temp > 65°F
but < 75°F and mean velocity through
mixed layer >^ 7.0 m/s.
Max temp < 65 "F or max temp >^ 65 "F
but < 75°F and mean velocity
through mixed layer < 7.0 m/s.
Max temps ^ 80° F.
Max temps ^ 70°F but < 80°F.
Max temps < 70° F.
en
-------
For the other stations, similar information can be obtained from
Table 7.
Because we are particularly Interested in knowing the meteorological
conditions existing on h1gh-ox1dant days, the meteorological characteristics
of the high met class depicted from AID decision trees are summarized in
Table 8. For example, the high-oxidant days at Newark, New Jersey, are
characterized by a high temperature >_ 85°F with low visibility < 8 mi
or by moderate temperature of 80°F to 85°F with low wind speed aloft
< 6 m/sec and high pressure .>! 012 mb. Among the 22 sites examined, as
many as 20 sites show high temperature as a key meteorological factor
for high-oxidant air pollution. Three other important meteorological
factors are low visibility, moderate-to-low wind speed, and south-to-
southwesterly wind. Surprisingly, mixing heights greater than 750 m are
associated with high oxidant levels at three sites. The high mixing
height probably does not indicate a necessary condition for high-oxidant
air pollution but simply means low mixing heights are not always associated
with high oxidant levels.
4.2.3 WE/HP Differences Within High-OX-Potential Class
This section discusses the WE/WD differences in daily maximum
oxidant concentrations on high-oxidant-potential days. Here, the high-
oxidant-potential days are determined solely by the meteorological condi-
tions existing on those days. Because AID decision trees are developed
from weekday data only, we can expect that the meteorological classes
derived from the AID decision trees provide a means of sorting each
individual day according to the meteorological oxidant formation potential
on that day.
58
-------
TABLE 8. METEOROLOGICAL CHARACTERISTICS OF
HIGH-OXIDANT CLASS FOR EACH SITE
Metropolitan ' Station Name
Area '
Washington,
D.C.
Baltimore
Philadelphia
Bethesda (BE)
Gaithersburg (GP)
Silver Spr (SS)
Suitland (SU)
Baltimore (BA)
Lancaster (LA)
Camden (CA)
Camden Co.(CC)
Norristown (NO)
Philadelphla(PH)
Asbury Park(AP)
• Babylon (BB)
Oanbury (DA)
New York-
Newark
Boston
Greenwich (GR)
Middletown(MI)
Mamaroneck(MA)
Newark (NE)
Welfare Is.(WT)
Medford (ME)
Quincy (QU)
Salem CSA)
T_,y i Vis
max
°F mi
• >80 <8
i "80 >8
"• 785 <8
785
(>85
>85
e >80
/>90
1180,90)
>75
/>90
1180,90)
/>90
.1 TJ30.90)
>80
X30
>75
>85
P5.85)
>85
180,85)
>85
175,85)
/>85
\TJ75.85)
1
<6
i6
<8
<8
<8
<8
<8
<10
>80
<~85 '
1^5.851
Worchester (WO) ' ~>80 J
MS
kts
<6
<8
<6
<8
<8
<10
7
m/sec
<4
>4
I6
13
<6
>5
WO
deg.
[-45,210)
[240,315)
[135,240)
[135,315)
[-45,240)
[-45,210)
MH
F P
m mb
^1500
>1250
>750
>1012
MOTE:
[ ~ • ImpHes endpolnt is Included 1n the range.
) — implies endpolnt is excluded from the range.
59
-------
In Table 9, the WE/WD differences in daily maximum oxidant levels
for all days and for high-oxidant-potential days only are shown for each
of the 22 stations. At 14 of the 22 stations, the WE/WD difference is
greater on high-oxidant-potential days. However, the percentage dif-
ference is significantly different only at a few sites. The average
percentage difference is about the same C-7%) between the two cases,
while the absolute WE/WD difference is slightly greater on high-oxidant-
potential days.
4.3 METEOROLOGICAL ADJUSTMENT OF WE/WD OXIDANT LEVELS
Fluctuations in meteorology cause most of the day-to-day variation
of oxidant concentration levels [Breiman and Meisel 1976]. Consequently,
it is desirable to account for the meteorological fluctuations by adjusting
the air quality data. One meteorological adjustment method is to compare
the WE/WD air quality levels under similar meteorological conditions, say,
the high met class developed in the preceding subsection. Another meteor-
ological adjustment method would be to compute a frequency weighted
average of the means of high, middle, and low classes for weekends and
that for weekdays. Zeldin and Meisel [1978] show that meteorological
normalization of air quality parameters (such as the arithmetic mean and
geometric mean) can be performed if one is able to determine meteorological
classes that explain a significant fraction of air quality variance. For
example, a meteorologically normalized arithmetic mean can be computed as
follows:
60
-------
TABLE 9. WE/WD DIFFERENCES IN DAILY MAX OX LEVELS
FOR ALL CASES AND FOR HIGH MET CLASS ONLY
(SS vs. MTWTF)
Metropolitan
Area
Washington
D.C.
Baltimore
Philadelphia
Newark -
New York
Boston
Station
Name
Bethesda
Gaithersburg
Silver Spring
Suit! and
Baltimore
Lancaster
Camden
Camden Co.
Norri stown
Philadelphia
Asbury Park
Babylon
Dan bury
Greenwich
Mamaroneck
Middle town
Newark
Welfare Is.
Medford
Quincy
Salem
Worchester
22 Station
Avg.*
WE Level
(ppb)
70
61
52
73
54
82
68
64
82
58
55
66
71
79
49
73
56
62
50
60
49
56
63
All Cases
WD Level
(ppb)
77
64
59
79
59
91
70
75
91
57
57
70
87
92
54
82
49
61
49
62
55
58
68
% Change
(WE-WD)
-9%
-4%
-13%
-8%
-9%
-8%
-2%
-14%
-9%
+2%
-4%
-7%
-18%
-15%
-8%
-11%
+13%
+1%
+2%
-4%
-12%
-3%
-7%
High
WE Level
(ppb)
105
81
71
107
75
104
95
112
108
93
82
107
88
102
79
100
87
93
62
87
69
74
90
Met. Cl
WD Level
(ppb)
114
81
86
107
88
114
103
102
120
81
89
101
127
125
87
124
77
94
69
89
86
79
97
ass
% Change
(WE-WD)
-8%
0%
-17%
0%
-15%
-9%
-8%
+10%
-10%
+15%
-8%
+6%
-31%
-18%
-9%
-19%
+13%
-1%
-10%
-2%
-20%
-6%
-7%
The station-average percentage, difference is computed from
the station averages of WE/WD concentration levels.
61
-------
k,Non» = CM P1
where C^ Norffl = the meteorologically normalized mean for a
pollutant (e.g., oxldant) 1n the kth period
(e.g. weekends),
M = the number of meteorological classes,
C^ = the sample mean of meteorological class "1"
(e.g., high class) during the k period,
P.J = the average frequency of occurrence of meteor-
ological class "1" during ajl_ periods (e.g.,
all days).
The following paragraphs describe a meteorological adjustment of the
actually observed WE/WD difference according to Equation (2). In making
such a meteorological adjustment, we utilized the three standardized
meteorological classes developed in the preceding subsection. Figure 16
is a bar graph showing the WE/WD mean concentrations in the high-, middle-,
and low-oxidant-potential classes at Lancaster, Pennsylvania. The figure
also shows the number of cases in each class on weekends and weekdays.
The meteorologically adjusted weekend mean for all cases represented
by the figure is given by
3
CWE,Norm = CWE,1 P1
62
-------
where 1 = 1, 2, 3 Indicates, respectively, high, middle, and low classes,
and CWE,i 1s the mean weekend oxidant level of the 1th class. P1 is the
average frequency of occurrence of each of the three classes during all
days, i.e., weekends and weekdays combined. Similarly, the meteorologi-
cally adjusted weekday mean is given by
3
CWD,Norm * CWD,i Pi • (2'b)
where CWQ ^ is the mean weekday OX level of the ith class.
X
As seen from Figure 16, the Lancaster station exhibits a consistently
lower oxidant level on weekends among the class means of each of the three
meteorological classes, the actual means, and the meteorologically ad-
justed means. Because of the consistently lower weekend levels among
all the five different WE/WD comparisons, one can say that the weekend
OX level is lower at this station. The above consistently lower type
of station is termed a "Type I station."
Figure 17 is a bar graph for the Newark, New Jersey, station. This
station exhibits a consistently higher weekend OX level among the five
different WE/WD comparisons. Therefore, we can expect that the weekend
OX level is higher at this station. The above type of station is termed a
"Type II station."
Figure 18 snows the bar graph comparisons of the WE/WD oxidant
levels at Baltimore, Maryland. Unlike the above two examples, the WE/WD
relationship 1s reversed from one comparison to another. While
the weekend OX level is lower 1n the high class, those in the middle and
63'
-------
.a
ex
Q.
0)
O
•i—
+J
2
4J
Q)
O
O
O
X
O
120 1
110 "
100 -
90-
70 •
60 •
4* 50
&
o 40
ie
* 30
20 -
10
62
WD
25
WE
High
Class
46
19
WD WE
Middle
Class
Station: Lancaster, Pa.
53
27
WD WE
Low
Class
71
WD WE
Actual
WD WE
Met. Adjusted
Figure 16. Example of Type I station, which exhibits a consistently lower
weekend OX level 1n every WE/WO comparison.
-------
CT)
.0
Q_
Q.
VI
c
o
ra
cu
o
o
o
X
o
X
rcs
tO
Q
0)
iio-l
100-
90-
80
70
60'
50-
40
30-
20-
10-
Station: Newark, N.J.
14
60
34
68
= 79
= 199
31
71
WD WE
High
Class
WD WE
Middle
Class
WO
Low
Class
WE
WD WE
Actual
WD WE
Met. Adjusted
Figure 17. Example of Type II station, which exhibits a consistently higher
weekend OX level in every WE/WD comparison.
-------
120-,
-§.
Q.
(/I
O
s
+J
Q)
O
O
X
o
iio-
100"
90-
80
70
60
50
40
30
20
10
60
18
28
74
WO WE
High
Class
WD WE
Middle
Class
Station: Baltimore, Md,
180
27
46
73
WD WE
Low
Class
WD WE
Actual
WD WE
Met. Adjusted
Figure 18. Example of Type III station, which does not exhibit a consistent relationship
amonq meteoroloa-ical -"•
-------
low classes are higher. While the actual WE/WD means show a lower week-
end level, the meteorologically adjusted WE/WD means indicate a slightly
higher weekend level. The Baltimore station does not show a consistent
WE/WD relationship among the five comparisons. Therefore, we cannot say
whether the weekend OX level is higher or lower than the weekday level
for this station. The above type of station is termed a "Type III
station."
In the next section, these classifications are used to categorize
the WE/WD differences in oxidant concentrations at each site.
4.4 CLASSIFICATION OF SITES
Table 10 summarizes the results of grouping the 22 oxidant monitoring
stations into the above-defined three station types. The table lists six
Type I stations that exhibit a consistently lower OX level on weekends,
and an additional seven stations having a consistently lower weekend level
with one exception. These 13 stations are all located either in the peri-
phery or outside the large metropolitan areas. Table lOb presents the
four Type II stations that exhibit a consistently higher OX level on
weekends: Philadelphia, Newark, Welfare Island, and Medford. These four
stations are all located in the most urban, source-intensive areas.
Table lOc lists the five Type III stations that do not exhibit'a con-
sistent relationship among the five different WE/TO comparisons. The
reason why, at these stations, the WE/WD oxidant relationship is altered
from one class to another is not known. However, one possible explanation
could be that these stations classed into Type III have only a weak WE/WD
67
-------
TABLE 10. CLASSIFICATION OF WE/WD DIFFERENCES
a. Type I stations, which exhibit a consistently loner OX level on weekends
Station
Bethesda, Md.
Silver Spring, Md.
Greenwich, Conn.
Mlddletown, Conn.
Lancaster, Pa.
Norrfstown, Pa.
With One Exception
Camden Co., N.J.
Danbury, Conn.
Hamaroneck, N.Y.
Qulncy, Mass.
Salem, Mass.
Suit! and, Md.
Uorchester, Mass.
Sample
Size XWE
286 31
248 „ 27
340 29
284 29
232 31
239 31
224 28
250 30
377 29
458 30
185 29
351 29
350 32
High Class
% Change
(WE-WD)
- 8
-17
-18
-19
- 9
-10
+10
-31
- 9
- 2
-23.
0
- 6
Middle Class
% Change
(WE-WD)
- 9
- 4
-12
- 1
- 4
- 6
-17
-16
-19
- 6
- 9
- 1
-10
Low Class
% Change
(WE-WD)
-20
- 8
- 2
- 2
- 7
-18
-12
+ 6
+16
+ 1
+19
- 4
+15
Actual
% Change
(WE-WD)
- 9
-13
-15
-11
- 8
- 9
-14
-18
- 8
- 4
-12
- 8
- 3
Met.
Adjusted
% Change
(WE-WD)
- 8
-12
-11
-10
- 6
- 3
- 8
-17
- 6
- 3
- 6
- 1
- 3
b. Type II stations, which exhibit a consistently higher OX level on weekends.
Newark, N.J.
Philadelphia, Pa.
With One Exception
Medford, Mass.
Welfare Is., N.Y.
278 28
293 30
415 30
362 29
+13
+15
- 7
- 1
+29
+20
+13
+ 2
+27
+23
+16
+ 8
+13
+ 2
+ 2
+ 1
+24
+17
+ 2
+ 2
c. Type III stations, which do not exhibit a consistent relationship
among meteorological classes.
Asbury Park, N.J.
Babylon, N.Y.
Baltimore, Md.
Camden, N.J.
Galthersburg, Md.
197 28
238 29
253 29
371 29
140 32
- 8
+ 6
-15
-8
0
-~9 ""'
-17
+ 9
+ 3
-15
+13
+19
+ 9
+20
+18
i -4" '
- 7
- 9
- 2
- 4
! + 1
+ 1
+ 1
+ 2
- 3
68
-------
difference in daily max OX levels so that the WE/WD oxidant relationship
can be altered by sample variation from one meteorological class to another.
Those stations classed into Type I and Type II have a strong enough WE/WD
difference in daily max OX levels so that the WE/WD oxidant relationship
cannot be altered easily by sample variation from one meteorological
class to another.
69
-------
5. DIFFERENCES IN OXIDANT PRECURSOR LEVELS
In Sections 3 and 4, WE/WD differences in daily maximum oxidant levels were
examined. This section discusses the WE/WD differences in oxidant precursor
levels. The 6-9 A.M. (EOT) average concentrations of nitric oxide (NO),
rr.
nitrogen dioxide (N02), nonmethane hydrocarbons (NMHC), and total hydro-
carbons (THC) are examined. The two major oxidant precursors are NO and
J\
NMHC. NOX is defined as the sum of NO and N02 concentrations. Most NOX
is emitted to the atmosphere in the form of NO and is subsequently oxidized
to NOg. Nonmethane hydrocarbons are defined as the difference between total
hydrocarbons (THC) and methane (CH4).
In addition to the four pollutants, the 6-9 A.M. averages of
(N0«/N0 ) ratios and carbon monoxide (CO) concentrations are also
£ /\
examined. Because NO works as a scavenger of OX while NO* works as a
promoter of OX, an ambient (N0«/N0 ) ratio provides an Indication of
£ A
net oxidant formation potential of the atmosphere [Jeffries et al. 1976].
CO is an inert gas that is mainly emitted by automobiles. Because
ambient CO levels may very well be an indicator of automotive-related
precursor levels, WE/WD differences in those data were also examined,
5.1 DIFFERENCES IN 6-9 A.M. AVERAGE NO AND NOp CONCENTRATIONS
Ambient NO and N02 data are not as complete as the oxidant data
analyzed in Sections 3 and 4. After checking the data completeness
at each air monitoring site, we found adequate NO and NOp data at the
eight stations whose names, site codes, and sample sizes are listed in
Table 11. The sample size is given for both the normal (SS vs. MTWTF)
and Sunday (Sun vs. TWTF) WE/WD definitions. The eight stations are
70
-------
TABLE 11. MONITORING STATIONS USED FOR ANALYZING
NO AND N02 AIR QUALITY DATA
Metropolitan
Area
Washington,
D.C.
Bal timore
Philadelphia
New York City
Station
Name
Bethesda
Silver Spring
Suit! and
Baltimore
Camden
Mamaroneck
Newark
Welfare Is.
SAROAD I.D.
210200005 F01
211480006 F01
211560001 F01
210120018 F01
310720003 F01
334100002 F01
313480002 F01
334680050 F01
EPA's
Site Code*
S-R
S
S
CC-C
S-R
CC-C
CC-C
CC
Number of Valid
Observation Dayst
SS" vs. MTWTF
115
121
in
194
202
224
194
213
Sun. vs. TWTF
85
87
82
137
149
166
140
151
Code key:
C - Commercial
CC - Center City
R - Residential
S - Suburban
A "valid observation day" is a day which has at least 2 out of 3 possible
observations during the 6-9 A.M. (EOT) period.
71
-------
all located within or near one of the large metropolitan areas (see
Figure 2).
5.1.1 WE/UP Differences in NO Concentrations
The WE/WD differences in 6-9 A.M. average NO concentration levels
are computed for both the normal and Sunday WE/WD definitions. The results
are summarized in Table 12. All the stations except for Bethesda exhibit
a large drop in 6-9 A.M. average NO levels on weekends. The weekend re-
ductions in 6-9 A.M. average NO levels range from 40% at Mamaroneck, New
York, to 83% at Camden, New Jersey, under the normal WE/WD definition.
Under the Sunday WE/WD definition, the drop is generally greater than that
under the normal definition. Also, the weekend reductions in the eight-
station 6-9 A.M. NO levels are 60% and 67% under the normal and Sunday
WE/WD definitions, respectively. The Sunday WE/WD definition (Sunday
versus Tuesday through Friday), therefore, yields a slightly larger change
between the WE/WD 6-9 A.M. average NO levels than the normal WE/WD
definition.
The NO behavior at the Bethesda station is different from that at
the other stations. In contrast to the large weekend drops at the other
stations, the Bethesda station exhibits a large increase in 6-9 A.M. average
NO concentrations on weekends. The increases are 134% and 252%, respec-
tively, under the normal and Sunday WE/WD definitions. A close examination
of the Bethesda NO data revealed that the weekend NO data contained a few
extremely high NO concentrations. The standard deviation of the weekend
NO data is much larger than that of the weekday NO data (5.3 pphm versus
1.4 pphm under the normal WE/WD definition and 7.6 pphm versus 1.4 pphm
72
-------
TABLE 12. WE/WD DIFFERENCES IN 6-9 A.M.
AVERAGE NO CONCENTRATIONS
CO
Station
Name
Bethesda
Silver Spring
Suitland
Baltimore
Camden
Mamaroneck
Newark
Welfare Is.
8-Station
Avg.
SS vs. MTWTF
Mean
WE level
(pphm)
1.5
5.3
0.5
1.0
0.9
1.9
2.2
1.4
1.8
Mean
WD level
(pphm)
0.7
12.8
1.3
4.6
5.4
3.2
6.3
2.4
4.6
% Change
(WE-WD)
+ 134%
- 58%
- 60%
- 79%
- 83%
- 40%
- 65%
- 43%
- 60%*
Sun. vs. TWTF
Mean
WE level
(pphm)
2.3
3.6
0,1
0.9
0.9
1.4
1.8
1,3
1.5
Mean
WD level
(pphm)
0.7
13.0
1.4
4.4
5.4
3.4
6.6
2.4
4.7
% Change
(WE-WD)
+ 252%
- 72%
- 93%
- 80%
- 84%
- 60%
- 72%
- 45%
- 67%*
The station-average percentage difference is computed from the
station averages of WE/WD concentration levels.
-------
under the Sunday WE/WD definition). Because of this large difference
between the WE/WD standard deviations, the results for the Bethesda
station should be viewed with cautton.
The WE/WD differences in 6-9 A.M. average NO levels are examined
by using the t-test. Figure T9 shows the significance level of the WE/WD
difference at each station under the normal and Sunday WE/WD definitions.
Among the eight stations examined, five—Silver Spring, Baltimore, Camden,
Newark, and Mamaroneck—exhibit significant WE/WD differences in 6-9 A.M.
average NO levels under both the normal and Sunday WE/WD definitions. The
remaining three stations, Bethesda, Suitland, and Welfare Island, show a
marginally significant difference under one of the two WE/WD definitions.
The t-test results are also presented in Figure 20, in which the
standardized values of Sunday/WD 6-9 A.M. average NO levels are plotted
on the x-y axes. The standardized value is a dimensionless number
given by a ratio of the mean to the square root of the pooled variance
(see Appendix A for definition). Therefore, the standardized values of
WE and WD can be compared directly, and their difference is exactly the
t-value. Figure 20 clearly shows that the WE NO level is lower than the
WD level at all the stations except Bethesda.
5.1.2 WE/WD Differences in NOp Concentrations
The WE/WD differences in 6-9 A.M. average N02 levels are examined
for the same eight stations as were used for the analysis of NO concen-
trations. The mean WE level, the mean WD level, and the percentage change
between weekends and weekdays are listed in Table 13. On the average for
the eight stations, the weekend N02 level is 42% lower than the weekday
74
-------
P-Value
1.0
0.5
0.10 . ^
(Marginally Significant Region)
0.05
0.002
0.001
(Insignificant Region)
*
0.02
01Q _ (Significant Region)
0.005
• .*
*
.. p-value under Sunday WE/WD definition
„ p-value under normal HE/HD definition
75
-------
(Tl
I
vo
0)
Ot
ID
I
0)
«o
•a
0)
N
•^
•o
to
1234567
Standardized Value* of Averaae.WE 6r9 A.M. NO
Standardized value = mean//pooled variance
Figure 20. Graphical presentation of t-test results for
WE/WD difference in 6-9 A.M. average NO concentrations
(Sun. vs. TWTF).
76
-------
TABLE 13, WE/WD DIFFERENCES IN 6-9 A.M.
AVERAGE N02 CONCENTRATIONS
Station
Name
Bethesda
Silver Spring
Suitland
Baltimore
Camden
Mamaroneck
Newark
Welfare Is.
8-Station Avg.
SS vs. MTWTF
Mean
WE level
(pphm)
1.3
3.4
3.2
3.4
1.8
1.5
2.9
1.8
2.4
Mean
WD level
(pphm)
3,1
5.6
5.7
6.6
3.3
2.0
4.0
3.2
4.2
% Change
(WE-WB)
- 59%
- 39%
- 44%
- 48%
- 47%
- 22%
- 27%
- 42%
- 42%*
Sun. vs. TWTF
Mean
WE level
(pphm)
1,1
3.0
3.0
3.0
1.7
1.2
3,0
1.7
2.2
Mean
WD level
(pphm)
3,0
5.6
5.7
6.7
3.3
2.0
4.2
3.3
4.2
% Change
(WE-WD)
- 64%
- 47%
- 47%
- 55%
- 47%
- 43%
- 28%
- 48%
- 48%*
The station-average percentage difference is computed from the
station averages of WE/WD concentration levels.
-------
level, under the normal WE/WD definition; while under the Sunday VIE/WD defini
tion, the mean weekend N02 level is 48% lower than the weekday level. The
percentage drop of 6-9 A.M. average N0« levels on weekends is slightly
smaller than that of 6-9 A.M. average NO levels under both the normal and
Sunday WE/WD definitions.
Unlike the NO case, the N02 concentration data for the Bethesda
station do not behave strangely. However; the weekend percentage drop of
6-9 A.M. average N02 levels is largest at the Bethesda station among
the eight stations examined. Under the normal definition, the Mamaroneck
station has the smallest percentage drop, 22%, on weekends, while under
the Sunday definition, the Newark station exhibits the smallest drop, 28%.
As seen from Figure 21, the t-test results of the WE/WD N02 data
indicate that, at all eight station sites, the WE/WD differences in
6-9 A.M. average N02 levels are significant under both the normal and
Sunday WE/WD definitions. The standardized values of WE/WD 6-9 A.M.
average N02 levels are plotted in Figure 22, showing that all the
stations have significantly lower N02 levels on weekends than on
weekdays.
5.2 WE/WD DIFFERENCES IN N0q/N0.. RATIOS
^^™^™^fc^^^"^"^™»» ^\
The preceding subsection has revealed that both NO and N02 concentra-
tions are significantly lower 1n the 6-9 A.M. period on the weekends than
1n the same period on weekdays. The remaining question 1s whether the
(N02/N0 ) ratio 1s also significantly different between weekends and
78
-------
P-Value
1.0
0.5
(Insignificant Region)
0.10
r
(Marginally Significant Region)
O.OZ
0.01
0.005
0.002
0.001
0.05
D
• „ p-value under Sunday HE/HP definition
r-j_- p-value under normal HE/HP
-------
(BA)*
(CA)*
1234567
Standardized Value* of Average WE 6-9 A.M. N02
"Standardized Value = mean// pooled variance
Fiaure 22. Graphical presentation of t-test results for WE/WD
Figure ;JV°?er.enceH1n 6.9 A.M. average NO concentrations
(Sun. vs. TWTF).
80
-------
weekdays. The important oxidant precursor, NO , consists mainly of NO
A
and N02. While NO scavenges ambient OX by reductive reactions, N02 pro-
motes OX formation by oxidation reactions. Therefore, the (N02/NOX)
ratio indicates how ambient NO works on ambient OX levels. Because N07
A t-
is formed by oxidation of NO, the (N0?/N0 ) ratio also may indicate the
^ A
age of the ambient '0 gas mixture. The greater the (N0?/N0 ) ratio, the
* £ A
older the pollutants measured in the air mass are likely to be. In actual
calculation of the ratio, we use CN02 + NO) as an approximate level of NOX.
In testing the WE/WD differences in 6-9 A.M. average (N02/NOX)
ratios, we encountered a difficulty: The t-test cannot be used to validate
such differences because the distribution of (N02/NOX) is a bounded one,
always lying between 0 and 1. Fortunately, the Wilcoxon rank sum test
(see Appendix B) is known to be efficient over a wide range of distribu-
tions, including a bounded one. We therefore applied the Wilcoxon rank
sum test to test observed WE/WD differences in 6-9 A.M. average (N02/NOX)
ratios at each of the eight stations.
The P-values of the Wilcoxon rank sum test for the eight stations
are plotted in Figure 23. Three stations, Camden, Newark, and Mamaroneck,
exhibit a significantly higher (N02/NOX) ratio on weekends, indicating
that the 6-9 A.M. weekend NOX gas mixture is more aged than the weekday
NO gas mixture at these sites. As in the case of NO concentrations, the
A
Bethesda station exhibits strange behavior in its 6-9 A.M. average
(N09/N0 ) ratios. Only this station shows that the weekend level of
£ A
6-9 A.M. average (N02/NOX) ratios is significantly lower than the weekday
level.
81
-------
00
ro
P-Value
1.0
0.5
0.10
0.05
0.02
0.01
0.005
0.002
0.001
0
O
(Nonsignificant Region)
(Marginally Significant Region)
*
HE level 1s higher than WO
WE level Is lower than WD
(Significant Region)
Figure 23. P-values for testing the WE/WD difference 1n N02/N0
ratio by the Wllcoxon rank sum test (Sun. versus TWTF)
-------
The WE/WD differences in 6-9 A.M. average (N09/NOV) ratios at each
u A
of the eight sta^ ^s are si^.narized in Tab! * 14. Because the (N02/NOX)
ratios are n t dis. 'ibuted normally, the mean values alone cannot indicate
the WE/WD d1 /erei. e in the (N09/NOY) ratios. Therefore, Table 14 lists
^ /\
both the mean and median values of 6-9 A.M. average (N02/NOX) on weekends
and weekdays.
Note th?t while the mean value at the Silver Spring station is lower
on weekends than on weekdays, the opposite is true for the median value.
The statistical test result shown in Figure 23 is consistent with the
WE/WD difference found in the median values. On the average, N02 accounts
for about 55% (59% in median) of NO on weekends and 59% (62% in median)
/\
on weekdays.
5.3 WE/WD DIFFERENCES IN 6-9 A.M. AVERAGE NMHC AND THC CONCENTRATIONS
Since methane is practically inert in photochemical reactions, non-
methane hydrocarbons (NMHC) are used as the organic precursor of photochemical
oxidants. Since adequate NMHC data are found only at four locations, total
hydrocarbon (THC) data at two other locations are also included in the analysis
to supplement the insufficient area coverage of NMHC data. THC data may be
an indicator of NMHC when viewed on a relative basis. Because of the known
unreliability of NMHC measurements at lower concentration levels and because
of the poor area coverage of hydrocarbon data, the results of this section's
analysis should be viewed with caution.
83
-------
TABLE 14. WE/WD DIFFERENCES IN 6-9 A.M. AVERAGE
(N02/NOX) RATIOS
Station
Name
Bethesda
Silver Spring
Suitland
Baltimore
Camden
Mamaroneck
Newark
Welfare Is.
8 -Station
Avg.
Sun vs. TWTF
Mean
VIE level
0.242
0.482
0.752
0.808
0.611
0.448
0.618
0.451
0.552
Mean
WD level
0.641
0.496
0.800
0.691
0.584
0.416
0.533
0.594
0.594
% Change
WE-UD
- 62%
- 3%
- 6%
+ 17%
+ 5%
+ 8%
+ 16%
- 24%
- 7%*
Median
WE level
0.163
0.500
0.911
0.889
0.728
0.433
0.732
0.342
0.587
Median
WD level
0.666
0.430
0.933
0.709
0.633
0.417
0.550
0.617
0.619
% Change
WE-WD
- 76%
+ 16%
- 2%
+ 25%
+ 15%
+ 4%
+ 33%
- 45%
- 5%*
The station-average percentage difference is computed from the
station averages of WE/WD concentration levels.
84
-------
5.3.1 WE/WD Differences in 6-9 A.M. Average NMHC Levels
Adequate NMHC data are found only at four monitoring sites:
Suit!and, Silver Spring, Baltimore, and Norristown. The first two
sites are in the vicinity of Washington, D.C. The last site is located
northwest of Philadelphia (see Figure 2). The weekend mean and weekday
mean of 6-9 A.M. average NMHC concentrations, as well as the percentage
difference of (WE-WD), are computed for the four sites under both the
normal and Sunday WE/WD definitions. The results are summarized in Table 15.
Under the normal WE/WD definition (SS versus MTWTF), all four stations
exhibit a considerable drop in 6-9 A.M. average NMHC levels on weekends,
ranging from 24% at Norristown to 65% at Suitland. Under the Sunday WE/WD
definition (Sun. versus TWTF), three stations, Suitland, Silver Spring,
and Baltimore, exhibit a measurable decrease in 6-9 A.M. NMHC levels
on weekends, while the Norristown station shows a slight increase on
weekends. The four-station average of weekend reductions in 6-9 A.M.
average NMHC concentrations is 48% under the normal WE/WD definition and
32% under the Sunday WE/WD definition.
Some of the NMHC data exhibit a large difference in variance
between the weekend and weekday data sets. The ordinary t-test is based
on the assumption of an equal variance for the two populations. The
assumption of equal variance is questionable for the NMHC data being
examined. Therefore, the Welsh t-test, which was designed to determine
the significance of differences between two populations with unequal
variances [Snedecor and Cochran 1976] is used to correct the bias of an
85
-------
TABLE 15. WE/WD DIFFERENCES IN 6-9 A.M. AVERAGE NMHC CONCENTRATIONS
Station Name
Suitland
Silver Spring
Baltimore
NorHstown
4-Stat1on
A.vg.
SS vs. MTWTF
# Valid
Obs. Dayst
191
207
199
213
Mean
WE level
(pphm)
30
49
56
87
55
Mean
VID level
(pphm)
87
69
153
114
106
% Change
(WE-WD)
- 65%
- 29%
- 64%
- 24%
- 48%*
Sun. vs. TWTF
# Valid
Obs. Dayst
141
148
146
159
Mean
WE level
(pphm)
31
56
60
121
67
Mean
WD level
(pphm)
74
61
154
109
99
% Change
(WE-WD)
- 58%
- 8%
- 61 X
+ 11%
. 32%*
00
A "valid observation day" is defined as a day which has at least 2 out of 3 possible observations
during the 6-9 A.M. period.
The station-average percentage difference is computed from the station averages of WE/WD
concentration levels.
-------
ordinary t-test caused by the large unequal variances, whenever (maximum
variance/minimum variance) >_ C. Here C is the critical value which depends
on the sample sizes and can be found from the above reference.
The results of statistical testing of the WE/WD differences in
6-9 A.M. average NMHC concentrations are summarized in Table -J6. At the
Suitland and Baltimore stations, the WE/WD differences in 6-9 A.M. average
NMHC concentrations are significant; at the other two stations, Silver
Spring and Norristown, the WE/WD differences are not statistically
significant. The above conclusions are consistent under both normal
and Sunday WE/WD definitions.
5.3.2 WE/WD Differences 1n 6-9 A.M. Average THC Levels
WE/WD differences 1n 6-9 A.M. average THC concentrations are
examined for only two sites, Bethesda, Maryland, and Camden, New Jersey.
The results are summarized 1n Table 17. The weekend mean of 6-9 A.M.
average THC concentrations is lower than the weekday mean by 13% at
Bethesda and 27% at Camden. The average percentage decrease (21%) in
THC over the two stations is smaller than that (48%) in NMHC, which
implies that ambient THC contains a considerable amount of methane,
whose concentration levels change little between weekends and weekdays.
The t-test indicates that the WE/WD difference in 6-9 A.M. average
THC levels is significant at both stations.
87
-------
TABLE 16. STATISTICAL SIGNIFICANCE OF WE/WD DIFFERENCES IN
6-9 A.M. AVERAGE NMHC CONCENTRATIONS
Station Name
SuWand
Silver Spring
Baltimore
Nor r 1s town
WE/WD Definition
New WE/WD
Old WE/WD
New WE/WD
Old WE/WD
New WE/WD
Old WE/WD
New WE/WD
Old WE/WD
Max, Var/M1n. Var *
5.132
11,565
1.660
1.072
6.014
6.960
1.192
1.156
Test
Welsh t-test
Welsh t-test
t-test
t-test
Welsh t-test
Welsh t-test
t-test
t-test
P-Value
0.005 (Significant)
0.000
0.832 (Non-Significant)
0.275
0.000 (Significant)
0.000
0.728 (Non-Significant)
0.283
*When Max. Var/M1n Var. > 1.8, the Welsh t-test 1s used to correct the bias of ordinary t-test
caused by large unequal""variances.
-------
TABLE 17. HE/WD DIFFERENCES IN 6-9 A.M.
AVERAGE THC CONCENTRATIONS
Station
Name
Bethesda
Camden
2-Station
A VOL
« Valid
Obs. Dayst
175
160
SS vs. *TW7F
Mean
WE level
(pphm)
174
185
180
Mean
WD level
{pphm)
201
255
228
'"r. Change
(HE-WD)/WDxlOO
- 13S
- 27%
- 212
KE/WD Difference
P-Value
0.04 (Significant)
0.0005 (Significant)
A "valid observation day" is defined as a day which has at least 2 out of 3
possible observations during the 6-9 A.M. period.
89
-------
5.4 WE/HP DIFFERENCES IN CO CONCENTRATIONS
Carbon monoxide (CO) is not an oxidant precursor but 1s an inert
gas emitted primarily by automobiles. Since ambient CO data are rela-
tively complete, the CO data are used here as an indicator pollutant of
the two precursor emissions, NO and NMHC, from mobile sources. Adequate
/\
CO data are found at as many as 17 monitoring sites, names and site
codes for which are listed in Table 18.
The WE/WD differences in CO concentrations are examined for three
overlapping time periods: 3-hr (6-9 A.M.), daytime (6 A.M.-9 P.M.),
and 24-hr. The intent 1s to estimate the temporal variation of WE/WD
differences in automotive emissions. The average CO concentrations
over the three different time periods are computed for weekends,
defined as Saturday and Sunday (SS), and for weekdays, defined as
Monday through Friday (MTWTF).
Table 19 summarizes the WE/WD differences in CO concentrations
computed for the 6-9 A.M., daytime (6 A.M.-9 P.M.), and 24-hr periods.
Except for the Camden County station, the stations exhibit a considerable
decrease in CO concentrations on weekends during all three time periods.
The Camden County station shows the smallest WE/WD differences in CO
concentrations among the 17 stations and exhibits a small increase (1%)
in the 24-hr average CO levels. The Camden County site is located on
the grounds of the New Jersey State Hospital, which could explain why
WE CO levels are not much lower than WD levels.
90
-------
TABLE 18. MONITORING STATIONS USED FOR ANALYZING
CARBON MONOXIDE AIR QUALITY DATA
Metropolitan
Area
Washington,
D.C.
Baltimore
Philadelphia
New York
City
Boston
Station
Name
Bethesda
Silver Spring
Suitland
Baltimore
Lancaster
Camden
Camden Co.
Norri stown
Philadelphia
Asbury Park
Babylon
Mamaroneck
Newark
Welfare Is.
Medford
Quincy
Worchester
SAROAD I.D.
210200005 F01
211480006 F01
211560001 F01
210120018 F01
394660007 F01
310720003 F01
310740001 F01
396540013 F01
397140026 F01
310060001 F01
330280002 F01
334100002 F01
313480002 F01
334680050 F01
221220003 F01
221880002 F01
222640012 F01
EPA's
Site Code*
S-R
S
S
CC-C
S-R
S-R
Ru-A
S-R
CC-C
CC-C
S-I
CC-C
CC-C
CC
S-I
S-I
CC-C
Number of Val id
Observation Dayst
266
283
277
299
236
224
219
202
213
210
145
272
198
247
433
377
441
Code key:
A - Agricultural
C - Commercial
CC - Center City
I - Industrial
R - Residential
Ru - Rural
S - Suburban
A "valid observation day" is a day which has at least 10 observations for
the 6 A.M.-9 P.M. period. For the 6-9 A.M. average concentration, fewer
observation days may result because of stricter criteria: 2 out of 3
possible observations.
91
-------
TABLE 19. WE/WD DIFFERENCES IN CO CONCENTRATIONS DURING
THE 3-HR (6-9 A.M.). DAYTIME (6 A.M.-9 P.M.),
AND 24-HR PERIODS
Metropolitan
Area
Washington,
D.C.
Baltimore
Philadelphia
New York
City
Boston
Station
Name
Bethesda
Silver Spring
Suit! and
Baltimore
Lancaster
Camden
Camden Co.
Norristown
Philadelphia
Asbury Park
Babylon
Mamaroneck
Newark
Welfare Is.
Medford
Quincy
Worchester
!7-Stat1on
Avg.
6-9 A.M. Average
Mean
WE level
(ppm)
0.4
0.9
1.1
0.6
0.6
1.3
1.6
0.9
1.5
2.5
1.2
2.6
2.3
1.8
1.7
1.9
1.9
1.46
Mean
WD level
(ppm)
0.8
1.3
2.2
1.9
1.1
1.9
1.7
1.9
2.9
2.9
1.8
3.2
3.6
2.0
3.3
2.6
3.0
2.24
% Change
(WE-WD)
- 50%
- 31%
- 50%
- 68%
- 46%
- 32%
- 6%
- 53%
- 48%
- 14%
- 33%
- 19%
- 36%
- 10%
- 49%
- 27%
- 37%
- 35%*
Daytime
Average
% Change
(WE-WD)
- 24%
- 33%
- 35%
- 42%
- 38%
- 25%
- 1%
- 43%
- 30%
- 12%
- 22%
- 15%
- 22%
- 14%
- 27%
- 14%
- 29%
- 25% *
24-Hr Avg.
% Change
(WE-WD)
- 19%
- 19%
- 11%
- 24%
- 23%
- 15%
+ 1%
- 29%
- 15%
- 5%
- 20%
- 8%
- 10%
- 6%
- 19%
- 10%
- 20%
- 15% *
The station-average percentage difference 1s computed from the
station averages of WE/WD concentration levels.
92
-------
The percentage reduction of CO concentrations on weekends is the
greatest during the 6-9 A.M. period and the smallest during the 24-hr
period. Of the station averages, the 6-9 A.M. average CO concentration
on weekends is 35% lower than that on weekdays; the daytime average CO
concentration is 25% lower than that on weekdays; and the 24-hr average
CO level is 15% lower.
All stations except Welfare Island and Silver Spring exhibit the
largest weekend drop in the 6-9 A.M. average CO concentration, followed
by the daytime average and then by the 24-hr average. At the Welfare Island
and Silver Spring stations, the weekend drop in daytime average concentra-
tions is slightly greater than the drop in the 6-9 A.M. average which,
in turn, is greater than that in the 24-hr average reductions. Because
of the smaller weekend drops in CO concentration during the non-6-9 A.M.
time periods at all the stations other than the above two, the weekend
decreases in 6-9 A.M. precursor-pollutant (NMHC, NO, and NO?) concentrations
should not be interpreted directly as the same decreases in precursor
emissions. The actual decreases in precursor emissions on weekends are
likely to be smaller than those found in the 6-9 A.M. average precursor
concentrations.
The t-test results of the WE/WD differences in CO concentrations
during the three time periods are plotted in Figure 24, using P-values.
As seen from Figure 24, the WE/WD differences in both the 6-9 A.M. average
CO concentrations and daytime average CO concentrations are significant at
15 out of the 17 sites examined. The two stations that do not exhibit a
93
-------
P-Value
1.0
0.5
0.01
0.05
0.01
0.005
0.001
—
(Noi
(Ma,
(Si
*
j
A •
• a
+ +
significant
ginally Sigr
inificant Re<
4 *
*
*
legion)
if leant Regi
Ion)
t •
on)
A
t *
A
A
a
. - ^
• -
a —
A
^
B
+
6-9 A.M. Av
Daytime Avg
24-hr Avg
1 1 1 1 1 1 1 I 1 1 1 1 1 1
gocuiae *£
-------
significant WE/WD difference in CO concentrations are Welfare Island,
New York, and Camden County, New Jersey. All 15 stations with signifi-
cant WE/WD differences have, of course, a lower CO concentration on
weekends.
In the 24-hr average CO concentrations, 11 out of 17 stations
exhibit a significant WE/WD difference, with two additional stations
having a marginally significant difference.
95
-------
6. RESULTS AND CONCLUSIONS
The preceding sections discuss the WE/WD differences in observed daily
maximum oxidant levels (Section 3), in meteorologically adjusted oxidant
levels (Section 4) and in 6-9 A.M. average concentrations of oxidant
precursors and carbon monoxide (Section 5). This section reviews the
analysis results of the preceding sections and discusses the implications
of those results for oxidant sensitivity to WE/WD changes in precursor
emissions. In Section 6.1, the oxidant-precursor relationship in urban core
areas is examined, and that in source/receptor areas is examined in Section 6.2.
In Section 6.3, the oxidant response to WE/WD emission changes is compared
with that of long-term emission changes.
6.1 OXIDANT-PRECURSOR RELATIONSHIP IN URBAN CORE AREAS
NO, N02> CO, and OX data are available for eight stations: Bethesda,
Silver Spring, Suitland, Baltimore, Camden, Newark, Welfare Island, and
Mamaroneck. Furthermore, three stations, Silver Spring, Suitland, and
Baltimore, have NMHC data. These eight stations are used to examine the
relationship among the WE/WD differences found in the five pollutants.
Table 20 summarizes means of daily maximum OX levels and 6-9 A.M.
average concentrations of NO, N02, NMHC, and CO under the normal WE/WD
definition. All eight stations listed in Table 20 are located within or
near the large metropolitan areas (see Figure 2). Therefore, the oxidant-
precursor relationships at the eight stations may be representative of
those in urban core areas. In Table 20, large WE/WD differences are found
96
-------
TABLE 20 . WE/WD MEANS OF DAILY MAXIMUM OX CONCENTRATIONS
AND 6-9 A.M. AVERAGE PRECURSOR AND CO
CONCENTRATIONS AT URBAN CORE SITES (SS vs. MTWTF)
Station
Name
Bethesda
Silver Spr.
Suit! and
Baltimore
Camden
Mamaroneck
Newark
Welfare Is.
8-Statlon
Avg.
Daily Max
OX (ppb)
WE WD
70
52
73
54
68
49
56
62
61
77
59
79
59
70
54
49
61
64
6-9 A.M. Avg.
NO (pphm)
WE WD
1.5
5.3
0.5
1.0
0.9
1.9
2.2
1.4
i.a
0.7
12.8
1.3
4.6
5.4
3.2
6.3
2.4
4.6
6-9 A.M. Avg.
NO? (pphm)
WE WD
1.3
3.4
3.2
3.4
1.8
1.5
2.9
1.8
2.4
3.1
5.6
5.7
6.6
3.3
2.0
4.0
3.2
4.2
6-9 A.M. Avg.
NMHC (pphm)
WE WD
49
30
56
45
69
87
153
103
6-9 A.M. Avg.
CO (ppm)
WE WD
0.4
0.9
1.1
0.6
1.3
2.6
2,3
1.8
1.4
0.8
1.3
2.2
1.9
1.9
3.2
3.6
2.0
2.1
-------
in 6-9 A.M. average concentrations of OX precursors and CO, but much smaller
WE/WD differences are found in daily maximum OX levels. Another tendency
is apparent: The higher the 6-9 A.M. NO level, the lower the daily maximum
OX level. This tendency is noted among stations but is not always noted
between weekends and weekdays.
The percentage changes in each pollutant concentration between
weekends and weekdays are computed and summarized in Table 21. On the
average, the weekend level of 6-9 A.M. average NO concentrations is 60%
lower than the weekday level, while the N02 weekend level is 42% lower.
Since NO concentration is approximately equal to the sum of NO and N09
A •
concentrations, the weekend level of 6-9 A.M. NOV concentrations should
J\
be lower than the weekday level by some similar percentage, say, 50%.
NMHC data are available for three stations only: Silver Spring,
Sultland, and Baltimore. The three-station average NMHC level in
6-9 A.M. concentrations is 56% lower on weekends than on weekdays.
Because fewer stations were used, an eight-station average decrease in
6-9 A.M. NMHC concentration on weekends may be different from the three-
station average value. In any case, the weekend precursor levels are
lower than the weekday levels by about 50% 1n 6-9 A.M. average NO con-
A
centrations and 50% to 60% lower in 6-9 A.M. average NMHC concentrations.
Under the above decreases in 6-9 A.M. precursor concentrations, the
eight-station average drop in daily maximum OX concentrations on weekends
1s only 5%.
98
-------
TABLE 21. PERCENTAGE CHANGES (WE-WD) IN DAILY
MAXIMUM OX CONCENTRATIONS AND 6-9 A.M.
AVERAGE PRECURSOR AND CO CONCENTRATIONS
AT URBAN CORE SITES (SS vs. MTWTF)
Station Name
Bethesda
Silver Spring
Suit! and
Baltimore
Camden
Mamaroneck
Newark
Welfare Is.
8-Stat1on
Avg.*
Daily Max
OX
- 9%
- 13%
- 8%
- 9%
- 2%
- 8%
+ 13%
+ 1%
- 5%
6-9 A.M. Average
NO
+134%
- 58%
- 60%
- 79%
- 83%
- 40%
- 65%
- 43%
- 60%
N02
- 59%
- 39%
- 44%
- 48%
- 47%
- 22%
- 27%
- 42%
- 42%
NMHC
- 29%
- 65%
- 64%
- 56%
CO
- 50%
- 31%
- 50%
- 68%
- 32%
- 19%
- 36%
- 10%
- 35%
*The station-average percentage difference is computed from the station
averages of WE/WD concentration levels listed in Table 20.
9.9
-------
Table 22 summarizes means of daily maximum oxidant concentrations
and of 6-9 A.M. average precursor concentrations for the Sunday WE/WD
definition. As noted in Figures 6 and 7, the WE/WD differences in
daily max OX levels are slightly greater for the Sunday definition than
for the normal definition. However, a similar magnification of the
WE/WD differences is not seen in the precursor-pollutant levels.
Table 23 briefly lists the percentage differences for oxidants
and precursors using the Sunday WE/WD definition. Except for NO at the
Bethesda station, the 6-9 A.M. precursor levels are consistently lower
on weekends among the eight stations. The daily maximum oxidant levels
are also consistently lower on weekends except for the Newark station,
which exhibits an increase (12%) in daily max OX levels. The station
average reductions on weekends are 9% for daily max OX levels and 67%,
48%, and 49%, respectively, for 6-9 A.M. average NO, N02, and NMHC levels.
Since only three stations have NMHC values, the average percentage reduc-
tion for NMHC, 49%, is a less reliable estimate than those for NO and N02-
A comparison of Tables 21 and 23 indicates that, in 6-9 A.M. average
NO and N0« levels, the percentage decreases for Sundays alone are 12% to
14% greater than those for Saturdays and Sundays combined. In daily maximum
OX levels, the percentage decrease for Sundays only is nearly twice as large
as that for Saturdays and Sundays combined. For NMHC, the percentage
decrease on Sundays is slightly smaller than that on Saturdays and Sundays
combined. However, because of the small number of stations (three) used,
such a difference may be due to sample fluctuation.
100
-------
TABLE 22 . WE/WD MEANS OF DAILY MAX OX AND OF 6-9 A.M.
AVERAGE LEVELS OF NO, NO,, AND NMHC:
SUN. VERSUS TWTF *
Station
Name
Bethesda
Silver Spr.
Suitland
Baltimore
Camden
Mamaroneck
Newark
Welfare Is.
8-Station Avg.
Daily Max
OX(ppb)
WE WD
64
51
74
48
68
46
56
57
58
76
59
79
57
69
54
49
63
63
6-9 A.M. Average Concentration
NO (pphm)
WE WD
2.33
3.63
0.10
0.87
0.85
1.37
1.84
1.32
1.54
0.66
13.00
1.43
4.38
5.39
3.39
6.58
2.39
4.65
N02 (pphm)
WE WD
1.09
3.00
3.05
2.97
1.75
1.16
3.02
1.69
2.22
3.00
5.61
5.73
6.66
3.31
2.05
4.18
3.27
4.23
NMHC (pphm}
WE WD
56
31
60
49
61
74
154
96
-------
TABLE 23. PERCENTAGE DIFFERENCE (WE - UD) IN DAILY MAX OX
AND IN 6-9 A.M. AVERAGE LEVELS OF NO, N0?) AND
NMHC: SUN. VERSUS TWTF
-
Station Name
Bethesda
Silver Spring
Suitland
Baltimore
Camden
Mamaroneck
Newark
Welfare Is.
*
8-Station Avg.
OX
- 17%
- 13%
- 6%
- 16%
- 2%
- 14%
+ 12%
- 9%
- 9%
NO
+252%
- 72%
- 93%
- 80%
- 84%
- 60%
- 72%
- 45%
- 67%
NO,
- 64%
- 47%
- 47%
- 55%
- 47%
- 43%
- 28%
- 48%
- 48%
NMHC
- 8%
-59%
-61%
-49%
The station-average percentage difference is computed from the station
averages of WE/WD concentration levels.
102
-------
Regardless of the WE/WD definition used, WE oxidant reductions at
the urban sites are not nearly as large as those associated with early-
morning precursors. To some degree, this may be expected. In the
source-intensive urban areas, emissions are generally high throughout
a weekday due to traffic. Because NO is a very effective scavenger of
oxidants, it is very likely that oxidant levels in the urban area may
actually be suppressed during the week. (Recall from Section 3 that
weekday maximum daily oxidant levels within urban areas were generally
lower than those outside.) The early-morning reductions in NO and the
daytime reductions in CO on weekends indicate that NO levels in the urban
areas may very well be lower throughout the weekend, leaving less NO
available for oxidant scavenging. Consequently, in the urban core, any
reductions in precursor emissions and subsequent oxidant formation upwind
of the urban area may be offset by the lowered scavenging effect on
weekends.
6.2 OXIDANT-PRECURSOR RELATIONSHIP IN DOWNWIND AREAS
The preceding subsection discusses the relationship between weekend
oxidant reductions and weekend precursor reductions at the eight urban
core sites. However, because a higher oxidant level is usually found in
the downwind area rather than in the urban source area, it would also be
useful to contrast weekend reductions in urban precursor levels with oxidant
reductions in downwind areas.
103
-------
By referring to Figure 2, we have formed source-receptor pairs in
Table 24 to compare the weekend precursor reduction at the source site
with the weekend oxidant reduction at the downwind receptor site. Because
ambient precursor data are available at fewer sites than oxidant data, each
source site is paired with more than one downwind receptor site. By
considering the location of each monitoring station and the availability
of precursor data, we chose to use the Baltimore station as the urban-
source site for the Washington, D.C.-Baltimore metropolitan area and the
Camden station for the Philadelphia metropolitan area. The Newark-New York
and Boston metropolitan areas are represented by the Newark and Medford
stations, respectively. The downwind receptor stations listed in Table 22
were selected by considering the geographical location and the WE/WD
oxidant behavior (a consistently lower weekend level is used as the
criterion).
Table 24 indicates that while the 6-9 A.M. weekend urban precursor
levels are 64% to 83% lower in the four metropolitan areas (five if the
Washington, D.C., and Baltimore areas are counted separately), the weekend
downwind oxidant levels are only 7% to 14% lower. On the average over the
four regions, the NO and NMHC levels are 75% and 64% lower, respectively, on
weekends. Those reductions in 6-9 A.M. precursor levels bring about
11%-12% reductions in daily maximum oxidant levels. In Table 24, percentage
reductions in 6-9 A.M. and daytime (6 A.M.-9 P.M.) CO levels are also
presented to facilitate some estimate of precursor reductions for those
104
-------
TABLE 24. WEEKEND REDUCTIONS IN 6-9 A.M. URBAN PRECURSOR
LEVELS AND DOWNWIND DAILY MAX OX LEVELS
(SS vs. MTWTF)
Metropolitan
Area
Washington D.C.-
Baltimore
Philadelphia
Newark-
New York
Boston
4-Region
Avg.a
% Changes in Urban Precursor Levels
Station 6-9 AM 6-9 AM 6-9 AM
Name NO NMHC CO
Baltimore
Camden
Newark
Medford
-79%
-83%
-65%
NA
-75%
-64%
NA
NA
NA
-64%
-68%
-32%
-36%
-49%
-45%
Daytiiud
CO
-42%
-25%
-22%
-27%
-27%
% Changes in Downwind
Station
Name
Bethesda .
Gaithersburg
Silver Spring
Suitlandb
Lancaster
Norristown
Camden Co.
Greenwich
Middletown
Dan bury
Mamaroneck
Salem .
Quincy
OX
-9%
-4%
-13%
-8%
-10% J
-9% )
-14% J
-15% )
-11% 1
-18%
-8%
OX Levels
Regional
Avg.a
-9%
(-14%)c
-14%
-12% \ -7%
-4% ) (-12%)c
-11% c
NA
a.
b.
c.
Data are not available.
The regional average percentage difference is computed from the station
averages of WE/WD concentration levels.
These stations are located somewhat off the typical downwind area.
Regional average value when nontypical downwind stations are excluded.
-------
stations at which data are not available for one or both of the two
precursor pollutants. The weekend reductions in CO levels are somewhat
smaller than those in the two precursors. However, their percentage
reductions are still much greater than those in downwind oxidant levels.
The ll%-to-12% reduction in downwind oxidant levels is small, compared
with the 64%-to-75% reduction in 6-9 A.M. urban precursor levels. However,
the percentage reduction in downwind oxidant levels is more than twice
as large as that in urban oxidant levels (see Table 21). These greater
responses of downwind oxidant levels to changes in precursor levels can
be seen from Table 25, where percentage changes in weekend oxidant levels
are computed for stations in each of three types of regions: upwind areas,
metropolitan areas, and downwind areas. Although the grouping of stations
into each of the three types of region is somewhat arbitrary, the following
general pattern emerges from such a spatial grouping of stations:
• The oxidant levels in urban source-intensive areas are
relatively low, and the weekend reductions are also small.
• The oxidant levels in areas downwind of a large metropolitan
area tend to be higher than those within the metropolitan area.
The weekend reductions in these downwind areas are also much
greater than those within the metropolitan area.
t The oxidant levels and weekend reductions in the upwind area
are slightly higher than those in the metropolitan area
itself, but are substantially lower than those in the
downwind area.
The above results should be viewed with some caution. The analysis
of precursor data indicated substantial reductions in 6-9 A.M. concentra-
tions in urban areas on weekends. These 6-9 A.M. concentrations are thought
106
-------
TABLE 25. WE/WD CHANGES IN DAILY MAXIMUM OXIDANT LEVELS
IN REGIONS UPWIND OF, DOWNWIND OF, AND WITHIN
THE LARGE METROPOLITAN AREAS (SS VS. MTWTF)
Region
Upwind
Areas
Metro-
politan
Areas
Downwi nd
Areas
Station Name
Bethasda (BE)*
Gaithersburg(GA)
Asbury Park (AP)
Worchester(WO)
Norristown(NO)*
Suitland (SU)*
Baltimore (BA)
Philadelphia(PH)
Camden (CA)
Newark (NE)
Welfare Is.(WI)
Mamaroneck(MA)*
Medford (ME)
Quincy (QU)*
SilverSpring(SS)
Lancaster (LA)
Camden Co.(CC)
Babylon (BB)*
Greenwich (GR)
Dan bury (DA)
Middletown(MI)
Salem (SA)
Regional Average
% Difference % Difference WE Level WD Level
(WE - WD) (ppb) (ppb)
-9
-4
-4
-3
-9
-8
-9 }
+2
-2
+13
+1
-8
+2
-4 '
-13
10
-14
-7
-15
-18
-11
-12 )
-7% 65 69
-2% 59 60
-12% 67 76
This station could be classified differently.
107
-------
to be indicative of the oxidant-forming potential of a city. However,
due to the limited spatial resolution of precursor data, the changes in
concentrations outside the urban area are unknown. Precursor emissions
occurring outside the urban area can affect the oxidant levels formed on
any given day. A second precautionary note is that the 6-9 A.M. reduc-
tions in weekend precursor levels may be misleadingly high as a result
of changes in temporal emission patterns. As Table 19 suggests, reductions
in weekend CO levels are considerably less than implied by the 6-9 A.M.
levels alone.
6.3 ME/WD OXIDANT CHANGE VS. LONG-TERM OXIDANT CHANGE
By comparing the WE/WD oxidant change with the long-term oxidant
change occurring in the Los Angeles region, this subsection gives a brief
overview of the effectiveness of simultaneous control of NOX and NMHC for
controlling oxidant air pollution. The hydrocarbon controls employed in
the Los Angeles area in the late 1960s brought about a remarkable oxidant
improvement in the source-intensive areas [Horie and Trijonis 1977]. The
152-to-18% reduction in NMHC and the 35%-to-36% increase in NO brought
A
over a 30% improvement in daily maximum oxidant levels during the 1965-1974
period. Over that 10-year period, the oxidant improvements in the source-
intensive western portion of the basin were greater than those in the
downwind, eastern portion of the basin [Trijonis et al. 1976].
On weekends, the eight-station average 6-9 A.M. NO and N02 levels are,
respectively, 60S and 42% lower than the weekday level, while the three-station
average 6-9 A.M. NMHC level is 56% lower than the weekday level (Table 21).
108
-------
These reductions in morning precursor levels brought about a ¥• (Saturdays
and Sundays) to 9% (Sundays) improvement in daily maximum oxidant levels in
the source-intensive metropolitan areas. A slightly greater improvement
(about 12%) occurred in areas downwind of the metropolitan areas.
Compared with the Los Angeles example, the weekend oxidant improvement
that is brought about by a simultaneous reduction in NO and NMHC is much
/\
less dramatic. In Los Angeles, there has been a reduction in NMHC and an
increase in NO , thereby resulting in a lower NMHC/NO ratio. These three
A -~' X
factors appear to have contributed synergistically to the 30% reduction in
daily maximum ox'dant levels. On the other hand, about the same percentage
reductions in both NMHC and NO occurred on weekends, thereby leaving the
J\
NMHC/NO ratio virtually unchanged. Therefore, the reduction in NMHC
^
alone contributed to lowering oxidant levels on weekends.
The above examples, however, should not be used to compare the
effectiveness of hydrocarbon controls and that of simultaneous control
of NO and NMHC. Because the Los Angeles area has higher oxidant levels
J\,
than are found in the northeast United States, a relative contribution of
transported oxidant to the ambient oxidant level would be smaller in the
Los Angeles area than in the northeast. As a result, the effect of the
same oxidant control on ambient oxidant level will show up more in the
Los Anqeles area than in the northeast United States.
109
-------
7. REFERENCES
Breiman, L. and W. Meisel (1976), "The Change in Ozone Levels Caused by
Precursor Pollutants: An Empirical Analysis," Proceedings of the EPA
Conference on Environmental Modeling and Simulation, April 19-22, 1976,
Cincinnati, Ohio.
California Air Resources Board (1974), "Weekday vs Weekend Oxidant Concen-
trations," California Air Quality Data, Vol. VI, pp. 5-6.
Cleveland, W. S., B. E. Kleiner and J. L. Warner (1974a), "Using Robust
Statistical Methods in Analyzing Air Pollution Data with Applications to
New York-New Jersey Photochemistry," Presented at the 67th Annual Meeting
of the Air Pollution Control Association, June 9-13, 1974, Denver, Colorado.
Cleveland, W. S., T. E. Graedel, B. Kleiner, and J. L. Warner (1974b), "Sunday
and Workday Variations in Photochemical Air Pollutants in New Jersey and
New York," Science, Vol. 186, pp. 1037-1038.
Cleveland, W. S. and J. E. McRae (1977), "Weekday-Weekend Ozone Concentrations
in the Northeast United States," Presented at the Conference on Air Quality,
Meteorology and Atmospheric Ozone, July 31-August 6, 1977, Boulder, Colorado.
Eldon, J. A., J. C. Trijonis and K. Yuan (1978), "Statistical Oxidant/Precursor
Relationships for the Los Angeles Region, Final Report: Creation of Empirical
Oxidant/Precursor Models," Prepared for the California ARB by Technology
Service Corporation, Contract #A5-020-87.
Elkus, B. and K. R. Wilson (1977), "Photochemical Air Pollution: Weekend-
Weekday Differences," Atmos. Environ., Vol. 11, pp. 509-515.
Horie, Y., A. S. Chaplin, and E. D. Helfenbein (1977), "Population Exposure to
Oxidants and Nitrogen Dioxide in Los Angeles, Volume II: Weekday/Weekend and
Population Mobility Effects," U.S. EPA Publication No. EPA-450/3-77-004b.
Horie, Y. and J. Trijonis (1977), "Population Exposure to Oxidants and
Nitrogen Dioxide in Los Angeles, Vol IV: Analysis and Interpretation of
Trends," U.S. EPA Publication No. EPA-450/3-77-004d, U.S. EPA/Office of
Air Quality Planning and Standards, Research Triangle Park, North Carolina.
Jeffries, H. E., J. E. Sickles, II, and L. A. Ripperton (1976), "Ozone
Transport Phenomena: Observed and Simulated," Presented at the APCA meeting
in Portland, Oregon, June 1976.
Levitt, S. B. and D. P- Chock (1975), "Weekday-Weekend Pollutant and Meteoro-
logical Studies of the Los Angeles Basin," Presented at the APCA Annual
Meeting, June 15-20» 1975.
no
-------
Ludwig, F. L. and E. Skelar (1977), "Ozone in the Northeastern United States,"
U.S. EPA Publication No. EPA 901/9-76-007, U.S. EPA/Office of Air Quality
Planning and Standards, Research Triangle Park, North Carolina.
Peterson, J. T. and E. C. Flowers (1977), "Interactions Between Air Pollution
and Solar Radiation," Solar Energy. Vol. 19, pp. 23-32.
Snedecor, G. W. and W. G. Cochran (1976), Statistical Method, Sixth Edition.
Ames, Iowa: The Iowa State Univ. Press, pp. 115-118.
Sonquist, J. A., E. L. Baker and J. N. Morgan (1973), "Searching for
Structure," University of Michigan, The Institute for Social Research.
South Coast AQMD (1976), "Preliminary Emissions Inventory and Air Quality
Forecast 1974-1995," South Coast Air Quality Management District, El Monte,
California.
Trijonis, J. C. et al. (1976), "Emissions and Air Quality Trends in the South
Coast Air Basin," EQL Memo No. 16, Calif. Inst. of Tech., Pasadena, California.
Trijonis, J. C., and K. Yuan (1977), "Visibility in the Southwest,"
Final Report prepared by Technology Service Corporation for the U.S. EPA,
Grant No. 802815.
Trijonis, J. C. and K. Yuan (1978), "Visibility in the Northeast,"
Final Report prepared by Technology Service Corporation for the U.S. EPA-
Grant No. 802815.
U.S. Dept. of Commerce (1973), 1970 Census of Population: Graphic Summary
of the 1970 Population Census, PC(Sl)-55, Bureau of the Census, Social and
Economic Statistics Administration, Washington, D.C.
U.S. EPA (1977), "National Air Quality and Emissions Trends Report, 1976,
U.S. EPA Publication No. EPA-450/1-77-002, U.S. EPA/Office of Air Quality
Planning and Standards, Research Triangle Park, North Carolina.
Wight, G. D., G. T. Wolff, P. J. Lioy, R. E. Meyers and R. T. Cederwall (1977),
"Formation and Transport of Ozone in the Northeast Quadrant of the U.S.,"
Proceedings of the ASTM Conference on Air Quality, Meteorology and Atmospheric
Ozone, August 3-6, 1977, Boulder, Colorado.
Wolff, G. T., P. J. Lioy, R. E. Meyers, R. T. Cederwall, G. D. Wight, R. E.
Pasceri and R. S. Taylor (1977), "Anatomy of Two Ozone Transport Episodes in
the Washington, D.C. to Boston, Mass., Corridor," Environ. Sci. Techno!.,
Vol. 11, pp. 506-510.
Zeldin, M. D., and W. S. Meisel (1978), "Use of Meteorological Data in Air
Quality Trend Analysis," U.S. EPA Publication No. EPA-450/3-78-024, U.S.
EPA/Office of Air Quality Planning and Standards, Research Triangle Park,
North Carolina.
Ill
-------
APPENDIX A
STUDENT T-TEST AND WELSH T-TEST
Let x-j and Xg be sample means of n. weekend values and ng weekday
2 2
values, respectively, and let s^ and s£ be sample variances of the weekend
values and the weekday values, respectively. Then, t corresponding to
the differences tn sample means is computed by
xl:" X2
2 J: *
s /n, + s /n«
V I L.
2
where s is the pooled variance, which, in turn, is given by
2
S =
Cn2-l)
The two-tailed probability for the t-value computed and the
(n-j + n« - 2} degrees of freedom, which can be found in most statistical
textbooks, is compared with the chosen confidence level (1-p).
If p £ 0.05, the difference between the two means x^ and x^ is significant,
i.e., not merely due to chance. If 0.05 < p <_ 0.10, the difference
between the two means is marginally significant. If p > 0.10, the
difference is insignificant, i.e., likely to be caused by mere chance.
The preceding test is based on the assumption that the two
population variances are the same. When the two population variances
112
-------
differ quite a lot, a Welsh t-test, instead of an ordinary t-test, has
to be used. The Welsh t-statistic is computed by
t1 =
xl " X2
V sl/nl + S2/P2
The significance level of t1 is, approximately,
(w1 + w2)
where w-, = s^/n-,, w2 = s2/n2, and t1 and t2 are the significance levels
of t found in th_ t-table for (n,-1) and (n2-l) degrees of freedom,
respectively.
113
-------
APPENDIX B
WILCOXON RANK SUM TEST
The significance probabilities (the P-value) of W£ can be obtained by
the use of a normal approximation. The mean of W can be approximated
by M = l/2n(n-Hn+l), and the variance of W can be approximated by
VAR = [mn(nH-n+l)]/12. Thus the P-value of Z = (Ws-M)/VvAR can be found
in the normal table.
114
-------
APPEND I'X C
CONFIDENCE INTERVAL FOR THE MEDIAN
/\
Let {Y..}, i=l , ... n be a sample of size n. Let Y be the median.
Then a robust confidence interval for the median can be computed by
the following procedure, suggested by Alan M. Gross ("Confidence
Interval Robustness with Long-Tailed Symmetric Distributions," JASA,
Vol. 71, June 1976, pp. 409-416) and simplified by John Tukey and
Mosteller "Data Analysis and Regression".
Procedure:
/s
1. Let Y be the median.
2. I = Interquantile range
= upper hinge-lower hinge.
3. Compute the standardized value
>\
v
4. Compute the asymptotic variance for the median:
A
Var Y = o sr
d-u2)4
5. The confidence interval, then, is
where ty is the t percentage with df
v = 0.7(n-l)
115
-------
APPENDIX D
THE AID DECISION-TREE PROGRAM
The purpose of the AID program is to illustrate the relationship
between a dependent variable A and independent variables x, y, z
by dividing the available data into classes (defined by the independent
variables) in such a way as to best explain the variance in A. The
data base for AID consists of a set of "n" measurements: A-.,...,A
and x.|,...,xn; y-j»...,yn; z-|,...,zn; etc. Each of the independent variables
is represented by discrete numbers: e.g., x might assume values of
1, 2, 3, or 4; y might assume values of 1, 2, 3, 4, 5, 6, or 7;
z might assume values of 1, 2, 3, 4, or 5; etc. To achieve discreteness,
the raw data for the independent variables are usually divided in
ranges. For example, if the variable z represents wind speed (WS),
one might define z = l if 0 .1 WS < 3 mph; z = 2 if 3 1 WS < 7 mph;
z = 3 if 7 1 WS < 15 mph.; z = 4 ff 15 1 WS < 25 mph; and z = 5 if
25 mph 1 WS.
The first step of AID determines the independent variable that
can be split into two groups (say, z = 1 and 4 versus z - 2, 3, and 5)
which maximize the residual sum of squares (RSS) explained in the
dependent variable. If the two groups (G1 and G2) have group means
AT and 7C", then the RSS explained by splitting the data is
116
-------
ARSS = RSSQ - RSS1 -
where
n _ _
Rssn = Z (A- - Ar = RSS for entire data set;
u -i=i J
RSS1 = (A, - fij") = RSS for GI ; and
RSS9 =L (A. - ATT = RSS for G9.
L G2 J L c
The program considers all possible partitions of the data among all the
independent variables and selects the specific independent variable and
specific partition that maximizes ARSS.
The partitioning process is repeated on each of the groups G^ and
Gp to yield subgroups. These and all further subgroups are divided
until either 1) no subgroup can be split to achieve a ARSS above a
preset lower bound, 2) further division will produce a subgroup with
number of elements less than a preset bound, or 3) the number of
terminal subgroups reaches a prespecified limit.
117
-------
APPENDIX E
BINARY DECISION TREES DEVELOPED FOR
CLASSIFYING DAILY MAXIMUM OX LEVEL BY METEOROLOGY
This appendix shows the 22 AID trees analyzed. Briefly listed are
meteorological variables and their corresponding symbols used 1n the tree
structure.
1. T - Maximum Temperature (°F)
max
2. T n - Average Temperature (°F)
avg
3. P - Surface Pressure (mb)
4. Vis - Visibility (miles)
5. WD - Wind Direction (°)
6. WS - Wind Speed (knots)
7. MH - Mixing'Height (meters)
8- V" - Mean wind speed through the mixing layer Cm/sec)
The other symbols Included 1n each decision tree are:
H: Indicating that the terminal node belongs to the high class.
M: Indicating that the terminal node belongs to the middle class.
L: Indicating that the terminal node belongs to the low class.
P.V.E.: Percentage of variance explained by the decision tree.
118
-------
MH<1250
/ \
-45fi:WD<240 240=sWD<-45
W
N = 74
Y = 45.6
1250
-------
/ \
-45
-------
W85
max<80
80
-------
210
-------
-45
-------
T <75
max
75
-------
01
8 < Vis
.Vis < 8
-45 < WD < 135 135 - WD < "45 -45 < WD < 135 135 < WD < -45
10 < WS 4 < WS < 10
/ \
M
240 < WD < -45
135 < WD < 240
M
0
-------
T <85
max
(2)
N « 50
Y * 56.3
8° 8°lTmax
(6)
N = 25
JT = 51.4
8
-------
Vis<8
-45
-------
W60
N = 70
Y = 69.5
/ 1
W70
MH<1000 10CQS1H
6
-------
H
Tmax < 75
75 i Tn,ax < 85
-45
-------
MH<1000 KWO
-------
-45
-------
(1)
N - 199
Y = 49.1
Tmax<85
W
N = 136
Y = 36.8>
63
75.7
T <75
max
751 T 185 8
-------
-45
-------
250
-------
CO
01
W75
-45
-------
T <85 85
-------
(1)
N = 179
Y = 59.6
/ \
T <85 85
-------
(1)
N = 248
Y = 78.9
/ \
T ' <85
max
85
-------
T < 75 T > 75
'max max -
8 < WS 4 < WS < 8
(12)
N = 56
Y = 27.2
W85
T <70 70
-------
Tmax<7° 70
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
1. REPORT NO.
EPA-450/4-79-013
3. RECIPIENT'S ACCESSION NO.
4. TITLE AND SUBTITLE
Weekend/Weekday Differences in Oxidants and
Their Precursors
5. REPORT DATE
March 1979
6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
Yuji Horie, Joseph Cassmassi, Larry Lai and
Louis Gurtowski
8. PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Technology Service Corporation
2811 Wilshire Boulevard
Santa Monica, California 90403
10. PROGRAM ELEMENT NO.
2AA635
11. CONTRACT/GRANT NO.
68-02-2595
12. SPONSORING AGENCY NAME AND ADDRESS
Air Management Technology Branch
Monitoring and Data Analysis nivision
Office of Air Quality Planning and Standards
U.S. Environmental Protection Agency
Research Triangle Park, North Carolina 27711
13. TYPE OF REPORT AND PERIOD COVERED
Final
14. SPONSORING AGENCY CODE
15. SUPPLEMENTARY NOTES
16. ABSTRACT
Differences between weekend and weekday oxidant and oxidant precursor concentra-
tions in the Northeastern United States are examined. Statistical tests are employed
to assess whether any differences in daily maximum oxidant concentrations are signi-
ficant. Weekend/weekday differences in meteorology are also examined for any potential
impact on weekend/weekday differences in oxidants. Oxidant concentrations are adjusted
in an attempt to account for variations in meteorology, and weekend/weekday differences
in the adjusted values are examined. Finally, statistical tests are used to determine
significant differences in early morning oxidant precursor concentrations. The analyse
revealed that oxidant concentrations downwind of major urban areas tend to be lower on
weekends. However, the reductions are not as large as those observed in precursor
concentrations within the major urban areas. Reductions in weekend oxidant concentra-
tions at locations within the major urban areas are not as great as those at downwind
locations.
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.IDENTIFIERS/OPEN ENDED TERMS C. COS AT I Field/Group
Air Pollution
Ozone
Oxidants
Meteorology
18. DISTRIBUTION STATEMENT
19. SECURITY CLASS (This Report)'
Unclassified
21. NO. OF PAGES
149
20. SECURITY CLASS (Thispage)
Unclassified
22. PRICE
EPA Form 2220-1 (Rev. 4-77) PREVIOUS EDITION is OBSOLETE
------- |