EPA-600/3-78-018
February 1978
Ecological Research Series
EMPIRICAL RELATIONSHIPS
BETWEEN ATMOSPHERIC NITROGEN
DIOXIDE AND ITS PRECURSORS
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park, North Carolina 27711
-------
RESEARCH REPORTING SERIES
Research reports of the Office of Research and Development, U.S. Environmental
Protection Agency, have been grouped into nine series. These nine broad cate-
gories were established to facilitate further development and application of en-
vironmental technology. Elimination of traditional grouping was consciously
planned to foster technology transfer and a maximum interface in related fields.
The nine series are:
1. Environmental Health Effects Research
2. Environmental Protection Technology
3. Ecological Research
4. Environmental Monitoring
5. Socioeconomic Environmental Studies
6. Scientific and Technical Assessment Reports (STAR)
7. Interagency Energy-Environment Research and Development
8. "Special" Reports
9. Miscellaneous Reports
This report has been assigned to the ECOLOGICAL RESEARCH series. This series
describes research on the effects of pollution on humans, plant and animal spe-
cies, and materials. Problems are assessed for their long- and short-term influ-
ences. Investigations include formation, transport, and pathway studies to deter-
mine the fate of pollutants and their effects. This work provides the technical basis
for setting standards to minimize undesirable changes in living organisms in the
aquatic, terrestrial, and atmospheric environments.
This document is available to the public through the National Technical Informa-
tion Service, Springfield, Virginia 22161.
-------
EPA-600/3-78-018
February 1978
EMPIRICAL RELATIONSHIPS BETWEEN ATMOSPHERIC
NITROGEN DIOXIDE AND ITS PRECURSORS
BY
John Trijonis
Technology Service Corporation
2811 Milshire Boulevard
Santa Monica, CA 90403
Contract No. 68-02-2299
Project Officer
Basil Dimitriades
Environmental Sciences Research Laboratory
Research Triangle Park, North Carolina 27711
ENVIRONMENTAL SCIENCES RESEARCH LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
RESEARCH TRIANGLE PARK, NORTH CAROLINA 27711
-------
DISCLAIMER
This report has been reviewed by the Environmental Sciences Research
Laboratory, U.S. Environmental Protection Agency, and approved for publica-
tion. Approval does not signify that the contents necessarily reflect the
views and policies of the U.S. Environmental Protection Agency, nor does
mention of trade names or commercial products constitute endorsement or
recommendation for use.
ii
-------
ABSTRACT
A two-part study is performed with ambient monitoring data for nitrogen
dioxide and its precursors (NO and hydrocarbons). Part I deals with a des-
rt
criptive analysis of the nationwide data base for N02; Part II involves
empirical models of the N02/precursor dependence.
Part I characterizes the statistical properties, geographical patterns,
and historical trends of ambient N02 concentrations. Included in Part I are
a survey and quality check of the nationwide N02 data base; a study of statis-
tical distributions for characterizing maximal N02 concentrations; a descrip-
tive analysis of present N02 air quality for both annual mean and one-hour
maximum concentrations; an examination of historical trends in N02 air
quality; and a study of the relationship between annual mean N02 and yearly
one-hour maximum N02.
Part II formulates, applies, and tests empirical models that indicate
the dependence of ambient N02 on NOX and hydrocarbon control. Although the
simple empirical models used are subject to uncertainties, the general con-
clusions of these models agree quite well with smog-chamber results and
historical air quality trends. Part II studies lead to the conclusions that
(1) with other factors held constant, annual mean and yearly maximum N02 are
essentially proportional to NOV input; (2) hydrocarbon control yields slight-
A
to-moderate reductions in yearly maximum N02; (3) hydrocarbon control yields
very slight, essentially negligible, benefits for annual mean N02; and (4)
the exact form of the N02/precursor relationship may vary somewhat from one
location to the next, depending on local conditions.
iii
-------
CONTENTS
Abstract iii
Figures vii
Tables xii
Acknowledgments xvii
1.0 Introduction and Summary 1
PART I: DESCRIPTIVE ANALYSIS OF THE NATIONWIDE N02 DATA BASE .... 11
2.0 Data Base Preparation . 13
2.1 SAROAD Printouts of Frequency Distributions 13
2.2 Data Quality Analysis 19
2.3 References 26
3.0 Statistical Distributions for Characterizing
Maximal Concentrations 27
3.1 A Method Based on the Lognormal Distribution 29
3.2 A Method Based on the Gamma Distribution 43
3.3 Summary: Uses of Mathematical Distribution Functions ... 48
3.4 References 52
4.0 Characterization of Present N0£ Air Quality Levels 54
4.1 Data Base for Describing Present N0£ Air Quality 54
4.2 Data Patterns Involving Monitor Environment 58
4.3 Nationwide Geographic Patterns in NO? Air Quality 65
4.4 Intraregional Patterns in N02 Concentrations 74
4.5 References 88
5.0 Trends in Nitrogen Dioxide Air Quality 89
5.1 Five- and Ten-Year Changes in N02 Air Quality 89
5.2 Year-to-Year Trends in N02 Air Quality 94
5.3 References 102
6.0 Relationship of Yearly One-Hour Maxima and Annual Means .... 103
6.1 Nationwide Patterns in the Maximum/Mean Ratio 103
6.2 Intraregional Patterns in the Maximum/Mean Ratio 109
6.3 Historical Trends in the Maximum/Mean Ratio 112
PART II: EMPIRICAL MODELS OF THE N02/PRECURSOR RELATIONSHIP .... 117
7.0 Empirical Analysis of the NOg/Precursor Dependence 119
7.1 Experimental Evidence of the N02/Precursor Dependence ... 120
7.2 Formulation of Empirical Models 128
7.3 References 139
-------
CONTENTS (Cont'd)
8.0 Preparation of Data Base for Empirical Modeling 140
8.1 Computer Tapes of Aerometric Data 141
8.2 Creation of the Processed Data Base 145
8.3 Data Quality Check 152
8.4 References "5°
9.0 Seasonal and Diurnal Patterns for N02 and 1ts Precursors . . . 157
9.1 Seasonal Patterns 157
9.2 Diurnal Patterns 163
9.3 Computer File of Dependent and Independent Variables . . 176
9.4 References 178
10.0 Empirical Models Applied to Downtown Los Angeles 179
10.1 Statistical Techniques for Empirical Modeling 180
10.2 Dependence of Daytime NO? on Precursors 185
10.3 Dependence of Nighttime N02 on Precursors 212
10.4 Predictive Models for Downtown Los Angeles 223
10.5 References 242
11.0 Empirical Models Applied to Various Cities 243
11.1 General Methodology 243
11.2 Control Models for Various Cities 247
11.3 References 269
12.0 Validation of Empirical Models Against Historical
Air Quality Trends 270
12.1 Central Los Angeles Area 271
12.2 Coastal Los Angeles Area 280
12.3 Inland Los Angeles Area 284
12.4 Denver 288
12.5 Chicago 293
12.6 Summary of Validation Studies ' 299
12.7 References 302
13.0 Comparison of Empirical Models Against Smog-Chamber Results. . 303
14.0 Conclusions of the Empirical Modeling Study 307
14.1 Summary of the 8-City Study 307
14.2 Confidence in the Results 310
Appendices
A. Station-Years with 75% Complete Data on SAROAD as of 3-6-76 . . 312
B. Derivation of Formulas for Distribution of Maxima 322
C. Data for Characterizing Present N02 Air Quality 325
D. Summary of Daytime and Nighttime Regression Models for
Lennox, Azusa, Pomona, Denver, Chicago, Houston/Mae, and
Houston/Aldine 335
vi
-------
FIGURES
Number Page
2.1 Example of SAROAD Printout of N02 Frequency Distributions ... 14
2.2 Statistical Technique for Identifying Outliers in Reported
Maxima ..... 21
3.1 Comparison of Theoretical Distribution of Maximal z Values
with Actual Data (m and s as Given in SAROAD) 35
3.2 Example Frequency Distributions for Hourly N02 Concentrations . 37
3.3 Comparison of Theoretical Distribution of z Values with
Actual Data (m* and s* Calculated from Mean and 99th
Percentile) A . . ? 38
3.4 Comparison of Theoretical Distribution of z Values with
Actual Data (Modified Lognormal Approach) 41
3.5 Comparison of Theoretical Distribution of s Values with Actual
Data (Gamma Distribution Approach) 47
4.1 Location of N02 Monitoring Sites in the U.S. (Includes sites
with at least one year of complete data during 1972-1974) . . 59
4.2 Location of N02 Monitoring Sites in California 60
4.3 Location of N02 Monitoring Sites in the Los Angeles Region . . 61
4.4 Location of NOg Monitoring Sites in the New York-New Jersey-
New England Area 62
4.5 Percentage of Urban Stations with Various Levels of Annual
Mean N02 Concentrations (1972-1974) 66
4.6 Annual Mean NO? Concentrations at Urban Stations in the United
States (1972-1974) 68
4.7 Percentage of Urban Stations with Various Levels of 90th
Percentile Concentrations (1972-1974) 70
4.8 90th Percentile N02 Concentrations at Urban Stations in the
United States (1972-1974) 71
4.9 Percentage of Urban Stations with Various Levels of Yearly
Maximum N02 Concentration (1972-1974) 72
4.10 Yearly One-Hour Maximum Concentrations at Urban Stations in
the United States (1972-1974) 75
4.11 Map of the Metropolitan Los Angeles AQCR 76
-------
FIGURES (Cont'd)
Number Page
4.12 Nitrogen Oxides Emission Density Map for the Los Angeles
Region ............................ 78
4.13 Annual Mean N0? Concentrations in the Los Angeles Region
(1972-1974) ......................... 79
4.14 90th Percentile NO? Concent- rations in the Los Angeles Region
(1972-1974) . 7 ...................... 80
4.15 Yearly One-Hour Maximum NO- Concentrations in the Los Angeles
Region (1972-1974) . . ? ................... 81
4.16 Map of the New York-New Jersey-New England Area ........ 83
4.17 NOv Emissions in Various AQCRs in the New York-New Jersey-New
England Area ......................... 84
4.18 Annual Mean NOg Concentrations in the New York-New Jersey-New
England Area (1972-1974) ................... 85
4.19 90th Percentile N02 Concentrations in the New York-New Jersey-
New England Area (1972-1974) ................. 86
4.20 Yearly One-Hour Maximum NO? Concentrations in the New York-New
Jersey-New England Area (1972-1974) ............. 87
5.1 N02 Air Quality Trends at 4 CAMP Sites (Denver, Chicago,
St. Louis, and Cincinnati) .................. 95
5.2 N02 Air Quality Trends at 2 New Jersey Sites (Bayonne and
Newark) ........................... 98
5.3 N0£ Air Quality Trends at 6 Sites in Coastal /Central Los
Angeles County ........................ 100
6.T Distribution of Maximum/Mean N02 Ratios for Urban Locations ... 104
6.2 Nationwide Geographic Distribution of Maximum/Mean N09 Ratio
at Urban Sites, 1972-1974 ............ .' ..... 107
6.3 Dependence of Maximum/Mean Ratio on Annual Mean N0?
Concentrations ................ , ....... 108
6.4 Maximum/Mean N02 Ratio at Monitoring Sites in the Los Angeles
Region, 1972-1974 ...................... HO
6.5 Maximum/Mean N02 Ratio at Monitoring Sites in the New York-New
Jersey-New England Area, 1972-1974
6.6 Trends in the Maximum/Mean N02 Ratio Averaged over 4 CAMP Sites
(Denver, Chicago, St. Louis, and Cincinnati) ......... 113
6.7 Trends in the Maximum/Mean N02 Ratio Averaged over 2 New Jersey
Sites (Bayonne and Newark) .................. 113
viii
-------
FIGURES (Cont'd)
Number Page
6.8 Trends in the Maximum/Mean NO? Ratio at 6 Sites in Coastal/
Central Los Angeles County (Burbank, Lennox, Long Beach,
Los Angeles, Reseda, and Westwood) .............. 114
6.9 Trends in the Maximum/Mean N02 Ratio at 5 High-Growth
Locations within the Los Angeles Basin (Anaheim, La Habra,
Azusa, Pomona, San Bernardino) ................ 115
6.10 Trends in the Maximum/Mean N0£ Ratio at 5 Locations in Central
California (Redwood City, Salinas, San Rafael, Santa Cruz,
Stockton) .......................... 116
7.1 Nitrogen Dioxide Ten-Hour Average Concentration vs. Initial
Oxides of Nitrogen for Urban Hydrocarbon Mix (Means of
Several Experiments), University of North Carolina Study ... 121
7.2 Nitrogen Dioxide Dosage as a Function of NOX at Various HC
Levels, Bureau of Mines Study ................ 122
7.3 Nitrogen Dioxide Dosages in the Irradiation of Multi component
Hydrocarbon/N0x Mixtures, General Motors Study ........ 122
7.4 Average N02 Concentration (Over Six Hours) vs. Initial NO
at Three HC Levels, HEW Study ............ *. . . . 123
7.5 Average NO? Concentration (During First Ten Hours) vs. Initial
NOX at Three HC Levels, HEW Study .............. 123
7.6 Stephens' Hypothesis of Effect of HC and NOX Control ...... 125
7.7 Nitrogen Dioxide Maximum Concentration vs. Initial Oxides of
Nitrogen (Means of Several Experiments), UNC Study ...... 126
7.8 Dependence of Nitrogen Dioxide Maximum Concentration on
Initial Nitrogen Oxides, Bureau of Mines Study ........ 126
7.9 Typical Diurnal Pattern for Nitrogen Dioxide .......... 129
7.10 Conceptual Diagram of Empirical Model for Daytime Peak One-
Hour N02 ........................... 131
7.11 Map of the Metropolitan Los Angeles AQCR ............ 135
9.1 Seasonal Pollutant Patterns for Denver (Monthly Averages of
Daily Max One-Hour Concentrations, 1969-1973) ........ 158
9.2 Seasonal Pollutant Patterns for Chicago (Monthly Averages of
Daily Max One-Hour Concentrations, 1969-1973) ........ 158
9.3 Seasonal Pollutant Patterns for Houston/Mae (Monthly Averages
of Daily Max One-Hour Concentrations, 1975-1976) ....... 159
9.4 Seasonal Pollutant Patterns for Hous ton/ Al dine (Monthly Averages
of Daily Max One-Hour Concentrations, 1975-1976) ....... 159
ix
-------
FIGURES (Cont'd)
Number Pa9g.
9.5 Seasonal Pollutant Patterns for Los Angeles (Monthly Averages
of Daily Max One-Hour Concentrations, 1969-1974) 160
9.6 Seasonal Pollutant Patterns for Lennox (Monthly Averages of
Daily Max One-Hour Concentrations, 1969-1974) 160
9.7 Seasonal Pollutant Patterns for Azusa (Monthly Averages of
Daily Max One-Hour Concentrations, 1969-1974) 161
9.8 Seasonal Pollutant Patterns for Pomona (Monthly Averages of
Daily Max One-Hour Concentrations, 1969-1974) 161
9.9 Diurnal Patterns at Denver (1969-1973) 165
9.10 Diurnal Patterns at Chicago (1969-1973) 166
9.11 Diurnal Patterns at Houston/Mae (1975-1976) 167
9.12 Diurnal Patterns at Houston/A!dine (1975-1976) 168
9.13 Diurnal Patterns at Downtown Los Angeles (1969-1974) 169
9.14 Diurnal Patterns at Lennox (1969-1974) 170
9.15 Diurnal Patterns at Azusa (1969-1974) 171
9.16 Diurnal Patterns at Pomona (1969-1974) 172
10.1 Mid-Mean and Percentiles of Daytime Peak N02 vs. 6-9 A.M. NOX . . 181
10.2 Output of COMPLIAR Program for DPKN02 vs. NMHCPR and NOX69,
Winter Season 184
10.3 Dependence of Residual Daytime N0£ on INTNO (NOX69 - N025),
Winter Season 195
10.4 Dependence of Residual Daytime N02 on INTNO (NOX69 - N025),
Summer Season 196
10.5 Residual Daytime N02 vs. INTNO at Various Hydrocarbon Levels,
Winter Season 200
10.6 Residual Daytime N02 vs. INTNO at Various Hydrocarbon Levels,
Summer Season 201
10.7 Residual Daytime N02 vs. INTNO at Various Hydrocarbon-to-NO
Ratios, Winter Season x. . . 202
10.8 Residual Daytime N02 vs. INTNO at Various Hydrocarbon-to-NO
Ratios, Summer Season x. . . 203
10.9 Residual Nighttime N02 vs. NITENO, Winter Season 217
10.10 Residual Nighttime N02 vs. NITENO, Summer Season 218
10.11 Residual Nighttime N02 vs. NITENO at Various Afternoon Ozone
Levels, Winter Season 219
-------
FIGURES (Cont'd)
Number Page
10.12 Residual Nighttime N0£ vs. NITENO at Various Afternoon Ozone
Levels, Summer Season 220
12.1 Total NOV Emission Trends in the Los Angeles Basin 272
A
12.2 Total Reactive Hydrocarbon Emission Trends in the Los
Angeles Basin 273
12.3 Geographical Distribution of Percentage Change in Population
in the Los Angeles Basin, 1965 to 1975 275
xi
-------
TABLES
Number Page
2.1
2.2
2.3
3.1
3.2
4.1
4.2
4.3
4.4
4.5
4.6
5.1
5.2
6.1
8.1
8.2
8.3
8.4
8.5
8.6
8.7
Sites Reporting at Least 75% Complete Data for Hourly N0£
Measurements
Monitoring Methods for Sites 'eporting at Least 75%
Complete Data
Results of Data Quality Check
Median z Values for the Maximum As a Function of
Sample Size
Variance in Yearly One-Hour N02 Maxima
Stations for Characterizing Present N02 Air Quality
Number of Sites in Various Categories of Monitor
Environment
N02 Air Quality for Various Categories of Monitor
Environment
Stations Exceeding the NAAQS for Annual Mean N0?
(5.3 pphm), 1972-1974 c
Monitoring Sites with 90th Percentile NO? Concentrations
Greater than 10 pphm (1972-1974)
Monitoring Sites with High Yearly Maximal One-Hour
Concentrations (1972-1974)
Five-Year Changes in Ambient NOg Concentrations
Ten-Year Changes in Ambient ^ Concentrations
Locations with Maximum/Mean NO? Ratios Exceeding
10.0, 1972-1974
Pollutant Data Used for Denver and Chicago
Format of Hourly SAROAD Data for CAMP Sites
Pollutant Data Used for Houston/Mae and Houston/Aldine ....
Pollutant Data Used for the 4 Los Angeles Sites
Format of Hourly APCD Data
Parameters Included in the APCD Meteorological "99 Cards"
for Downtown Los Angeles
New Format for Pollutant Variables
17
18
23
33
50
55
b8
64
67
70
73
91
93
105
141
142
143
144
144
145
146
XII
-------
TABLES (Cont'd)
Number page
8.8 Number of Days Meeting Each Criterion 150
8.9 Deletions Made in Processed Data Bases for Chicago
and Denver 155
9.1 Variables for the Empirical Modeling Analysis 177
10.1 Glossary of Variables for the Daytime Analysis 186
10.2 Correlation Coefficients Between Morning Percursor
Variables 189
10.3 Logarithmic Regression Coefficients for Pairs of
Morning Pollutant Variables 190
10.4 Values of A, B,, and B? foiFRegressions According
to Equation (12) . 194
10.5 Hydrocarbon Regression Coefficient for Logarithmic
Regressions of Daytime N02 vs. NOX69 and HC69 198
10.6 Results of Stepwise Regressions According to
Equation (14) or (15) 206
10.7 Results of Logarithmic Regressions Between Daytime
N02 and Weather Variables 208
10.8 Linear Correlation Coefficients Between Weather
Variables and Precursor Variables 210
10.9 Effect of Including Weather Variables in the Linear
Regressions According to Equation (15) 213
10.10 Results of Nighttime Regression Analysis According
to Equation (16) 215
10.11 Results of Nighttime Regression Analysis According
to Equation (18) 221
10.12 Percentage Changes in Winter Daytime Average N02 at
Downtown Los Angeles as a Function of NO and
Hydrocarbon Control 227
10.13 Percentage Changes in Summer Daytime Average N02 at
Downtown Los Angeles as a Function of NO and
Hydrocarbon Control 229
10.14 Percentage Changes in Summer Nighttime Average N02 at
Downtown Los Angeles as a Function of NOX and
Hydrocarbon Control . 232
10.15 Percentage Changes in Summer Nighttime Average N02 at
Downtown Los Angeles as a Function of NOX and
Hydrocarbon Control 233
• • •
xm
-------
TABLES (Cont'd)
Number Page
10.16 The Effect of NO and Hydrocarbon Control on Annual
Average N02 at Downtown Los Angeles ............. 235
10.17 Days in the Processed Data Base with Extreme One-Hour
N02 Levels in Downtown Los Angeles (1969-1974) ....... 237
10.18 Percentage Changes in Winter Yearly Peak One-Hour N02
as a Function of NOV and Hydrocarbon Control ........ 239
t\
10.19 Percentage Changes in Summer Yearly Peak One-Hour N02
as a Function of NOV and Hydrocarbon Control ........ 239
A
10.20 Yearly One-Hour Maximum N02 Levels in Downtown Los
Angeles as a Function of NOX and Hydrocarbon
Control .......... ..... . ... ......... 240
11.1 Assumptions to Convert Equation (28) into a Control
Model for Daytime N02 .................... 246
11.2 Assumptions to Convert Equation (29) into a Control
Model for Nighttime N02 ................... 247
11.3 The Effect of NOX and Hydrocarbon Control on Annual
Mean N0£ at Lennox ..................... 249
11.4 The Effect of NOX and Hydrocarbon Control on Yearly
Maximum One-Hour N02 at Lennox ............... 251
11.5 The Effect of NOX and Hydrocarbon Control on Annual
Mean NO at Azusa ...................... 252
11.6 The Effect of NOX and Hydrocarbon Control on Yearly
Maximum N02 at Azusa .................... 254
11.7 Predicted Yearly Maximum NO? Concentrations at Azusa
as a Function of NOX and Hydrocarbon Control ........ 255
11.8 The Effect of NOX and Hydrocarbon Control on Annual
Mean N02 at Pomona ..................... 257
11.9 The Effect of NOX and Hydrocarbon Control on Yearly
Maximum N02 at Pomona ............ ........ 258
11.10 The Effect of NOX and Hydrocarbon Control on Annual
Mean N02 at Denver ..................... 259
11.11 The Effect of NOX and Hydrocarbon Control on Yearly
Maximum N02 at Denver .................... 261
11.12 The Effect of NOX and Hydrocarbon Control on Annual
Mean N02 at Chicago ..................... 262
11.13 The Effects of Hydrocarbon Control on Yearly
Maximum One-Hour N02 at Chicago ............... 264
xiv
-------
TABLES (Cont'd)
Number Page
12.1 Best Estimates of Nine-Year NOV and NMHC Trends at
DOLA, Burbank, and Reseda .x 278
12.2 Test of DOLA Empirical Control Model for Annual
Mean N02 279
12.3 Test of DOLA Empirical Control Model for Yearly
Maximum One-Hour N02 279
12.4 Best Estimates of Nine-Year NOX and NMHC Trends at
Lennox, Long Beach, and West LA 282
12.5 Test of Lennox Empirical Control Model for Annual Mean
N02 283
12.6 Test of Lennox Empirical Control Model for Yearly
Maximum One-Hour N02 283
12.7 Best Estimates of Nine-Year NOX and NMHC Trends at
Azusa and Pomona 286
12.8 Test of Azusa and Pomona Control Models for Annual
Mean N02 286
12.9 Test of Azusa and Pomona Control Models for Yearly
One-Hour Maximum N02 287
12.10 Estimates of Hydrocarbon and NOX Emission Trends for the
Denver Region 290
12.11 Best Estimates of Five-Year NOX and NMHC Trends at
Denver 291
12.12 Test of Denver Control Model for Annual Mean N02 292
12.13 Test of Denver Control Model for Yearly Maximum
One-Hour N02 293
12.14 Estimates of Hydrocarbon and NOX Emission Trends
for Chicago 295
12.15 Best Estimates of Eight-Year NOX and NMHC Trends at
Chicago 297
12.16 Test of the Chicago Control Model for Annual Mean N02 .... 297
12.17 Test of the Chicago Control Model for Yearly Maximum
One-Hour N02 298
12.18 Summary of Historical Precursor Trends and Ambient N02
Trends for the 4 Study Areas Experiencing Significant
Hydrocarbon Control 301
13.1 Predicted Impact of a 50% Hydrocarbon Reduction on
Daytime N02 in the Winter 305
xv
-------
TABLES (Cont'd)
Number Page
13.2 Predicted Impact of a 50% Hydrocarbon Reduction on
Daytime N0 in the Summer .................. 305
14.1 Predicted Impact of a 50% Hydrocarbon Reduction on
Annual Mean N02 and Yearly One- Hour Maximum N02 ....... 309
xvi
-------
ACKNOWLEDGMENTS
The preparation of this report has benefited from the assistance of
numerous people at Technology Service Corporation (TSC) and the Environmental
Protection Agency (EPA). TSC's air quality programmers, in particular, Matt
Jolley, Saul Miller, and Eric Helfenbein, provided creative solutions for
several difficult programming problems and worked diligently on the many
computer runs required in the study. Drs. William Meisel and Leo Breiman
deserve thanks for the valuable guidance they contributed on data analysis
methods and other technical issues. The work, and patience, of Patty
Mickelsen, Carolyn Sink, Susan Feder, and Marian Branch, in preparing the
manuscript is also gratefully acknowledged.
Basil Dimitriades, the EPA Project Officer, provided sound technical
guidance and helped to solve problems in the project while leaving us the
degree of freedom necessary for creative thinking and fruitful research.
Several EPA reviewers, including Edwin Meyer, Gerald Akland, Robert
Frankhauser, Joseph Bufalini, Thomas McCurdy, Walter Stevenson, and John
McGinnity, deserve thanks for their interest in the project and their helpful
comments on draft reports.
The EPA National Air Data Branch, California Air Resources Board,
Southern California Air Quality Management District, and Texas Air Control
Board are sincerely thanked for providing air quality data in an expeditious
manner.
xvn
-------
1.0 INTRODUCTION AND SUMMARY
The National Ambient Air Quality Standards presently include one
o
standard for nitrogen dioxide, 100 ug/m (0.053 ppm) for the annual mean
concentration. In the near future, EPA will revise the "Air Quality
Criteria for Nitrogen Oxides." This revision may lead EPA to supplement
the long-term standard for N02 with a short-term (e.g., one-hour) standard
and to consider new control strategies. To support this regulatory program,
there is a need for empirical analyses of ambient monitoring data for
nitrogen dioxide and its precursors. These analyses should consider both
annual mean N02 concentrations and maximal short-term N02 concentrations.
At least two types of empirical studies are required. The first in-
volves a descriptive analysis of ambient N02 concentrations. There is a
need to identify regions of the United States that may exceed the annual
mean standard and/or a proposed short-term standard for N02> Statistical
properties of N02 frequency distributions and trends in N02 air quality
should be quantified. Also, data should be assembled for assessing whether
the annual standard or a proposed short-term standard is the binding
constraint for control strategy formulation.
The second type of study involves empirical modeling of the relation-
ship between N02 concentrations and precursors. It is generally agreed
that N09 concentrations should be proportional to NOV concentrations
L~ _ .._ . ._ . ... yv. _ ... —
with all other factors held constant, but there is substantial uncertainty
concerning the impact of hydrocarbon control on ambient N02 levels. Empiri-
cal modeling techniques, applied to ambient monitoring data, should provide
a better understanding of these relationships.
-------
This report is organized in two parts corresponding to the two
types of empirical studies. Part I (Chapters 2 through 6) involves a
descriptive analysis- of the nationwide N02 data base; Part II (Chapters 7
through 14) deals with empirical models of the N02/precursor relationship.
The remainder of this chapter provides a summary of the report.
1.1 SUMMARY OF PART I: DESCRIPTIVE ANALYSIS OF THE NATIONWIDE N02
DATA BASE
The objective of Part I is to describe the statistical properties,
geographical patterns, and historical trends of ambient N02 concentrations.
The findings and conclusions of this descriptive analysis are summarized
in the paragraphs below. For convenient referencing, the summary is organ-
ized according to the order of the chapters (2 through 6).
Data Base Preparation
• As of March 1976, the National Aerometric Data Bank contained 462
station-years of hourly N02 data that met EPA's 75% completeness
criterion. Most of these data, 302 station-years, are from California.
Since 1972., there has been a sharp increase in the number of sites pro-
viding complete data sets for N02, especially in the number of sites
outside California.
t Data quality checks should precede all statistical studies of air
monitoring data. In this study, data verification procedures focus
on reported yearly one-hour maxima. As a result of the quality check,
42 of the 462 station-years of data are found to require
correction. It is remarkable that no errors were uncovered in the
California N02 data.
Statistical Distributions for Characterizing Maximal Concentrations
• Lognormal distributions which are fit to the entire range of one-hour
concentration data overpredict yearly one-hour N02 maxima, typically
by about 50%. If lognormal distributions are fit to the upper range
of the hourly concentration data (e.g., to the arithmetic mean and
99th percentile), the overprediction of the maximum is reduced to only
10%-20%. Some of the overprediction may be due to autocorrelations
-------
(e.g., dependent sampling) in the actual hourly data. A modified log-
normal approach to predicting maxima involves reducing the theoretical
yearly sample size; this approach can account for the autocorrelations
in a very approximate way.
t The Gamma distribution seriously underpredicts yearly one-hour N02
maxima. The lognormal distribution, fit to the upper range of the
hourly concentration data, seems preferable to the Gamma distribu- \
tion for the purposes of characterizing expected yearly maxima.
t There are four potential uses for mathematical distributions in
analyzing maximal N02 concentrations: to identify outliers for the
data quality check, to estimate the random variance in yearly maxima,
to adjust yearly maxima for incomplete sampling, and to characterize
patterns in yearly maxima using expected (predicted) maxima rather
than measured maxima. For the purposes of this study, the lognormal
distribution is appropriate for the first three uses. No distribu-
tion is appropriate for the fourth use; it is best to characterize
spatial and temporal patterns in yearly maxima by using the actually
measured maxima.
Characterization of Present N02 Air Quality Levels
• There are 123 monitoring sites'which provide at least one year of
data from 1972 to 1974. Of these, 120 can be classified as
urban; the other 3 are rural/power plant sites. Averages for
various categories of urban sites (center city vs. suburban, or indus-
trial vs. commercial vs. residential vs. mobile) all show about the
same level of annual mean, 90th percentile, and yearly maximum N02
concentrations. The 3 rural/power plant sites are atypical be-
cause of their low annual means and very high maximum-to-mean ratios.
• Eighteen of the 120 urban monitoring sites exceed the federal standard
for annual mean N02- Thirteen of these sites are in the Los Angeles
basin; the other five locations are Baltimore, Md.; Springfield, Mass.;
Chicago, 111.; Newark, N.J.; and Elizabeth, N.J. Fourteen nationwide
sites exhibit 90th percentiles exceeding 10 pphm. Forty-seven of the
-------
120 urban monitoring sites have yearly one-hour maxima which exceed
25 pphm, but only 4 sites experience yearly one-hour maxima that exceed
50 pphm. The sites with the highest yearly maximal and 90th percentile
concentrations are generally the same as the sites with the highest
annual mean concentrations; the Los Angeles basin is the national hot
spot for annual mean, 90th percentile, and one-hour maximal N02
concentrations.
• Within the Los Angeles basin, the spatial distribution of N02 concen-
trations generally corresponds well with the distribution of NOX
emissions. In parts of the Los Angeles region, on a scale of about
50 km, there is evidence that both transport and local effects are
important for N02. N02 concentrations do not follow consistent overall
patterns in the New York-New Jersey-New England area. On the scale of
this multi-state region (approximately 500 km), localized emissions
seem to be more important than regional transport for nitrogen dioxide
pollution.
Trends in Nitroflen Dioxide Air Quality
t Five-year trends (1969-1974) in N02 concentrations show the following:
no change for Los Angeles County (a slow-growth part of the Los Angeles
basin), about a 30%-50% increase in Orange County (a high-growth part
of the Los Angeles basin), about a 10% decrease at other California sites
and in New Jersey, and a large increase (about 30%-50%) in Chicago. Ten-
year concentration trends (1964-1974) show about a 10% to 20% increase in
Los Angeles County, with larger increases (about 40%) at Stockton, Calif.
and Chicago, 111. For most sites, yearly maximal N02 concentrations
increase less than (or decrease more than) annual mean N02 concentrations.
• For the most part, year-to-year trends in N02 concentrations at CAMP
sites, New Jersey sites, and Los Angeles sites can be explained in
terms of source growth and changes in emission factors. Historical
N02 trends in Los Angeles show an earlier rise, and then an earlier
decline, than N02 concentrations at CAMP sites; this reflects the new-car
-------
emission control program which started two years earlier in California
and included an NO emission standard two years earlier in California
/\
than nationwide.
Relationship of Yearly One-Hour Maxima and Annual Means
• Eighty-five percent of the 102 urban monitoring sites have maximum/mean
N02 ratios between 4 and 8. One location exhibits a maximum/mean
ratio slightly less than 4. Only six percent of the locations have
maximum/mean ratios greater than 10.
t If a one-hour N02 standard were set at 25 pphm, and if maximal and mean
N02 concentrations responded equivalently to emission changes, the one-
hour standard (rather than the present annual mean standard) would be
the binding constraint at 88% of the urban locations. If a one-hour N02
standard were set at 50 pphm, the annual mean standard would be the
binding constraint at 94% of the urban locations.
• There are no broad nationwide patterns in the maximum/mean N02 ratio.
Also, the maximum/mean ratio for urban sites does not depend
significantly on the annual mean concentrations. No consistent
patterns in the maximum/mean ratio appear on an intraregional scale
in Los Angeles or the New York-New Jersey-New England area.
• The maximum/mean ratio shows a strong downward trend in coastal/central
Los Angeles County over the past decade. This area has experienced
significant hydrocarbon control. The empirical models of Part II indi-
cate that hydrocarbon control preferentially reduces maximum N02 con-
centrations over mean N02 concentrations. Changes in the spatial
distribution of emissions may also lead to reductions in the maximum/
mean ratio. CAMP sites and other sites in California also show a de-
creasing maximum/mean ratio. The maximum/mean ratio has remained
approximately constant at New Jersey locations and in high-growth
areas of the Los Angeles basin.
»
1.2 SUMMARY OF PART II: EMPIRICAL MODELS OF THE N02/PRECURSOR RELATIONSHIP
The objective of Part II is to develop, apply, and test empirical
models that indicate the dependence of N02 on hydrocarbon and NOX control.
-------
6
The findings and conclusions of Part II are summarized below; the summary
is organized according to the order of the chapters (7 through 14).
Empirical Analysis of the N02/Precursor Dependence
• Historically, experimental studies with smog chambers have provided
most of our understanding of the N02/precursor dependence. The various
smog-chamber studies agree concerning the proportional dependence of
N07 (average or peak concentrations) on NO . Although the individual
L- X
chamber studies disagree concerning the effect of hydrocarbons on NC^,
a consensus would be that hydrocarbon control yields slight to
moderate benefits in terms of maximal N02 and produces essentially
no effect on annual mean N02. Because of disagreements among the
chamber studies and because of uncertainties in extrapolating experi-
mental simulations to the atmosphere, there is a need for empirical
models that extract information on the NOp/precursor dependence from
ambient monitoring data.
t This report develops empirical control models by combining statistical
(regression) equations with simple physical assumptions. The empirical
modeling analysis is disaggregated by time of day and by season. The
statistical equations for daytime N02 use early-morning (6-9A.M.) hydro-
carbons and NOY as precursor variables. Evening NOV and late-afternoon
A A
oxidant are considered as precursors of nighttime N02. The final control
models for annual mean N02 and yearly maximum N02 are based on a synthesis
of submodels for each time of day and each season, with linkages between
the daytime and nighttime models to account for carryover N02 from one
part of the day to another.
• The simplified empirical approach followed here is subject to several
limitations: the omission of meteorological variables, the neglect of
transport phenomena, and the assumption that precursor changes produced
by variance in meteorology can be used to model the effect of control
strategies. These limitations indicate a need to compare the empirical
models against smog-chamber results and historical air quality trends.
-------
Preparation of Data Base for Empirical Modeling
t The empirical models are applied to 8 locations: 2 CAMP sites
(Denver and Chicago), 2 Houston sites (Mae and Aldine), and 4
Los Angeles sites (Downtown Los Angeles, Lennox, Azusa, and Pomona).
The historical monitoring data for each site are processed, reformu-
lated, checked, and edited before statistical models are attempted.
Seasonal and Diurnal Patterns for N02 and its Precursors
• Concentrations of the photochemical precursors (NMHC and NOY) tend to
J\
be greatest during the winter. Oxidant concentrations, however, are
greatest during the summer because of increased solar radiation and
temperature. Seasonal patterns for NC^ vary from location to location
and apparently reflect competition between higher primary contaminant
concentrations in the winter and greater photochemical activity in the
summer.
• Diurnal patterns for primary contaminants exhibit two peaks—one in
the morning at about 8:00 A.M. and the other in the evening, anywhere
from 6:00 P.M. to midnight, depending on the site. Oxidant concentrations
exhibit a single maximum, usually between 1:00 P.M. and 5:00 P.M. At most
locations, N02 concentrations Ifiow two maxima—one at about "9YOO~A~7W^
to 10:00 A.M., the other anywhere From 6:00 P.M. to midnight. Although
the diurnal patterns for NMHC, NOX, and N02 do exhibit two maxima during^
the day, the concentrations at other times of the day and night are by
no means negligible compared with the peaks.
• For the purposes of the empirical modeling study, the seasonal patterns
indicate that a two-season division, winter (October-March) and summer
(April-September), is appropriate. The diurnal pollutant patterns lead
us to define "daytime" and "nighttime" modeling periods as 6:00 A.M. to
4:00~ P.M. and 4:00 P.M. to~6:00 A.M. of the following day, respectively.
Empirical Models Applied to Downtown Los Angeles
• A wide variety of statistical techniques are used to explore the data
base for Downtown Los Angeles. These techniques all yield the same
-------
8
qualitative conclusions concerning the dependence of N02 on precursors.
The most important contributors to daytime N02 (both peak and average)
are nighttime carryover N02 (N02 at 6:00 ATR.) and initial NO (N0y from
6:00 A.M. to 9:00 A.M/mTnus N02 at~67d6~O!T.~ Hy^d>Tca~rbbn reductions yield a
small, but statistically significant, benefit in terms of daytime N02.
The hydrocarbon effect is greater for peak N02 than average N02, is
greater in winter than summer, and is greater at high NOX levels than
low NO levels.
„
t Three factors are found to be contributors to nighttime NQ9--carryover
4_
N02 from the afternoon, early-evening NO, and afternoon oxidant (which
presumably combines with evening NO). Hydrocarbon control should de-
crease afternoon oxidant (assumed proportional to the NMHC/NOX ratio).
This effect is counterbalanced because hydrocarbon control apparently
increases carryover N02 from the afternoon.
• Statistical analyses involving meteorological parameters indicate that
the observed hydrocarbon effect may be partially an artifact produced
by unaccounted for weather variables. This finding reinforces the
need to check the empirical control models against smog-chamber results
and historical trends.
Empirical Models Applied to Various Cities
• The formulation of empirical models for the 8 selected locations
proceeds smoothly with the exception of nighttime models for the 2
Houston locations. Complete empirical control models for annual mean
N02 and yearly one-hour maximum N02 are developed for the 6 non-Houston
sites.
• The empirical models for all SHIbcations (as well as the daytime models
for the 2 Houston sitesT indicate that, with other factors held constant,
both annual mean N02 and yearly maximum N02 are essentially proportional
to NOX input. The slight deviations away from proportionality that
sometimes occur in the empirical models are all in the direction of a
less -than-proportional relationship.
-------
t The empirical models for various cities show that hydrocarbon control
yields slight, essentially negligible4 effects on annual mean N02. The
models predict that 50% hydrocarbon control decreases annual mean
N02 by 6% at Downtown Los Angeles, 2% at Lennox, 2% at Azusa, 11%
at Pomona, and 0% at Chicago, and increases annual mean N02 by 5% at
Denver.
• The empirical models indicate that hydrocarbon reductions yield slight-
to-moderate benefits for yearly maximum N02« Fifty-percent hydrocarbon
control reduces yearly maximum N02 by 25% at Downtown Los Angeles, 10%
at Lennox, 6% at Azusa, 19% at Pomona, 0% at Chicago, and 8% at Denver.
Yearly maximum N02 occurs under winter/daytime conditions at Downtown
Los Angeles, Lennox, Denver, and Houston/Mae; under summer/daytime
conditions at Chicago; and under winter/nighttime conditions at Azusa,
Pomona, and Houston/Aldine (the 3 downwind receptor sites studied).
Validation of Empirical Models Against Historical Air Quality Trends
• Validation studies for the empirical N02 control models are conducted
at 10 monitoring sites: 3 in the central Los Angeles area", 3
in the coastal Los Angeles area, 2 in the inland Los Angeles area, 1
in Denver, and 1 in Chicago. Verification of "the models against trends
at individual monitoring sites attains a mixed level of success, with
best results obtained in the central and coastal Los Angeles area.
The less successful tests at Azusa and Pomona may indicate that omitting
transport relationships in the empirical models is inappropriate for
these two sites (as it also seemed to be for Houston/Aldine).
• In an aggregate sense, historical air quality trends confirm the
general results of the empirical modeling study. Viewed as a whole,
the trends are consistent with the conclusions: (1) a proportional
relationship exists between N02 and NOX; (2) a slight to moderate
hydrocarbon effect exists for yearly maximum N02; and (3) a very slight
(if any) hydrocarbon effect exists for annual mean N02.
-------
10
Comparison of Empirical Models Against Smog-Chamber Results
• The general conclusions of the empirical modeling study agree quite
well with conclusions based on smog-chamber simulations. Agreement
exists with respect to the impact of NOY and/or hydrocarbon control
J\
on both annual mean and yearly maximum NO/> concentrations.
• The variations in the empirical models from city to city can be due
either to errors in the individual models or to real variations in the
N02/precursor dependence from one location to the next. The differences
in the N02/precursor relationship found in different smog-chamber
studies indicate the latter case is certainly a possibility. However, we
are more sure of the general conclusions of the empirical modeling study
than we are of the specific models for individual cities.
Conclusions of the Empirical Modeling Study
• The empirical models, in conjunction with smog chamber studies and
historical trend analysis, lead us to a basic understanding of the
dependence of ambient N02 concentrations on precursor control. Although
all three approaches involve uncertainties, they all are consistent
with the same general conclusions:
1. With other factors held constant, yearly maximal and annual
mean N02 concentrations are essentially proportional to NOV input.
A
2. Hydrocarbon control yields slight to moderate benefits in
yearly maximal one-hour N02; reducing hydrocarbons by 50% should
decrease yearly maximal N02 by about 10% to 20%.
3. Hydrocarbon control yields very slight, essentially negligible,
benefits in annual mean N02>
4. The exact form of the N02/precursor relationship may vary some-
what from one location to the next, depending on climatic conditions,
reaction times, and the existing hydrocarbon/NO ratio.
A
-------
11
PART I:
DESCRIPTIVE ANALYSIS OF THE
NATIONWIDE N02 DATA BASE
-------
13
2.0 DATA BASE PREPARATION
The objective of Part I of this report is to characterize nitrogen
dioxide air quality in the United States, with particular emphasis on
annual mean N02 concentrations and maximum one-hour N02 concentrations.
The data base available for performing this characterization consists of
yearly frequency distributions of all hourly N02 data in the EPA SAROAD*
system. This chapter describes how the available data were processed,
checked, and modified for the purposes of the study.
2.1 SAROAD PRINTOUTS OF FREQUENCY DISTRIBUTIONS
The available data for analyzing nationwide N02 air quality are
printouts of yearly frequency distributions for all hourly N02 data in
the EPA SAROAD system as of 6 March 1976. Figure 2.1 presents an example
of this type of printout. As illustrated in Figure 2.1, the SAROAD printout
for each site provides the site code, the name of the monitoring agency,
regional population statistics, general information on the site location
(address, city, county, state, air basin, and type of surrounding environ-
ment), and specific locational parameters (latitude, longitude, UTM coor-
dinates, and elevation). For each year of monitoring activity, the SAROAD
printout indicates the 10th, 30th, 50th, 70th, 90th, 95th, and 99th percen-
tile concentrations and the yearly maximum one-hour concentration in units
of yg/m3. The total number of hourly measurements and the monitoring method
are also listed each year. For those years with data for at least 75% of
SAROAD = Storage and Retrieval of Aerometric Data
-------
0«T«
»€«I»LY FREQUENCY DISTRIBUTION
t-y
«rnr
COUWTY ti*?oi: Jtrrrssoi en
L"N«;iTi)pr_; us «• 15 p. IT
CITY
STATION TYPE t«3l: CENTER CITt .
1} LOUI5»I|IE
UTH NOPTMIMCf
»tn r« ST 1161 — nonnmn
pnttitTion CB
: I.OWISVII.LE,
or
fr.
nmnn coutTY --
co««!Cnrs: MC»VY TRAP. orN5tTTi senr INO., CONSIPEW^LC nrsiocncrs WITHIN i-
«-tTt
POtLI'T*frf-'»ETHO9 C0f>r N1W »IN
PERCENT ILES
N*I »H1TM
«« ••• tta
«rn"€T^IC
UMTTS
TI
BIO»I? ofi/c» Mrrrn in r»
47.
75. |13.
160. 3*S.
;n«Trn."iENT»l.
73
I-HOUR
ei
BlOXlOf
-------
15
all hours, the arithmetic mean, geometric mean, and geometric standard
deviation are listed. For the example presented here, Louisville site
011601, data for 75% of all hours were reported only during 1973 and 1974.*
In this study, data are included only for those stations and years
that meet the 75% completeness criterion. The reasons for excluding data
which fail the 75% criterion are threefold. First, N02 concentrations at
many sites follow distinct seasonal patterns. There is a danger that in-
complete sampling might be performed only during certain seasons. This
could seriously bias the measured frequency distribution. Second, the
arithmetic mean, geometric mean, and geometric standard deviation are inn
portant for statistical calculations performed in this study. These para-
meters are not provided for data sets which fail the 75% criterion. The
parameters could be estimated from the reported percentile concentrations,
but this would introduce another source of error. Third, a quality check
is planned for all data to be used in the study. To include the numerous
cases which fail the 75% criterion would dilute the project resources avail-
able for the quality check. The impact on the quality check program would
be especially significant because there appears to be a positive correlation
between incomplete data sets and apparently anomalous data sets.
The SAROAD printout of 6 March 1976 included 462 station-years of N02
data that met the 75% completeness criterion. In order to facilitate statis-
tical computations, the data for these 462 station-years were punched on
computer cards. The information put on each card included the site code,
In many cases, the 75% completeness criterion would have been met in
1975 except that some 1975 measurements were not yet reported to SAROAD.
-------
16
monitoring method, year, number of observations, percent!le concentrations,
arithmetic mean, geometric mean, and geometric standard deviation. Appen-
dix A presents a printout of these cards (as modified by the data quality
analysis discussed later in this chapter).
Table 2.1 lists the number of sites, by year, that meet the 75% com-
pleteness criteria. It is obvious that the number of sites reporting 75%
complete data to SAROAD has increased substantially over the past twelve
years. The increase in the number of sites with "complete" data is greatest
for the period of 1971 to 1974, with a particularly large jump occurring from
1973 to 1974. The growth in the number of monitoring sites is especially sig-
nificant outside California; the number of non-California monitoring sites
multipled by nearly a factor of seven from 1971 to 1974.
Table 2.1 also lists the average percentage of data for the sites which
meet the 75% completeness criterion. The average percentage of data re-
ported by these sites has undergone a steady increase over the years. Thus,
not only have more sites attained the 75% criterion in the past few years,
but these sites have attained better completeness ratios. The sites in
California tend to show higher completeness ratios than sites outside Calif-
ornia. This is one indication of the higher quality of the California data
base.
Table 2.2 lists the number of sites by year and by monitoring method.
There are four monitoring methods: colorimetric-Lyshkow (mod.), chemilumin-
escence, colorimetric-Griess-Saltzman, and coulometric. Although none of
these methods has yet been approved by EPA as a reference method, none of
-------
Table 2.1 Sites Reporting at Least 75% Complete Data for Hourly N02 Measurements
NATIONWIDE SITES
Total Number of
Year Sites with "Complete"
(75%) Data
1962 4*
1963 18
1964 18
1965 19
1966 22
1967 32
1968 26
1969 29
1970 29
1971 33
1972 52
1973 58
1974 112
Total Number 452
Station-Years
Average Percentage
Data for these
Sites
CALIFORNIA SITES
Number Average
of Sites Percentage Data
NON-CALIFORNIA SITES
Number Average
of Sites Percentage Data
80. 4X
81.0
81.4
83.7
84.5
85.1
86.5
87.6
87.7
87.7
88.7
88.8
88.3
0
13
14
13
17
23
20
23
21
25
39
37
57
302
81. 2%
81.2
83.3
84.9
85.2
87.2
88.2
88.5
88.0
89.1
90.0
90.6
4
5
4
6
5
9
6
6
8
8
13
31
55
160
80.42
80.5
82.0
84.4
83.1
84.8
84.0
85.4
85.7
87.1
87.2
87.4
85.8
There are several monitoring sites 1n California which meet the 75% completeness
criteria for 1962 and prior years. However, California has reported data to SAROAD
only for years starting 1n 1963.
-------
Taole 2.2 Monitoring Methods for Sites Reporting at Least 75% Complete Data
NUMBER OF SITES IN OPERATION WITH EACH MONITORING METHOD
Year
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
Total Nuntoer
Station- Years -
1
Color1metr1c-Lyshkow(Mod. )
(SAROAD 14260211)
0
12 (12)
13 (13)
13 (13)
17 (17)
24 (23)
21 (20)
23 (23)
22 (21)
27 (25)
43 (39)
57 (37)
83 (56)
355
2
Chemi 1 unri nescence
(SAROAD #4260214)
0
0
0
0
0
0
0
0
0
0
0
2
19 (1)
21
3
Col ori metric -
Gri ess-Sal tzman
(SAROAD #4260212)
4
6(1)
5(1)
6
5
8
5
6
7
6
8
7
7
80
4
Coulometrlc
(SAROAD #4260213)
0
0
0
0
0
0
0
0
0
0
1
2
3
6
00
Values 1n parentheses are for California only
-------
19
them has yet been designated "unacceptable." Table 2.2 reveals that the
colorlmetric-Lyshkow (mod.) method accounts for nearly all the California
measurements and over 75% of the nationwide measurements. The colorimetric-
Griess-Saltzman method accounts for much of the remaining data; this is the
method used in the past in the EPA CAMP program.
2.2 DATA QUALITY ANALYSIS
A cursory examination of the original SAROAD printouts indicated likely
errors in the data, especially in the maximum one-hour concentrations. For
example, the reported frequency distribution for one station was as follows:
Maximum
Percentile - 10% 30% 50% 70% 90% 95% 99% one-hour
Concentration (ppm)- - .005 .02 .03 .04 .05 .07 .11 5.05 (!)
Although the anomalies in the reported data were not always as blatant as the
example above, there seemed reason to question the validity of at least 70
of the 462 station-years of data. The maximum one-hour concentration was the
only suspicious value in nearly all of these questionable cases. In a very
small number of cases, the percentile concentrations (10% to 99%), as well as
the maximum one-hour, appeared dubious.
A data quality check was performed to correct and upgrade the data base.
The quality check was guided, in part, by the use of a statistical technique
which predicted maximum one-hour concentrations for each station-year based
on the arithmetic mean and 99th percentile concentrations for that station-year.
This technique, which we call the "modified lognormal" approach, is described
at length in the next chapter. Its use in the data quality check was to identify
outliers by comparing the reported one-hour maxima with the predicted maxima.
-------
20
Figure 2.2 demonstrates how this was done. Figure 2.2 compares the predicted
distribution of "lognormal z values" for yearly one-hour maxima to the histogram
of actual z values for the 462 reported one-hour maxima. The data sets corre-
sponding to the right-hand tail of the histogram were considered questionable.
- • w
All reported maxima which yielded z values greater than 4.3 were subjected
to data verification procedures.
The statistical technique identified 60 potential outliers which
were submitted to data verification procedures. Several station-years of
data, other than those flagged by the statistical technique, were also selected
for the verification process. Some of these other data sets were the ones
identified as potential problems by staff members of EPA's Office of Air
Quality Planning and Standards[1,2]. Other data sets were chosen for verifi-
cation based on a visual scan of the data base in search of anomalies. Since
all of the CAMP data were available to us on magnetic tape, every year of
CAMP data was subjected to verification procedures.
The procedures for checking the data were as follows: For each year of
CAMP data, all hourly measurements of NOg, NO, and OX were printed out for the
day of maximal one-hour NOg concentration. The hour-by-hour pattern of NO^
concentrations was checked for internal consistency and compared with the pat-
terns of NO and OX concentrations. So that one erroneous NOp maximum would
not be replaced with a second-highest value that was also in error, the
second, third, and fourth days of highest N02 concentration were checked by
similar procedures. For data other than from CAMP sites, the state or local
monitoring agency was contacted, and the reported one-hour maximum was
checked against the local data logs. The records of the monitoring agency
-------
.4i—-
Probability Density Function of z
Values for Yearly Max One-Hour
Concentration Based on Modified
Lognormal Approach
TtfUmtllffi-HWm
Histogram of z Values for the
Actual Maxima Based on all 462
Station-Years in the Unconnected
Data Base
Data Subject to Quality Check
t\s
4 5
Lognormal z Value of Yearly Maximum
Figure 2.2 Statistical Technique for Identifying Outliers in Reported Maxima
-------
22
did not agree with the SAROAD output in several instances, implying a trans-
cription error between the local agency and SAROAD. If the records of the
local agency did agree with SAROAD, further checks were conducted at the
convenience of the monitoring agency. These checks involved examining the
diurnal pattern of N02 and other pollutants on the day of the yearly maximum
one-hour concentration. An especially intensive check of N0« data for
1974 in "up-state" New York was conducted because of anomalies pointed
out by EPA personnel[2].
It should be emphasized that the data quality check was basically di-
rected toward eliminating large errors which appear as .outliers in frequency
distributions of the hourly data. The techniques used to flag outliers
would miss small errors or moderate-size errors involving a constant fac-
tor (such as a calibration factor). Identifying the latter types of errors
would require a major data quality program and, even then, might be impos-
sible.
Table 2.3 lists the errors discovered in the data quality check.
Forty-two station-years of data needed modifications. Thirty of these in-
volved corrections to the reported yearly maximum one-hour concentration;
the other 12 station-years had to be deleted from the data base. It is
striking that no corrections were necessary for California data, even though
California data accounted for 65% of the station-years in the data base, and
even though several California data sets were flagged for the verification
procedures.
-------
Table 2.3 Results of Data Quality Check
SITE (SAROAD CODE) YEAR MONITORING METHOD CORRECTIVE ACTION TAKEN
REMARKS
Phoenix, Arizona
(030600002-G01)
1967
Colorimetric-
Lyshkow (Mod)
Delete data for this year
Conversation with Marlcopa County Health Department re-
vealed anamolous data for 1967 and 1968. There may have
been calibration and other procedural difficulties during
these first two years of instrument operation.
1968
Delete data for this year.
1973
SAROAD reports 1.00 ppm as yearly max
one hour. Correct value is .22 ppm.
Transcription error between Maricopa County records
and SAROAD system.
New Britain,
Connecticut
(070680002-F01)
1973
SAROAD reports .22 ppm as yearly max
one hour. Correct value is .104 ppm.
Transcription errors between Conn. Dept. of Environmental
Protection records and SAROAD system.
1974
SAROAD reports 1.40 ppm as yearly max
one hour. Correct value is .090 ppm.
Washington, D.C. 1965 Colorimetric- SAROAD reports .42 ppm as yearly max
(090020002-A10) Greiss-Saltzman one hour. Correct value is .23 ppm.
Examination of hourly CAMP data reveals that the highest
hour recorded in 1965 is probably invalid. The value used
here is the second highest recorded hour in 1965.
Washington, D.C.
(090020003-A05)
1974
Chemi 1 unti nescence
SAROAD reports .42 ppm as yearly max
one hour. Correct value is .17 ppm
Correction recommended by Robert Faoro of EPA-OAQPS.
Chicago, Illinois
(141220002-A10)
1964 Colorimetric- SAROAD reports .47 ppm as yearly max
Greiss-Saltzman one hour. Correct value is .33 ppm.
Examination of hourly CAMP data reveals that the recorded
.47 ppm is obviously in error (possibly in a decimal
point). The .33 ppm value is the second highest re-
ported value for 1964.
1971
Delete data for this year.
Examination of hourly CAMP data reveals long strings ef
high N02 values In July 1971. These'data are probably
all in error. The July values affect the 99% as well as
the yearly max and thus the year is not salvageable.
1973
SAROAD reports .45 ppm as yearly max
one hour. Correct value is .36 ppm.
Examination of hourly CAMP data indicates that the .45ppa
value Is dubious. The second highest hour In 1973 1s
.36 ppm.
ro
co
Kansas City,
Kansas
(171800001-H01)
1973
Coulometric
Delete data for this station.
SAROAD values disagree with records of the Kansas CltyAIr
Quality Division for both the yearly max and 99th percen-
tile. The Kansas City Air Quality Dviision is unable to
supply a correct value for the 99th percentile.
1974
Louisville, Kentucky
(182380017-A05)
1974
Chemiluminescence
SAROAD reports .39 ppm as yearly max
one hour. Correct value is .17 ppm.
Correction recommended by Robert Faoro of EPA-OAQPS
Minneapolis,
Minnesota
(242260022-H01)
Bellefontaine,
Missouri
(260200002-G01)
1972
1974
Coulometric
Colorimetric-
Lyshkow (Mod)
Delete data for this station.
SAROAD reports .382 ppm as yearly max
one hour. Correct value is .236 ppm.
City of Minneapolis Air Pollution Control Division does
not consider the data to be reliable. The reported year-
ly maximum followed a period of instrument failure.
Correction recommended by Robert Faoro of EPA-OAQPS.
-------
SITE (SAROAO CODE)
YEAR
Table 2.3
MONITORING METHOD
Results of Data Quality Check (Cont'd)
CORRECTIVE ACTION TAKEN
REMARKS
St. Louis, Missouri
(264280007-H01)
1973
Colorimetric-
Lyshkow (Mod)
Delete this station from the data base.
The max one hour value appears too high compared to the
rest of the frequency distribution and to max values for
other years. The St. Louis City Division of Air Pollu-
tion Control can provide no help in determining the real
max value.
St. Louis, Missouri
(264280061-H01)
1973
The max one hour value and other aspects of the reported
frequency distribution do not make sense for this year.
St. Louis, Missouri
(204230062-H01)
1973
St. Louis, Missouri
(264280063-H01)
1973
Rosebud, Montana
(271360028-F03)
1974
Colorimetric-
Lyshkow (Mod)
All reported values for this station are below the mini-
mum detectable (.005 ppm). The Montana Air Quality Bu-
reau indicates that the ambient N02 levels are actually
that low. The data are deleted since there is no infor-
mation for ascertaining the frequency distribution of
various concentrations.
Reno, Nevada
(290480005-101)
1973
Coulometric
SAROAD reports 1.11 ppm as yearly max
one hour. Correct value is .182 ppm.
Transcription error between Uashoe County records and
SAROAD system.
1974
SAROAD reports 4.56 ppm as yearly max
one hour. Correct, value is .335 ppm.
ro
PhilUpsburg,
New Jersey
(314240002-F01)
Rochester, New York
(335760004-F01)
1972
Colorimetric-
Griess-Saltzman
SAROAD reports .39 ppm as yearly max
one hour. Correct value is .17 ppm.
Examination of hourly values reveals that the reported
maximum is very dubious. It is replaced with the second
highest value for the year.
1973
SAROAD reports .328 ppm as yearly max
one hour. Correct value is .19 ppm.
Buffalo, New York 1974 Colorlmetric-
(330660005-F01) Lyshkow (Mod)
Buffalo, New York 1974
(330660007-F01)
Kingston, New York 1974 "
(333500002F01)
Niagra Falls, 1974
New York
(334740006-F01)
SAROAD reports .38 ppm as yearly max
one hour. Correct value is .17 ppm.
SAROAD reports .59 ppm as yearly max
one hour. Correct value is .13 ppm.
Also correct 99th percent! le from
.107 ppm to .097 ppm.
SAROAD reports .25 ppm as yearly max
one hour. Correct value is .09 ppm.
SAROAD reports .31 ppm as yearly max
one hour. Correct value is .17 ppm.
Also correct 99th percent! le from
.107 ppm to .097 ppm.
Scan of hourly data reveals that the reported maximum is
very dubious. It is replaced by the second highest value
for the year.
Scan of hourly data reveals that the reported maximum and
several other high values are very dubious.
Scan of hourly data reveals that the reported maximum and
several other high values on the same day are very dubious
Scan of hourly data reveals that the reported maximum and
several other high values are very dubious.
1974
II It
SAROAD reports .33 ppm as yearly max
one hour. Correct value is .11 ppm.
Scan of hourly data reveals that the reported maximum and
two other high values are very dubious.
-------
Table 2.3 Results of Data Quality Check (Cont'd)
SITE (SAROAO CODE) YEAR MONITORING METHOD
CORRECTIVE ACTION TAKEN
REMARKS
Schenectady, New York
(336020003-fOl)
1974 Colorimetric- SAROAD reports .20 ppm as yearly max
Lyshkow (Mod) one hour. Correct value is .09 ppm.
Scan of hourly data reveals that the reported maximum is
very dubious. It is replaced by the second highest value
for the year.
Syracuse, New York
(336620011-F01)
1974
SAROAD reports 1.70 ppm as yearly max
one hour. Correct value is .13 ppm.
The value of 1.70 ppm is a transcription error between
New York State'Department of Environmental Conservation
and the SAROAD system. The second highest value, .23 ppm,
was also invalidated by a scan of the hourly W>2 data.
Utica, New York
(336880004-F01)
Cincinnati, Ohio
(361220003-A10)
Lancaster City,
Pennsylvania
(394660007-F01)
Philadelphia,
Pennsylvania
(397140002-AOS)
Philadelphia,
Pennsylvania
(397140004-H01)
Providence,
Rhode Island
(410300005-F01)
Providence,
Rhode Island
(410300007-F01)
Nashville, Tennessee
(442540010-G01)
1974
1964
1974
1973
1972
1972
1973
1972
1973
1974
1974
II II II
Colorimetric-
Griess-Saltzman
Chemi 1 umi nescence
n ii it
Colorimetric-
Lyshkow (Mod)
Colorimetric-
Griess-Saltzman
II II 1
II t
II II II
t II I
Colorimetric-
Lyshkow (Mod)
SAROAD reports .24 ppm as yearly max
one hour. Correct value is .13 ppm.
SAROAD reports .34 ppm as yearly max
one hour. Correct value is .24 ppin.
SAROAD reports .272 ppm as yearly max
one hour. Correct value is .084 ppm.
SAROAD reports .65 ppm as yearly max
one hour. Correct value is .29 ppm.
SAROAD reports 5. 05 ppm as yearly max
one hour. Correct value is .25 ppm.
SAROAD reports .45 ppm as yearly max
one hour. Correct value is .26 ppm.
SAROAD reports .531 ppm as yearly max
one hour. Correct value is .175 ppm.
SAROAD reports .35 ppm as yearly max
one hour. Correct value is .16 ppm.
SAROAD reports .39 ppm as yearly max
one hour. Correct value is .13 ppm.
SAROAD reports .465 ppm as yearly max
one hour. Correct value is .205 ppm.
Delete this station from the data base.
Scan of hourly data reveals that the reported maximum and
several other high values on the same day are very dubious.
Examination of hourly CAMP data reveals that .34 ppm value
is obviously in error. The .24 ppm value is the second
highest for the year.
The reported .272 ppm value has been invalidated by the
Pennsylvania Department of Environmental Resources.
Correction recomnended by Jerry Ackland of EPA-
NERC/RTP.
Transcription error between Philadelphia Air Management
Services and SAROAD system.
Transcription error between Rhode Island Department of
Health and SAROAD system.
ii ii H i
tl Ii n it
II II ii ft
n n H n
Metropolitan Nashville/Davidson County Health Department
indicates that all data from this station have been in-
ro
01
validated.
-------
26
2.3 REFERENCES
1. R. Faoro, "1974 NO2 Maximum Values," Memorandum to W.F. Hunt and
J. McGinnity, EPA Office of Air Quality Planning and Standards, RTP,
North Carolina, 26 January 1976.
2. D. Iverach, EPA Office of Air Quality Planning and Standards, Personal
communication concerning ,N@2 data*quality for stations in New York state,
December 1977.
-------
27
3.0 STATISTICAL DISTRIBUTIONS FOR CHARACTERIZING MAXIMAL CONCENTRATIONS
One of the major objectives of the present study is to characterize
yearly maximum one-hour N02 concentrations. The simplest and most direct
way of performing this characterization is to base it on the actually
measured yearly maxima for various stations and various years. An alter-
native is to base the characterization on expected yearly maxima, with
the expected maxima determined by fitting statistical functions to the
concentration frequency distribution for each station-year.
The latter approach, involving calculated expected values, offers
four basic advantages. First, calculating expected maxima facilitates
the data quality check since a comparison of the expected maxima to
recorded maxima helps to identify potential outliers in the recorded data.
Second, the statistical methods used to compute the expected maxima also
provide an estimate of the variance in yearly maximum concentrations at
each location. The variance in the yearly maximal concentrations can be
estimated from as little as one year of data. Using actually measured
maxima to estimate the variance in yearly maxima requires several years of
data and is subject to errors caused by the confounding of long-term
trends with year-to-year stochastic fluctuations.
A third advantage is that the expected maxima are calculated assuming
a full year of sampling, 8760 hours. Unlike the expected maxima, the mea-
sured maxima depend on the number of samples taken per year, ranging from
around 6600 to 8600 for the data base in question. In this regard, the
statistical techniques used to determine the expected maxima offer a side
benefit: Utey provide a method for adjusting the measured maxima in order
-------
28
to account for incomplete (less than 100%) sampling during the year. Such
an adjustment will be made to all the measured maxima in this study.
A final advantage is that expected maxima, determined from the en-
tire concentration frequency distribution, are statistically more "robust"
i ,.'<.
than measured yearly maxima; i.e., they are based on a larger number of
measurements. The robust nature of the expected maxima may help in the
identification of geographic and temporal patterns in N02 concentrations.
Geographic and temporal patterns are often difficult to discern in measured
yearly N02 maxima because of the large random variance (standard deviation
typically +_ 20%) associated with these once-per-year events. Statistical
parameters of N02 air quality that are associated with a large number of
measurements, e.g., annual mean concentrations, show less variance from
year to year, typically £ 11%.
The potential advantages of using calculated expected values to
characterize yearly maximal N02 concentrations justify an attempt to
formulate a statistical method of determining expected maxima. This
chapter describes the effort made in the present study to develop methods
of predicting expected maxima. Section 3.1 deals with a method based
on the lognormal distribution, while Section 3.2 describes a method
based on the Gamma distribution. Section 3.3 discusses the usefulness of
these statistical methods for the purposes of this study. It is concluded
that the statistical approach involving expected maxima is very useful for
analyzing data quality, estimating the random variance in yearly maxima,
-------
29
and adjusting measured maxima according to yearly sample size. However,
the simple approach involving measured maxima (adjusted to a common yearly
sample size) is preferred for analyzing geographic and temporal patterns
in maximal N02 concentrations.
«
3.1 A METHOD BASED ON THE LOGNORMAL DISTRIBUTION
The mathematical function most often used to analyze air pollutant
frequency distributions is the lognormal distribution function popularized
by Larsen and his co-workers [1,2,3,4,5,6]. Both theoretical consider-
ations and empirical evidence indicate that the assumption of lognormality
is a good approximation for air pollutant concentrations in many situ-
ations [7,8,9,10,11,12]. When properly used, the lognormal distribution
can be a valuable tool in studying air quality data. However, important
questions have been raised concerning the Larsen techniques and the assump-
tion of lognormality [13,14,15,16]. A degree of caution should be observed
whenever the lognormal distribution is used to analyze air quality data, es-
pecially in the case of a reactive pollutant such as nitrogen dioxide [11].
3.1.1 The Lognormal Distribution Function
The assumption that a pollutant concentration variable, C, follows
a lognormal distribution means that the natural logarithm of the concentra-
tion, £nC, follows a normal distribution. If the probability density function
for a normal distribution (with mean, y, and standard deviation, a) is de-
noted by f(y),
-------
30
(D
then the probability density function for a lognormal distribution is
£nC
- £nm ^2
2" '
f (C)dC = f (£nC)d£nC = dC , (2)
where \i =tnm = In (geometric mean)
and as-615 = £n (geometric standard deviation).
The cumulative frequency for the normal distribution is
= I f(x)dx ,
-00
F(y) - / f(x)dx , (3)
and for the lognormal distribution,
A
F(C) = J f(x)dx
0
•/
f(x)dx = F(JlnC) . (4)
Above, the notation for the lognormal distribution is kept closely
tied to the notation for the normal distribution because useful mathe-
matical tables are readily available for the latter. In particular, if
we introduce the change of variable
-------
31
then the distribution functions depend only on the variable z and not on the
parameters wig and sg. Tabled cumulative frequencies for the normal dis-
tribution (or in this case, the lognormal distribution) are commonly available
in terms of the parameter z. '
3.1.2 Maximal Values from Sampling Lognormal Distributions
In this study, we are interested in using the lognormal distribution func-
tion to predict expected yearly maximal one- hour concentrations of N02. Assuming
that we know m and s for the concentration frequency distribution, and
9 g
assuming that a yearly sample consists of N = 8760 independent measurements,
then the distribution of the yearly maxima can be readily calculated (see
Appendix B). For large N, the cumulative frequency distribution for the
yearly maxima, c * is (approximately)
-NCI - F(C )] -N[l - F(£n C )]
H(y = e m - e m , (6)
where F is the cumulative distribution function for the normal distribution.
The median value for the yearly maximum is found by setting M(Cm) equal to
TF; i.e., the median of the yearly maxima for given m and s is obtained by
simply solving
_,. r x , , 0.693 ...
F(£nCm) = 1 -nr= 1 - -fp- , (7)
using values for F found in common mathematical tables. The z value
which is the solution to Equation (7) for N = 8760 is z = 3.78. That is,
-------
32
for samples of size 8760 drawn independently from a lognormal distribution,
the median of the maxima of those samples would be determined from
3.78 =
The expected value of the maximum for a sample of size 8760 corresponds to
*
a z value of 3.82. Thus, for the expected maximum
3.1.3 Adjusting Measured Maxima for Incomplete Sampling
As an aside, we note that the above results are useful for a special
task—adjusting measured maxima to account for incomplete sampling. Since
the sample sizes in our data base can range from 6570 to 8760 hours per
station-year, the recorded maxima are always less than or equal to the
actually occurring maxima during all 8760 hours. The lower the number of
sampling hours, the more likely it is that the recorded maximum is less
than the actual maximum. The recorded maxima should be adjusted upward
to account for incomplete sampling. To make this adjustment, Equation (7)
can be solved for the median z value, call it z1 , for the distribution of
maxima corresponding to the actual sample size, N1. The recorded maximum
can then be adjusted upward by a factor of
The expected maximum is computed by integrating the function C
d M(CJ
nr
m d cm
Note that Larsen's approximate formula for the expected maximum is m
r 3.81
Cm = mg sg •
-------
33
3.78 - z1
.
ragsg
= s.
(10)
Table 3.1 provides estimates of z1 for various sample sizes.
Table 3.1 Median z Values for the Maximum As a
Function of Sample Size
All
Sample Size, N1
8760
8000
7000
6000
5000
4000
3000
Median z1 Value
3.78
3.76
3.72
3.68
3.64
3.58
3.50
yearly maximal values reported in this and subsequent
have been adjusted according to Equation (10). The adjustment
chapters
factors
were not of great consequence; they ranged from around 1.005 to 1.07 for
the various station-years. The results of this study should be insensi-
tive to the specific assumptions (e.g., lognormality) which were made in
deriving the correction factors for incomplete sampling.
3.1.4 Test of Theory for Predicting Expected Maxima
Equations (6) through (9) provide a means for predicting expected yearly
maximal N02 concentrations based on the entire frequency distribution of
-------
34
concentrations. Before this method is accepted as valid, it should be
verified by comparing the predicted maxima with the actual maxima. Equation
(6), which forms the foundation for the method, can be tested simultaneously
against all 450 station-years in the data base. This is accomplished by
using the z parameter which puts the lognormal distribution in a universal
mathematical form, independent of m and s . Equation (6) will predict the
y y
theoretical distribution of the z values for the maxima. This theoretical
distribution can be verified by comparison with the distribution of z values
for the actual maxima.
The z values for the actual maxima are calculated according to Equation
(5),using the (adjusted) recorded maxima, geometric mean, and geometric
standard deviation specific to each individual station-year. In testing
the methodology, we start by using the geometric mean (m ) and geometric
standard deviation (s ) for each station-year as listed on the SAROAD output.
These listed values for m and s are the geometric mean and geometric standard
deviation calculated from all the measured concentrations each year.
Figure 3.1 compares the theoretical distribution of z values for the
maxima (assuming 8760 independent samples per year) with the histogram of
z values for the actual maxima. The disagreement is obvious. The histogram
for the actual maxima is slightly more spread out than the theoretical
distribution, and the median of the histogram (z = 2.99) is substantially
lower than the median of the theoretical distribution (z = 3.78). The
predicted maxima based on the lognormal theory would tend to be greater
than actually occurring maxima. For typical geometric standard deviations,
ranging from 1.5 to 2.5, the lognormal theory would tend to overpredict
the maximum by 30% to 100%.
-------
-Q
O
Q.
1.2-
1.0-
.8-
.6.
.4 -
.2 -
Jl
I
Theoretical Probability Density
Function of z Values for Yearly
Maximal One-Hour Concentration
Assuming Independent Sampling,
Sample Size of 8760, and Lognormal
Concentration Distribution.
Histogram of 2 Values for the
Actual Maxima Based on all 450
Station-Years and on MQ and s_
as Given in SAROAD. *
2.0
3.0 4.0 5.0
z Value of Yearly Maximal Concentration
6.0
co
CJ1
>7.0
Figure 3.1 Comparison of Theoretical Distribution of Maximal z Values with Actual Data
(m and s as Given in SAROAD)
y 9
-------
36
Much of the disagreement in Figure 3.1 may be due to the assumption
that the entire concentration frequency distribution is lognormal. Hourly
N02 concentrations often follow an "s-shaped" curve when plotted on log-
probability paper rather than a straight (lognormal) line. Figure 3.2
presents examples of this type of deviation from lognormality. Because
of the "s-shape" phenomenon, a lognormal distribution fit to the entire
range of yearly concentrations will tend to overpredict the maxima. This
error can be partially corrected for by using a lognormal distribution that
is fit only to the upper end of the concentration frequency distribution.
For instance, a lognormal distribution can be defined by the geometric mean
and 99th percentile concentration, the geometric mean and 95th percentile,
the geometric mean and 90th percentile, the 90th percentile and 99th per-
centile, the arithmetic mean and 99th percentile, etc. Several of these
alternate methods of fitting a lognormal distribution were tried; all of
the methods yielded about the same level of improvement over the log-
normal distribution that was fit to the entire range of concentrations.
The method based on the arithmetic mean and 99th percentile was chosen
for further study.
Figure 3.3 compares the theoretical distribution of z values for the
maxima to the histogram of z values for the actual maxima, with the histo-
gram now based on a geometric mean and a geometric standard deviation cal-
culated from the arithmetic mean and 99th percentile. The two distributions
*
The lognormal distribution specified by the arithmetic mean (m) and
99th percentile (€99) has the following geometric mean and geometric standard
deviation:
m* = m exp[- \ In2 s*]
(/ c
2.33 - A.332 - 2 £n-~
-------
37
1
J
4
>
4
3
2
1
a
i
7
«
S
4
3
2
1
b
.t.
.
I
. j
Louis, Missouri
X
L. ..
(1971)
x;
Ac
•«-
tua
X*
VI^BHfl
1
X
•>••*•
Fre
rvi
qu<
j
x^«
^~
jnc)
r x
x^
' t
)1s
t-fn:
V
\
p>
jx
1
trit
K-^SS
X* '
)Ut1(
X
XV. •
^ ~^
>n
^x
~ 1 I
Lognormal Distribution
X1
,x
^
^ *"
X^
*^
Fit to
X*
^
^x
^
x-
Data
t
7
«
S
4
3
2
1
»
1
7
*
1
4
1
2
1
Long Beach, California
(1972
)
Actual Frequency Distrlbutlpn
S-
r'>
X
X*
M^
X
x
C*
\
V
^
fV
p^
•*-Lognormal
••"•(
'S
yr~~
i<
,
-------
1.4
1.2
1.0
.8
.6
.4
.2
2.0
Theoretical Probability Density Function
of Z Values for Yearly Maximal One-Hour
Concentration Assuming Independent
Sampling, Sample Size of 8760, and
Lognormal Concentration Distribution.
co
CO
Histogram of z Values for the Actual
Maxima Based on all 450 Station.Years
and on n| and s| Computed from the
Arithmetic Mean and 99th Percentile.
5.0
z Value of Yearly Maximal Concentration
6.0
7.0
Figure 3.3 Comparison of Theoretical Distribution of z Values with Actual Data
(m* and s* Calculated from Mean and 99th Percentile)
-J
-------
39
have approximately the same width, and the median of the histogram (3.58)
is not far from the median of the theoretical distribution (3.78). Since
the theoretical distribution is still centered around a higher z value than is
the histogram, predicted maxima based on the lognormal theory would still
tend to be greater than actually occurring maxima. The overprediction of
the maxima would typically be about 10% to 20%.
3.1.5. A Modified Lognormal Approach
The method for predicting expected maxima that was developed in Section
3.1.2 and tested in Section 3.1.4 is based on the assumption that each of
the 8760 hourly measurements in a year is independent, i.e., that no auto-
correlations exist among the hourly N02 data. This assumption is obviously
incorrect because of the persistence of meteorology over a span of a few
hours to a few days and because of the consistent diurnal and seasonal
patterns in NOo concentrations. The autocorrelations in NOg concentrations
may explain why measured maxima tend to be lower than the maxima predicted
by the lognormal theory. Because of the autocorrelation, the number of
"independent" conditions that are being sampled are, in effect, less than
the assumed value of 8760. This decreases the chance of attaining very
high concentrations.
The above observation suggests a way to improve the method of predicting
expected N02 maxima. It might be possible to discount for autocorrelation
in the data by reducing the sample size used to compute the theoretical
distribution of yearly maxima. To provide a fit to the nationwide air
quality data, the "effective" sample size can be chosen so that the median
-------
40
of the theoretical distribution in Figure 3.3 matches the median of the
histogram (which is 3.58). This value for the sample size turns out to
be 3990.
The results of this "modified lognormal approach" are presented in
Figure 3.4. The median of the theoretical distribution has been force fit
to the median based on the nationwide data. It is somewhat encouraging,
however, that the shape of the histogram appears to agree fairly well with
the shape of the theoretical curve. A Kolmogorov-Smirnov (K-S) test was run to
determine if the two distributions are significantly different. The K-S test
rejected the hypothesis that the two distributions are the same at a signifi-
cance level of 5%; i.e., there is less than a 5% chance that the two distribu-
tions are identical. Since the sample size is large (450), it is not obvious
if a statistically significant difference between the distributions is really
of practical importance; i.e., the difference between the distributions may be
very small but still statistically significant.
3.1.6 Predicting Expected Maxima
The modified lognormal approach described in the previous section can
be used to predict expected maxima for any station*year as follows:
Input Data:
1. Arithmetic mean NOp concentration, m
2. 99th percentile NO,, concentration, Cgg.
Calculations:
1. Compute m* and s* corresponding to m and Cgg (see footnote
on page 36 for formulas).
-------
.Q
10
-Q
O
Q.
Theoretical Probability Density Function
for the Modified Lognormal Approach
(Assumed Sample Size Reduced to 3990
to Match Median with Histogram)
Histogran of z Values for the Actual
Maxima Based on all 450 StatiorvYears
and on m* and si Computed from the
Arithmetic Mean and 99th PercentHe.
4.0 5.0
z Value of Yearly Maximal Concentration
Figure 3.4 Comparison of Theoretical Distribution of z Values with Actual Data
(Modified Lognormal Approach)
-------
42
2. Calculate the expected value of the maximum according to
Alternately, the median of the maximum can be used. The
median is slightly lower than the expected value. For the
median,
C = m* s*3'58. (12)
m g g
)
The expected maxima calculated by this approach may be useful for
certain applications, such as providing estimates of the yearly maximum
when very little data (less than 3000 measurements) exist. However, for
the purposes of the present study, using expected maxima does not seem
worthwhile for at least two reasons. First, we are interested in describing
spatial and temporal patterns in maximal N02 concentrations, including
special locations with unusual distributions. The predicted-maxima approach
would involve the assumption that N02 concentrations at all locations follow
the same type of distribution. Forcing all locations into the mold of a
single type of distribution would distort some of the special and interesting
situations. The importance of this problem is evidenced by testing the log-
normal approach for individual locations. Kolmogorov-Smirnov tests indicate
that 13 of the 149 individual locations have distributions of maximal z
values which are different from the modified lognormal distribution at a
1% significance level. Thirty-one of the 149 locations deviate from the
modified lognormal distribution of maxima at a 5% significance level. Thus,
for a large percentage of the locations, adopting the modified lognormal
distribution would be a significant distortion. These "special" locations
-------
43
occur throughout the country and are characterized by either unusually high
or unusually low maximal z values.
The second reason for not using expected maxima in this study is even
more fundamental. The one basic advantage that expected maxima might have
over actually measured maxima is less variance. The expected maximum is
based on two statistics (the annual mean and 99th percentile) that are
more robust than the measured yearly maximum. Thus, the expected maximum
should fluctuate less from year to year than the measured maximum. This
decreased variance should make it easier to discern geographic and tem-
poral patterns in yearly N02 maxima. In reality, this advantage is incon-
sequential. For a given location, the standard deviation of the expected
maximum is typically +_ 18% from year to year. This is almost as large as
the standard deviation of the measured maximum (+_ 20%). The reason for
the large variance in the expected maximum appears to be a compounding of
the variance in the annual mean (+_ 11%) with the variance in the 99th per-
centile (+_ 13%). In any case, using expected maxima does not achieve the
anticipated decrease in random, year-to-year variance.
3.2 A METHOD BASED ON THE GAMMA DISTRIBUTION
The basic problem in using the (unmodified) lognormal distribution to cal-
culate expected maximal concentrations for nitrogen dioxide is overprediction.
The lognormal distribution appears to have a "heavier tail" than actual
frequency distributions of NOp concentrations. Other mathematical functions,
with lighter tails, might provide better predictions of maximal N02 concen-
trations. Light-tail mathematical functions that have been recommended in
the literature are the Gamma distribution, the Weibull distribution,and
the exponential distribution.
-------
44
The resources allocated to this study do not permit a thorough investi-
gation of several alternative mathematical distributions as applied to N02
data. However, it is worthwhile to test at least one "light-tailed" distri-
bution. The Gamma distribution was selected for this test.
3.2.1. The Gamma Distribution
The probability density function for a pollutant concentration variable
following a Gamma distribution is
ra-l -C/S
9 (C) = ' * . (13)
r(a) ea
where a > 0,
6>0,
and r(a) = Gamma function of a.
The cumulative frequency of the Gamma distribution,
C
G (C) = J g(x) dx, (14)
0
is listed in mathematical tables. Unlike the normal distribution, which
can be put in a universal form by a change of variable to the z parameter,
no change of variable exists which makes the Gamma distribution independent
of both a and 3. A partial normalization is accomplished with
t = C/0. (15)
Tables for the cumulative frequency distribution are typically found in terms
of the variable t. However, a separate table is required for each value of a.
For the purpose of predicting expected yearly maxima, the best results
for the lognormal distributions were obtained when the distribution was fit
to the "upper end" of the actual concentration data. Specifically, the
-------
45
lognormal distribution was fit to the arithmetic mean (m) and the 99th per-
centile (Cgg). To be consistent, the Gamma distributions will also be fit
to the arithmetic mean and 99th percentile. This can be done by choosing a
according to the following table:
Cgg/m a cgg/m a
2.25 5.5 3.02 2.5
2.32 5.0 3.32 2.0
2.41 4.5 3.78 1.5
2.51 4.0 4.61 1.0
2.64 3.5 6.64 0.5
2.82 3.0 oo o
and by choosing B as
3 = m/a . (16)
3.2.2 Maximal Values from Sampling Gamma Distributions
Assuming that a yearly sample consists of N = 8760 independent
measurements, then the distribution of yearly maxima can be readily
calculated (see Appendix B). For large N, the cumulative frequency
distribution for the yearly maximum, Cm, is
where G is the cumulative distribution function for the Gamma distribution.
This distribution can be shown to be approximately (see Appendix B)
M(Cm) - e"e S (18)
C
where s * -~- - A, (19)
P
and A is the solution to
a-l -A
N •
1 Aa-l -A _ 1 /9n\
A e - • {20)
-------
46
Using Equation (18), the median value of the maximum can be shown to be
C =3 [A- £n(£n2)]. (21)
The expected value of the maximum is
Cm = 6 (Y + A), (22)
where y = Euler's Constant = 0.5772
3.2.3 Test of Theory for Predicting Expected Maxima
Equations (18) through (22) provide a means of predicting expected
yearly maximum concentrations by using a Gamma distribution. Equation (18),
which is the basis for the method, can be tested simultaneously against
all 450 station-years in the data base. Equation (18) predicts the theo-
retical distribution of the "s parameter." This theoretical distribution
can be compared with the distribution of actual "s parameters" for all
station-years in the data base. The actual s values are calculated ac-
cording to Equations (19) and (20), using the actual C , N = 8760, and
a and g determined from the arithmetic mean and 99th percent! le for each
station-year.
Figure 3.5 presents the results of testing the Gamma distribution
against the actual data. The agreement is very poor compared with the
equivalent test for a lognormal distribution (Figure 3.3). The median of
the theoretical distribution of s values is well below the median of the
histogram of actual s values, implying that the Gamma distribution would
underpredict yearly maximum concentrations. Also, the theoretical distri-
bution has much less spread than the histogram. This means that the Gamma
-------
.4 -
.3
$ .2
o
a.
Probability density function of s values for
yearly one-hour maximum assuming independent
sampling, sample size of 8760, and gamma
concentration distribution.
.1 -
Histogram of s values for the actual maxima
based on all 450 station-years and on a and
6 computed from the arithmetic mean and 99th
percentile.
s Parameter for the Yearly Maximum
Figure 3.5 Comparison of the Theoretical Distribution of s Values with Actual
(Gamma Distribution Approach)
Data
-------
48
distribution also underestimates the variance in the yearly maxima. These
observations indicate that the Gamma distribution is too "light-tailed" com-
pared with actual frequency distributions of N02 concentrations.
Section 3.1 revealed that the lognormal distribution (with a tail
«/ C"1 e"£n2C) was slightly "heavy-tailed" compared with actual N02 frequency
distributions. The present section shows that the Gamma distribution (with
a tail'vC01'1 e"C) is very "light-tailed" compared with actual N02 concentrations.
Perhaps other distributions, such as certain forms of the Weibull distribu-
tion, may provide a compromise between the lognormal and Gamma and may result
in a better fit to the actual data. However, further investigation of math-
ematical distributions is not in line with the main purposes of this study.
Only limited use will be made of mathematical distributions in this report.
For the purposes of this study, the lognormal distribution appears sufficient.
3.3 SUMMARY: USES OF MATHEMATICAL DISTRIBUTION FUNCTIONS
The introduction to this chapter identified four potential uses for
mathematical distributions in analyzing maximal N02 concentrations:
1. To identify outliers for the data quality check;
2. To estimate the random variance in yearly maxima;
3. To adjust yearly maxima for incomplete sampling; arid
4. To characterize patterns in yearly maxima, using
expected (predicted) maxima.
Based on our investigation of mathematical distributions, we conclude that
the first three uses are appropriate in this study but the fourth use is not.
A summary of our results and conclusions concerning each use follows.
-------
49
Data Quality Analysis
The "modified lognormal approach" for predicting expected maxima served
as a useful method to identify potential outliers among the recorded maxima
(see Chapter 2). Station-years for which the actual maximum deviated a large
amount from the expected maximum were subjected to a data verification pro-
cess. The statistical method of identifying outliers was part of a more
comprehensive procedure for noting anomalous data that included a visual
scan of all the frequency distributions and a detailed examination of hourly
CAMP data.
Variance in Yearly Maxima
The modified lognormal approach can provide an estimate of the random
variance in yearly one-hour maxima. Using the theoretical distribution func-
tion in Figure 3.4, the cumulative frequency range from 16% to 84% is assumed
to represent +_ 1 standard deviation. The z values for the cumulative 16th
and 94th percentiles are 3.32 and 3.91, respectively. Thus, the standard
deviation away from the expected maximum (z = 3.62) is
, ,? + 0.29
Cm = m* s*3'62 - 0.30 . (23)
The percentage variance in the expected maximum depends on the geo-
metric standard deviation. Table 3.2 presents results based on Equation (23)
for values of s* from 1.3 to 2.3 (nearly all station-years of data have
«/
values of s* in the range 1.5 to 2.0).
-------
50
Table 3.2 Variance in Yearly One-Hour NO Maxima
*£ Standard Deviation of Yearly Maxima
1.3 +8%
- 8%
1.5 +12%
- 11%
1.7 + 17%
- 11%
1.9 + 20%
- 18%
2.1 + 24%
- 20%
2.3 + 27%
- 22%
Since the average s* for all stations is 1.8, Table 3.2 indicates that the
typical variance should be +_ 17 or 18%. It is encouraging that this re-
sult agrees with the actual variance in yearly maxima. The actual stan-
dard deviation of maxima for individual stations (determined for the years
1970 to 1974) averages around +_ 20%.
Adjustment for Incomplete Sampling
The sample sizes in our data base, ranging from around 6600 to 8600 hours
per year, are all less than 100% complete (8760 hours per year). Thus, the
recorded maxima are less than or equal to the actual maxima during all 8760
hours. As discussed in Section 3.1.3, the lognormal distribution can be
used to calculate adjustment factors which account for incomplete sampling.
These adjustment factors have been applied to the maximum for each station-
year in the data base.
-------
51
As applied to the present data base, the adjustment factors for in-
complete sampling are quite small. This is because all station-*years in
our data base are at least 75% complete. The results of this study should
be insensitive to the specific assumptions used in deriving adjustment fac-
tors. However, caution should be observed before applying this method to
data which are less than 75% complete. The underlying assumptions will be
more important for cases requiring large adjustments to the recorded max-
ima.
Patterns in Yearly Maxima
One of the basic reasons for our investigation of mathematical distri-
butions was to calculate expected maxima for each station-year. It was hoped
that the expected maxima would exhibit less random variance that the actual
maxima. Eliminating some of the variance would facilitate identifying spa-
tial and temporal patterns in NOg maxima.
As it turned out, the benefit gained in terms of reduced variance was
insignificant. The year-to-year fluctuations in the expected maxima (cal-
culated from the mean and 99th percentile using the modified lognormal ap-
proach) were nearly as great as the fluctuations in the actual maxima. For
this reason, and because of the danger that calculating expected maxima
might distort some of the interesting special cases, it was concluded that
the best approach for characterizing N02 maxima is to use actually measured
maxima. The measured maxima will be used in this study.
-------
52
3.4 REFERENCES
1. C. E. Zimmer and R. I. Larsen, "Calculating Air Quality and Its Control,"
Journal of the Air Pollution Control Association, Vol. 15, p. 565, 1965.
2. R. I. Larsen, C. E. Zimmer, D. A. Lynn, and K. G. Berne1, "Analyzing Air
Pollution Concentration and Dosage Data," Journal of the Air Pollution
Control Association, Vol. 17, p. 85, 1967.
3. R. I. Larsen, "A New Mathematical Model of Air Pollutant Concentration
Averaging Time and Frequency," Journal of the Air Pollution Control
Association, Vol. 28, p. 24, 1969.
4. R. I. Larsen, A Mathematical Model for Relating Air Quality Measure-
ments to Air Quality Standards, Publication AP~89. U.S. Environmental
Protection Agency, Research Triangle Park, North Carolina, 1971.
5. R. I. Larsen, "An Air Quality Data Analysis System for Interrelating
Effects, Standards, and Needed Source Reductions," Journal of the
Air Pollution Control Association, Vol. 23, p. 933, 1973.
6. R. I. Larsen, "An Air Quality Data Analysis System for Interrelating
Effects, Standards, and Needed Source Reductions - Part 2," Journal of
the Air Pollution Control Association, Vol. 24, p. 551, 1974.
7. F. A. Gifford, "The Form of the Frequency Distribution of Air Pollution
Concentrations," Proceedings of the Symposium on Statistical Aspects
of Air Quality Data. EPA Document 650/4-74-038. EPA Office of Research
and Development, 1974.
8. H. D. Kahn, "Note on the Distribution of Air Pollutants," Journal of
the Air Pollution Control Association, Vol. 23, p. 973, 1973.
9. J. B. Knox and R. Lange, "Surface Air Pollutant Concentration Frequency
Distributions: Implications for Urban Modeling," Journal of the Air
Pollution Control Association, Vol. 24, p. 49, 1974.
10. N. D. Singpurwalla, "Extreme Values from a Lognormal Law with Applica-
tions to Air Pollution Problems," Technometrics. Vol. 14, p. 703, 1972.
11. H. E. Neustadter, S. M. Sidik, and J. C. Burr, Jr., "Statistical
Summary and Trend Evaluation of Air Quality Data for Cleveland, Ohio
in 1967 to 1971: Total Suspended Particulate, Nitrogen Dioxide, and
Sulfur Dioxide," NASA TN D-6935, Lewis Research Center, Cleveland, 1973.
-------
53
12. D. B. Turner, "A1r Quality Frequency Distributions from Dispersion
Models Compared with Measurements," Proceedings of the Symposium
on Statistical Aspects of Air Quality Data, EPA Document 650/4-74-038,
EPA Office of Research and Development, 1974.
13. N. R. Patel, "Comment on a New Mathematical Model of Air Pollution
Concentration," Journal of the Air Pollution Control Association,
Vol. 23, p. 291, 1973.
14. R. E. Barlow, "Averaging Time and Maxima for Air Pollution Concentration,"
NTIS #AD-729413, 1971. !
15. T. C. Curran and N. H. Frank, "Assessing the Validity of the Lognormal
Model when Predicting Maximum Air Pollution Concentrations," 68th
Meeting of the Afr Pollution Control Association, Boston, 1975.
16. D. T. Mage and W. R. Ott, "An Improved Statistical Model for Analyzing
Air Pollution Concentration Data," 68th Meeting of the Air Pollution
Control Association, Boston, 1975.
-------
54
4.0 CHARACTERIZATION OF PRESENT N02 AIR QUALITY LEVELS
This chapter summarizes present nitrogen dioxide air quality
in the United States. The results are based on data for the years 1972,
1973, and 1974. The discussion includes three indices of N02 air quality—
the annual arithmetic mean, the 90th percent!le of hourly concentrations,
and the yearly one-hour maximum. Geographical patterns in these indices
are illustrated nationwide; intraregional patterns are examined within
the Los Angeles area and the New York-New Jersey-New England area. This
chapter also investigates the effects of local environment (urban vs, rural,
commercial vs. industrial, etc.) on N0« concentration distributions.
4.1 DATA BASE FOR DESCRIBING PRESENT N02 AIR QUALITY
From the overall data base of 450 station-years, data for the years
1972, 1973, and 1974 are chosen for the purpose of describing present air
quality levels. Each station with at least one year of complete data from
1972 to 1974 serves as a measurement point for present N02 air quality. For
those stations with two or three years of data from 1972 to 1974, air
quality indices are averaged over the two or three years. There are two
advantages in using data from three years rather than from a single year.
First, including more years increases the number of locations in the analysis,
Second, averaging over two or three years, where possible, provides more
robust estimates of the air quality indices.
Table 4.1 lists the 123 stations which have at least one year of
complete data from 1972 to 1974. The arithmetic mean, 90th percentile,
-------
55
TaMe 4:1 Stations
1. Phoenix, Arizona
(002A01)
2. Anaheim, California
(001101
3. Azusa, California
(002101)
4. Bakersfleld, California
(003F01)
5. Barstow, California
{001101)
6. Burbank, California
(002101)
7. Camarlllo, California
(001101)
8. CMco, California
(001F01)
9. Chlno, California
(001101)
10. Concord, California
(001101)
11. Costa Mesa, California
(001101)
12. El Cajon, California
(001101)
13. Eureka, California
(002F01)
14. Fresno, California
(002F01)
15. Indie, California
(001101)
16. La Habra, California
(001101)
17. Lancaster, California
(001101)
Tor Character!zing "Present KCTg Air Qua 1 ity
18. Lennox,.California
(001IOf)
35. Pittsburgh, California
(001101)
36. Pomona, California
(001101)
19. Livermore, California
(002101)
20. Long Beach, California 37- R^"f? California
(002101) (002F01)
21. Los Alamltos, California 38- **£?"??• California
(001101) (001101)
22. Los Angeles (Downtown), CA39- Redwood City, California
(001101) (001101)
23. Los Angeles (Westwood), CA 40. Richmond, California
(002101) . (003101)
24. Los Angeles (Reseda), CA 41. Riverside. California
(001101)
25. Lynwood, California
(001101)
26. Modesto, California
(001101)
27. Monterey, California
(001101)
28. Napa, California
(003101)
29. Newhall, California
(001101)
30. Norco, California
(001101)
31. Oakland, California
(003G01)
32. Ojal, California
(001101)
(003F01)
42. Rubidoux, California
(001101)
43. Sacramento, California
(003F01)
44. Salinas, California
(001 roi)
45. San Bernadino, California
(001101)
46. San Diego, California
(004101)
47. San Francisco, California
(003101)
48. San Jose, California
(004A05)
49. San Luis Obispo, California
(001F01)
33. Palm Springs, California »• San Rafael, California
(001101) (001101)
34. Pasadena, California
(004101)
51. Santa Barbara, California
(002F01)
-------
56
Table 4.1 Stations for CharacteH zi nn^esent~^~Ai r Quality (Continued)
52. Santa Barbara, California 69. Chicago, Illinois
(004F01) (023A05)
53. Santa Cruz, California 70. Ashland, Kentucky
(001101) (008F01)
54. Santa Rosa, California 71. Louisville, Kentucky
(002101)
55. Stockton, California
(002F01)
"56. Sunnyvale, California
(001101)
57. Upland, California
(003101)
58. Upland, California
(004F01)
59. Vallejo, California
(003101)
(011601)
72. Louisville, Kentucky
(017A05)
73. Newport, Kentucky
(001F01)
74. Ohio, Kentucky
(006N02)
75. Owensboro, Kentucky
(008F01)
76. Baltimore, Maryland
(018F01)
86. St. Louis, Missouri
(002A10)
87. St. Louis, Missouri
(006G01)
88. Las Vegas, Nevada
(009G01)
89. Reno, Nevada
(005101)
90. Bayonne, New Jersey
(003F01)
91. Camden, New Jersey
(003F01)
92. Elizabeth, New Jersey
(004F01)
93. Newark, New Jersey
(002F01)
60. Victorvllle, California 77. Silver Spring, Maryland 94. Phil!ipsburg, New Jersey
(001101) (006F01) (002F01)
61. Vlsalia, California
(001F01)
62. Whlttier, California
(001101)
78. Sprlngield, Massachusetts 95. Buffalo, New York
(005A05) (005F01)
79. Detroit, Michigan
(020A05)
63. Yuba City, California 80. Lansing. Michigan
(001F01) (002F01)
64. Denver, Colorado
(002A05)
81. Saglnaw, Michigan
(002F01)
65. New Britain, Connecticut 82. Afton, Missouri
(002F01) (001601)
96. Buffalo, New York
(007F01)
97. Glens Falls, New York
(003F01)
98.' Hempstead, New York
(005F01)
99. Kingston, New York
(002F01)
66. Washington, D.C.
(003A05)
67. Atlanta, Georgia
(001A05)
68. Chicago, Illinois
(002A05)
83. Belle Fontaine Neighbors, 100. Mamaroneck, New York
Missouri (002601) (002F01)
84. Clayton, Missouri
(001601)
85. St. Ann, Missouri
(001801)
101. New York City, New York
(006A05)
102. New York City, New York
(OSOF01)
-------
57
Table 4.1 Stations for Characterizing Present N Air Quaffty (Continued)
103. New York City, New York
(061A05)
104. Niagara Falls, New York
(006F01)
105. Rensselaer, New York
(001F01)
106. Rochester, New York
(004F01)
107. Schenectady, New York
(003F01)
108. Syracuse, New York
(005F01)
109. Syracuse, New York
(011F01)
110. Utlca, New York
(004F01)
111. Akron, Ohio
(013H01)
112. Cincinnati, Ohio
(019A05)
113. Portland, Oregon
(002F01)
114. Lancaster City, Pennsylvania
(007F01)
115. Philadelphia, Pennsylvania
(002A05)
116. Philadephia, Pennsylvania
(004H01)
117. Scranton, Pennsylvania
(006F01)
118. Providence, Rhode Island
(005F01)
119. Providence, Rhode Island
(007A05)
120. Memphis, Tennessee
(027N02)
121. Stewart, Tennessee
(005N02)
122. Salt Lake City, Utah
(001A05)
123. Alexandria, Virginia
(009H01)
-------
58
yearly maximum, and ratio of yearly maximum to annual mean for each of
the 123 stations can be found in Appendix C. The information in Appendix C
will serve as the basis for characterizing present N02 air quality.
Figure 4.1 shows the locations of the stations on a map of the United
States. Because of the high density of sites in California, the Los Angeles
area, and the New York-New Jersey-New England region, separate maps are pre-
sented for those areas (see Figures 4.2 through 4.4). The numbers plotted on
the maps correspond to the stations numbers listed in Table 4.1.
4.2 DATA PATTERNS INVOLVING MONITOR EMVIROIWEffT
The SAROAD printout classifies the general environment of each monitor
according to "center city," "suburban," and "rural." For the urban and
suburban classes, a subcategorization can be made according to four local
environments: commercial, industrial, residential, and mobile station.
For the rural class, four other subcategories are possible: commercial,
near urban, agricultural, and power plant. Table 4.2 indicates the distri-
bution of the 123 sites among these categories.
Table 4.2 Number of Sites in Various Categories of Monitor Environment
Center City
Commercial ...
Industrial ...
Residential ...
Mobile
Total
57
6
9
2
74
Suburban
Commercial
Industrial
Residential ...
Mobile
Total
17
7
17
2
43
Rural
Commerical ' ...
Near Urban
Agricultural ...
Power Plant ...
Total
1
1
1
3
6
-------
Station numbers are as listed
1n Table 4.1
For stations marked with •, see
Figures 4.2, 4.3, and 4.4
for station numbers
cn
Figure 4.1 Location of N02 Monitoring Sites in the U.S. (Includes sites
with at least one year of complete data during 1972-1974)
-------
60
Station numbers are as listed in Table 4.1.
For stations marked with • , see Figure 4.3
for station numbers.
Figure 4.2 Location of N02 Monitoring Sites in California
-------
! 32
29
34 ' /
23 22
57/58
62
45
*° ^ j jf n
C 9fy30 ^ 41
25 ;' 16 V /
38
Station numbers are as listed in Table 4.1.
^
w._
Figure 4.3 Location of N02 Monitoring Sites in the Los Angeles Region
-------
Station numbers are as listed
1n Table 4.1
Figure 4.4 Location of N02 Monitoring Sites 1n the
New York-New Jersey-New England Area
-------
63
It is interesting to determine if the different classifications of
monitor environment are associated with different levels of N02 air
quality. Table 4.3 lists average N02 air quality for each station clas-
sification. The only category of stations that stands out from the rest
is the rural/power plant class. The three rural/power plant stations
(Ohio Co.-Kentucky, Memphis-Tennessee, and Stewart Co.-Tennessee) exhibit
very low annual means and 90th percentiles. The yearly maximum for these
stations is only moderately low, leading to an extremely high ratio of
yearly maximum to annual mean. This type of concentration distribution,
low mean and a high maximum-to-mean ratio, is no surprise for these stations.
The rural/power plant sites are subjected to near background N02 concen-
trations except for infrequent fumigation by power-plant plumes.
The most striking feature of the rest of the categories is their
sameness. No substantial differences exist among the eight categories of
center city and suburban stations. Even the three sites labeled as rural/
commercial, rural/near urban, and rural/agricultural are not significantly
different from the center city and suburban sites. Perhaps some of the
uniformity is due to unrealistic classification. The rural/commercial site
(Rubidoux, Ca.) and the rural/agricultural site (Norco, Ca.) are well inside
the boundaries of the Los Angeles basin. The rural/near urban site is in
St. Louis.
In the next section, maps will be presented which illustrate nationwide
patterns in urban N02 air quality. These maps will be based on data
-------
Table 4.3 N02 Air Quality for Various Categories of Monitor Environment
Average N02 Air Quality for Station of Each Type (1972-1974)
Type of Site
Center City/ Commercial
Center City/Industrial
Center City/Residential
Center City/Mobile
Suburban/Commercial
Suburban/Industrial
Suburban/Residential
Suburban/Mobile
Rural /Commercial
Rural /Near Urban
Rural /Agricultural
Rural /Power Plant
Number of
Stations
57
6
9
2
17
7
17
2
1
1
1
3
Annual
Mean
(pphm)
3.5
4.0
3.4
3.0
4.3
4.1
3.2
3.5
2.7
3.0
2.8
0.8
90th
Percent! le
(pphm)
6.2
7.0
6.4
5.2
7.8
7.6
5.8
6.5
5.0
6.0
5.0
1.3
Yearly
Maximum
(pphm)
23.6
20.3
22.2
13.8
27.8
24.5
21.2
26.2
20.3
33.6
22.4
14.4
Ratio of
Maximum
to Mean
6.7
5.1
6.5
4.6
6.5
6.0
6.6
7.5
7.5
11.2
8.0
18.0
-------
65
from all 123 sites listed in Table 4.1, except for the 3 rural/power plant
sites. The rural/power plant sites are atypical and will be treated separately.
The 3 rural/commercial, rural/near urban, and rural/agricultural sites will be
included among the urban locations.
4.3 NATIONWIDE GEOGRAPHIC PATTERNS IN N02 AIR QUALITY
The existing National Ambient Air Quality Standard (NAAQS) for nitrogen
dioxide is 100yg/m3 (approximately 5.3 pphm), annual arithmetic mean. If a
short-term standard is promulgated for nitrogen dioxide, it may be a one-
hour standard, or it may be based on a percentile concentration such as
the 90th percentile. This section provides information which allows a
comparison between present N0« air quality levels nationwide and the NAAQS,
including the annual mean standard and potential one-hour or 90th percentile
standards.
A drawback in characterizing nationwide air quality is the limited
number of monitoring sites. During the 1972-to-1974 period, only 120 urban
sites (58 outside California) provided 75% complete data on hourly N02 con^
centrations. As shown in Figure 4.1, the only areas of the country with
good spatial coverage are California and the northeast sector (Illinois to
New England). Thus, we cannot make definitive conclusions concerning the
status of N02 air quality in all urban areas. We will, however, attempt
to identify broad regions of the country with the potential for exceeding
N02 air quality standards. A better assessment of nationwide air quality
for nitrogen dioxide should be possible in the future as more stations come
on line and as data quality improves from existing stations.
-------
66
4.3.1 Annual Mean N02 Concentrations
Figure 4.5 illustrates the distribution of annual mean N02 concentrations
for the 120 urban stations in the data base. Most of the stations have annual
mean N02 concentrations in the range 1 pphm to 5 pphm. Only 18 of the stations,
15% of the total, exceed the NAAQS for annual mean N02 (5.3 pphm). Five of
the 58 sites outside of California exceed the standard. Within California,
21% of the locations 03 out of 62) violate the standard; all of the California
violations occur in the Metropolitan Los Angeles AQCR.
30% -,
20% -
10% -
12345
L , • ,
6 7 8 9 10
Annual Mean N02 (pphm)
Figure 4.5 Percentage of Urban Stations with Various Levels of Annual
Mean N02 Concentrations (1972-1974)
-------
67
Table 4.4 lists the 18 locations which exceed the national standard
for annual mean N02. Los Angeles and Pasadena head the list at 7.3 pphm,
nearly 50% above the standard. Ten of the top 11 sites are in Los Angeles
County, and 13 of the 18 sites are in the Los Angeles basin. The prepon-
derance of Los Angeles sites in the table is partly due to the intense
photochemical smog problem in Los Angeles and partly due to the large num-
ber of monitoring locations in that air basin.
The 5 non-California sites exceeding the standard are headed by Baltimore
at 6.4 pphm. The other 4 non-California sites are all less than 20% in excess
of the standard.
Table 4.4 Stations Exceeding the NAAQS for Annual Mean N02 {5.3 pphm), 1972-1974
Mean N02 Mean N0?
Station (pphm) Station (pphm)
Los Angeles, Ca. 7.3 Los Angeles (Reseda),Ca. 6.3
Pasadena, Ca. 7.3 Azusa, Ca. 6.2
Burbank, Ca. 7.1 Upland, Ca. 6,0
Pomona, Ca. 6.9 Springfield, Mass, 5.9
Los Angeles (Westwood),Ca. 6.8 Chicago, 111. 5.7
Long Beach, Ca. 6.7 La Habra, Ca. 5.6
Whittier, Ca. 6.5 Newark, N.J. 5.6
Lennox, Ca. 6.4 Lynwood, Ca, 5.5
Baltimore, Md. 6.4 Elizabeth, N.J. 5.3
Figure 4.6 illustrates the nationwide geographic pattern of annual
average N02 concentrations. To avoid cluttering the map, not all of the
120 stations are plotted. Where there are two or more monitoring sites in
close proximity, only the site with the highest annual mean is included in
the map. For instance, only 1 site represents Los Angeles County, only
-------
All concentrations 1n pphm
CO
Figure 4.6 Annual Mean N02 Concentrations at Urban Stations in the United States (1972-1974)
-------
69
4 sites represent the Los Angeles basin, 2 sites represent the St. Louis
area, 1 site represents New York City, etc.
The Los Angeles area stands out in Figure 4.6 with the highest
annual mean N02 concentration in the nation. Several cities in the
Northeast and Chicago in the Midwest also exceed the NAAQS. None of the
sites in the Southeast and none of the sites west of the Mississippi (except
for Los Angeles) violates the federal standard, although Atlanta is close at
4.8 pphm. Because of the sparsity of stations in the Southeast and the
West (except for California), we cannot be sure that the standard is
attained everywhere in those areas. However, since some of the largest
cities in those areas (such as Portland, Salt Lake City, Denver, Phoenix,
and Atlanta) are represented, it seems unlikely that there would be signifi-
cant violations among the unmonitored cities in the West and Southeast. The
main problem areas in the nation with respect to attainment are Los Angeles
and a few cities in the Northeast and Midwest.
4.3.2 90th Percentile N02 Concentrations
Figure 4.7 presents a histogram of 90th percentile N02 concentrations for
the 120 stations in the data base. Most of the sites, 73% of the total, have
90th percentiles below 8 pphm. Only 14 sites, 12% of the total, have 90th
percentile concentrations exceeding 10 pphm.
To point out the sites of the greatest N02 concentrations, Table 4.5
lists the 14 stations with 90th percentile concentrations that exceed 10 pphm.
Eleven stations from the Los Angeles basin (10 from Los Angeles County) head
the list. Of the other 3 sites, 2 are in Maryland, and 1 is in Massachusetts.
-------
70
Percentage of Urban Stations
— i ro co
o o o
&3 5-S IS
I i i
1 1
2 4 6 8 10 12 14 16 18 20
Figure 4.7 Percentage of Urban Stations with Various Levels of
90th Percent!le Concentrations (1972-1974)
Table 4.5 Monitoring Sites with 90th Percentile N07 Concentrations
Greater than 10 pphm (1972-1974) *
Station
90th Percentile
(pphm)
Burbank, Ca. 13.0
Los Angeles, Ca. 12.3
Los Angeles (Westwood), Ca. 12.3
Pasadena, Ca. 12.0
Long Beach, Ca. 12.0
Los Angeles (Reseda), Ca. 11.7
Azusa, Ca. 11.4
Station
Whittier, Ca.
Lennox, Ca.
Pomona, Ca.
Upland, Ca.
Baltimore, Md.
Springfield, Mass.
Silver Spring, Md.
90th Percentile
(pphm)
11,
11,
11
11
11
11
3
0
0
0
0
0
10.0
-------
All concentrations 1n pphm
Figure 4.8 90th Percentile N02 Concentrations at Urban Stations in the United States (1972-1974)
-------
72
Figure 4.8 illustrates the nationwide geographic distribution of 90th
percentile NOp concentrations. Again, to avoid cluttering, only the site with
the highest 90th percentile is listed when there are two or more monitors in
close proximity. The pattern in Figure 4.8 is similar to the pattern for
annual means (Figure 4.6). Los Angeles has the highest concentrations in the
nation, but the rest of the West has relatively low concentrations. A few
cities in the Northeast and Midwest (Springfield, Baltimore, Silver Spring,
New York, Newark, Chicago, and Owensboro) have notably high concentrations.
4.3.3 Yearly Maximal Concentrations
Figure 4.9 illustrates the distribution of yearly maximal onei.hour
concentrations of N02 among the 120 urban stations. Forty-seven of
co
c
o
J3
s_
ZD
<+-
O
O)
CD
It)
-4->
c
O)
o
s_
QJ
Q_
30%-
205L
i — •
o
&?
1 I
10 20 30 40
Yearly Maximum One-Hour N02 (pphm)
50
60
Figure 4.9 Percentage of Urban Stations with Various Levels of
Yearly Maximum N02 Concentration (1972-1974)
-------
73
the stations have yearly maxima which exceed 25 pphm (the California one-hour
standard). Only 4 of the stations experience yearly maxima which exceed 50 pphm.
Table 4.6 lists the 19 stations with yearly NOp maxima exceeding 36 pphm.
Again, the Los Angeles basin dominates the list; 4 of the top 5 and 14 of the
top 19 locations are in the Los Angeles basin. Baltimore and Silver Spring,
Maryland are also repeaters from the lists of "worst stations" for the annual
mean and 90th percentile. Table 4.6 includes three other locations, Barstow,
CA, Ashland, KY, and Denver, CO, that did not appear in Tables 4.4 or 4.5.
Table 4.6 Monitoring Sites with High Yearly Maximal One-Hour Concentrations
(1972-1974)
Station
Los Angeles (Westwood), CA
Los Angeles, CA
Baltimore, MD
Whittier, CA
Pasadena, CA
Barstow, CA
Silver Spring, MD
La Habra, CA
Ashland, KY
Azusa, CA
Yearly One-Hour
Maximum (pphm)
55.8
54.6
51.9
50.6
47.7
47.7
45.1
42.9
41.6
41.0
Station
Yearly One-Hour
Maximum (pphm)
Anaheim, CA
Lennox, CA
Denver, CO
Upland, CA
Long Beach, CA
Lynwood, CA
Chino, CA
Los Angeles (Reseda),
CA
Los Alamitos, CA
40.7
40.7
40.2
39.7
37.7
37.7
37.7
36.7
36.3
-------
74
Figure 4.10 shows the nationwide geographic distribution of yearly
maximal N02 concentrations. Only the station with the highest yearly
maximum is listed on the map when there are several stations in close
proximity. Los Angeles again stands out as having the greatest N02 con-
centrations in the country. The Baltimore area also appears as a "hot-
spot". Other areas with high maximal N02 concentrations include Denver,
CO, Ashland, KY, Owensboro, KY, and New York City.
4.4 INTRAREGIONAL PATTERNS IN N02 CONCENTRATIONS
There are at most four regions in the country where the monitoring
sites in our data base are sufficiently dense to describe spatial patterns
of N02 concentrations within the region. These regions are the Los Angeles
basin, the San Francisco Bay area, the St. Louis region, and the New York-
New Jersey-New England area. Two of these areas, Los Angeles and New York,
are particularly interesting because they exceed the NAAQS for annual mean
N02< Intraregional patterns for the Los Angeles and New York will be dis-
cussed below.
4.4.1 Metropolitan Los Angeles AQCR
Figure 4.11 presents a map of the Metropolitan Los Angeles region. The
map shows the six counties that are within or partially within the region. It
also notes the location of major cities in the region. When analyzing air
pollution in the Los Angeles region, it is important to note that the area
-------
All concentrations 1n pphm
01
Figure 4.10 Yearly One-Hour Maximum'Concentrations at Urban Stations
in the United States (1972-1974)
-------
RIVERSIDE
Anaheim
ORANGE
Figure 4.11 Map of the Metropolitan Los Angeles AQCR
-------
77
of highest traffic density and greatest industrial/commercial activity
is the central/coastal area. This is demonstrated in .Figure 4.12 [l],which
shows the distribution of NOX emissions within the region; a major portion
of regional emissions occur in the western/central parts of Los Angeles
County. It is also important to recognize that the meteorology of Los Angeles
is dominated by a daytime sea breeze during much of the year. Typically,
there is a transport of air pollutants from the coast toward the inland
areas. Dilution occurs along with the transport. It is noteworthy that
mixing heights are lowest at the coast and greatest inland.
Figures 4.13, 4.14, and 4.15 illustrate the spatial patterns of N02
concentrations within the Los Angeles basin. The three figures are for annual
mean, 90th percentile, and yearly maximum concentrations, respectively. The
patterns in all three figures are similar. The greatest N02 concentrations
occur in the area of greatest NO emission density, i.e., the coastal and
^\
central parts of Los Angeles County. The stations at Westwood, Downtown,
Pasadena, and Burbank show particularly high concentrations. The eastern/
inland stations show moderately high concentrations, while the Ventura and
Santa Barbara stations record relatively low concentrations.
Figures 4.13 through 4.15 indicate that the N02 problem in Los Angeles
is partly regional in nature. Stations such as Azusa, Pomona, and Upland that
are directly downwindnaf the most source-intensive area experience fairly
high concentrations even though they are located in areas of relatively low
emission density. Stations which rarely experience transport from the central
areas, such as Santa Barbara or Ventura County stations, show the lowest
concentrations. There is also evidence that the N02 problem is partly
-------
I
1 Dot = 12 Tons/Day NOX Emissions
00
Figure 4.12 Nitrogen Oxides Emission Density Map for the Los Angeles Region
Source: Reference [1]
-------
\ *I
6'
6.2
6.8 7.3
6.0/4.9
6.9/
6.5
5.5 _(. a4r 2.8 5.0
;~5.6 xy"
\
All concentrations in pphm
vo
Figure 4.13 Annual Mean N02 Concentrations in the Los Angeles Region (1972-1974)
-------
All concentrations In pphm
Figure 4.14 90th Percentile N02 Concentrations in the Los Angeles Region (1972-1974)
-------
\
20
\ 23
\
\
f' 37 35
41
/
„ .
48 i 40/30
56 55 34/
All concentrations 1n pphm
,41 37 ,43
38 *6 41
51 f38 ,' 22 20;
X..
— 00
Figure 4.15 Yearly One-Hour Maximum N02 Concentrations in the Los Angeles Region (1972-1974)
-------
82
localized in nature. Norco and Rubidoux in Riverside County show lower
concentrations than surrounding stations, and the Lynwood site in Los
Angeles County has significantly lower NO,, levels than adjacent stations.
4.4.2 New York-New Jersey-New Engtancf Kfea
Figure 4.16 presents a map of the New York-New Jersey-New England
area. This area has been studied extensively by Cleveland, Kleiner,
and their associates at Bell Laboratories [2,3,4,5]. An NO emission
A
density map, prepared by the Bell Labs group, is shown in Figure 4.17.
The most striking feature of the emission density map is the high level of
emissions in the northeast New Jersey and New York City areas.
Figures 4.18, 4.19, and 4.20 present the spatial pattern of N0« concen-
trations for the region. The three figures represent annual mean, 90th
percentile, and yearly maximum concentrations, respectively. On each of the
three figures, the New York City/ northern New Jersey area shows relatively
high concentrations. This makes sense in light of the emission density map.
No other consistent patterns emerge. Springfield, Mass, has a very high annual
mean, even though it is in a region of low emission density. Between Spring-
field and New York (the high-concentration sites) are two exceptionally clean
stations. The lack of consistent patterns is partly due to the relatively
large scale of the region (~ 500 km). On this scale, local emissions are
probably much more important than transport effects, at least for nitrogen
dioxide. Another possible reason for the lack of patterns may be inconsis-
tencies between measurement methods and monitor siting criteria used by the
various state and local agencies in the area.
-------
83
Glens Falls
Phillipsburg
'* Hartford
*New Britain
hiladelphia
* Camden
Figure 4.16 Map of the New York-New Jersey-New England Area
-------
84
All Emissions are 1n 1000 Tons Per Year
Figure 4.17 NO,, Emissions in Various AQCRs in the New York-
New Jersey-New England Area
Source: Reference [5]
-------
85
All concentrations in pphm
Figure 4.18 Annual Mean NOg Concentrations 1n the New York-
New Jersey-New England Area (1972-1974)
-------
86
All concentrations 1n pphm
Figure 4.19 90th PercentHe N0« Concentrations 1n the New York-
New Jersey-New England Area (1972-1974)
-------
87
All concentrations 1n pphm
Figure 4.20 Yearly One-Hour Maximum NOg Concentrations In the
New York-New Jersey-New England Area (1972-1974)
-------
4.5 REFERENCES
1. J. Trijonis, G. Richard, K. Crawford, R. Tan, and R. Wada. An Implemen-
tation Plan for Suspended Particulate Matter in the Los Angeles Region,
TRW Environmental Services, EPA Contract No. 68-02-1384, Redondo Beach, Ca.,
March 1975.
2. S.M. Bruntz, W.S. Cleveland, I.E. Graedel, B. Kleiner, and J.L. Warner,
"Ozone Concentrations in New Jersey and New York: Statistical Association
with Related Variables," Science. Vol. 186, p.257, 1974.
3. W.S. Cleveland and B. Kleiner, "Transport of Photochemical Air Pollution
from Camden-Philadelphia Urban Complex," Environmental Science and Tech-
nology. Vol. 9, p. 869, 1975.
4. W.S. Cleveland, B. Kleiner, and J.L. Warner, "Robust Statistical Methods
and Photochemical Air Pollution Data," Journal of the Air Pollution Control
Association, Vol. 26, p. 36, 1976.
5. W.S. Cleveland, B. Kleiner, J.E. McRae, and J.L. Warner, "The Analysis of
Ground-Level Ozone Data from New Jersey, New York, Connecticut, and
Massachusetts: Transport from the New-York^WHrbpolitan Area," Bell
Laboratories, Murray Hill, New Jersey, 1975.
-------
89
5.0 TRENDS IN NITROGEN DIOXIDE AIR QUALITY
This chapter examines recent historical trends in ambient N0? con-
centrations. As in the previous chapter, the focus is on three air quality
indices: annual mean, 90th percentile, and yearly one-hour maximum. Five-
and ten-year changes in N02 concentrations are examined, using the data base
described in Chapter 2. Year-to-year trends are investigated with an ex-
panded data base. The year-to-year trends are discussed in terms of emis-
sion-factor changes and source growth.
5.1 FIVE- AND TEN-YEAR CHANGES IN N02 AIR QUALITY
A convenient way of determining overall air quality trends is to fit
a regression line to the year-to-year levels of air quality. The change in
ambient concentrations over a period of interest is defined by the values
of the regression line at the beginning and end of the period. This method
is applied here to determine net changes in N02 concentrations for individual
stations over the periods 1969-1974 and 1964-1974.
The data base described in Chapter 2 serves as the basis for this trend
s
study. To ensure adequate data for the trend estimates, five-year trends are
determined only for those stations with at least two years of complete data
from 1969 to 1971 and at least two years of complete data from 1972 to 1974.
/
Ten-year trends are computed only for those stations with at least three
years of complete data from 1964 to 1968 and at least three years of
complete data from 1970 to 1974. For a given location, data are included
only if they have been taken with the same monitoring method each year.
Stations are excluded from the analysis if the site has been relocated
-------
90
or if there has been a change in the monitoring agency. With these
restrictions, 19 sites in the data base qualify for the five-year trend
analysis, and 10 sites qualify for the ten-year analysis.
Table 5.1 summarizes five-year trends at various monitoring sites.
Net percentage changes in annual mean, 90th percentile, and yearly
maximum NOp concentrations are listed for each site. The sites are
grouped according to geographical area.
Caution should be followed in drawing inferences from five-year
trend estimates. Year-to-year meteorological variance can play havoc
with air quality trends over a span of five years. Substantially dif-
ferent results occur, depending on whether the first (or last) couple of
years were good years for air quality or bad years. With the 90th per-
centile concentrations there is an additional problem, round-off error.
For most locations, the percentile concentrations are reported only to
the nearest pphm, and the error in round-off can be significant. The large
upward trend in 90th percentile concentrations at Portland may, in part,
be due to this type of error. The reported 90th percentiles at Portland
from 1970 to 1974 are 4, 4, 4, 5, and 5 pphm, respectively.
With these caveats in mind, we make the following observations con-
cerning five-year trends. Essentially no overall change in N02 concentra-
tions occurred in Los Angeles County over the five years. Orange County,
a rapidly growing part of the Los Angeles basin, experienced a substantial
increase in NOp concentrations. Other California locations and New Jersey
From 1965 to 1974, population grew at 4.3% per year in Orange County
and only 0.3% per year in Los Angeles County. Traffic levels increased by
7.5% per year in Orange County and only 2.8% per year in Los Angeles County [!]•
-------
91
Table 5.1 Five-Year Changes in Ambient N02 Concentrations
STATIONS
LOS ANGELES BASIN SITES
Orange County: Anaheim
(rapid growth) La Habra
Average for Orange County
Los Angeles County: Azusa
(slow growth) Lennox
Los Angeles
L.A. (Westwood)
L.A. (Reseda)
Average for Los Angeles County
OTHER CALIFORNIA SITES
Oakland
Pittsburg
Redwood City
Salinas
San Rafael
Santa Cruz
Stockton
Average for Other California Sites
NEW JERSEY SITES
Bayonne
Camden
Newark
Average for New Jersey Sites
OTHER SITES
Chicago, IL
Portland, OR
NET PERCENTAGE CHANGE
TRATIONS
Annual
Mean
+ 9%
+99%
+54%
+17%
- 7%
+ 3%
+ 8%
- 4%
+ 3%
- 7%
- 8%
-24%
- 1%
+ 5%
+15%
- 3%
- 3%
-27%
- 9%
- 5%
-14%
+32%
- 4%
FROM 1969 TO
90th
Percent! le
+ 5%
+60%
+33%
+ 7%
-11%
- 2%
+11%
-10%
- 1%
- 9%
- 4%
-25%
- 1%
0%
-24%
-44%
-15%
-18%
- 7%
0%
- 8%
+51%
+44%
IN N02 CONCE
1974
Yearly
Maximum
+13%
+72%
+43%
+ 6%
+ 1%
-31%
+32%
-13%
- 1%
-14%
-12%
- 9%
+27%
0%
-27%
-21%
- 8%
-36%
-52%
+17%
-24%
+94%
-16%
-------
92
locations witnessed a moderate improvement in N02 air quality. The Chicago
site evidently underwent a substantial worsening of N02 air quality.
Table 5.2 presents ten-year changes in ambient N02 concentrations.
Over the ten years, Los Angeles County, Stockton, and Chicago all show sig-
nificant increases in annual mean N02 concentrations. This increase is
presumably due to increases in NO emissions over the decade. NO emissions
X A
increased because of traffic growth and because the controls initially used
to reduce HC and CO in new cars had the side effect of raising NO emissions.
A
A recent study indicates that NOV emissions increased by about 29% in Los
X
Angeles County from 1964 to 1974 [1]. This is only slightly above the 22%
increase in annual mean N02 concentrations for the county.
A very interesting feature of Table 5.2 is the trend in 90th percentile
and yearly maximum concentrations in Los Angeles County and Stockton, Cali-
fornia. In contrast to the increase in annual mean N02 concentrations, the
90th percentiles showed little change over the decade, and the yearly maxima
showed a moderate decrease. We are not sure why this is the case. The most
plausible explanation involves HC controls. California, in particular
Los Angeles County, has achieved significant HC control over the decade.
The decrease in HC levels may have an amelioratory effect on ambient NOp
levels, especially maximum N02 concentrations. This hypothesis is supported
by a study which showed that daily maximum N02 concentrations increased less
than daily maximum NOX concentrations in the Los Angeles basin over the
past decade [1]. A second explanation involves changes in the spatial
-------
93
Table 5.2 Ten-Year Changes in Ambient N02 Concentrations
STATIONS
LOS ANGELES BASIN SITES
Los Angeles County: Azusa
Burbank
Lennox
Long Beach
Los Angeles
L*A. (Westwood)
L.A. (Reseda)
Pomona
Average for Los Angeles County
OTHER CALIFORNIA SITES
Stockton
NON-CALIFORNIA SITES
Chicago, IL
NET PERCENTAGE CHANGE
CENTRATIONS FROM 1964
Annual
Mean
+66%
+28%
- 9%
+18%
+18%
+27%
+18%
+ 9%
+22%
+45%
+40%
90th
Percent! le
+32%
+21%
-25%
+ 5%
+ 3%
+12%
+ 7%
0%
+ 7%
- 2%
+60%
IN N02 CON-
TO 1974
Yearly
Maximum
+39%
+ 5%
-35%
-23%
-20%
+31%
-21%
+ 7%
-17%
-17%
+46%
distribution of emissions. On both a local and regional scale, source
growth occurs in a spreading fashion. As the spatial distribution of
emissions becomes more spread out, maximal concentrations may be reduced
relative to mean concentrations. This second hypothesis fails, however, to
explain the historical decreases in maximal NQ2 relative to maximal NOX.
-------
94
5.2 YEAR-TO-YEAR TRENDS IN N02 AIR QUALITY
The previous section presented overall air quality trends for five- and
ten-year periods. A given overall trend can occur in a variety of ways, i.e.,
a variety of year-to-year patterns. The year-to-year pattern in the trend
is important in relating air quality changes to source growth and control
strategies. This section discusses the year-to-year changes in NC^ air
quality for several regions.
With the original data base that was subject to the 75% completeness
criterion, it is difficult to examine year-to-year trends at certain stations
because many years of data fail the completeness test. The data base for
trend analysis can be expanded considerably by including years with at least
50% complete data, and by interpolating for years with less than 50% complete
data. This expanded data base has been assembled for several areas of the
country. All the data which have been added have been subjected to the quality
control procedures of Chapter 2, and all reported maxima have been adjusted
for sample size according to the method described in Chapter 3.
5.2.1 Trends at CAMP Sites
Figure 5.1 presents year-to-year trends averaged over 4 CAMP sites
(Denver, Chicago, St. Louis, and Cincinnati) from 1964 to 1973.* Yearly
values and three-year moving averages are plotted for the annual mean, the
90th percentile, and the annual one-hour maximum. Since the SAROAD printout
did not include annual means for several of the years, the 50th and 70th
Of the 6 CAMP sites, Washington D.C. is not included because of a
change in station location, and Philadelphia is excluded because of the
lack of data for 1972 and 1973.
-------
95
40 -J
30 -
Q.
D.
c
o
JO
I 2°
O
C
o
o
CM
O
10 -
-C
Q.
Q.
C
O
C
O)
u
c
o
o
CM o
O '-
5 -
4
3
1 -
YEARLY ONE-
HOUR MAXIMUM
-i 1 1 1 1 1 1 1 1 r
64 65 66 67 68 69 70 71 72 73
90th
PERCENTILE
APPROXIMATE
ANNUAL
MEAN
Yearly Values
Three-Year
Moving Average
64 65 66 67 68 69 70 71 72
Figure 5.1 N02 Air Quality Trends at 4 CAMP Sites
(Denver, Chicago, St. Louis, and Cincinnati)
-------
96
percentiles were averaged each year to provide an estimate of the
annual mean.
The three-year moving averages are approximately constant at the CAMP
sites from 1965 to 1968 for all three air quality indices. An increase in
N02 concentrations, especially for the annual mean and 90th percentiles,
occurs from 1968 to 1972. Little of the increase in N02 concentrations
can be attributed to growth in VMT (vehicle miles travelled). The 4
CAMP urban areas are low-growth areas [2]. Slow growth would especially
prevail in the center-city parts of the areas where the monitors are lo-
cated. The increase is most likely due to the rise in NOV emissions for
A
1968-1972 model-year light-duty vehicles. Those model-years were subject
to HC and CO emission standards but no NO standard, and the technology
A
used to attain the HC and CO standards increased NO emissions. The
A
leveling off of the annual mean and 90th percentile from 1972 to 1973 might
be partly due to the federal emission standard for NO that came on line in
A
1973.
The net change in the three-year moving average at the 4 CAMP sites from
1965 to 1972 was +16% for the annual mean, +20% for the 90th percentile con-
centration, and +7% for the yearly one-hour maximum. The lower increase in
the yearly maxima compared with the annual means may be an anomaly caused by
random variance. However, it does follow the pattern noted previously among
California sites, where maximal concentrations increased much less than an-
nual mean concentrations. As we hypothesized for California, hydrocarbon
control may have yielded the side benefit of reduced maximal N02 concentrations.
Overall, the average of 50th and 70th percentiles provided a quite
good estimate of the annual mean.
-------
97
Significant decreases in hydrocarbon (and oxidant) concentrations at CAMP
sites have recently been documented [3]. Another explanation for the
lesser increase in maximal concentrations could be the spreading-out of
emissions (see discussion on pages 92 and 93):.
5.2.2 ..Trends at New Jersey Sites
Figure 5.2 summarizes trends averaged over 2 New Jersey sites
(Bayonne and Newark) from 1966 to 1974. In this case, there was a slight
improvement for all three air quality indices. Three-year averages de-
creased 12% for the annual mean, 9% for the 90th percentile, and 13% for
the yearly maximum from 1967 to 1973.
For the two New Jersey sites, the increase in automotive emissions
from 1968 to 1972 is not apparent in the air quality trends. We are not
sure why. Possibly, reductions in stationary area source NO emissions,
/\
caused by conversions to natural gas, may have compensated for the increase
in vehicular emission factors. It is also noteworthy that northern New
Jersey is a low-growth area; there may have actually been negative growth
in the environs of the monitor. A striking feature of the trends is the
decrease in N02 concentrations from 1973 to 1974. This may be largely
due to the energy crisis and the associated reduction in VMT that occurred
in 1974.
As with California sites and CAMP sites, yearly maximal N02 levels
decreased by more than annual mean levels in New Jersey. However, the
difference in the trends (-13% vs. -12%) is certainly not statistically
significant.
5.2.3 Trends in Coastal/Central Los Angeles County
Los Angeles County provides high-quality aerometric and emission data
that are very suitable for trend analyses. The coastal/central areas of
-------
40
i: 30
o
o
CJ
o
20 -
10 _
10
9
0 -
"S. 7
Q.
c;
o
c
o
o
o
1 _
98
YEARLY ONE-
HOUR MAXIMUM
I I I i 1 I I I I I i
64 65 66 67 68 69 70 71 72 73 74
v'
Yearly Values
ANiIUAL
MEAN
Three-Year
Movinq Average
64 65 66 67 68 69 70 71 72 73
Figure 5.2 NCL Air Quality Trends at 2 New Jersey Sites
(Bayonne and Newark)
-------
99
Los Angeles County are particularly interesting because of the reductions
in hydrocarbons that have been achieved in those areas. Figure 5.3 sum-
marizes N02 air quality trends at 6 coastal/central stations in Los Angeles
County from 1964 to 1974.
For all three air quality indices, three-year moving averages of NOp
concentrations increased slightly from 1965 to 1970 and decreased from 1970
to 1973. This reflects changes in automotive-emission factors. NO emis-
/\
sions increased substantially in 1966 to 1970 model-year vehicles due to
the "leaning out" of engines for HC and CO control. California established
emission standards for NO starting in 1971. Growth in traffic has not
J\
had great effect on trends in this part of the Los Angeles basin. VMT grew
at 2.8% per year in Los Angeles County from 1965 to 1974 [1]. However,
most of the VMT growth occurred in portions of the county away from the
older, well-established, central business districts where the 6 monitors
in question are located.
The net changes in three-year moving averages of N02 concentrations
from 1965 to 1973 are +13% for the annual mean, +2% for the 90th percentile,
and -8% for the yearly one-hour maximum. The increase in the annual mean
N02 concentrations is slightly less than the increase in mean NOX concen-
trations at these 6 stations (approximately +20%) [4]. Increases in 90th
percentile N02 concentrations and yearly maximum N02 concentrations are even
less than the increase in annual mean N02 levels. The varied trends may be
due to hydrocarbon control. It is possible that hydrocarbon control has
*These stations are Burbank, Lennox, Long Beach, Downtown Los Angeles,
Westwood,and Reseda.
-------
100
60 -i
50
40 -J
c
u
OJ
o
10 -
Yearly Values
64
I
65
66
I
67
90th
PERCENTILE
ANNUAL
MEAN
Three-Year
Moving Average
68 69
I
70
I
71
72 73
I
74
Figure 5.3 N02 Air Quality Trends at 6 Sites in-Coastal/Central
Los Angeles County
-------
101
slightly reduced annual mean N02 levels relative to annual mean NOX levels.
Even more plausible is the contention that HC reductions have yielded sig-
nificant benefits with respect to maximal NOo concentrations. Part II of
this report, which involves empirical models of the N02/precursor relation-
ship, should shed more light on these issues. Models are developed for both
annual mean and yearly maximum N02 concentrations.
-------
102
5.3 REFERENCES
1. J.C. Trijonis, T.K. Peng, 6.J. McRae, and L. Lees, "Emissions and
Air Quality Trends in the South Coast Air Basin," EQL Memorandum
No. 16, Caltech Environmental Quality Laboratory, Pasadena, California,
January 1976.
2. U.S. Bureau of the Census, Statistical Abstract of the United States;
1975, Washington, D.C., 1975.
3. A.P. Altshuller, "Evaluation of Oxidant Results at CAMP Sites in the
United States," Journal of the Air Pollution Control Association,
Vol. 25, p. 19, 1975.
4. California Air Resources Board, Ten-Year Summary of California Air
Quality Data: 1963-1972, Sacramento, January 1974.
-------
103
6.0 RELATIONSHIP OF YEARLY QNF-HQUR MAVTMfl flMn ANN1IAI MFANS
If the existing long-term air quality standard for N02 (5 pphm-annual
mean) is supplemented with a one-hour standard, it will be important to
know when and where each of the standards is the binding constraint for
control strategy formulation. Under certain conditions, one of the standards
may be obviously binding; under other conditions, both standards may have to
be considered. Since a specific one-hour standard has not yet been chosen,
we cannot completely solve the problem of determining the binding constraint.
Rather, this chapter will provide the information that is required to ad-
dress the problem once a standard has been set. The required information
is based on an analysis of spatial and temporal patterns in the ratio of
one-hour maximum to annual mean.
A similar problem (determining the binding constraint) will arise if
a short-term air quality standard is set for 90th percentile concentrations.
The information needed to solve that problem can be assembled in a manner
entirely parallel to the present analysis. The key variable would then
be the ratio of the 90th percentile concentration to the annual mean
concentration.
6.1 NATIONWIDE PATTERNS IN THE MAXIMUM/MEAN RATIO
The data base described in Chapter 4 and Appendix C provides information
on present ratios of maximum-to-mean N02 concentrations. That data base
includes yearly one-hour maxima and annual means, averaged from 1972 to
1974, for 120 urban stations and 3 rural/power plant stations. The
-------
104
distribution of maximum/mean ratios for the 120 urban stations is shown
in Figure 6.1.
t/o
O)
CD
ra
-------
105
There does not appear to be any general geographical pattern in Table 6.1.
Seven of the seventeen sites are in California, but this is not unusual
because half of the 120 urban monitoring sites are in California.
Table 6.1 Locations with Maximum/Mean Ratios Exceeding 8, 1972-1974
Station Maximum/Mean Ratio Station Maximum/Mean Ratio
Ojai, CA
Phoenix, AZ
Bar stow, CA
St. Louis (006), MO
Chi no, CA
Ashland, KY
Costa Mesa, CA
San Diego, CA
Denver, CO
13.4
12.2
12.0
11.5
11.1
11.0
10.2
9.6
9.2
Glen Falls, NY
Silver Spring, MD
Reno, NV
Las Vegas, NV
Baltimore, MD
Los Angeles,
(Westwood), CA
San Jose, CA
Redwood City, CA
9.0
8.8
8.5
8.4
8.1
8.1
8.0
8.0
Figure 6.1 provides some clues as to whether a one-hour or annual mean
standard would be the binding constraint. If a federal one-hour standard
were established at the level of the California one-hour standard (25 pphm),
and if the maximum and mean responded equivalently to emission control,
-------
106
then a maximum/mean ratio of 5 would be the dividing point for a binding
one-hour standard vs. a binding annual mean standard. Figure 6.1 indicates
that the one-hour standard would be binding for 38% of the urban locations.
If a federal one-hour standard were set at 50 pphm, and if the maximum
and mean responded equivalently to emission changes, then the critical
maximum/mean ratio would be 10. For this case, the present annual mean
standard would be the binding constraint for 94% of the locations. Before
too much is read into this simplistic analysis, we should note that the
assumption of the maximum and mean responding equivalently to emission
control seems to be a poor one. As noted earlier, the maximum-to-
mean N02 ratio evidently changes with time. In Part II we will find evidence
that this occurs because HC control reduces maximal N02 levels preferen-
tially over mean f^ levels. If emission control can significantly alter
the maximum/mean ratio, then it may be best to consider both the annual
mean and one-hour standard for every location in formulating strategies
for attainment and maintenance of the NAAQS.
Figure 6.2 illustrates the national geographic pattern of maximum/
mean N02 ratios. To avoid cluttering the map, not all of the 120 urban
stations are included. Where there are several stations in close proximity,
the stations with the highest and lowest ratios are recorded on the map to
illustrate the range in the ratio. Figure 6.2 reveals no broad nationwide
patterns in the maximum/mean N0£ ratio. Both the western and eastern
sites show about the same maximum/mean ratio, typically ranging from 5
-------
Figure 6.2 Nationwide Geographic Distribution of Maximum/Mean N0
Urban Sites, 1972-1974
Ratio at
-------
108
to 12; also, no discernable gradient in the ratio is apparent from north to
south.
It is interesting to determine if there is a relationship between the
maximum/mean ratio and the overall level of NO,, concentrations. Do sites
with higher N02 concentrations tend to have higher or lower maximum/mean
ratios? Figure 6.3 shows the average maximum/mean ratio for sites with
various levels of annual mean N02. For the 120 urban stations, there appears
to be essentially no dependence of the maximum/mean ratio on the annual mean.
Sites with annual mean concentrations from 1 pphm to 4 pphm have an average
maximum/mean ratio of 6.5, while sites with annual mean concentrations from 4
pphm to 8 pphm have an average maximum/mean ratio of 6.4. Figure 6.3 also
demonstrates the anomaly of the 3 rural/power plant sites. These 3 sites
have annual means of about 0.8 pphm, and the average maximum/mean ratio
among these sites is nearly 20. As noted above, the high maximum/mean ratio
is expected for these sites because they are subject to infrequent, but
rather intense, fumigations by power-plant plumes.
6.2 INTRAREGIONAL PATTERNS IN THE MAXIMUM/MEAN RATIO
In this study, two areas have been selected for the purpose of investi-
gating intraregional patterns in N02 concentrations. These are the Los
Angeles air basin and the New York-New Jersey-New England area. Figures
6.4 and 6.5 illustrate the spatial patterns of the maximum/mean N02 ratio
within these regions.
No consistent spatial gradients appear in Figures 6.4 and 6.5. The most
populated portion of the Los Angeles basin, the central/coastal area,
shows about the same average ratio (approximately 7) as the downwind
-------
109
20-
15-
10 _
5 _
3 Rural/Power Plant Stations
T
2
Urban Stations
~T
3
T
4
T
5
Annual Mean
6 7
Concentration, pphm
8
Figure 6.3 Dependence of Maximum/Mean Ratio on Annual Mean
N00 Concentrations
-------
Figure 6.4 Maximum/Mean N02 Ratio at Monitoring Sites in the Los Angeles Region, 1972-1974
-------
Ill
Figure 6.5
Maximum/Mean N02 Ratio at Monitoring Sites
in the New York-Hew Jersey-New England Area,
1972-1974
-------
112
eastern/inland areas and the isolated northwestern counties (Santa Barbara
and Ventura). New York City has about the same ratio (approximately 6) as
Philadelphia, northern New Jersey, New Britain, CT, Springfield MA, and
Providence, RI.
6.3 HISTORICAL TRENDS IN THE MAXIMUM/MEAN RATIO
It is important to investigate historical trends in the maximum/mean
N0£ ratio. If it can be shown that the maximum/mean ratio is essentially
constant over time at all locations, then it may be safe to take a simplistic
approach in determining binding air quality standards. For instance, the
California one-hour N02 standard (25 pphm) could be considered binding
over the federal annual standard (5 pphm) for all locations with a maximum/
mean ratio greater than 5. If, on the other hand, the maximum/mean ratio
shows significant trends, then the binding standard may change with time.
In this case, both standards should always be considered in formulating
and evaluating control strategies.
Figures 6.6 and 6.7 illustrate recent historical trends in the
maximum/mean N02 ratio averaged among 4 CAMP sites and 2 New Jersey sites,
respectively. The data base used to compute these trends is described
in Section 5.2. There is a slight decline in the maximum/mean ratio for
the CAMP sites. Essentially no overall change occurs at the New Jersey
sites from 1966 to 1974.
Figure 6.8 illustrates the trend in maximum/mean ratio at 6 sites
in the central/coastal part of Los Angeles County. A persistent decline
in the ratio is evident; the three-year moving average decreases by 19%
from 1965 to 1973. As previously discussed, a possible explanation for
-------
113
7
£ 6
cv, 5
o
'x
3
2^
1 -
Three-Year
Moving Average
Yearly Values
64 65 66 67 68 69 70 71 72 73 74
Figure 6.6- Trends in the Maximum/Mean N02 Ratio Averaged over 4 CAMP
Sites (Denver, Chicago, St» Louis, and Cincinnati)
03
CC.
CM
O
5 -
§ 4^
Three-Year Moving Average
3 -
2 -
1 ~
i
Yearly V,
64 65 66 67 63 69 70 7"l 72 ^3 74
Figure 6.7 Trends in the Maximum/Mean N02 Ratio Averaged over 2
New Jersey Sites (Bayonne and Newark)
-------
114
o 1 ~\
•r-
-------
115
ro
CC
CVJ
o
c
(O
-------
116
Figure 6.10 illustrates trends in the maximum/mean ratio at 5 locations
in central California. The ratio shows a substantial rate of decline; the
three-year moving average decreases by 17% from 1968 to 1973. Again, this
may be related to HC control, although HC trends are not well documented
for these locations.
8
7-
6-
CM _ ,
S 5 ~
c
10
4 ~
•I 3
X
2
1 H
64
65
Three-Year Moving Average
66
67
68 69
70
71
72 73
74
Figure 6.10 Trends in Maximum/Mean NOo Ratio at 5 Locations
in Central California (Redwood City, Salinas, San
Rafael, Santa Cruz, Stockton)
-------
117
PART II:
EMPIRICAL MODELS OF THE
N00/PRECURSOR RELATIONSHIP
-------
119
7.0 EMPIRICAL ANALYSIS OF THE N02/PRECURSOR DEPENDENCE
Determining the impact of control strategies or new emission sources
on air quality requires a method of translating emission changes into air
quality changes. The conventional method for nitrogen dioxide is to model
total NOX as an inert primary pollutant and to assume that N02 concentra-
tions are directly proportional to ambient NO concentrations, with the
/\
proportionality constant equated to the existing atmospheric ratio of
N02 to NOX. This approach has some merit, because it is generally agreed
that ambient N02 levels should be approximately proportional to ambient
NO levels, with all other factors held constant. However, other factors
A
are not always invariant. In particular, hydrocarbon emission reductions
may affect ambient N02 concentrations. If we are to predict changes in
N02 air quality with more confidence, we must know the dependencies of
N02 concentrations on both photochemical precursors, hydrocarbons as well
as NO.
/\
Experimental studies with smog chambers have provided most of our
present understanding of the N02/precursor dependence. The various chamber
studies agree on some aspects of the N02/precursor dependence, but they
disagree on other aspects. Because of these disagreements and because of
uncertainty in extrapolating experimental studies to the real atmosphere,
there is a need for empirical models that extract information about the
N02/precursor dependence from ambient data. The purpose of Part II of
this report is to develop and apply such empirical models.
This chapter serves as an introduction to Part II. Section 7.1
reviews the results of various experimental studies and summarizes existing
-------
120
knowledge of the M^/precursor dependence. Section 7.2 presents the
conceptual framework for empirical models. The remaining chapters
develop empirical models for various cities and check these models against
historical trends and smog-chamber results. Models for both annual mean
NOp and yearly one-hour maximum NOp are included.
7.1 EXPERIMENTAL EVIDENCE OF THE N02/PRECURSOR DEPENDENCE
Several researchers have used experimental test chambers (smog
chambers) to investigate the dependence of nitrogen dioxide concentrations
on the levels of precursor inputs. These experimental studies have pro-
vided most of the present understanding of the N02/precursor dependence.
Before we formulate and apply empirical methods for determining the N02/
precursor dependence, it is useful to review the results of the smog-
chamber experiments. Because both the empirical approach and the smog-
chamber approach involve significant uncertainties, it will be important,
in the end, to compare the results of both approaches.
Our review of experimental studies will consider results from five
smog chamber projects:
• The University of North Carolina (UNC) stddy using an 11,000-cubic-foot
outdoor Teflon chamber, a simulated urban hydrocarbon mix, and
twelve-hour irradiations[l];
t The Bureau of Mines study, using a 100-cubic-foot aluminum-glass
chamber, auto-exhaust hydrocarbons, and six-hour irradiations[2,3];
t The General Motors study, using a 300-cubic-foot stainless steel-
glass chamber, a simulated Los Angeles hydrocarbon mix, and six-
hour irradiations[4];
-------
121
• The HEW study using a 335-cubic-foot chamber, auto-exhaust hydro-
carbons, and up to ten-hour irradiation time[5]; and
• The HEW study using a 335-cubic-foot chamber, toluene and m-xylene,
and six-hour irradiations[6].
7.1.1 Average N02 Concentrations
The various smog-chamber studies apparently yield consistent results
concerning the dependence of average N02 yield (or N02 dosage) on NOX
input. With other factors held constant, average N02 concentrations tend
i
to be directly proportional to initial NO . The proportional relationship
.Ai
for average N02 is illustrated in Figures 7.1 through 7.5.
a.
3
O
3C
O
c
O
CJ
0)
O)
<0
OJ
.30
.25
.20
.15
.10
.05
.00
.00
.10
I ' I
I ' I
I T
>,
^Small Sample Sizes
1
_L
_L
1
_L
1
.20
.30 -40 .50 .60
Initial Nitrogen Oxides, ppm
.70
Fiaure 7 1 Nitrogen Dioxide Ten^Hour Average Concentration
'vs. Initial Oxides of Nitrogen for Urban Hydro-
carbon Mix (Means of Several Experiments),
University of North Carolina Study[l]
.80
-------
122
I.
Q.
o
a
ss
240
200
160
120
80
40
(HO,: 5.0 ppmC
O.I - 2.3 ppmC
O.Z
0.4
0.6
0.8
ppm
1.0
Figure 7.2 Nitrogen Dioxide Dosage as a Function of NO
at Various HC Levels, Bureau of Mines Stud/12]
15
1.2
0.9
O)
cn
CM
Q
06
0.3
0 •—- Varied HC Levels
o.:
0.6
0.2 0.3 0.4 0.5
Initial NOX Cone., ppm
Figure 7.3 Nitrogen Dioxide Dosages in the Irradiation
of Multicomponent Hydrocarbon/N0x Mixtures,
General Motors Study [4]
-------
123
s-
3
o
1.0
'1.5 ppm To!uene
o
c
o
o
(O
O!
.5 H
4 ppm Toluene
^/_ ^ — -^". ^~' 3 ppm Toluene
i.o
1.5
0"xfdes of Nftrdgen Concentration, pprfi
Figure 7.4 Average N02 Concentration (Over Six Hours)
vs. Initial NOY at Three HC Levels,
HEW x
Q.
Q.
in
•3
O
o:
i
o
o
o
C\'.
o
o
o>
£
1.0 -n
.5 J
12 ppm (Auto
Exhaust)
6 ppm (Auto
Exhaust)
3 ppm (Auto
Exhaust)
I I I I
.5 1.0 1.5 2.0
UXfcfes of Nitrogen uonceritratibn, ppm
Figure 7.5 Average N02 Concentration (During First
Ten Hours) vs. Initial NOX at Three HC
LevelSs HEW Study[5]
-------
124
The dependence of average N02 concentrations on hydrocarbons is less
understood. Stephens has hypothesized that reductions in hydrocarbon
concentrations should tend to increase average N02 concentrations because
the hydrocarbon reductions would delay and suppress the reactions that
consume N02 after it reaches a peak [7J. Figure 7.6 presents a schematic
illustration of "this hypothesis. The Bureau of Mines chamber results are
consistent with Stephens' hypothesis (see Figure 7.2) [2]. However, three
other chamber studies indicate that hydrocarbons produce no consistent
effect on average NO, concentrations [4,5,6].
In direct contradiction to Stephens' hypothesis, the UNC outdoor
chamber experiments found that a 50% reduction in hydrocarbons produced about
a 20% decrease in average N02 [1]. However, in defense of the hypothesis,
it should be noted that the UNC chamber runs were of ten-hour duration
and that the NO,, levels at the end of the experiments were greater when
hydrocarbons were reduced. The extra N02 remaining after the ten-hour
period could cause an increase in 24-hour average N02, even though average
N02 was reduced during the first ten hours.
7.1.2 Maximal N02 Concentrations
As was the case with average N02, the various chamber experiments
yield consistent results with respect to the dependence of one-hour maximal
N02 on NOX input. With other factors held constant, maximal N02 concentra-
tions tend to be directly proportional to NOV input[l,3,4]. This proper-
A
tional effect is illustrated in Figures 7.7 and 7.8.
There is less agreement with respect to the dependence of maximal
N02 concentrations on hydrocarbon input. The Bureau of Mines study
-------
125
C
0
N
C
E
N,
R
A
T
I
0.
N
PRESENT
03 (OZONE) 0
N
0
F
HOURS
HOURS .
CONTROL OF HYDROCAR'BON
|
I
NO
HOURS
HOURS
D
E
Figure 7.6. Stephens' Hypothesis of Effect of
HC and NOX Control[7]
-------
.00
.00 .10 .20 .30 .40 .50 .60
Initial Nitrogen Oxides, ppm
.70 .80
Figure 7.7 Nitrogen Dioxide Maximum Concentration vs. Initial Oxides
of Nitrogen (Means of Several Experiments) UNC Study [1]
Initial NOX, ppm
Figure 7.8 Dependence of Nitrogen Dioxide Maximum Concentration
on Initial Nitrogen Oxides, Bureau of Mines Study [3]
-------
127
found that maximal N02 concentrations are essentially independent of initial
hydrocarbon input [3]. However, two other studies imply that hydrocarbon
reductions decrease maximal N02 concentrations. The UNC outdoor chamber
results indicate that 50% hydrocarbon control tends to decrease maximal N02
concentrations by about 10% to 20% [1]. The General Motors chamber
studies indicate that 50% hydrocarbon control reduces maximal N02 by about
25% [4]. These latter two studies also show that maximal N09 is rela-
tively more sensitive to hydrocarbon reductions at higher NOV levels.
X
7.1.3 Summary of Chamber Results
All of the chamber experiments agree concerning the proportional
dependence of N02 (average or peak concentrations) on NO . These studies
also concur that hydrocarbon control will reduce maximal N02 concentra-
tions relative to average N02 concentrations. The disagreement concerns
exactly how this relative change in maximal and mean N02 will occur. The
Bureau of Mines study (and Stephens' hypothesis) indicate that hydrocarbon
control would leave maximal N02 unchanged but would increase average N02.
The UNC and General Motors studies indicate that hydrocarbon control would
reduce maximal N02 but would yield no change (or a slight benefit) in average
N02.
Considering the results of all the chamber studies, it is possible to arrive
at an overall best estimate of the effect of hydrocarbon control on N02
concentrations. The consensus based on existing chamber results would
appear to be as follows: Fifty-percent hydrocarbon control would have little
effect on average N02 concentrations, a change of + 10%, but would yield
moderate benefits in terms of maximal N02, a reduction of about 10% to Z0%.
-------
128
7.2 FORMULATION OF EMPIRICAL MODELS
Empirical models, based on statistical analysis of ambient data,
should be able to further our present understanding of the N02/precursor
dependence. Where the various chamber studies appear to reach a consensus,
empirical models can verify that the conclusions are representative of the
real atmosphere. Where the individual chamber studies disagree, empirical
models may help to resolve the discrepancies.
Developing empirical models for annual average N02 and yearly one-
hour maximum NOo is a complex procedure. Some of the complications become
apparent if the typical diurnal pattern of nitrogen dioxide, shown in Figure 7.9,
is considered. Figure 7.9 demonstrates that ambient N02 concentrations tend
to peak twice during the day--once in the late morning and once in the evening.
The exact times and relative strengths of these peaks vary from day to day
and depend on the season and geographic location. The yearly maximum one-
hour concentration in the morning is often about the same as the yearly one-
hour maximum in the evening. Thus, in general, an empirical model relating
precursors to yearly one-hour maximum N02 should consider both the morning
and evening peaks. Figure 7.9 also demonstrates that the minimal N02 con-
centrations, which occur in the early morning and late afternoon, are not
negligible compared with the maximal concentrations. This phenomenon warrants
the conclusion that an empirical model for annual average N02 must include all
hours of the day, not just the times of peak concentrations.
-------
129
o
•I—
Also, leftover N02 from the nighttime
period may significantly affect the N02 levels of the subsequent "daytime"
period.
-------
130
7.2.1 Alternative Model Formulations
Figure 7.10 presents a conceptual diagram for an empirical model of
daytime peak one-hour NOg. It 1s assumed that daytime peak N02 concentra-
tions depend on only two types of factors: (D 6:00-9:00 A.M. concentra-
tions of precursors (NMHC and NOJ; and (2) meteorological factors that govern
A
the concentration of N02 produced from the precursor concentrations. The
empirical models will be based on relationships between day-to-day changes in
6:00-9:00 A.M. precursor concentrations and corresponding changes 1n day-
time peak one-hour N0« concentrations. Day-to-day changes 1n precursor
concentrations are produced by several processes, Including variance 1n
overnight and early-morning dispersive conditions, weekday/weekend emission
changes, variance in overnight air mass trajectories (and associated
stationary source areas), and changes in vehicular emission factors induced
by variance 1n temperature and humidity. The first process, dispersion, is
the dominant factor changing precursor concentrations from day to day. The
last two processes are notable because they affect the NMHC/NO ratio as
X
well as overall NMHC and NO concentrations. The empirical approach followed
A
here implicitly assumes that daily changes in precursor concentrations,
produced mostly by overnight and early-morning meteorological variance, can
be used to model the effect of changes in precursor concentrations that
would result from control strategies.
The most simplistic statistical analysis that could be performed on
the system in Figure 7.10 would be to determine the function
Daytime Peak One-Hour N02 = F^NMHC, NOX) » (1)
where NMHC = 6:00-9:00 A.M. NMHC concentration,
and NO = 6:00r9:00 A.M. NO concentration.
^ n
-------
I. Morning Precursor
Concentrations
6:00-9:00 A.M. NOX
6:00-9:00 A.M. NMHC
I t
Previous-Evening,
Overnight, and
Early-Morning
Factors
Meteorology
Emissions
II. Meteorological Factors
Governing N02 Concentrations
Produced from Morning Pre-
cursor Concentrations.
(Post 9:00 A.M.)
Solar Radiation
Mixing Height
Temperature
Wind Speed
Etc.
CO
Late-Morning
Peak One-Hour N02
Figure 7.10 Conceptual Diagram of Empirical Model for Daytime Peak One-Hour N02
-------
132
The function, F,, would form the basis for an empirical model for daytime
peak N02 by indicating the percentage change 1n peak one-hour N02 that would
be attained from various percentage changes 1n NMHC and NOV concentrations.
A
One of the major drawbacks of the simplistic approach is that the relation-
ship between peak one-hour N02 and precursors (i.e., Equation (1)) might be
spurious in the sense that it is due to mutual correlations with unaccounted
for weather factors. For instance, NMHC concentrations might be positively
correlated with solar radiation, which 1n turn has a positive relationship
to peak N02 concentrations. These effects can be partially discounted for
by a more complex analysis that explicitly Includes the weather factors.
In this case, the statistical analysis would determine the equation
Daytime Peak One-Hour N02 = F2(NMHC,NOX,W15...,WN) , (2)
where Wp...,W^ are the daily values of N weather parameters that govern
N02 concentrations produced from the precursor concentrations. Equation (2)
would form the basis for an empirical model by Indicating the net effect
of precursor changes on N02 under various types of meteorological conditions.
The analysis for daytime peak N02 can also be made more complex by
including 5:00 A.M. N02 concentration as an independent variable in
Equation (1) or Equation (2). This would allow the carry-over effect
of previous-day N02 to be accounted for. In this case, the basic
empirical equation would be
Daytime Peak One-Hour N02 = F3(NMHC,NOX,N02), (3)
or if weather variables are included,
Daytime Peak One-Hour N02 = F4(NMHC,NOX>W1,...,WN,N02) (4)
*
where N02 = 5:00 A.M. N02 concentration.
-------
133
Empirical models based on Equation (3) or (4) would require coupling
with a model for overnight N02. That is, the dependence of 5:00 A.M. N02
on previous-day precursors (NMHC and NOX) would have to be included before
Equation (3) or (4) could be used to represent the full dependence of
daytime peak N02 on primary pollutants.
i
Empirical models for daytime average N02 are obtained simply by
taking average N02 rather than one-hour peak N02 as the dependent
variable in Equations (1) through (4). Similar empirical models can
be formulated for nighttime peak one-hour N02 and nighttime average
N02. For the nighttime case, the averaging times for precursor con-
centrations and weather variables would, of course, be different from
the averaging times for the daytime case. Also, late-afternoon oxidant
might be included as a "precursor" variable for nighttime N02. An assumed
relationship of oxidant versus NMHC and N0¥ would then be required to trans-
/\
late the dependence of N02 on oxidant into a dependence of N02 on primary
precursors.
7.2.2 Study Areas
The empirical modeling analysis will be conducted for 8 locations.
Two of these, Denver and Chicago, are center-city CAMP sites, operated by
the United States Environmental Protection Agency. Two other sites are in
Houston, Texas: the Mae Drive site (near the main source area of Houston)
and the Aldine site (about ten miles downwind of the main source area). The
other 4 sites are in Los Anqeles County and are operated by the
Southern California Air Quality Management District. The Los Angeles County
sites were selected so that 1 (Downtown Los Angeles) is in the center
of the county, 1 (Lennox) is in the coastal upwind portion of the county,
-------
134
and 2 (Azusa and Pomona) are in the inland downwind portion of the county.
These 4 sites and the typical wind patterns in the Los Angeles basin are
shown in Figure 7.11
The complexity of the empirical models selected for each location
(e.g., whether meteorological parameters are included), will depend on
data availability. For the present study, comprehensive data are
available for pollutant and weather variables at 1 location (Downtown
Los Angeles). Several empirical models with varied degrees of complexity
will be applied to that location. The results of the alternative models
will be compared, and an assessment will be made of the adequacy of very
simple models (e.g., Equation (1)). Applications to locations other than
Downtown Los Angeles will be restricted to simple models because meteoro-
logical data are not readily available for the other locations.
7.2.3 Combination of Submodels
In this study, the empirical modeling analysis will be disaggregated
by season. As discussed in Chapter 9, diurnal patterns for nitrogen di-
oxide show marked seasonal changes, especially from summer months to winter
months. It is interesting to determine if the N02/precursor relationship
also undergoes substantial seasonal changes. Disaggregating the analysis
by seasons also tends to keep weather factors more uniform in each analysis.
This disaggregation should reduce the problem of spurious relationships due
to hidden correlations between precursor concentrations and weather factors
that govern NC^ production from the precursors.
To construct complete empirical models for annual average N02 and
yearly peak one-hour N02 requires a synthesis of the individual models for
-------
SANTA BARBARA Ci&NTY
^^^^^^^
\
VENTURA
COUNTY
\LOS ANGELES, COUNTY^
/ -' • SAN BERNARDINO COUNTY
/ "Azusa/ >
CO
RIVERSIDE
COUNTY
Figure 7.11 Map of the Metropolitan Los Angeles AQCR
-------
136
daytime and nighttime for each season. Using the definition of "daytime"
and "nighttime" given above, the daily average for each season is given by
Daily Average NOg = -ir ' Daytime Average + -i4 • Nighttime Average- (5)
Yearly average N02 is just a linear combination of the daily averages for
the individual seasons. The empirical model for yearly one-hour maximum NOg
will be the peak one-hour model for the particular season and time of day
when the yearly one-hour maximum occurs. If the yearly maximum can occur
in more than one season or more than one time of day, then two or more
submodels for peak one-hour M^ will have to be considered.
7.2.4 Limitations of Approach
The specific empirical models proposed here for determining the NC^/
precursor dependence suffer from several limitations. It is implicitly
assumed that daily changes in precursor concentrations, produced mostly by
variance in overnight and early-morning meteorology, can be used to model
the effect of control strategies. The validity of this assumption has not
been resolved.
As noted previously, the simple models that omit meteorology may result
in correlations which are not representative of causality. The more complex
models require a detailed meteorological data base. Data requirements may
present a problem even with the simple models, because measurements are needed
each day for N02» NOX> NMHC, and oxidant. Because of missing values for
one or more pollutants, two to three years of ambient data are often necessary
to provide an appropriate sample size for the statistical analyses. The data
requirements are worsened by the need to sample over a wide range of NMHC/NOx
ratios.
-------
137
Another limitation of the empirical models used here is the neglect
of precursor emissions that occur after the time when ambient precursors
are measured. For example, in the daytime N02 model, the 6:00-9:00 A.M.
precursor concentrations represent emissions only up to 9:00 A.M. This limita-
tion may not be extremely serious because accumulated overnight and early-
morning emissions (up to 9:00"A.M.) are substantially larger than the total
amount of late-morning emissions. Also, the day-to-day variations in 6:00-
9:00 A.M. concentrations may be somewhat representative of day-to-day variations
in precursor concentrations resulting from late-morning emissions.
Perhaps the most serious drawback of our approach is the neglect of
transport. Nitrogen dioxide concentrations will be related to precursor
concentrations measured at the same location but at an earlier time. If
significant transport occurs, the nitrogen dioxide measurements and pre-
cursor measurements will be associated with totally different air masses.
This could destroy the possibility of obtaining the desired relationships.
For the case of peak N02 concentrations in the Los Angeles basin, there is
reason for encouragement because the times between the precursor measure-
ments and N02 peaks (approximately 9:00 A.M. to 10:00 A.M. and 7:00 P.M. to
10:00 P.M.) tend to be periods of stagnation[8,9]. For other cities and for
average N02 concentrations, transport may be a very important problem. This
problem should be kept in mind when reviewing the results of the models ap-
plied in this study.
It is possible to formulate more complex empirical models that can
take into account emissions from all hours[10,11] and that do include pollu-
tion transport[ll,12]. However, formulating and applying these complex
models requires much greater effort and is beyond the resources of the
present investigation.
-------
138
The limitations of the empirical approach taken here can have consider-
able impact on the relationships observed between ambient N02 and ambient
precursor concentrations. For example, assume that N02 concentrations are,
in actuality, directly proportional to NOV input. If transport were a very
A
significant factor, regressions of daytime N02 concentrations versus early-
morning NO concentrations may show little or no dependence because the two
/\
measurements are associated with different air masses. In this case, the
statistical relationship between N0? and NO would entirely misrepresent the
C. X
causal dependence.
Because of the limitations in our approach, it may not be possible to
arrive at purely statistical formulas that precisely represent the depen-
dence of N02 on its precursors. At the minimum, however, the empirical
models should indicate the important qualitative aspects of the N02/precur-
sor relationship (such as whether a hydrocarbon dependency exists). These
conclusions can be checked against historical trends in precursors and N02-
In the end, control strategy analysis might best be performed by combining
the results of the empirical models with the findings of smog-chamber tests.
-------
139
7.3 REFERENCES
1. H. Jeffries, D. Fox, and R. Kamens, "Outdoor Smog Chamber Studies:
Effect of Hydrocarbon Reduction on Nitrogen Dioxide," prepared for
EPA Office of Research and Development by University of North
Carolina, EPA-650/3-75-011, June 1975.
2. B. Dimitriades, "On the Function of Hydrocarbons and Nitrogen Oxides
in Photochemical Smog Formation," Bureau of Mines Report of Investi-
gations #7433, September 1970.
3. B. Dimitriades, "Oxidant Control Strategies. Part I. An Urban Con-
trol Strategy Derived from Smog Chamber Data," paper submitted for
publication in Environmental Science and Technology.
4. J. M. Heuss, "Smog Chamber Simulation of the Los Angeles Atmosphere,"
General Motors Research Publication GMR-1802, Warren, Michigan,
February 1975.
5. M. W. Korth, A. H. Rose, and R. C. Stahman, "Effects of Hydrocarbon
to Oxides of Nitrogen Ratio on Irradiated Auto Exhaust," Journal of
the Air Pollution Control Association. Vol. 14, May 1964.
6. A. P. Altshuller, e_t al., "Photochemical Reactivities of Aromatic
Hydrocarbon-Nitrogen T5x~ide and Related Systems," Environmental
Science and Technology, Vol. 4, January 1970.
7. E. R. Stephens, "Proceedings of the Conference on Health Effects of
Air Pollution," U.S. Senate Committee on Public Works, U.S. Government
Printing Office Stock No. 5270-02105, 1973.
8. M. Neiburger and J. Edinger, "Meteorology of the Los Angeles Basin,"
Report No. 1 of the Air Pollution Foundation, Southern California
Air Pollution Foundation, 1954.
9. M. Neiburger, N. Renzetti, and R. Tice, "Wind Trajectory Studies of
the Movement of Polluted Air in the Los Angeles Basin," Report No. 13
of the Air Pollution Foundation, Southern California Air Pollution
Foundation, 1956.
10. G. Tiao, M. Phadke, and G. Box, "Some Empirical Models for the
Los Angeles Photochemical Smog Data," Journal of the Air Pollution
Control Association, Vol. 26, p. 485, 1976.
11. L. Breiman and W. Meisel, "The Change in Ozone Levels Caused by
Precursor Pollutants: An Empirical Analysis," Proceedings of the
Conference on Environmental Modeling and Simulation. EPA 60D/9/76-
016, April 1976.
12. J. Trijonis, "Economic Air Pollution Control Model for Los Angeles
County in 1975," Environmental Science and Technology. Vol. 8, p. 811,
1974.
-------
140
8.0 PREPARATION OF DATA BASE FOR EMPIRICAL MODELING
The objective of Part II of this project is to determine how peak
and average NOo concentrations depend on precursor concentrations. The
data base consists of hourly readings of NO, N02, NOX, HC, CH4, NMHC,
and oxidant at 2 CAMP monitoring sites (Chicago and Denver), 2
Houston sites (Mae and Aldine), and i Los Angeles basin sites (Down-
town Los Angeles, Lennox, Azusa, and Pomona). At 1 of the Los Angeles
sites, Downtown Los Angeles, detailed meteorological data are also in-
cluded. The Chicago and Denver data were obtained from EPA's SAROAD
system; the Houston data, from the Texas Air Control Board (TACB); and
the Los Angeles data, from the Southern California Air Quality Management
District (SCAQMD).
Chicago, Denver, and Los Angeles were selected as three cities
providing long-term air quality data and representing a range of climatic
conditions. Although Houston has only a short history of air quality data,
it was included because of the possibility of special conditions in
Texas[l,2]. Because of its numerous air monitoring stations, the Los Angeles
area is very suitable for study of intraregional patterns in the N02/pre-
cursor dependence; therefore, we have included 4 SCAQMD sites in the
analysis.
This chapter documents the procedures used to process and check the
raw data. Section 8.1 describes the original data base; Section 8.2
indicates how the raw data were organized into a processed data base; and
Section 8.3 discusses the data quality check. These efforts culminated
in the creation of an edited data base with a convenient format for
statistical studies.
-------
141
8.1 COMPUTER TAPES OF AEROMETRIC DATA
Magnetic tapes of hourly air quality data from the Denver and Chicago
CAMP sites were obtained from the SAROAD system. Although these tapes
covered the period 1963 to 1973, only five years of data (1969 to 1973)
were employed in the statistical study. Restricting the analysis to five
years provided sufficient data for the empirical models and limited the
cost of data processing. Table 8.1 lists the pollutants and monitoring
methods for Denver and Chicago.
Table 8.1 Pollutant Data Used for Denver and Chicago
SAROAD Code
Pol 1utant Method (Pollutant-Method)
NO Colorimetric 42601-11
N02 Colorimetric-Griess-Saltzman 42602-12
NOX (NO + N02)
OX Colorimetric Neutral KI 44101-14
HC Flame lonization 43101-11
CH4 Flame lonization 43201-11
NMHC (HC - CH4)
The hourly data on the SAROAD tapes was listed in an 80-column (card-
image) format as described in Table 8.2. Missing values were represented
by blanks. The original SAROAD tapes were organized according to the
following order: station, pollutant, year, and day. The units of all the
pollutants were ppm.
-------
142
Table 8.2 Format of Hourly SAROAD Data for CAMP Sites
Column
2,3
4-7
8-10
11
12,13
14
15,16
17,18
Entry
State
Area
Site
Agency type
Project classification
Sampling time interval
Year
Month
Column
21,22
23-27
28,29
30,31
32
••^•••ton*
33-36
37-40
Etc.
77-80
Entry
Sampling start hour (stan-
dard time), either 0:00
or 12:00
Parameter identification
Method identification
Unit code
Decimal locator
Observed values, in
chronological sequence.
Position of entry indi-
cates time of observation
(0:00-1:00, 1:00-2:00,
etc.) in standard time
The data for the Houston/Mae and Houston/A!dine sites were provided
through the courtesy of the Texas Air Control Board. These data covered
the years 1974 through 1976. However, monitoring for NO (and total NOY)
/\
began in March 1975 at the Houston sites. Thus, data from only March 1975
to December 1976 were useful for the present study. Table 8.3 lists the
pollutants and monitoring methods for the Mae and Aldine locations.
-------
143
Table 8.3 Pollutant Data Used for Houston/Mae and Houston/A1dine
n „ ,. ^ SAROAD Code
p°.'.lutant Method (Pollutant-Method)
NO Chemiluminescence 42601 - 14
N02 Chemiluminescence 42602 - 14
NOX (NO + N02) 42603 - 14
03 Chemiluminescence 44201 - 11
HC Flame lonization 43101 - 11
CH4 Flame lonization 43201 - 11
NMHC (HC - CH4) 43102 - 11
The hourly data for Houston were organized in the SAROAD format illus-
trated in Table 8.2. The units of all pollutants were ppm. The Houston
data were organized according to the following order: year, station,
pollutant, day.
Data tapes for Los Angeles sites had been obtained earlier by
Technology Service Corporation from the Los Angeles section of the Southern
California AQMD. Although these tapes covered the period 1955 through
August 1974, only data taken after 1969 were used in the statistical
study. Table 8.4 lists the pollutants used in the statistical analysis.
Table 8.5 presents the format for the hourly APCD data. As with the
SAROAD tapes, missing data were represented by blanks for the Los Angeles
sites. The original APCD data were organized according to pollutant,
station, year, and day.
*Most of these data, except for methane, are also available from
SAROAD or from the California Air Resources Board.
-------
144
Table 8.4 Pollutant Data Used for the 4 Los Angeles Sites
Pol blunt
NO
N02
*°x
OX
HC
CH4
NMHC
CO
SO,
NfrtHod
APCD
ColorlmtHc 25
Colorimetric 21
(NO + N02)
Colorlmetric KI 39
Flame lonization 31
Flame lonlzation 32
(HC - CH4)
Nond1spers1ve Infrared 15
Coulometric 18
Table 8.5 Format of Hourly APCD Data
Column
Entrv
Column
1
2-4
5-8
9-12
13-14
15,16
17,18
19
20
21
Dele. Code
Variable
Station
Year
Month
Days in month
Day
Day of week
Holiday
No-data day
22-24
25-27
28-30
*
91-93
94-97
98-100
101-103
104-106
107,108
*
Entry
Hourly readings. 3 spaces each
The position of an entry de-
fines the time of the reading,
0:00-1:00, 1:00-2:00,...,
23:00-24:00, standard time.
Daily average
Number of hourly readings
Maximum hourly reading
Instantaneous maximum
Hour of occurrence of
inst. maximum
"Eacfi"3ay~con?Tnues"as"a66ve"
until the end of the month.
Then there is a list of
various averages and other
statistics pertaining to
that month.
-------
145
Meteorological data tapes, in the form of the APCD "99 Cards," were also
available for Downtown Los Angeles. Table 8.6 lists the parameters included
on the meteorological data tapes.
Table 8.6 Parameters Included in the APCD Meteorological
"99 Cards" for Downtown Los Angeles
Parameter
1. Maximum oxidant value in Los
Angeles County and station
where maximum oxidant occurred
2. Maximum degree of eye irri-
tation and time of occurrence
3. Minimum recorded visibility
and related data
4. Minimum relative humidity
from 6:00 to 19:00
5. Maximum temperature
6. Average wind speed, 6:00-12:00
7. Hourly wind directions
8. Average wind speed, 6:00-9:00
Parameter
9. Inversion base height at 4:00
or 7:00
10. Various parameters describing
the 4:00 or 7:00 inversion
11. Calculated maximum mixing
height for the day
12. Parameters describing 850
pressure level
13. Pressure gradient (LAX to
Palmdale) at 7:00
14. Temperature gradient (LAX to
Palmdale) at 7:00
15. Accumulated solar radiation,
7:00-12:00
8.2 CREATION OF THE PROCESSED DATA BASE
The first part of the data-processing task was to reorganize the ori-
ginal data into a more practicable format for the statistical studies. Since
the original tapes were organized first by pollutant and then by day, the
air pollution readings for any given day were scattered over the tapes.
The data were reorganized so that all pollution data for each day are
-------
146
grouped together. In this new format, each day on the tape is followed by
the subsequent day. The new, reorganized data tapes were then used with a
simple data retrieval program to generate printouts in the format illustrated
in Table 8.7. With the new format, one could quickly visually examine
all pollutant data for a given day. Also, the new tape format greatly sim-
plifies data retrieval for the statistical analysis.
Table 8.7 New Format for Pollutant Variables
STATION
DATE Year | Month-1 Day
READING NUMBER ... 1 2 3 .... 24
(STANDARD TIME)... (0:00-1:00) (1:00-2:00) (2:00-3:00) .... (23:00-24:00)
NO
N02
HC
CH4
OX
NOX
NMHC
*
so2
**
CO
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
A • * * *
A • * * •
A • • • *
A • • • •
A • • • •
A • • * «
A • • * *
A • • • •
A • • • t
X
X
X
X
X
X
X
X
X
**
APCD 99 CARD VARIABLES
*
Los Angeles stations only
**
Downtown Los Angeles only
-------
147
The "first generation" processed data base included many days with in-
complete data. The next task was to develop a "second generation" processed
data base which excluded days with grossly incomplete data. For this task,
we composed six completeness criteria and determined, for each site, the
number of summer and winter days meeting each criterion.* We hoped to choose
a criterion that would strike a balance between quantity and quality; i.e.,
we wished to retain as many days as possible while restricting ourselves to
days with rather complete data.
The selection criteria were based on our interest in certain times of
the day for which we needed precursor or N02 data for the empirical models.
the periods (in civil time) were
t hydrocarbons (preferably nonmethane): 6-9 A.M. (readings 7,8? & 9
in standard time
readings 6,7 & 8
daylight time)
t oxides of nitrogen: 6-9 A.M. plus 3-7 P.M.
• oxidant or ozone: 2-5 P.M.
• nitrogen dioxide: 6 A.M. of the first day to 6 A.M. of the next
day with emphasis on readings at 4-6 A.M.,
9-12 A.M., and 4-7 P.M.
These times can be called the "fields of interest."
The first selection criterion (1A) required essentially complete data
within the fields of interest and allowed only one-hour gaps in the N02
record for the day. This strict criterion involved the following specific
restrictions:
^Summer" was taken as April-September, "winter" as October-March.
-------
148
1A. NMHC: all 3 readings from 6-9 A.M.
OX: at least 2 readings from 2-5 P.M.
N02: all 3 readings from 9-12 A.M.
all 3 readings from 4-7 P.M.
at least 1 reading among 4-6 A.M.
Not more than 1 consecutive missing value from 6 A.M. of
the first day to 6 A.M. of the next day
NOX: all 3 readings from 6-9 A.M.
all 4 readings from 3-7 P.M.
All times were in civtT time ("dayTight time from May through October and
standard time from November through April).
Criterion IB was the same as 1A except the NMHC restriction was changed
to a THC restriction. A separate criterion was formulated for THC because
preliminary investigations indicated that some CAMP sites might have con-
siderably more THC data than NMHC data.
Criteria 2A and 2B require that most (but not all) of the data in the
fields of interest be present. The specific criteria were:
2A. NMHC: at least 2 readings from 6-9 A.M.
OX: at least 2 readings from 2-5 P.M.
N02: at least 2 readings from 9-12 A.M.
at least 2 readings from 4-7 P.M.
No more than 3 consecutive missing values from 6 A.M.
to 6 A.M. the next day
NOV: at least 2 readings from 6-9 A.M.
A
at least 3 readings from 3-7 P.M.
-------
149
2B._ Same as 2A, except THC is required instead of NMHC.
The weakest pair of criteria, 3A and 3B, allow substantial data gaps
in the field of interest, as long as the data are complete enough to per-
mit reasonable "interpolation." Thus,each completeness requirement within
the field of interest is replaced by a weaker one, and specifications are
added for data which will allow interpolation. Criteria 3A is as follows:
3A. NMHC: at least 1 reading from 6-9 A.M.
at least 2 readings from 5-10 A.M.
OX: at least 1 reading from 2-5 P.M.
N02: at least 1 reading from 9-12.A.M.
at least 2 readings from 8 A.M.-i P.M.
at least 1 reading from 4-7 P.M.
at least 2 readings from 3-8'P.M;
No more than 4 consecutive readings missing from
6 A.M. to 6 A.M. the next day.
NO : at least 1 reading from 6-9 A.M.
-------
150
Table 8.8 Number of Days Meeting Each Criterion
Criterion Number
Chicago
summer
winter
Denver
summer
winter
Houston/Mae
summer
winter
Hous ton/ A1 dine
summer
winter
Los Angeles
summer
winter
Lennox
summer
winter
Azusa
summer
winter
Pomona
summer
winter
TA
68
79
133
193
74
41
47
60
471
400
419
181
522
429
529
464
II
271
288
143
222
74
41
47
60
474
403
424
184
523
429
532
472
2A
138
124
221
330
94
58
63
77
713
636
624
520
683
591
752
662
2B
398
386
242
345
94
58
63
77
716
639
632
523
684
592
752
663
3A
162
140
277
427
105
62
67
83
839
783
746
652
742
637
801
677
3B
455
415
305
436
105
62
67
83
842
785
757
655
742
637
801
677
Note: "Summer" is defined as April through September.
"Winter" of a given year is January through March,
plus October to December.
-------
151
2 to 3). With the exception of Chicago, each site provided nearly as many
days meeting the NMHC criteria as the HC criteria.. However, because of spotty
methane data, Chicago had many more days meeting the "B" criteria than
the "A" criteria. The low number of days in Houston for all criteria
results from the limited duration of sampling (March 1975-December 1976).
To create the processed data base, we decided to choose criterion 2A
for all sites except Chicago,where we selected criterion 2B. Choosing
criterion 2 allowed us to maintain a sufficiently large data base for the
empirical models. Although criterion 3 would have yielded an even larger
data base, it was rejected as permitting too much interpolation and lowering
the quality of the data. Criterion 1 was rejected as leaving too little
data for certain cities.
One subtlety in the compilation of the data base was the distinction
between ozone (03) and oxidant (OX). The 2 Houston sites measure ozone
according to the chemiluminescence method. Although the 6 CAMP and Los
Angeles sites measure total oxidant by the potassium iodide (KI) method,
the 2 CAMP sites actually report 03 by correcting for NO and NO Interference
according to the equation
[03] = [OX] -0.2[N02] -0.2[NO]. (6)
During the years of interest, the oxidant monitors at the CAMP sites were
fit with S02 scrubbers, and the above interference correction is appropriate
for such monitors[3,4,5].*
The Los Angeles oxidant monitors are not equipped with S02 scrubbers.
A different interference correction is appropriate for these sites[4,6]:
*Note that the S02 scrubbers convert NO to N02<
-------
152
[03] = [OX] -0.2[N02] + [S02] (7)
This correction has not been applied by the Los Angeles AQMD because N02
and S02 are generally negligible when ozone is high[7]. However, at
night and at other low-ozone periods, the negative contributions of S02 and
the positive contribution of N02 can be quite significant. For consistency
with Chicago and Denver, we decided to correct all Los Angeles area oxidant
readings for N02 and S02 interference, using Equation (7). During the.
afternoon oxidant peak, this correction appears reasonable; however, one
must use caution in correcting the low oxidant readings of 1 pphm (the minimum
reported) encountered at other times. In these cases, the number generated by
adding S02 to oxidant may exceed the actual ozone level. For example, an
ozone value of 2 pphm and a S02 reading of 4 pphm would cause a minimum oxi-
dant reading of 1 pphm to be corrected to 5 pphm, more than twice the actual
ozone level. Thus, in cases where oxidant is reported as 1 pphm, there is
uncertainty as to the real ozone level.
8.3 DATA QUALITY CHECK
The Los Angeles and Houston air quality data are subject to extensive
quality control procedures and are thoroughly screened before publication[8,9].
In contrast, the post-1969 CAMP data for Denver and Chicago have been subject
to little quality control beyond a cursory inspection[5]. Therefore, our
quality control and editing efforts focused on the Denver and Chicago data
bases.
We screened all of the Houston data and some of the Los Angeles data
ourselves and found no severe anomalies. We also found that the daily
pollutant patterns made good sense from a physico-chemical viewpoint.
-------
153
The first step in the data quality check for Denver and Chicago was
to list, for each pollutant, the diurnal patterns for the five days per year
with the highest pollution levels. We scanned these visually to determine:
0) whether the maximum concentrations reported for each pollutant were reason-
able and were consistent with other pollutants; and (2) whether there were
any unusually abrupt concentration changes between consecutive readings.
Next, the entire processed data base (grossly incomplete days excluded
by the criteria chosen in Section 8.2) was reviewed for reasonableness and
consistency among pollutant readings. The following checks were then
applied:
• The hourly pollution values were scanned, and any sharp rises or
drops between consecutive readings were scrutinized. Deletions or changes
were made when appropriate.
• Oxidant values were compared against the normal pattern of low
nighttime levels and higher afternoon concentrations. Days with (inexplicably
high (>10 pphm) nighttime OX values were deleted.
• The relationships between NO and OX levels were noted. Since these
pollutants should not coexist at high concentrations, we calculated the
product [OX] x [NO]. Where [OX] x [NO] exceeded 100 pphm2, the OX and NO
values were regarded as suspect [10].
t Days with high N02 levels were examined to verify that these were
preceded by moderate or high NO levels.
• We deleted a few days in which the interim'ttency of readings caused
us to suspect the validity of the data.
-------
154
• A check was made for negative NMHC values, which represent the
physically impossible situation of the CH4 concentration exceeding that
of total hydrocarbons. Any day with several NMHC readings of less than
-2 pphm was deleted as having suspicious HC data. All negative NMHC readings
in the "field of interest" were deleted.
• Days with a substantial majority of zero entries for one or more
pollutants were deleted on the grounds that the zeroes might actually be
missing data.
Table 8.9 lists the days eliminated from the processed data base along
with the justification for their deletion. In addition to these deletions,
we altered one reading, the 9:00-10:00 A.M. N02 reading at Denver on 24
February 1971. Most of the N02 readings that day were less than 7 pphm, ex-
cept for a single value of 37 pphm. We changed that value to 4 pphm, the
mean of the preceding and following levels.
It should be noted that Table 8.9 applies only to the processed data
base and cannot be considered as a complete list of corrections to the
Chicago and Denver CAMP data. There were also obvious problems on many of
the days that were eliminated from the processed data base according to
the selection criteria. Since these days were already excluded from our
study, we did not subject them to the data quality check.
-------
155
Table 8.9. Deletions Made in Processed Data
Bases for Chicago and Denver
City
Chicago
Date
70-4-22
70-6-22
70-12-16
71-2-18
71-4-27
72-4-10
72-5-18
73-2-4
Reason
NMHC « 0
All entries = 0
HC - 0
High 03 x NO
Low NO?, simultaneously high and rising NO +
OX peaks at 11 P.M.
NO erratic; NO • OX > 100; OX peak 10 P.M.
High NO; NO • OX > 100
Denver
72-9-25
72-9-4
72-10-31
73-7-29
73-8-22
73-8-28
73-8-29
73-8-30
73-8-31
NO falls while HC is high, level,and largely missing;
no photochemical activity to account for NO falling.
High N02 without prior NO precursor.
CH* and OX - many O's - could be blanks
MH H it n n ii ii ii
ii
n
n
n
it
n n
n n
n n
it ii
ii n
n n
n n
n it
it n
it n
n n
n ii
n ii
it n
n n
n
n
n
n
n
-------
156
8.4 REFERENCES
1. 6. K. Tannahill, "The Hydrocarbon/Ozone Relationship in Texas,"
presented at the Air Pollution Control Association Conference on
Ozone/Oxidants, Texas Air Control Board, Dallas, March 1976.
2. E. L. Meyer, Jr., C. 0. Mann, 6. L. Gipson, and 0. C. Bosch, "A Review
of the Air Quality and Emission Data Base for Ozone and its Precursors
in Selected Texas Cities," U. S. Environmental Protection Agency,
Research Triangle Park, North Carolina, November 1975.
3. A. P. Altshuller, "Evaluation of Oxidant Results at CAMP Sites in
the U.S.," Journal of Air Pollution Control Assn., Vol. 25, p. 19,
1975.
4. J. A. Hodgeson, "Review of Analytical Methods for Atmospheric
Oxidant Measurements," International Journal of Environmental
Analytical Chemistry. Vol. 2, p. 95,
5. 6. Ackland, EPA Office of Research and Development, personal communi-
cation, June 1976.
6. Los Angeles A.P.C.D., "Interferences with Ozone Measurement Made
With Neutral Buffered KI Method," 4 pages, 1972.
7. J. L. Mills, W. D. Holland, I. Chernack, "Air Quality Monitoring
Instruments and Procedures," Los Angeles APCD, 1974.
8. J. Foon, Los Angeles Air Pollution Control District, personal communi-
cation, Sept. 1976.
9. J. Price, Air Quality Evaluation Division of the Texas Air Control Board,
personal communication, May 1977.
10. T. Curran, EPA Office of Air Quality Planning and Standards, personal
communication, June 1976.
-------
157
9.0 SEASONAL AND
Before the empirical modeling analysis is performed, the seasonal
breakdown for the analysis and the averaging times for the pollutant
variables must be specified. This chapter arrives at those specifications
through an examination of seasonal aad diurnal patterns of ambient N02,
NMHC, NOX, and oxidant concentrations. Section 9.1 describes seasonal
patterns and provides a preliminary recommendation as to the seasonal
breakdown. Section 9.2 discusses diurnal patterns during each quarter
of the year; this discussion leads to final selections concerning sea-
sons and averaging times. Section 9.3 explains how these selections are
used as the basis for a computer file of dependent and independent varia-
bles.
9.1 SEASONAL PATTERNS
Figures 9.1 through 9.8 present seasonal pollutant patterns for Denver,
Chicago, Houston/Mae, Houston/Aldine, Los Angeles, Lennox, Azusa, and Pomona,
respectively. For each location, the monthly averages of daily maximum one-
hour concentrations are plotted for N02, oxidant, NOX, and NMHC (divided by
ten). The Denver and Chicago plots represent averages over the period 1969
to 1973; the Houston plots, averages over 1975 to 1976; and the Los Angeles
plots, averages over 1969 to 1974.
For each location except Chicago, the primary photochemical precursors
(NO and NMHC) show pronounced peaks during the winter (1st and 4th quarters),
A
typically reaching a maximum during November or December. These high
*In the seasonal and diurnal patterns, oxidant measurements (not corrected
for interference) are used for the 4 Los Angeles sites. In the empirical
models, corrected values representing 03 will be used.
-------
158
III I I in.ll Ml'li U.lLl LJjJ
I! 111 11111J 1111 11 n !
CX
So
•»- tn
h §
8
0 m rpmprm rmrn 1111111111 irrpn ri 111 mrm trn
1 ?. 3 4 S 6 7 6 9 10 U 12
MONTH
ii ! LuJj-ULLlJaJ il JJjJj.LljJj.J.-ljJlJJ jJUuill.! 11 1 liLI I
MIX /;
TTTTpTrrTrrrrnTrrrr rrrnrri 11 itriTrn prmiTtTTiTn
123156789101112
MONTH
Figure 9.1 Seasonal Pollutant Patterns for Denver
(Monthly Averages of Daily Max One-Hour Concentrations, 1969-1973)
vn _|ilMllUl.luuilUliu.Ll.t.lllxl.llulllllLJillillJl.hjll.J.... O ...
N33
. o
„,
rm 171 M i'i i j rrrn i rrnrmp rn pi rrrrnTrn nji'n H—
1 2 3 M S 6 7 8 9 10 11 12
MONTH
i i.i 1111.1 iliiuLuji! ii 11 Ij Lii.luj.11 u i ill i.uLi. Li.l.i 1.11
—-'' ^*\ ^^ '''
""'•- -..,/ "
-- ^^
' '~ - - W-IC/10
vnnnjrnrj-TT^
231567891011
MONTH
Figure 9.2 Seasonal Pollutant Patterns for Chicago
(Monthly Averages of Daily Max One-Hour Concentrations,1969-1973)
-------
159
P
fi"
I.L LJ I.I I111 J.U uilJJ.lJj UJ LmljLUjllU.lil I II I III
11 rjrTTTTnrnirnp 1i ipi ii|imTmTjTTTTTi
3 '< 5 C 1 a 9 10 11 12
MONTH
tLlluljXlllj.J.M.LlUjJu-LI-LujJjJJJ.I LLU.I LU.1.L
NOX
NMHC/1C
1 Z 3
s e
e 9
[ rrn iT
10 u
MONTH
Figure 9.3 Seasonal Pollutant Patterns for Houston/Mae
(Monthly Averages of Daily Max One-Hour Concentrations, 1975-1976)
' i U JjJ.U J LIJJ.I J.1J lllIJ llxil-liuiaJ UH-llU.1 111.!
i TIT] iTTf] m 11 n IT] n 11 j i nrpn rq rr 11 ] rrrrj) TTTJ n i r
? 3 M E C 7 B 9 10 ! 1 1
_. 1JL) I.LJ IILhojlJjil-i.il I llLu.lllilljL I1 I ' • .uJj.Jixl.IIU.
NNHC/10
nn j ITI tTrtTTTrmp i n pn iynrq irn 111 rr| ITTT p 'Ti
] 2 UN SC 1 0 9IOU1
MONTH
Figure 9.4 Seasonal Pollutant Patterns for Houston/Aldine
(Monthly Averages of Daily Max One-Hour Concentrations, 1975-1976)
-------
160
i - o _
I"
& .
8u, J
I I I 1 I I 1 I I I I I I HI ill II I
X
rac"
I I I 1 I I t I I I I 1.11 I I M. 1.1 I 11 11 I I I I I I I 11 limlJ.LL 1-Ll Lll 1.11 I I I j
ITT ITT [ 11 I 11 I I ITTt TTI' 1TTTIIIH 1111 M 111TI I 11 11 ITT
2 3 M S 6 7 B 9 10 11 18
MONTH
NMHC/10 :
" 11111111 riTrrrjriTTrTiTTjrrrrp 'iii|in I|IITTJTIM |iirr -
1 g 3 M 66 1 B 9 10 U 1!?
MCNTH
Figure 9.5 Seasonal Pollutant Patterns for Los Angeles
(Monthly Averages of Daily Max One-Hour Concentrations, 1969-1974)
_ ILuJ 111 1.1 JJ.1 J i l.lllij ulll Llil.U.: 1 J.LU 1J Ul 11 11 l.LlJ.LL
I- -•
-I"'
ixuL.u :it u LJJ u> :J.ujJai j 11111111111J I..LU! i; • L!.L 1.11
tMHC/10
2 3 1 S E V 8 9 JO U 18
M3NIH
IS3M567B9iOil 1!.',
Figure 9.6 Seasonal Pollutant Patterns for Lennox
(Monthly Averages of Daily Max One-Hour Concentrations, 1969-1974)
-------
161
111 n 1111 ii 111111 M n i n M 111111111 in n i r
> I I I I I I II
1 II 1 1 1 1 1
11 inli
III lIllllliLuJuil ..
NJX
WHC/1D
II I I I I I I M 11 I I 1 I I I I II I I il II I I I I I I I I M I I I I I I I I I I I I I
1 2 3 M S 6 7 8 8 10 11 12 1 e 3 M S 6 7 8 3 10 11 IE
HONTH MONTH
Figure 9.7 Seasonal Pollutant Patterns for Azusa
(Monthly Averages of Daily Max One-Hour Concentrations, 1969-1974)
.111 tLuiJ Ll I ! llll I !iu.U-Ll-U l! I i:jllJiilj.-UJ.il I U ll.L
[llll[im 11 11111111111111 It H j 111 IjTTTI pTfT JTT
i iiil 111 i.Lijiliuj.LuaiLu LI Li 111 hiii I.ULiJ.i
tax
1 8
o
NMHC/13
mrpmjnrrp-n i[ 1111 j 111i-[-nTrprrr| n 11 jTTTT-p-rn
1 2 .1 M S 6 7 8 9 10 U 12 1 Z 3M SB 1 8 9 10.1111?
MONTH
W3NTH
Figure 9.8 Seasonal Pollutant Patterns for Pomona
(Monthly Averages of Daily Max One-Hour Concentrations, 1969-1974)
-------
162
winter values for the primary contaminants are most likely due to intense
nocturnal inversions that tend to occur in the winter. In Chicago, there
is no seasonal pattern for NOV, and NMHC appears to peak in the summer.
A
The last conclusion may not be reliable because NMHC data are very sparse
in Chicago.
At all 8 locations, oxidant concentrations are greatest during the
summer (2nd and 3rd quarters), with peak oxidant values usually occurring
in July or August. Elevated temperature and high solar-radiation intensity
are largely responsible for higher oxidant in the summer. In the Los Angeles
region, the subsidence inversion which persists throughout the day in the sum-
mer also contributes to high oxidant in that season.
Seasonal patterns of nitrogen dioxide concentrations are not consistent
among the various locations. Denver and the coastal Los Angeles station
(Lennox) experience distinctly higher NOg concentrations during the winter.
Los Angeles, Azusa, Pomona, and the Houston sites show practically no sea-
sonal pattern in N02 levels, although a very minor peak seems apparent in
the 4th quarter. Chicago shows a marked peak in NC^ concentrations during
the summer.
It is interesting to note that the seasonal patterns in NO^ appear to
reflect competition between two factors: dispersion and photochemical acti-
vity. In the winter there are higher concentrations of NO available to
f\
produce N02, but in the summer there is greater photochemical activity. At
Denver and Lennox, primary contaminants show a strong peak in the winter, while
oxidant shows a relatively weak summer peak. This may account for NO/> reaching
-------
163
a peak during the winter at those two stations. At Los Angeles, Azusa,
Pomona, and the Houston sites, both the winter primary contaminant peak
and the summer oxidant peak are pronounced. This balance may account for
the lack of a seasonal N02 pattern for those 5 locations. At Chicago,
there is no seasonal pattern for NOX, but the summer oxidant peak still
exists. This seems consistent with N02 reaching a summer peak in Chicago.
The month-to-month patterns in N0?, oxidant, NO , and NMHC concentrations
£. ^
suggest that at least two seasons can be distinguished. The winter (1st and
4th quarters) is marked by high levels of primary contaminants, while the
summer (2nd and 3rd quarters) is marked by high oxidant levels. The empiri-
cal modeling analysis should be divided at least once, according to these two
seasons. The seasonal division will help to keep weather factors more uni-
form in the analysis and will also permit an investigation of seasonal changes
in the N02/precursor dependence.
9.2 DIURNAL PATTERNS
This section analyzes diurnal pollutant patterns for each of the 6
study areas. The diurnal patterns are examined individually for each quarter
of the year.* The purpose of the analysis is twofold: (l) to determine if
further seasonal breakdowns (beyond the summer/winter division) are called
for, and (2) to select appropriate averaging tUne* for the pollutant variables
to be included in the empirical models.
*The quarters are defined as (1) Jan.-Feb.-Mar,, (2) Apr.-May-June,
(3) July-Aug.-Sept., and (4) Oct.-Nov.-Dec.
-------
164
Figures 9.9 through 9.16 present diurnal patterns for each quarter
of the year at Denver, Chicago, Houston/Mae, Houston/Aldine, Los Angeles,
Lennox, Azusa, and Pomona, respectively. Averages for each hour of the
day, reported in pphm, are given for N02, oxidant, NOX and NMHC (divided
by ten). The 1st and 4th quarters are reported according to standard time,
while the 2nd and 3rd quarters are reported in daylight time. Since the
hourly data from all 8 cities represent averages from midnight-l:00 A.M.,
1:00 A.M.-2:00 A.M., 2:00 A.M.-3:00 A.M., etc., the hourly values are
plotted on the half hour, starting at 0:30 A.M.
As evidenced by Figures 9.9 to 9.16, the primary contaminants (NOX
and NMHC) exhibit two peaks during the day. At all the stations and during
all seasons, the morning peak tends to occur around 7:30 A.M. or 8:30 A.M.
(the 7:00-8:00 A.M. or 8:00-9:00 A.M. hourly average). The morning peak
is due to rush-hour traffic and to the low level of atmospheric dispersion
that often exists in the early morning. In Denver and Chicago, the evening
peak in NOV and NMHC tends to occur around 5:00-6:00 P.M., reflecting the
A
evening rush hour. The Houston and Los Angeles sites exhibit much later
evening peaks, often as late as midnight. The precursor peak occurs this
late at Los Angeles sites because atmospheric mixing is quite good in
Los Angeles during the afternoon rush hour. The sea breeze is at full
strength in the late afternoon, and the inversion is elevated by ground
heating. It is not until later in the night, when the sea breeze termi-
nates and a nocturnal inversion begins to take hold, that primary contami-
nants reach their evening peak in Los Angeles. The late-evening peak at
Houston might be explained by similar conditions at that coastal city.
-------
165
8" -
§..:
FIRST QUflRTER
iiiliuliiiliiiliiiliiil.nl. ..I.,. I. .,|m|
*«... .*....
ll|lll[ll l[l II |IM|II l[lll [IM|lll|lll[IM[llr
o e M e
8
c i
SECOND QURRTER
III 111 llnilinli.ili.il HI I,., lii.lml,., I,,,
ri
i~
I |MI| II l[lll |lll| IM|I II |UI| II l[l |M|M| III
10 12 11 16 IB 20 22 2H D 2 M 6 B 10 12 It IB 18 20 22 2
N0«
NHMC/10
^ /
|™~ • —•
£•"---.,-** v~~ _BHP/}°
rrriTTi | 11111111111 j 1111111111111 > 11 f 111111111
0 e M 6 6 10 12 11 16 IB ZO 82 2* D 2 N a S 10 12 II 16 18 20 22 2V
ETWOARO TINE DBTLIWT TIME
FOURTH QUflRTER
t I I I I t I I I I I I I i i i i t I I iI I I 11 11t11 11iIiiI 1111111i
N32
— - ID —
I
— ° Al —
THIRD QUflRTER
lllllilillllilllltlllllllllllillllllllllltlllM
H[IM| Ml[l ll[m[IM|m Jill] M I[MI |l M|
TTT
111 r 1111111 u 11 n 1111111111 n 1111 j i M 111111
0 I M e 6 10 12 11 IB 18 20 22 2» 0 I M 6 6 10 12 14 IB IB en 22 2H
II Illll II ill II ll| I I I I ill Illl 'ill "I' I'l '"!'
~ S
wwc/io
b I
N0<
0 2 M I. I 10 12 I'l 16
« «
6 8 10 12 11 1C IB
0»m.li»iT rim
2,
Figure 9.9 Diurnal Patterns at Denver (1969-1973)
-------
166
FIRST QURRTER
11111111 i i L 111111111111111111111111111111111111
SECOND QURRTER
NOS
OX
g -
is
rrrrm ru n 111111 rri M i n 11 nTrrrrpTT*
M 6 6 10 12 1M IE 18 JD 22
I I I I 1 M 1 I 1 i I I 1 I I I 1 J I I I \ 1 I I \ I I n I I t t 1 I i
NBX
NMHC/10
rfs -
,— tn —
TTrr
0 8
rjl I I |l 11 1111 [III11II|I7l[II 111 II11>I| 11
M E « 10 12 11 IE 18 20 22
EFRNOPRO TIME
FOURTH QUflRTER
~~ rlTTTT I I I I I I I I I j I / I I I I I I I II I I II I I I I I I M I I ( I I I I I
o e M E e 10 12 11 IB IB co ss i
I I 1 I I I » I I I I I I I I I I 1 I I I I I I I I I i I 1.1 I I I I I ! I I I I I 4 I l_ I I )
NMHC/10
ii
~ in
I,
N32
OX
11111111114 j 111 j rn~p 11111111111 i 11 pi 111 M 11
£ M 6 8 ID 12 11 16 IB 20 22 SI
DAYLIGHT TIME
THIRD QURRTER
I 1 I I I I I I i I 1 I I 4 I I 1 4 I I I I I I 4 I I I I I I I 4 I I I I I I I I t I I I I I
N02
OX
TJTTI 1111[ 111"[ i n [ n 11111 [ 111 [ 111 [ 111 [ i n [ i T r
o ? M e B in 12 11 IB is 30 22 et o e H 6 e 10 12 m IE IB ?o 22 2
_ i 11 11111 n 111 111 n 111111111 111111111 i u I n i I
\ n |in j n
16 IB »n r-r1 :'t
r!s
r§s
NOX
me/10
pTT-jTfq mp rrjr?rprriiii [nrjr
n ? M s D 10 iz 11 IB IB M 2?
Oi.fltrHT TIM
Figure 9.10 Diurnal Patterns at Chicago (1969-1973)
-------
167
8
I.
. . , FIRST QURRTER
iiilmlmlmliiiliiiliiiliiil...I.,.I..,!..,
Li. j
SECOND QURR1ER
11J I I > I I I I I I I I I I I I I I ! I I I ! I I I , . I I , , i | | , , | | t | | , ,
»••—'^
0 2
6 8 10 12 14 16 IB 20 22 24
Illlllllll ill II lull n iliiiliiilnilinlnilii
— O _
III [III] II l[l I I|IM| Illjl II ||M[ II l|l II |IM| M I
0 e M 6 B 10 12 14 16 16 £0 22 i
I II hill II ill lllnilinliiilii.lniliiilnilni
rs- -
I II | I I I | I I l| I II [ II II II III I I I I III II III II | I M|
0 I M 6 8 10 12 14 16 IS 20 22 24
STRNOPRO TINE
FOURTH QURRTER
o i I I 11 111 I I I 11 11 I 11 11 I i 111 i i I I 111 11 111 n i; i 11 11 I o
NMHC/IO
NOX
N02
: §_
s
-I. -1
iii|iii|fii|iii]ffi|iii|iii|iM|ri iii 11 11 M i i
2 M 6 8 10 12 14 16 18 20 22 £4
DWLIGHT TIME
THIRD QURRTER
11111111111111 i_l 111111111111111111111111111111
ri 1 '
,' «3--
n M 1111111111 M 1111111111 n T nr
0 2 M 6 8 10 12 14 16 18 20 22 24 0 t M 6 I 10 It H IE IB 20 n PH
i i|linln ill ii hill 11 ill ii lull n ill n htJ I m_ .
£ w
Z"
~ TTTrrrr [ 111111111111111 [ n T [ 111 [ 111 p'' |!
0 2 M B 8 10 12 14 16 IB 20
SrHNDPRD TIMf
„ o —
NMHC/IO
IOX
pi qrrr
0 2 M 6 8 10 12 14 16 in 20 n PI
ORVLtBIT TIMt
oirtnutriL' i •"*. / t /\^rm i/\-?/*\
Figure 9.11 Diurnal Patterns at Houston/Mae (1975-1976)
-------
168
FIRST QURRTER
"""""• flO """"'
N02
-.81-—
: fi
— cu —
T I I 1 I M I I I I 1 I I I I I f I I I I I I I 1 I M 1 I ] I I I 1 I 1 TTTTtI I'I
0 Z t 6 8 10 1? It 16 IB BO 22 21
111 11 111 I I I I i i I I 11 11 11 111 11 11 11 I i I 111 11 11 i 11 i 1 I
SECOND QURRTER
inlitil iiilniliiil iiiliiilniliiilni Iliiluj
03
8-
NMHC/10
I I I I I M I I I f I I rT|lM|llljlll|IM|lll|llTTIH| I I I
0 E >t 6 B 10 12 1H 16 IB 80 22 2;; HM
: I
NOX,
NMHC/10
11111 r i j 111 j 1111t11111111111111111111 M 11 r
0 e M B e 10 12 11 16 IB ?0 2?
STBNDSRO T1MF DBYUGHT TIME
Figure 9.12 Diurnal Patterns at Houston/Aldine (1975-1976)
-------
169
FIRST QURRTER
~ 8
CU
n 11 u 1111111111 m 111111 1111 M j n 111111 m I u
0 2 1 6 a 10 12 11 16 IB 20 ZZ 24
I I I I I 1 ll I 1 I I II I I I I I I I lill Illl! I i I 1 I I . ! ! I I , .
SECOND DUflRTER
NEK
NHHC/10
"T'T"!"^'''!"'!'"!'"!"'!"'!'"!"1
0 E 1 6 8 10 12 11 16 18 20 22 i
STRNDRRD TIME
FOURTH QURRTER
i n |iii| 111| i n jm | n 111111 m | MI| 11111 M 1111
o e M E e ID 12 H 16 IB ?n zz i
- O | I II I. I.I I. ll| .ill.. I. I lll.lll.llM.il.. ll.l I I.,
NHHC/10
HOt
ox
I I |I I I| I I I [ I I I | I II | I I I | I I I |l M | I I I |I I I |NI| I I I
0 2 M 6 8 1C 12 It 16 IB 20 ZZ i
I II Illll H
-------
170
FIRST QUflRTER
JUjJLJJ. > i 1 i l i ll i i I i 111 11 LIllll lvi.l.iXii ..l-Li-LlA --ill LI
SECOND QUflRTER
uo.
0 2 M 6 6 10 12 11 IE 18 20 22 24 0 2 4 6 8 10 12 11 16 111 E() 22 2<4
§,
in! ii ill ii [ml 11 il 111 liiiliuli 111
n I 111
NNHC/10
i i i I i i i I i i i I i M I i t t I 11 i I i i i I i t i I i i i i i n 11
0 e M E 8 10 12 11 IE 18 20 22 21
STBNDWD TINE
FOURTH QUflRTER
1111111111111111111111 i 11111111111111111111111
8
I II I I M I M I | I I I | I M | I I 1 | I ! I | M I | I I I | I I I | I M | II I |
0 E M 6 8 10 12 11 16 18 20 22
DHYUGHT TIME
THIRD QUflRTER
i ii lii Li. n .i i M i I M 11 11 i I i n 111 i I 11 ill i i I i ' vljj
I-
TTrpTTprijMri.MjiiTTMUM.i.MinTi i-rrprr
0 e M E 8 10 12 11 16 IB JO 22 i
0 Z M
0 I I 6 8 in 12 11 1C IB ?!) ?2
r.rtiMiitf n I \t*-
r ° ~-
Lic, j
£„ _,
r ^
_ o _
6 B 10 12 1H 16 IB 20 22
11111111111111 L i 11111111 M LO.LJ.U,
NOX
•
NMNC/JO
0 ?
in |rn-[irfpn-jTTTpTiiri': |i
6 6 10 ir 11 16 18 iO
onV' :'i'T nur.
21
Figure 9.14 Diurnal Patterns at Lennox (1969-1974)
-------
171
PI
s -
i
8
FIRST QURRTER
-J-l I Ml I I I I I I I I I tt I I I I I I I I I I I . ! I I I , I . , , 1 , ( , ) ,
SECOND QUflRTER
lillmlllllllllmlliili.ilinlnilinllnlni
* n....
I'M j in [ I I I [ I I I [ I I I | I I Ij I I I [ I I I [ I I lj I I I I III I I I I
0 t t 6 8 10 12 11 16 IB 20 22 2< 16 IB 80 23 2")
ml n ill n Iml n ill 11 lnil n il Mi )m|n
NMHC/10
TTT pTTJ II I | I'I I |'U l[ I I I |l II | N l| I I l|l H | I M| in
0 '! M 6 8 10 12 1H 16 IB ED ZZ 21
5TRHOPRD TIME
FOURTH QUflRTER
NO*
NHHC/ID
i
NO?
l l l 11 l l | l l l j : i l | > . . | i i 11 . . i f . . . | . . i j . i . | i r i | i-
P 2 M 6 B 10 12 m 16 IB 80 28 ?.t
DRYU91T TIME
THIRD QUflRTER
11! i i 1111111111111111111111 i i 11111111111111111
I I I I * I I I I
i ii 1111| i n [i n| 111 jm |i 1111! i |ii i |i 1111 r
M E B 10 12 H 16 IB 80 C8 21 0 Z M B 8 10 12 II 16 IB 10 g?. 21
14 i I | i i I i i I 11 I I I i 1 l I I i l I l I I I I I I I I I I I I I I I I 1 I I 1 I l I
f.
I i i I n i I l l i b i i | i l l I i i l I I l l I l I 11 1 I J 1 11 i 11J I mL- =r _
NB« "" ~*
rls
NMNC/IC
rrjTTi Trrt-|TrrjTTrpiTTTrr-|
a in i? n "' 18 eo 2Z
MliMHHO fl!1
- ni
N0» t-
•JMnC/IO
0 2 1 b B in 12 PI It' 18 !1
Figure 9.15 Diurnal Patterns at Azusa (1969-1974)
-------
172
FIRST QURRTER
• ulinl ii ill I ilLuJ-U.LluJLl.U-llujljJ.JLiiJ^ilXi .
ox
TTT i n 1 ir-i
SECOND OUflRTER
lLlJ ItJ 11 " I n ' 1^ UL! IU I I III U-l i.
M
0 2 M 6 B 10 18 H IB 18 ZO 22 21 0 8 M B B 10 12 11 IB IB £0 22 81
I I I I I I . I I . I , I ) , , I 1 I I 11 I I . I . I j 1 111W LLLLmL.jjJ lnil.llllllluilllllllll.LLl.lMllllllllllmlllll
HNKC/tO
I I I I I M I II H ' " p'l I I I I I I I I I |l M I II I I I I I [ I M I I I 1
0 2 M 6 1 10 12 11 16 18 20 22 2M
BTRNDHRO TIME
FOURTH QURRTER
i,,1,.,I,,,I,,,I,,,!,,,I,M|,,,!,,,I,,,I,,,!,,,
h
i
t—
s
1I2 -
i
"'|"'lr 111 i i > 11' 11111111111 • 111111 rr T| 1111111
0 £ M 6 8 10 12 11 16 18 20 22 21
ORVUGHT TIME
THIRD QURRTER
I I • I 1 i 1 111 ll1 I 1 I it 1 I I I I Ii i I 1 I t I I I I I I I I t I J t I I I I t
..,|,,,|.,,|,,,|,,,,,TT|,,,|,,,|,,,|,,,|,,,| ITT
0 ? M 6 B 10 IP IN IB 16 EC ?2 2
I I i t I 1 I I I I I I I I i I 1 I I I i 1, I 1 I I t I ) I M i I 1 i 1 1 1 i i IJ I I
0 I 1 E 6 10 12 11 16 IB 80 22 2'1
I I I 1 I I III II ill I I 111 1 I I ! ill n I II I | , ! | | | ! 1 | | 1 , I IJ ] J
MIX
- .-•-* NHIIC/IO
i n jinTrnTTTTTninniTiTrrTTTTTTn trirrrpi i
0 2 H 6 a 10 1C 11 16 IB »fl M ?'l
[IRVI I'll! IIMl
Figure 9.16 Diurnal Patterns at Pomona (1969-1974)
-------
173
The diurnal patterns for the primary contaminants are similar during
the 1st and 4th quarters, and during the 2nd and 3rd quarters. The two
winter quarters tend to have stronger nighttime N0¥ and NMHC peaks than
/\
the two summer quarters, especially in the case of the Los Angeles sites.
This reflects stronger nocturnal inversions during the winter.
At all sites and during all quarters, oxidant reaches maximal concen-
trations in the afternoon. The oxidant peak tends to occur around 1:30 P.M.
at Denver, Chicago, Los Angeles, and Lennox and slightly later, around 2:30-
3:30 P.M. at the 2 Houston sites and 2 downwind Los Angeles sites (Azusa
and Pomona). The 2nd and 3rd quarters are distinguished by higher oxidant
levels than the 1st and 4th quarters.
With respect to diurnal patterns for nitrogen dioxide, at each site the
1st quarter is similar to the 4th quarter, while the 2nd quarter is similar
to the 3rd quarter. At Denver, the winter quarters show two nearly equal
NOo peaks, one at 9:30 A.M. and one at 5:30 P.M. In the summer, N02 peaks
are lower at Denver, and the nighttime maximum occurs later (at about 10:30 P.M.).
Chicago shows very little diurnal variation in N02 concentrations during the
winter, although a single, minor peak is evident at about 4:30 P.M. N02 con-
centrations are higher during the summer in Chicago, and the peak at 4:30 P.M.
is much more pronounced.
At Houston Mae, the winter quarters exhibit two nearly equal N02 peaks,
one at 8:30 A.M. and one at 6:30 P.M. The evening N02 peak occurs later (about
11:30 P.M.) during the summer at Houston/Mae. Houston/Aldine shows a pronounced
nighttime N02 Peak, about 6:30 P.M. in the winter and 9:30 P.M. in the sunroer.
-------
174
The summer N02 maximum is nearly the same as the winter N02 maximum at both
Houston sites.
Los Angeles and Lennox exhibit a single major peak in N02 at about 10:30 A.M.
during all seasons. At Los Angeles, the morning peak in the summer quarters ex-
ceeds the morning peak in the winter quarters. At Lennox, the morning peak is
higher in winter than in summer. The downwind Los Angeles sites (Azusa and
Pomona), show two N02 peaks during the day, one in the morning (at about
9:30 A.M.) and one in the evening. In the winter, the evening peak occurs
around 6:30 P.M. and is larger than the morning peak. In the summer, the
evening peak occurs around 9:30 P.M. at a level close to that of the morning
peak. The summer maxima in Azusa and Pomona have about the same strength as
the winter maxima.
The above observations indicate that diurnal patterns for each pollu-
tant are similar in the 1st and 4tTi quarters, and 1n the 2nd ana 3rd
quarters. This suggests that multiple seasonal divisions, according to indi-
vidual quarters of the year, are not necessary. A single seasonal division
(summer vs. winter) appears adequate for the empirical modeling study.
The diurnal patterns also suggest averaging times for the variables to
be included in the empirical models. The dependent variable, nitrogen dioxide,
usually reaches two minima at around 5:30~~A~.M. and 3:30 P.M. Thus, it appears ap-
propriate to select 6:00 A.M. to 4:00 P.M. for "daytime average" N02 and 4:00 P.M.
to 6:00 A.M. for "nighttime average" N02. The daytime peak N02 will be taken as
the peak hour between 6:00 A.M. and 2:00 P.M., while the nighttime peak N02 will
be the peak hour between 4:00 P.M. and 2:00 A.M. The only exception to these
rules is Chicago, which attains a single N02 peak in the late afternoon. For
-------
175
consistency with other sites, the daytime and nighttime averages will be de-
fined the same in Chicago as elsewhere. However, the daytime N02 peak will be
taken as the peak hour from 6:00"A.M. to 4:00 P.M. in Chicago.
The early-morning precursor levels might best be taken at the time of
maximal precursor concentrations, say 7:00 A.M. to 9:00 A.M. However, for consis-
tency with the convention established by other researchers[l,2,3,4], a 6:00 A.M.
to 9:00 A.M. average will be used for morning precursors, NO and NMHC (or HC).
A
An average from 4:00 P.M. to 7:00 P.M. will be used to measure evening NOV as
A
a precursor of nighttime N02. In Denver and Chicago, this is the period of
the evening maximum in NOX concentrations. For the Houston and Los Angeles
sites, the evening NOV maximum occurs much later. However, it seems best to
A
use the 4:00 P.M. to 7:00 P.M. average for the Houston and Los Angeles sites as well
since this average is a measure of precursor levels at the beginning of the
nighttime period. In the empirical models, day-to-day fluctuations in pre-
cursors, rather than overall precursor levels, are the key to obtaining the
desired relationship. Thus, it is not mandatory that the precursors be mea-
sured during the period when they reach a maximum.
Afternoon ozone will also be considered as a precursor to nighttime N02-
The averaging time for oxidant will be taken as 2:00 P.M. to 4:00 P.M.
The above selections of averaging times for the dependent and independent
variables are somewhat arbitrary. Alternative arguments can be made which
would suggest different seasonal breakdowns and different averaging times than
the ones we have chosen. In our selections, we have attempted to strike a
balance between the need for detail to represent a varied and complex phenomenon,
-------
176
and the need for simplicity to facilitate application of the empirical modeling
techniques.
9.3 COMPUTER FILE OF DEPENDENT AND INDEPENDENT VARIABLES
To facilitate the empirical modeling analysis, computer files were created
which listed, for each site, the values of the dependent and independent varia-
bles. Separate files were established for summer (April-September) and winter
(October-March). The variables in the files are summarized in Table 9.1. As
indicated in Table 9.1, "initial conditions" for N02 at the beginning of each
daytime and nighttime period were included in the files. These initial condi-
tions, as well as the "precursor variables," might be important in explaining
peak and average NO,, concentrations.
A special computer file was created for Downtown Los Angeles. This
file includes seven weather parameters as well as the pollutant variables.
The seven meteorological parameters are:
t calculated maximum mixing height for the day (HM);
t maximum temperature for the day (TM);
• minimum relative humidity from 6:00 to 9:00 (RH);
• average wind speed from 9:00 to 12:00 (WS);
a accumulated solar radiation from 7:00 to 12:00 (SR);
• pressure gradient from LAX to Palmdale (PG); and
• temperature gradient from LAX to Palmdale (TG).
Another variable, 6:00-9:00 A.M. carbon monoxide concentration, was also
added to the Downtown Los Angeles file. This variable, which is not a causal
precursor of N02, should be useful for assessing how the intercorrelations
between morning precursors affect the statistical results.
-------
Table 9.1 Variables for the Empirical Modeling Analysis
Dependent Variables
Independent Variables
Initial Conditions
DAYTIME ANALYSIS
Peak One-Hour N02
(from 6:00 A.M. to 2:00 P.M.)"
Average N02
(6:00 A.M.-4:00 P.M.)
Morning NOX
(6:00 A.M.-9:00 A.M. average)
Morning NMHC
(6:00 A.M.-9:00 A.M. average)
Morning HC
(6:00 A.M.-9:00 A.M. average)
Six Weather Variables
(Downtown Los Angeles only)
Early-Morning NO?
(5:00 A.M.-6:00 A.M. average)
NIGHTTIME ANALYSIS
Peak One-Hour NO?
(from 4:00 P.M. to 2:00
Average N02
(4:00 P.M.-6:00 A.M.)
Evening NOX
(4:00 P.M.-7:00 P.M. average)
**
Afternoon 03
(2:00 P.M.-4:00 P.M. average)
Afternoon N0£
(3:00 P.M.-4:00 P.M. average)
**
For Chicago, this period is 6:00 A.M. to 4:00 P.M.
CAMP oxidant data were obtained already corrected for interferences.
Los Angeles oxidant data were adjusted for interference in this study to represent
-------
178
9.4 REFERENCES
1. E. A. Schuck, A. P. Altshuller, D. S. Barth, and G. B. Morgan, "Relation-
ship of Hydrocarbons to Oxidants in Ambient Atmospheres," Journal of the
Air Pollution Control Association. Vol. 20, May 1970.
2. J. R. Kinosian and J. Paskind, "Hydrocarbons, Oxides of Nitrogen, and
Oxidant Trends in the South Coast Air Basin, 1963-1972," California Air
Resources Board—Division of Technical Services, Internal Working Paper,
3. J. C. Trijonis, "Economic Air Pollution Control Model for Los Angeles
County in 1975," Environmental Science and Technology, Vol. 8, p. 811,
1974.
4. E. A. Schuck and R. A. Papetti, "Examination of the Photochemical Air
Pollution Problem in the Southern California Area," EPA Internal Working
Paper, May 1973.
-------
179
10.0 EMPIRICAL MODELS APPLIED TO DOWNTOWN LOS ANGELES
Before performing empirical studies for all six cities, it is useful
to conduct an exploratory analysis with the detailed data base for Down-
town Los Angeles. This exploratory analysis should point out the most
important variables and should indicate the most promising statistical modeling
techniques. Since meteorological data are available for Downtown Los
Angeles, the effect of including weather variables in the empirical models
can also be investigated. It is important to include meteorology, if possible,
to avoid spurious N02/precursor relationships which could result if the pre-
cursors were correlated with weather factors that govern N02 production.
Section 10.1 describes the various statistical techniques that are
used to investigate the data from Downtown Los Angeles. Since all these
techniques are applied to the same data base, they all yield similar qualita-
tive conclusions concerning the N02/precursor dependence. The qualitative
conclusions concerning daytime N02 are presented in Section 10.2. Included
are discussions of the role of NO , the importance of initial conditions
A
(e.g., 5:00 A.M. N02), the apparent role of hydrocarbons, and the effect
including weather parameters. Section 10.3 presents conclusions concerning
nighttime N02-
Passing from qualitative conclusions concerning the N02/precursor
dependence to a quantitative model that can be used to predict the impact
of precursor control is a difficult and tenuous step. The limitations of
our particular approach (see Section 7.2.4 and Section 10.2.4) imply that
there will be some uncertainty in the quantitative predictions. Section
10.3 does formulate a predictive model, but this model should at
-------
180
present be regarded as an educated hypothesis that explains certain observed
phenomena rather than a definitive tool. The model should be checked by
quantitative comparisons with smog-chamber results and with historical air
quality trends before full confidence can be placed in it. Such comparisons
will be conducted in later chapters of this report.
10.1 STATISTICAL TECHNIQUES FOR EMPIRICAL MODELING
This section discusses some of the statistical techniques that were used
to investigate the Downtown Los Angeles data base. The discussion does not
delve into the statistical theory behind the methods. Rather, the intent
is to provide a brief description of the techniques and to familiarize the
reader with the type of outputs available to us.
Graphical Technique Using Mid-Means
In order to provide graphical illustrations of the relationship between
a "dependent" variable and an "independent" variable, a program was developed
based on a "moving mid-mean" technique. The solid line in Figure 10.la or 10.Ib
illustrates the output from this program. In this case, the independent
variable is 6-9 A.M. NOV (NOX69); the dependent variable is daytime peak
A
one-hour N02 (DPKN02). The plotted values for NOX69 represent the average
of 30 daily measurements for NOX69, while the plotted values of DPKN02
ic
represent mid-means of corresponding measurements for DPKN02. A "moving-
window" technique is used which examines the data according to ascending
order of NOX69. The window (containing 30 data points) is moved 10 data
points to generate each point on the graph.** The mid-mean of the dependent
variable (DPKN02) is plotted against the mean of the 30 data points for
the independent variable (NOX69).
The mid-mean is the average of all values between the 25th and 75th
percentile.
**
In some cases with small amounts of data, the window is moved only
5 data points in each step.
-------
181
-i-L,..ili I i , ilililj -U-ilim)lj,.ll.;i|; ;L.
Q.
Q.
•»
CM
S -:
90th Percentlle
. WINTER
t
X,
10th Percentile :
• : ' 'l"11^! ';i'|inijT
0 !C 20
30 tO 50 63 70 30
NOX69, pphm
Q.
a.
CVJ
I ! | I I I I | I I I I P IT
30 »0
NOX69, pphm
Figure 10.1 Mid-Mean and Percent!les of Daytime Peak
N02 vs. 6-9 A.M. NOX
-------
182
Mid-means are used for the independent variable (DPKN02) because they
are less sensitive to outliers than are averages [1]. That there is con-
siderable scatter in the individual values of DPKN02 for any given level
of NOX69 is illustrated by the dashed lines which give the 10th and 90th
percent! les for DPKN02 as a function of NOX69.
Multiple Linear Regression
A common technique used to investigate the relationship between variables
is multiple linear regression. In essence, multiple linear regression com-
putes coefficients A and B-, ,...,B that give the best least-squares fit of
the form
A + B-jX-j + ... + Bnxn (8)
for a dependent variable (y) and independent variables (x, ,...,x ). Since
application of the graphical mid-mean technique revealed that many depen-
dencies appear linear, extensive use was made of multiple linear regression.
In some cases, nonlinearities were introduced by choosing an independent
variable in the regression as a nonlinear function of precursor variables,
for instance x = NMHC-NOY or x = NMHC/NOY.
A A
The specific computer program used in this study was the SPSS (Statistical
Package for the Social Sciences) multiple regression program. As well as
using the regression coefficients (A,B1 ..... Bn), we employed the following
outputs from that program:
• matrix of partial correlation coefficients
• total correlation coefficient (R)
-------
183
• percentage of variance explained in the dependent variable (R2)
• standard error in the regression coefficients
• F-statistic
• residuals of the regression (y . , - y _,. )
VJfactual ^predicted''
Multiple Logarithmic Regression
In some cases, multiplicative rather than additive relationships were
explored. This was done by performing a linear regression of the form
+ B. Jinx. + ... + B
II n
or
B1 B2 B
y = A x] x2 ... xnn . (9)
In such cases, the square of the correlation coefficient measures the per-
centage of variance explained in £ny, not the percentage of variance explained
in y. A separate program was written to determine the percentage of variance
explained in the original dependent variable.
TSC COMPLIAR Program
The TSC COMPILAR program is a multivariate nonlinear regression technique.
It represents the relationship between the dependent and independent variables
with continuous oiecewise linear functions (hyperplanes). Using an iterative
technique, the program selects hyperplanes that define regions where certain
characteristic relationships exist. The iterations are directed at maxi-
mizing the percentage of variance explained in the dependent variable.
Figure 10.2 gives an example of output from the COMPLIAR program. The
relationship between winter DPKN02 and morning precursors is indicated by
-------
37
Surface of
DPKN02, pphm
NOTE: NMHCPR is an
estimate of NMHC
calculated from total
hydrocarbon measurements
Figure 10.2 Output of COMPLIAR Program for DPKN02 vs. NMHCPR and NOX69, Winter Season
-------
185
four hyperplanes. This particular example is interesting because it repre-
sents the particular case (winter season and peak N02) where the most
significant hydrocarbon dependence was found. For high values of NO (or,
J\
more appropriately, for low NMHC/NOXratios), winter DPKN02 appears to be quite
sensitive to hydrocarbons.
In this study, we used the COMPLIAR program to obtain a qualitative picture
of the relationships in the data and to check the conclusions yielded by other
techniques. The predictive models for assessing control strategies were based
on simpler regression forms.
10.2 DEPENDENCE OF DAYTIME N02ON PRECURSORS
This section discusses conclusions concerning the daytime N02/precursor
relationship at Downtown Los Angeles. These conclusions are based on statis-
tical analyses involving the variables listed in Table 10.1, which serves as
a glossary for this discussion. All four statistical techniques described
in the previous section were used to explore the data. These techniques
were employed with various combinations of the variables and with various
functional forms.
Since each statistical technique was applied to the same data base,
each yielded the same qualitative conclusions concerning the N02/precursor
dependence. The qualitative aspects of the findings will be the subject
of this section. The final section of this chapter will use a specific
statistical technique to arrive at a quantitative model.
The relationship between daytime N02 and precursors turned out to be very
complex. For instance, the effect of hydrocarbons was different on average
N02 than peak N02; was dependent on the season; and was different for high
NO levels than for low NOV levels. The observed dependence on hydrocarbons
x x
-------
186
was also sensitive to including weather factors in the analysis. Because
of this complexity and because of the large number of variables in-
volved, the investigation of the data was an iterative learning process.
It is impractical to describe all of the specific analyses that
Table 10.1 Glossary of Variables for the Daytime Analysis
(All Hours are in Civil Time)
Dependent Variables
DPKN02 Peak one-hour N02 from 6:00 A.M. to 2:00 P.M. in pphm
DAVN02 Average N02 from 6:00 A.M. to 4:00 P.M. in pphm
Independent Variables (Pollutants)
NOX69 6:00-9:00 A.M. average NOX concentration in pphm
N025 4:00-5:00 A.M. N02 concentration in pphm (if the 4:00-
5:00 A.M. reading was missing, 3:00-4:00 A.M. or
2:00-3:00 A.M. was used)
INTNO NOX69-N025, representing overnight NO plus morning
injection of NO
NMHC69 6:00-9:00 A.M. nonmethane hydrocarbon concentration in pphraC
HC69 6:00-9:00 A.M. total hydrocarbon concentration 1n pphmC
NMHCPR (HC69-100)/2, approximate value of NMHC calculated
from total HC concentration
C069 6:00-9:00 A.M. CO concentration
Independent Variables (Meteorology)
HM calculated daily maximum mixing height
TM maximum daily temperature
RH minimum relative humidity (6:00 A.M.-7:00 P.M.)
WS average wind speed (9:00-12:00 A.M.)
SR accumulated solar radiation (7:00-12:00 A.M.)
PG pressure gradient from LAX to Palmdale
TG temperature gradient from LAX to Palmdale
-------
187
were performed. What follows is a selected sample of results that best
illustrates the relationships, and lack of relationships, that exist in
the data.
10.2.1 Morning Precursor Variables
The original intent in the empirical modeling study was to use NOX69
and NMHC69 as the morning precursor variables for daytime N02. There was
some concern, however, about the accuracy of the NMHC data. NMHC values
are obtained by subtracting methane(CH^) measurements from total hydrocarbon
(HC) measurements. A recent study indicates that methane and total hydro-
carbon data tend to be of uncertain reliability [2]. NMHC values, obtained
by subtracting one uncertain measurement from another one of comparable
magnitude, are especially suspect. A further problem in the case of Los
Angeles NMHC data is round-off error. HC and CH^ are both reported only
to the nearest ppm. Thus, an individual hourly NMHC measurement can only
assume values of 0, 100 pphm, 200 pphm, etc. This is an extremely gross
resolution considering that average NMHC concentrations in Los Angeles are
less than 100 pphm.
Because of the concern about the NMHC data, total hydrocarbons (HC69)
were also included in the data base. If the NMHC data proved of little use,
it might be possible to conduct the analysis with the total hydrocarbon
measurements.
Another concern was the colinearity problem with morning precursor
variables. The intercorrelations between the precursors might make it
difficult to separate out the individual effects of NMHC and NOX on daytime
N02. To assess this problem, C069 was included in the data base. This
-------
188
variable bears no causal relationship to N02, and it is interesting to
determine if the statistical techniques can find that result.
Table 10.2 illustrates the correlation coefficients between the morning
precursor variables (NOX69, NMHC69, HC69, and C069). As expected, high
intercorrelations exist between the 6:00-9:00 A.M. concentrations because the
pollutants tend to rise and fall together, depending on dispersive condi-
tions. It is notable that the smallest correlations occur when NMHC is
one of the variables. As previously remarked, the NMHC data are considered
the least reliable.
As a test of the relative importance of the variables as N02 precursors,
logarithmic regressions were conducted between daytime N02 (both peak and
average) and pairs of the precursor variables. For example, the regression
for DPKN02 vs. NOX69 and NMHC69 was of the form
B] JlnNOX69 + B2 £nNMHC69
Bl B?
or DPKN02 = A-NOX69 '. NMHC69 L . (10)
Table 10.3 lists the regression coefficients (B, and B2) for each pair of
independent variables.
Table 10.3 reveals that the coefficient for NOX69 tends to dwarf the
coefficient for any other variable paired with it. In particular, C069 tends
to be assigned insignificant importance when it is paired with NOX69, even
though C069 is the morning pollutant variable most highly correlated with
NOX69 (see Table 10.2). The dominance of NOX69 is not surprising; we
expect daytime N02 to be most strongly dependent on NQX69. Part of NOX69
-------
189
Table 10.2 Correlation Coefficients Between Morning Precursor Variables
Summer
NOX69
HC69
NMHC
C069
NOX69
HC69
NMHC69
C069
NOX69 HC69 NMHC69
1.00 0.78 0.65
1.00 0.78
1.00
Winter
NOX69 HC69 NMHC69
1.00 0.81 0.74
1.00 0.81
1.00
C069
0.79
0.77
0.66
1.00
C069
0.86
0.81
0.79
1.00
-------
Table 10.3 Logarithmic Regression Coefficients for Pairs of Morning Pollutant Variables
Day
Peak
N02
Day
Average
N02
Summer
NOX69 HC69* NMHC69 C069
0.63** 0.05
0.68 0.03
0.67** 0.06
0.44** 0.12**
0.26** 0.50**
NOX69 HC69* NMHC69 C069
0.63** -0.01
0.63** -0.01
0.64** -0.02
**
0.38 0.09**
0.21 0.43**
0.07** 0.50**
Winter
NOX69 HC69* NMHC69 C069
0.50** 0.25**
0.63** 0.04
0.67** 0.15
0.61** 0.09**
0.07 0.66**
NOX69 HC69* NMHC69 C069
0.49** 0.19**
0.60** 0.02
0.62** 0.00
0.55** 0.08**
0.38** 0.35**
0.08** 0.56**
*In the logarithmic regressions HC69-80 (units are 1n pphmC) 1s used to avoid singularities
of the logarithm function.
^^
Coefficients significant from zero at 99X confidence level.
vo
o
-------
191
is already N02 (which contributes to the daytime peak and average N02),
while the rest of NOX69 is NO (which is a source of further N02). However,
it is encouraging that the regression analysis does discover the importance
of NOX. This indicates that the interconnections between the variables are
not so high as to make the real precursor (NOX) indistinguishable from the
surrogate (CO).
For the regressions in Table 10.3, the percentage of variance explained
in £nDPKN02 and £nDAVN02 tended to be around 55% to 65% (R = 0.75 to 0.80)
when NOX69 was one of the independent variables. When NOX69 was not included,
the percentage of variance explained dropped to around 30% to 50% (R = 0.55
*
to 0.70). This again indicates the particular importance of NOX69 as a
precursor variable.
Another interesting feature of Table 10.3 is that HC69 is the only
variable which appears to retain some importance when it is paired with NOX69.
In the winter, the HC69 coefficients for both peak and average N02 are
highly significant. The NMHC69 coefficients, on the other hand, are always
insignificant when N0¥ is included and are small even when NMHC69 is paired
j\
with HC69 or C069. This is a further indication that the NMHC data for
Los Angeles are not as useful as the HC data for empirical modeling.
That the dependence of N02 on HC69 is not solely due to interconnection
with NOX69 can be seen by graphing N02 vs. morning NOX, while stratifying for
hydrocarbon levels. Such plots (presented later) show that higher HC69
The percentage variance explained in the original dependent variables,
DPKNO? and DAVN02» was slightly less than,for the logarithms, typically about
5% to 10% less.
-------
192
levels tend to yield higher N02 concentrations for fixed values of morning
NOV. When similar plots are prepared, stratified by NMHC69, little, if any,
J\
hydrocarbon effect is evident. Again, this is probably a reflection of the
poorer quality of the NMHC data.
Because of the questions concerning the NMHC data, it was decided to use
HC69 instead of NMHC69 as a precursor variable for daytime N02. To allow
a basis for comparison with other studies using NMHC data, the HC69 were
adjusted to be approximately representative of NMHC values. This new
variable, denoted by NMHCPR, is defined by the formula
NMHCPR • HC69 - 10° (11)
with units in pphmC.;
*
A set of field measurements by the California Air Resources Board [3]
arrived at a formula,
NMHC = HC " c35 •
I .00
A regression applied with our data base yields the formula
NMHC = HC"19 .
We chose the constant "100" in Equation (11) to avoid negative values of
NMHCPR (the minimum reported value for HC69 is 100). The constant "2" in
Equation (11) is somewhat arbitrary; one-digit significance is chosen as an
indication of the uncertainty in that constant.
-------
193
10.2.2 Importance of Initial N02
The previous section indicated that NOX69 is an important variable
in explaining daytime peak and average N02. Part of NOX69 consists
of N02 leftover from the previous night. This initial N02 can be
distinguished from the remainder of NOX69,which consists of NO leftover
from the night plus the injection of morning NO emissions. Since it al-
ready starts out as N02, initial N02 may have special significance.
To examine the importance of initial N02, NOX69 was split into two variables
N025, N02 at 5:00 A.M., and INTNO, NOX69 - N025. Multiple linear regressions
were run of the form
DPKN02 = A + B^NOgS + B2 • INTNO . (12)
(or DAVN02)
The results of these regressions are summarized in Table 10.4. The high
t\
values for percentage variance explained (R*) are encouraging. These regressions
indicated that both N025 and INTNO are highly significant (as measured by
the F-statistic). Thus, it seemed important to distinguish initial N02 in
the empirical modeling analysis.
•Most of the regression coefficients (B-j and B2) in Table 10.4 make
sense physically. Initial N02 contributes more to peak N02 than to average
N02, because peak N02 occurs early in the daytime period (i.e., closer to
the time of the N025 measurement). The contribution of INTNO, as measured
by B7, is much greater in summer than winter because photochemistry is
^ *
more active in summer, leading to a greater conversion of morning NO into N02-
The only result in Table 10.4 that seems unreasonable is the fact that
the contribution from N025 (Bi) for the winter peak exceeds 1.0. However,
it is only slightly in excess of unity.
-------
194
The regressions according to Equation (12) also demonstrate that the
constant "A" is substantial. This constant might be considered a measure
of the amount of daytime N02 that is not relatable to N025 or to INTNO, such
as N02 resulting from post 9:00 A.M. emissions.
Table 10.4 Values of A, B^and B2 for Regressions
According to Equation
A B1 Bz
(Constant Term) (NOgS Coefficient) (INTNO Coefficient)
WINTER
DPKN02
DAVN02
SUMMER
DPKN02
DAVNOp
0.75
0.83
0-77
0.80
0.57
0.68
0.60
0.64
0.2 pphm
0.6 pphm
2.5 pphm
2.3 pphm
1.18
0.83
0.78
0.53
0.18
0.11
0.43
0.21
To check whether daytime N02 (peak or average) actually depends on
INTNO in a linear fashion, the contribution of "nonrelatable" N02 (the con-
stant A) and the contribution of initial N02 (B,• N025) were subtracted from
total daytime N02 to yield "residual daytime N02,"
Residual Daytime N02 = Daytime N02 - A - B|-N025. (13)
This was plotted vs. INTNO using the mid-mean graphical technique. The
results for winter are shown in Figure 10.3, and for summer; in Figure 10.4.
These graphs indicate that the dependence of daytime NOp (both peak and average)
on INTNO is essentially linear.
-------
195
20
30 40
INTNO, pphm
50
T
60
70
Figure 10.3 Dependence of Residual Daytime N02
on INTNO (NOX69 - N025), Winter Season
-------
196
O-
Q.
evi
o
ro
3
T3
20 _
15-
10.
5-
I
10
I
20
30
INTNO, pphm
I
40
50
i
o.
0.
CM
O
>o
3
•o
«r-
tn
INTNO, pphm
Figure 10.4 Dependence of Residual Daytime
N02 on INTNO (NOX69 - N025),
Summer Season
-------
197
10.2.3 Dependence of Daytime N02 on Hydrocarbons
The dependence of daytime N02 on hydrocarbons at Downtown Los Angeles
was investigated using all four statistical techniques: graphical analysis,
linear regression, logarithmic regression, and TSC's COMPLIAR program. These
techniques consistently pointed toward several general conclusions:
0 For fixed NOX69 (or for fixed INTNO), hydrocarbons appeared to be
positively related to peak and average daytime N02; i.e., hydro-
carbon reductions would tend to decrease both peak and average
daytime NOp. However, the hydrocarbon dependence was of
secondary importance compared with the NOX dependence.
• The hydrocarbon dependence is greater for peak N02 than for average
N02.
• The hydrocarbon effect appears to be greater in winter than in
summer.
• The hydrocarbon effect is greater at high NOX levels
(NOX69 * 20 pphm) than at low NOX levels.
Some of these conclusions are illustrated in Table 10.5. Table 10.5
lists the hydrocarbon regression coefficient for logarithmic regressions of
daytime N02 vs. NOX69 and HC69. The data are split by season and for
NOX69 < 20 pphm and NOX69 * 20 pphm. The regression coefficients for HC are
greater for DPKN02 than for DAVN02; are higher in winter than in summer; and
are negligible for NOX69 < 20 pphm.
-------
198
Table 10.5 Hydrocarbon Regression Coefficient for Logarithmic Regressions
of Daytime N02 vs. NOX69 and HC69 *
NOX69 < 20
NOX69 > 20
WINTER
DAY PEAK N02
DAY- AVG. N02
0.04
0.04
0.61**
0.45**
SUMMER
DAY PEAK N02
DAY AVG. N02
0.01
-0.03
0. 30**
0.13**
*Actually, the variable (HC69-80) is used for logarithmic regressions.
**Significant at 95% confidence level.
It is encouraging to note that these results agree qualitatively with
recent smog-chamber tests and with expectations based on theoretical argu-
ments. A smog-chamber study of the N02/precursor dependence[4] indicated
that both peak and average N02 were related positively to hydrocarbon input,
that the hydrocarbon dependence was secondary compared with the NOV dependence,
/\
and that the effect of hydrocarbons was relatively greater at higher values
of initial NOX. Physical arguments have been advanced that hydrocarbon re-
ductions would decrease peak N02 more than average N02 [5] and that the
hydrocarbon effect should be greater in winter than in summer[6].
It has also been argued that (for fixed NOX) morning hydrocarbons should be
negatively correlated with N02 levels late in the day. As will be shown later,
this effect is also evident in the aerometric data.
-------
199
For investigating the hydrocarbon dependence quantitatively, it seemed
best to use residual daytime N0£ (as defined by Equation (13) as the depen-
dent variable. The contribution of initial N02 and of N02 not directly re-
latable to 6:00-9:00 A.M. precursors would already be subtracted out. We would,
in effect, be examining the extra N02 brought about by INTNO and NMHCPR.
In Figures 10.5 and 10.6, the effect of hydrocarbons is taken into
account by plotting residual N02 vs. INTNO, with the data stratified by
hydrocarbon (NMHCPR)level. Figure 10.5 is for winter, while Figure 10.6 is
for summer. The vertical distance between the curves represents the impact
of hydrocarbons on daytime NO,,. These results show graphically some of the
conclusions alluded to earlier: The hydrocarbon effect is of secondary impor-
tance, is greater for the daytime peak than the daytime average, and is
greater in winter than in summer.
An alternative way of examining the effect of hydrocarbons is to use
the hydrocarbon-to-N0tf ratio, NMHCPR/NOX69. Figures 10.7 and 10.8 give
A
plots of residual N02 vs. INTNO, with the data stratified by the hydrocarbon-
to-NO ratio. These plots are interesting because they indicate that residual
A
daytime N02 may be proportional to INTNO, with the proportionality constant
depending on the hydrocarbon-to-NOx ratio.
Our hypothesis is that morning hydrocarbons (NMHCPR) impact daytime
N02 by governing the amount of INTNO converted to N02. In effect, the
constant "B2" in Equation (12) depends on hydrocarbons. After consider-
able thought, it was decided that an appropriate way to quantify the
-------
8_
E
CL
Q.
CXJ 6 _
O
z:
Q.
O
Residual
ro -p»
1 1
(
8
6 _
0.
•k
00 .
o 4 -
1
1 2 "
0
0
200
I I 1 1 1 i
245 pphmC
.*
/
.* * «**« *
•" *•*
x 140 pphmC
Aj' v v'A'vX
f\r^\^ 70 pphmc
i i ,i ii i
3 10 20 30 40 50 60 7C
INTNO, pphm
I 1 1 1 1 1
.....•••'' 245 pphmC
/
,<"^140 pphmC
A />>-->:'" .^.//"
v- v'
/ y — • 70 pphmC
1 1 1 1 1 1
10 20 30 40 50 60 /
—
!
)
^4*
•!•
M>|
'0
INTNO, pphm
Figure 10.5 Residual Daytime N02 vs. INTNO at Various
Hydrocarbon Levels, Winter Season
-------
201
20
Q. 15
o.
•o
OJ
10 H
5-4
-•-119 pphmC
55 pphmC
220 pphmC
10
~T ~F
20 30
INTNO, pphm
"T"
40
50
10
8j
a.
a.
fM
O
TO
119 pphmC
10
20 30
INTNO, pphm
220 pphmC
40
Figure 10.6 Residual Daytime N02 vs. INTNO at Various
ngure iu. Hydrocarbon Levels, Summer Season
-------
202
Q.
0.
CM
O
to
3
"O
• r-
l/)
O)
oc
12-
10-
8_
6_
4-
2_
NMHCPR
NOX69
= 8.6
\
10
T
1
20
I
30
INTNO, pphm
40
I
50
60
O.
O.
•«
C\J
O
4 -
t/1
oi
a:
NMHCPR
NOX69
= 2.7
I
10
I
20
I
30
INTNO, pphm
I
40
I
50
60
Figure 10.7 Residual Daytime N02 vs. INTNO at Various
Hydrocarbon-to-NO Ratios, Winter Season
-------
203
15-
10-
5-
NMHCPR
1TOX69" "\10.8
r
20 .25
INTNO, pphm
T~
30
35 40
6-
4-
2-
10.8
^
5.2
V
NHHCP.R *
NOX69 " z-7
r
25
10
15 20
INTNO, pphm
30 35
40
Figure 10.8
Residual Daytime N0? vs. INTNO at Various
Hydrocarbon-to-NOx Ratios, Summer Season
-------
204
hydrocarbon dependence would be to conduct a linear regression,
y = CQ + C]x1 + C2x2, (14)
with y = DPKN02 - A - B] • N025 - B2« INTNO (14a)
(or DAVN02)
as the dependent variable (A, B-,, and B2 taken from Table 10.4) and with
x] = [RATIO - RATIO] • INTNO (14b)
and
*2= [NMHCPR - NMHCPR] • INTNO (14c)
as the two independent variables.* The x, term allows the conversion of
INTNO to depend on the hydrocarbon-to-NO ratio. Noting that NMHCPR is
/\
just RATIO • NOX69, we can see that the x2 term allows the effect of RATIO
to change with the level of NOX69.
The regression according to Equation (14) will yield a constant term,
CQ, and two regression coefficients, C, and C2> The predictive equation for
daytime N02 would then be
DPKN02 = (A + Cfl) + B1 • N025 + INTNO - [B'2 + C,-RATIO + C2 - NMHCPR] ,
(or DAVN02) (15)
where B2 = B2 - C].RATIO - C2 • NMHCPR
As will be shown in Section 10.4, this regression form is convenient for
estimating the effect of precursor control on daytime N02.
RATIO = NMHCPR/NOX69. The "" represents average values.
-------
205
The reader may note that it would seem equivalent to run linear regres-
sion with daytime N02 (peak or average) as the dependent variable and with
N025, INTNO, INTNO • RATIO, and INTNO • NMHCPR as four independent variables.
However, the last three of these independent variables are highly inter-
correlated (partial correlation coefficient = 0.9) because they all involve
the parameter INTNO. Because of this intercorrelation, it would be dan-
gerous to attach physical meanings to the relative sizes of the regression
coefficients. In particular, the existence of a real hydrocarbon effect
from the INTNO • RATIO and INTNO • NMHCPR variables would be in doubt because
these variables are highly correlated with the INTNO variable, which includes
no hydrocarbon dependence. The method we have chosen (Equation (14)) re-
stricts the problem of intercorrelation to only two terms (x, and x2), both
of which involve hydrocarbons.* Thus, although some doubt remains as to
the relative importance of these two terms, we avoid confounding of terms
which involve hydrocarbons with terms which do not involve hydrocarbons.
Table 10.6 presents the results of stepwise regressions according to
Equation (14). For each case (summer vs. winter and peak vs. average), the
F-statisties indicated that both INTNO • RATIO and INTNO • NMHCPR are
significant at a 95% confidence level. The results show a positive
hydrocarbon effect that is greater for DPKN02 than for DAVN02.
It is interesting to note that the variables NMHCPR and RATIO are
not highly intercorrelated (correlation about 0.2). However, when both
variables are multiplied by INTNO, the correlation rises to about 0.9.
-------
Table 10.6 Results of Stepwise Regressions According
to Equation (14) or (15)
Dependent
Variable
WINTER
DPKN02
DAVN02
SUMMER
DPKN02
DAVN02
Total
Correlation
Coefficient
R
0.78
0.84
0.79
0.81
Percentage
Variance
Explained
R2
61%
71%
62%
65%
Constant
Term
A + C0
0.54
0.64
2.81
2.42
N025
Coefficient
Bl
1.18
0.83
0.78
0.53
INTNO
Coefficient
B2
-0.05
0.00
O.T7
0.13
INTNO NMHCPR
INTNU. NQX69
Coefficient
Cl
0.029
0.015
0.027
0.009
INTNO. NM'HCPR
Coefficient
C2
0.00040
0.00018
0.00043
0.00011
*Units of all variables are in pphm. Note that all regression coefficients are significant from zero
at a 95% confidence level.
-------
207
Table 10.6 also shows the percentage variance explained (R2) for the
entire predictive equation. The high percentage variance explained is en-
couraging considering the potential errors in aerometric data and the fact
that transport has been neglected in the analysis.
10.2.4 The Effect of Including Weather Variables
For daytime N02 in Downtown Los Angeles, data are available for seven
meteorological parameters: maximum mixing height (HM), maximum daily tempera-
ture (TM), 9:00-12:00 A.M. wind speed (WS), minimum relative humidity (RH),
7:00-12:00 A.M. solar radiation (SR), pressure gradient from LAX to Palmdale
(PG), and temperature gradient from LAX to Palmdale (TG). By including these
variables in the empirical modeling analysis, an investigation can be made of
the possibility that the observed relationships between the precursor variables
and daytime NO^ are spurious. Spurious relationships could result if a
precursor variable were highly correlated with the weather parameters that
govern the amount of N02 produced from the precursors. In such a case, the
precursor variable might act as a surrogate for the weather variables.
To determine the most important meteorological parameters, logarithmic
regressions were run between daytime N02 and the seven weather variables.
Table 10.7 summarizes the results of these regressions. This table indicates
that there are three key weather variables in winter (HM, TM, and WS), while
there are only two key weather variables in summer (HM and TM). The signs
of the dependencies are as expected, negative for HM and WS and positive for
TM.
-------
208
The logarithmic regressions with weather variables were also run
according to a stepwise procedure. The stepwise regressions produced
similar conclusions; i.e., HM, TM and WS are the three important variables
in winter, while HM and TM are the two important variables in summer.
Table 10.7 Results of Logarithmic Regressions Between Daytime N02 and
Weather Variables
Correlation .Coefficient, R,
for Logarithm of Daytime NO?
Percentage Variance Explained
in Logarithm of Daytime N02,
R2
Meteorological Variables
HM
TM-45
WS
RH
SR
PG+40
T6+20
WINTER
DPKN02 DAVN02
0.72 0.73
52% 53%
SUMMER
DPKN02 DAVN02
0.65 0.55
42% 30%
Logarithmic Regression Coefficients
-0.54** -0.53**
0.58** 0.46**
-0.59** -0.57**
0.08 0.05
0.20* 0.13
0.00 0.00
-0.02 0.05
-0.48** -0.34**
0.66** 0.50**
0.06 -0.08
-0.27* -0.15
0.06 0.03
0.00 0.00
0.00 0.00
*Significant at 99% confidence level
**Significant at 99.99% confidence level
-------
209
An interesting result of the weather regression analysis is that
temperature has very high significance, while solar radiation is of little
significance and wind speed is important only in winter. The importance
of TM can be partially explained by the hypothesis that elevated tempera-
tures enhance the photochemical reactions that convert NO to N0?. It is
also possible that the variable TM encompasses some of the effects of WS
and SR. Table 10.8 lists the linear correlation coefficients among the
meteorological variables. This table shows that TM is negatively correlated
with WS and positively correlated with SR. Thus,TM may partially act as
a surrogate for WS and SR.
It is notable that the explanatory capability of all seven weather
variables combined tends to be less than that of the two precursors, NOX69
and HC69. Table 10.7 indicates that the meteorological variables explain
30% to 52% of the variance in the logarithm of the daytime N02- Logarithmic
regressions of daytime N02 vs. the two precursor variables (as summarized
in Table 10.2) explain 58% to 65% of the variance.* This conclusion was
supported by other types of regressions. The nonlinear regression program,
COMPLIAR, explained about 30% to 40% of the variance in daytime N02 in terms of
the two or three key weather variables. However, COMPLIAR was able to explain
about 60% to 70% of the variance in daytime N02 in terms of N025, INTNO, and
NMHCPR.
that for the logarithmic regressions, the percentage variance
-
oe a
explain SnKN (orv) was about 5% to 10% less than the percen-
tage variance explained in ZnDPKN02 (or *nDAVN02).
-------
210
Table 10.8 Linear Correlation Coefficients Between Weather Variables and
Precursor Variables
WINTER
HM
HM 1.00
TM
WS
RH
SR
TG
PG
NOX69
NMHCPR
TM WS RH SR
-0.00 0.25 -0.26 0.36
1.00 -0.31 -0.54 0.39
1.00 0.10 0.12
1.00 -0.64
1.00
0
0
-0
-0
0
1
TG
.13
.27
.15
.46
.28
.00
-0
-0
0
0
-0
-0
1
PG
.02
.68
.29
.68
.41
.64
.00
NOX69
-0
0
-0
-0
0
0
-0
1
.16
.36
.34
.32
.10
.19
.39
.00
NMHCPR
-0.
0.
-0.
-0.
0.
0.
-0.
0.
1.
32
40
39
21
03
08
34
80
00
SUMMER
HM
HM 1.00
TM
WS
RH
SR
TG
PG
NOX69
NMHCPR
TM WS RH SR
-0.31 0.08 -0.23 -0.03
1.00 -0.20 -0.45 0.38
1.00 -0.02 0.27
1.00 -0.50
1.00
0
-0
0
-0
0
1
TG
.67
.20
.14
.36
.01
.00
-0
-0
0
0
-0
-0
1
PG
.16
.58
.03
.62
.36
.48
.00
NOX69
-0
0
-0
-0
0
0
-0
1
.04
.46
.04
.40
.29
.12
.50
.00
NMHCPR
-0
0
-0
-0
0
0
-0
0
1
.22
.53
.08
.35
.19
.34
.46
.78
.00
-------
211
To investigate whether inclusion of the meteorological parameters in
the empirical modeling analysis would affect the observed relationships
between daytime N02 and the precursor variables (i.e., to check whether the
observed N02/precursor relationships might be spurious), two tests were
used. The first test was based on "weather discounted" dependent variables.
Weather effects were subtracted by defining new dependent variables as
DPKN02/DPKN02 and DAVN02/DAVN02> where DPKN02 and DAVN02 are predicted
values based on various weather regression formulas. In one case, stepwise
logarithmic weather regressions were used to define DPKN02 and DAVN02- This
analysis indicated that N025 and INTNO retained their significance as pre-
cursor variables, but that hydrocarbon variables (NMHCPR or NMHCPR/NOX69)
*
lost their apparent positive effect on N02. In a much more general analysis,
/s. /\
COMPLIAR regressions were used to determine DPKNO/, and DAVN02, and further
COMPLIAR regressions were then run between the weather discounted variables
and the precursors. This general analysis indicated that hydrocarbon
variables, as well as N025 and INTNO, retained their importance.
The second test was to include the significant weather parameters as
independent variables in various regressions that had previously been run
with precursor variables only. It was found that these new regressions,
with weather added, attributed about the same importance to NOX variables
(such as N025, INTNO, or NOX69) but reduced the importance assigned to
There was reason to suspect the method based on the simple logarithmic
regressions. The residuals of the logarithmic weather regressions contained
a strong bias. It is possible that this bias could serve to mask the hydro-
carbon dependence.
-------
212
hydrocarbon variables (such as NMHCPR and NMHCPR/NOX69). For instance, Table
10.9 shows the effect that including weather variables has on the hydro-
carbon coefficients in the linear regression according to Equation (14). it
is apparent that inclusion of weather parameters reduces the size of the
hydrocarbon coefficients (especially.in summer).
All in all, the results of including weather variables in the empiri-
cal modeling analysis are inconclusive. On one hand, it can be contended
that the observed effect of hydrocarbons is partly spurious. The hydro-
carbon effect may be overstated because of intercorrelations between hydro-
carbons and the weather factors governing N02 production, especially TM
(see Table 10.8). On the other hand, the observed hydrocarbon effect may
be real. A plausible argument can be made that including weather factors
in the statistical analysis could mask the actual effect of hydrocarbons.
It is encouraging that the most general method of including weather variables
(using the COMPLIAR program) retained the significance of hydrocarbons.
Perhaps the best use of the analyses with weather factors is to place
a caveat on our results. We will proceed with the empirical model (e.g.,
Equation (15) that was derived without including weather variables. How-
ever, the possibility should be kept in mind that this model may overstate
the relationship between hydrocarbons and N02. This caveat stresses the
need to conduct quantitative checks of the empirical model against smog-
chamber results and against historical air quality trends.
10.3 DEPENDENCE OF NIGHTTIME N02 ON PRECURSORS
The second part of the empirical study for Downtown Los Angeles involves
the dependence of nighttime N02 on precursors. The dependent variables for
the nighttime period are night peak one-hour N02 (NPKN02) and night average
-------
Table 10.9 Effect of Including Weather Variables in the
Linear Regressions According to Equation (15)
Dependent Variable
WINTER
DPKN02
DAVN02
SUMMER
DPKN02
DAVN02
Without Weather
(With Weather)
Without Weather
(With Weather)
Without Weather
(With Weather)
Without Weather
(With Weather)
Ci, Coefficient
1 n£ NMHCPR
UT NOX69
0.029
(0.028)
0.015
(0.013)
0.027
(0.015)
0.009
(0.003)
C2, Coefficient
Of NMHCPR
0.00040
(0.00034)
0.00018
(0.00013)
0.00043
(0.00015)
0.00011
(0.00005)
ro
CO
-------
214
N02 (NAVN02) as defined in Section 9.3. The two basic precursor variables
are: 4:00-7:00 P.M. NOX (NOXPM) and 2:00-4:00 P.M. (O/FT).
The analysis for the nighttime period turned out to be much more straight-
forward than the analysis for the daytime period. The main reason for this
was the insigificant correlation between the nighttime precursor varibles;
the correlation coefficient between NOXPM and OgAFT was only about - 0.07'.
This resulted in a simplification over the daytime case which had involved high
correlations between the independent variables.
Taking a hint from the daytime analysis, we decided to include initial
*
conditions by dividing NOXPM into two parts :
N0216 = N02 at 3:00-4:00 PM
and
NITENO = NOXPM - N0ol6
To investigate the N02/precursor dependence, simple linear regressions were
run of the form
NPKN02 = A + B^NOglG + B2'NITENO + B3«03AFT . (16)
(or NAVN0)
The correlations of 03AFT to both these parts were small.
-------
215
As summarized in Table 10.10, these regressions produced excellent results.
The multiple correlation coefficients ranged from 0.76 to 0.90 (variance
explained = 58% to 81%), and the regression coefficients for all three in-
dependent variables were highly significant. The most significant variable,
as measured by the F-statistic, was N0216.
Table 10.10 Results of Nighttime Regression Analysis According to
Equation (16)*
Dependent Variable
WINTER
NPKN02
NAVN02
SUMMER
NPKN02
NAVN02
Multiple
Correlation
Coefficient
0.90
0.84
0.85
0.76
Percentage
of Variance
Explained
81%
70%
72*
58%
Regression Coefficients
CONSTANT NO,! 6 NITENO OoAFT
A B^ B2 B3
-0.24 0.92
1.23 0.58
0.70 0.88
0.65 0.59
0.29 0.31
0.12 0.16
0.51 0.09
0.38 0.08
*Units of all pollutant variables are 1n pphm.
**A11 three independent variables were highly significant in each case. All
t values were greater than 6, I.e.. ^-statistics weregreater than 36. An F-
statlstlc of 4 Is necessary for a 95* significance level.
-------
216
To obtain a better understanding of the form of the relationships, residual
nighttime N02 was calculated according to
Residual Nighttime N02 = Nighttime N02 - A - B^NOglG (17)
and plotted vs. NITENO. Figures 10.9 and 10.10 illustrate such plots for
winter and summer, respectively. These graphs indicate that the form of
the dependence on NITENO is approximately linear (note that the fluctuations
in the graphs are due to statistical noise).
Figures 10.11 and 10.12 present similar plots stratified by the level
of afternoon ozone. For fixed ozone level, the dependence of residual N02
on NITENO tends to be approximately linear. There is, however, an obvious
shift from one ozone level to the next. To account for a linear dependence
on NITENO that shifts with the ozone level, regressions were run of the form
NPKN02 = A + B^NOgie + NITENO-(B2 + B3-03 AFT) (18)
= A + B1.N0216 + B2-NITENO + B3'NITENO-03 AFT .
These regressions did yield some improvement in percentage variance explained
over Equation (18); the results are summarized in Table 10.11. This par-
ticular regression form will be used in the predictive models formulated in
the final section of this chapter.
There is a potential problem in the regression form represented by
Equation (18). The last two terms are intercorrelated because they both
contain the variable NITENO. However, the intercorrelation is not extremely
-------
217
Q.
Q.
CM
O
«O
&
esj
O
2:
res
3
TO
10 15
NITENO, pphm
20
25
Figure 10.9 Residual Nighttime N02 vs. NITENO,
Winter Season
-------
218
3_
Q-
Q.
CVJ
1 *•
r—
-------
219
0.
0.
CM
O
<0
O)
oc
5J
21
.0,AFT =11.3
»...„
/\ / ^O-AFT = 5.
2.0 pphm
10
15
NITENO, pphm
20
25
CM
O
3 _
/VOAFT = 11.3
03AFT « 5.0
2.0 pphm
25
Figure 10.11 Residual Nighttime NO, vs. NITENO at Various
Afternoon Ozone Levels, Winter Season
-------
220
Q.
OL
CM
I/I
a;
oe
4-
3-
2 -
1 ~
0-AFT - 17.3
OgAFT - 3.8 pphm
I i i
234
NITENO, pphm
3 -
2 -
CNJ
1 _
1/1
'3AFT = 17.3
03AFT«3.8pphm
I I 1
2 3 4
NITENO, pphm
Figure 10.12 Residual Nighttime N02 vs. NITENO at Various
Afternoon Ozone Levels, Summer Season
-------
221
high (about 0.6). Also, the relative importance assigned to the oxidant
term in Equation (18) turns out to be about the same as in Equation (16)
(which did not involve the colinearity difficulty). Thus, the inter-
correlation does not appear to affect the results in Table 10.11 significantly.
The results in Table 10.11 indicate that at least three variables are
important in explaining nighttime N02 concentrations, the initial N02 (N0216),
the remainder of NOXPM (NITENO), and afternoon oxidant (OgAFT). To construct
a model relating nighttime NO? to primary precursors (NOY and NMHC), an
t» A
assumption must be made concerning the dependence of O^AFT on primary precursors.
This assumption will be made in Section 10.4.
Table 10.11 Results of Nightrfetme Regressrron Analysis According to Equation (18)*
Dependent Variable
WINTER
NPKN02
NAVN02
SUWER
NPKN02
NAVN02
Multiple
Correlation
Coefficient
0.91
0.84
0.86
0.77
Percentage
of Variance
Explained
82%
71*
73%
60%
TC
Regression Coefficients
INSTANT N0216 NITENOJ NITENO. OoAFT
A B.J B2 I B3
1.38 0.91 0.09 0.052
2.08 0.57 0.02 0.026
1.51 0-89 0.19 0.036
1.34 0.59 0.13 0.028
*Un1ts of all pollutant variables are 1n pphm.
-------
222
There is one other subtlety in the nighttime analysis, the dependence of
N0216 on the primary precursors. The simplest assumption would be that N0216
is proportional to NOY. However, theoretical reasoning [5] and smog-chamber
/\
evidence [4] indicate that N02 late in the day should bear an inverse rela-
tionship with hydrocarbons or with the hydrocarbon-to-NOX ratio. An in-
verse relationship should exist because hydrocarbon reductions can suppress
the photochemical reactions that consume N02 after it has reached a peak
(see Figure 7.6). This effect might also account for our conclusion that
day peak N02 is more sensitive to hydrocarbon reductions than is day average
N02. To test for this effect, we ran a linear regression for each season
between N0216 and the morning hydrocarbon-to-NOX ratio. The results were as
follows:
Winter: N0216 =9.7 pphm 1- 0.025
(19)
Summer: N0216 = 6.7 pphm I 1- 0.018 ^59
The dependence of N0216 on the hydrocarbon-to-NOx ratio was very significant
(as measured by the F-statistic) for both summer and winter.
It is possible that Equation (19) does not actually represent a causal,
photochemical relationship between afternoon N02 and the hydrocarbon/NOx ratio.
Rather, the observed relationship may be an artifact produced by the positive
correlation which exists between N0216 and NOX69 and, in turn, the negative
correlation which exists between NOX69 and NMHCPR/NOX69. We will assume that
the relationship is causal and will include Equation (19) in the predictive
models developed in the next section. This assumption is not very critical
because the effect represented by Equation (19) is not one of the dominant
aspects of the predictive models.
-------
223
10.4 PREDICTIVE MODELS FOR DOWNTOWN LOS ANGELES
The previous two sections analyzed aerometric data for N02 and its precur-
sors and discussed general conclusions concerning the N0,/precursor dependence.
fm ' ----"--— -*••—•' — — —' " ' —J -
The important precursor variables were identified, and their impact on N02 con-
centrations was illustrated using graphical techniques and regression equations.
The present section develops empirical models which predict the impact that pre-
cursor control would have on N02 concentrations in Downtown Los Angeles. These
models are based on a combination of the regression equations with certain simple
physical assumptions.
The empirical models formulated here are directed toward the question:
If hydrocarbon and NO concentrations in Downtown Los Angeles are changed by
/\
certain amounts, how would peak and average N02 concentrations change? Our
answer to this question implicitly assumes that the general diurnal pattern
of the precursor concentrations is not drastically altered when overall pre-
cursor levels are changed. Also, the empirical models do not address the
question as to how precursor emission changes are related to precursor concen-
tration changes. That is a separate problem which can be answered by diffusion
models for the primary contaminants, or by rollback models if the spatial
distribution of emissions is assumed constant.
It should be emphasized that the empirical models formulated below
are based on N02/precursor relationships observed when weather variables are
not included in the analysis. As noted earlier, the inclusion of weather
variables indicates that the observed hydrocarbon effect might be partially
spurious. Thus, it is possible that the quantitative models presented here
may overstate the real effect of hydrocarbons.
Considering the limitations in our approach and the potential for spurious
relationships in the regression equations, we are, in a sense, stretching our
-------
224
results by formulating predictive models based on the regressions. However,
the reason for extending the results is not to derive a quantitative tool that
is ready for application without qualifications to control strategy analysis.
Rather, the reason is to put our conclusions in a form that can be checked
quantitatively by comparison with historical air quality trends and with pre-
dictions of smog-chamber models, and to provide control guidelines that are
consistent with observations of aerometric data.
10.4.1 Predictive Model for Annual Mean N02
The model for annual mean N02 must be constructed from submodels for the
two seasons and the two times of day. Because of the importance of initial
conditions, these submodels must include linkage between the daytime and
nighttime periods. The following discussion will first deal with each of
the four submodels individually. These submodels will then be synthesized
into a single predictive model for annual mean N0?.
Daytime Average N02, Winter Season
Section 10.2.3 developed regression formulas which indicated the depen-
dence of daytime average NO,, on precursors. The winter regression formula
(summarized in Equation (15) and Table 10.6) included four terms,
DAVNO- = 0.64 + 0.83NOo5 + INTNO 0.015'^^+0.00018 NMHCPR
£ C. \ liV/AO.7 I
J
= 0.64 + 0.83N025 +0.015 INTNO- +0.00018 INTNO-NMHCPR, (20)
with all pollutant variables in units of pphm.
Substituting in average pollutant values for the winter season
(N025 = 6.6 pphm, NOX69 = 30.1 pphm, INTNO = 23.5 pphm, and NMHCPR = 149 pphm),
-------
225
this equation yields
1 II III IV
DAVN02 = 0.64 + 5.48 + 1.74 + 0.63 (21)
8.49 pphm
for the mean value of daytime average N02 in the winter season. As expected,
this is close to the actual winter daytime mean of 8.54 pphm.
A predictive model for daytime average N02 can be formulated by making
assumptions as to how each of the four terms in Equation (21) will change
when the precursors, NOX and hydrocarbons, are controlled. We will make the
following assumptions:
Term I: This is the remainder term that we did not explain in terms of
NOX69 and NMHCPR. It is presumably due to factors (such as postr
9:00 A.M. emissions) that were not accounted for in our analysis.
Fortunately, this is not a large term, and our assumption will
not be critical. We will make the simple assumption that it is
directly proportional to NO control and independent of hydrocarbon
X
control.
Term II: This is the initial N02 term. It will depend on the effect that
precursor control has on overnight N02> Thus, it requires a coup-
ling with the nighttime models. As will be shown later (Table 10.14),
nighttime average N02 is proportional to NOX control and essentially
independent of hydrocarbon control Thus, we conclude that initial
N02 should be proportional to NOX control and independent of hydro-
carbon control.
-------
226
Term III: As indicated by Equation (20), this term involves the precursors in
the form INTNO-NMHCPR/NOX69. The effect of NOX control on this term
should be zero, a cancelling of proportionality between INTNO in the
numerator and NOX69 in the denominator. Thus, this term should be
directly proportional to hydrocarbon control.
Term IV: As indicated by Equation (20), this term involves INTNO-NMHCPR. Thus,
it should be proportional to the product of NO and hydrocarbons.
X
With these assumptions, we can calculate the effect that given amounts of
precursor control will have on daytime average N02 in the winter season. For
instance, assume we control NO by 20% and hydrocarbons by 50%, then the four
A
terms would change as follows:
I: 0.64 x 80% 0.51
II: 5.48 x 80% 4.38
III: 1.74 x 50% 0.87
IV: 0.63 x 80% x 50% 0-25
New Daytime Average = 6.01
Percentage Change = 6'°V *j'49 = -29%
O • ^i/
Using this method, we can calculate the impact that various degrees
of precursor control have on daytime average NOp in the winter season.
Table 10.12 presents the results in terms of percentage changes in winter
daytime average N02- The model should not be used for very large degrees
of control (^ 80% or more) since we would be extrapolating beyond the degree
of variation which we observed in the morning precursor levels. To predict
winter daytime average N02 at various degrees of control, the percentage
reductions in Table 10.12 should be applied to the actual value for daytime
average NOp (8.54 pphm) rather than the computed value (8.49 pphm).
-------
227
Table 10.12 Percentage Changes in Winter Daytime Average N02
at Downtown Los Angeles as a Function of NO and
A
Hydrocarbon Control
CONTfWL
**
r>.
o>
i— •
a*
10
o>
s_
*t-
«/»
O)
O)
f»
(O
^^
o
a*
o>
-------
228
Daytime Average NO,,, Simmer Season
The regression equation for daytime average N02 in the summer season
(summarized in Equation (15) and Table 10.6) was of the form
DAVN02 = 2.42 +0.53N025 + 0.13 INTNO + 0.009 INTNO - +0.0001.1 INTNO-NMHCPR
(22)
with units of all pollutant variables in pphm. Substituting in average
pollutant values for the summer season (N025 =5.3 pphm, NOX69 =18.9 pphm,
INTNO = 13.6 pphm, and NMHCPR = 120 pphm) in Equation (22) yields
I II III IV V
DAVN02 = 2.42 + 2.81 + 1.77 + 0.78 + 0.18 (23)
= 7.96 pphm
This is close to the actual summer daytime mean N02, 7.94 pphm.
To form a predictive model, we make the following assumptions for each
term:
Term I: This term is assumed proportional to NO and independent of hydrocarbons
J\
(see earlier discussion).
Term II: As indicated by coupling with the nighttime model, this term should
be proportional to NO and independent of hydrocarbons (see earlier
J\
discussion).
Term III: This term involves the variable INTNO in Equation (22). it should
also be proportional to NOV and independent of hydrocarbons.
A
Term IV: This term should be proportional to hydrocarbons and independent of
NOX (see earlier discussion).
Term V: This term should be proportional to the product of NO and hydro-
/\
carbons (see earlier discussion).
-------
229
Following the procedures outlined for average N02 in winter, the im-
pact of NOX and hydrocarbon control on summer daytime average N02 was cal-
culated. Table 10.13 presents the results. Again, the effect of NOX is
more important than the effect of hydrocarbons. A 60% NOV reduction (with
A
no hydrocarbon control) decreases summer daytime average N02 by 54%, while
a 60% hydrocarbon reduction (with no NOX control) decreases N02 by only 7%.
Compared with Table 10.12, Table 10.13 indicates that N0¥ control is slightly
A
more important in summer than in winter, and that hydrocarbon control is less
important in summer than in winter.
}
Table 10.13 Percentage Changes in Summer Daytime Average N02
at Downtown Los Angeles as a Function of NO and
»|U f\
Hydrocarbon Control
NO CONTROL
(Percentage Changes from 1969-1974 Level)
01
1—
CT>
ID
_) CT*
HA $••
5 §
£•»
o
0)
O)
to
c
£
' of
+20%
0%
-20%
-40%
-60%
+20% 0%
+21% +2%
+18% 0%
+16% -2%
+13% -5%
+11% -7%
-20%
-16%
-18%
-20%
-23%
-25%
-40%
-34%
-36%
-38%
-41%
-43%
-60%
-52%
-54%
-56%
-58%
-61%
*To calculate summer daytime average N02 levels, these percentage
changes should be applied to 7.94 pphm.
-------
230
Nighttime Average N02, Winter Season
Section 10.3 develops regression equations which indicate the depen-
dence of nighttime average N02 on precursors. The winter regression equa-
tion (summarized by Equation (18) and Table 10.11) included four terms,
NAVN02 = 2.08 + 0.57 N0216 + 0.02 NITENO + 0.026 NITENO • 03AFT (24)
with all pollutant variables in units of pphm.
If we substitute in average values for the pollutant variables in win-
ter (N0216 =8.3 pphm, NITENO = 7.0 pphm, and OgAFT = 4.9 pphm), this
equation yields
I II III IV
NAVN02 = 2.08 + 4.71 + 0.14 + 0.89
= 7.82 pphm.
This calculated value is close to the actual nighttime N02 average in
winter, 7.66 pphm.
A predictive model for nighttime average N02 can be formulated by making
the following assumptions concerning the dependence of each term on the
primary precursors:
Term I: This is the remainder term that we did not relate directly to
the precursors. The assumption for this term is somewhat arbitrary.
We will make the simple assumption that it is directly proportional
to NOV control and independent of hydrocarbon control.
A
Term II: This is the initial N02 term for the nighttime period. As indi-
cated by Equation (19), initial N02 for the nighttime period
bears a slight inverse relation with the hydrocarbon-to-NO ratio.
«
-------
231
Using existing hydrocarbon-to-NOX ratios as a starting point, this
effect can be approximately accounted for by taking this term to be
[HC 1
1 - 0.12^0-J , where NOX and HC represent the control
A\
variables. This formula should only be valid for moderate levels
of control, i.e., NOX between -60% to +60% and HC between -60%
to +60%.
Term III: This term involves the precursor, NITENO. It should be directly
proportional to NO and independent of hydrocarbons.
A
Term IV: As indicated by Equation (24), this term involves NITENO-03AFT.
An assumption is required as to the dependence of OoAFT on the
control variables. This assumption will not be critical since
Term IV is relatively small. We will make the assumption that
OgAFT in Downtown Los Angeles is proportional to the hydrocarbon/
NOV ratio. Thus, Term IV would be directly proportional to hydro-
S\
carbons and independent of NOV (the NOV effect is cancelled by
A A
multiplying NITENO times the hydrocarbon/NOx ratio).
Following procedures outlined previously, the above assumptions can be
used to calculate the effect of precursor control on winter nighttime average
N02. Table 10.14 presents the results. Table 10.14 indicates that changes
in nighttime average N02 are almost directly proportional to NOX control.
Hydrocarbon control is slightly beneficial, but the effect is essentially
negligible. It appears that the N02 decreases that hydrocarbon control
brings in Term IV (through oxidant reductions) are neutralized by the N02
-------
232
increases that hydrocarbon control brings in Term II (through increased
afternoon NCL levels).
Table 10.14 Percentage Changes in Summer Nighttime Average N02
at Downtown Los Angeles as a Function of NOV
A
and Hydrocarbon Control*
1
2
s
J
c
E
t
rca
03
7 +20%
O^
^O
o^
E Q%
o
$ -20%
c
at
5 -40%
OJ
o>
03
§ -60%
$-
-------
233
As was the case with winter nighttime average N02, the summer nighttime
average is almost directly proportional to N0¥ control. Hydrocarbon con-
A
trol yields almost negligible.benefits.
Table 10.15 Percentage Changes in Summer Nighttime Average NO,
at Downtown Los Angeles as a Function of NO and
^\
Hydrocarbon Control*
r^"* <^v
§! (Percentage Changes from 1969-1974 Level)
— i
«st"
Ol
oj +20%
3 2
C "^
I 1 0%
3 M-
c «»
i | -20%
10
CJ
1 "40%
c
1 § -60%
+20% 0%
+20% +1%
+19% 0%
+18% -1%
+17% -2%
+17% -3%
r a?
-20%
-18%
-19%
-20%
-21%
-22%
-40%
-37%
-38%
-39%
-40%
-41%
.-60%
-57%
-57%
-58%
-59%
-60%
*To calculate summer nighttime average NO, levels, these
percentage changes should be applied to 5
PP
nrn.
Annual Average N02
To arrive at an empirical model for annual average N02, we must combine
the results for daytime and nighttime and for the two seasons. For each
level of precursor control, the predicted N02 level for each season and
-------
234
time of day is calculated by applying the percentage reductions listed
in Tables 10.12 through 10.15 to the existing N02 average (1969-1974) for
that season and time of day. The annual average is then computed according
to the tautology
Annual Average 5 |[^DWA + ^NWAJ + -^DSA + ||NSAJ , (25)
where
DMA = daytime winter average N02
NWA = nighttime winter average NOp
DSA = daytime summer average N02
NSA = nighttime summer average N02
The weights of one-half are used for the two seasons, because each
season represents six months. The 10/24 and 14/24 weights are used for daytime
and nighttime, respectively, because the daytime average represents 10 hours
while the nighttime average represents 14 hours.
The results for annual average N02 at Downtown Los Angeles are summarized
in Table 10.16. Table 10.16a lists percentage changes in annual average N02,
while Table 10.16b lists predicted annual average N02 levels. These results
indicate that changes in annual average N02 will be almost directly propor-
tional to NO changes, with a slight beneficial impact due to hydrocarbon
A
reductions. The relationship to NO is not exactly proportional because
A
NO reductions would have the side effect of increasing the HC/NOV ratio.
x *
To attain the federal air quality standard for annual average N02 would
require approximately a 31% reduction in NO levels if hydrocarbons remained
A
constant. If hydrocarbons were reduced by 60%, only a 23% reduction in NOX
levels would be required for attainment.
-------
235
As indicated by Tables 10.12 through 10.15, the beneficial impact
of hydrocarbon control would be accrued almost entirely during the day-
time period. A 60% reduction in hydrocarbons (with no NO control) would
y\
result in a 12% decrease in daytime average N02 but only a 2% decrease in
nighttime average N02. The impact of hydrocarbon control would also be
significantly greater for daytime average N02 in winter than for daytime
average N02 in summer.
Table 10.T6 The Effect of NOY and Hydrocarbon Control on
A.
Annual Average N02 at Downtown Los Angeles
Table 10.16a Percentage Changes in Annual Average N02
J
3
£
3^
5
j
j
c
>
?
r
.>
(U
_J
7 +20*
us
o>
E 0%
2
1-
S -20%
C
10
£••
« -40%
«*
§ -60%
V
Q-
I1W krVIV 1 I\U1. " ' .1 .-^,.
(Percentage Changes from 1969-1974 Level)
+20% 0% -20% -40% -60%
+21% + 2% -16% -34% -53%
+18% 0% -18% -36% -55%
+16% - 2% -20% -38% -57%
+13% - 5% -23% -41* -59%
+11% - 7% -25%, -43% -61%
-------
236
Table 10.16b Annual Average N02 Levels, pphm
p (Pi
O)
(U
B +20%
Ji
i
> CT>
"* VO
i * o%
3 §
^ £
1 " -20%
0)
O)
c
(O
5 -40%
Ol
o>
fO
1 -60%
w
Q_
HU LUMIKUL ^*
srcentage Changes from 1969-1974 Level)
+20% 0% -20% -40% -60%
8.9 7.5 6.2 4.8* 3.5*
8.7 7.3 6.0 4.7* 3.3*
8.5 7.2 5.8 4.5* 3.2*
8.3 7.0 5.7 4.4* 3.0*
8.2 6.8 5.5 4.2* 2,9*
*Attains federal standard of 5.3 pphm (100 yg/m )
10.4.2 Predictive Model for Yearly One-Hour Maximum
In examining yearly one-hour maximum N02> it appears sufficient to
restrict the analysis to the daytime period. Table 10.17 lists the days
in our processed data base with the five highest one-hour N02 concentrations
for winter and summer, and for daytime and nighttime. It is evident that
the most extreme one-hour levels of NOp tend to occur during the daytime
period.
-------
237
Table 10.17 Days in the Processed Data Base with Extreme One-Hour N02
Levels in Downtown Los Angeles (1969 to 1974)
DAYTIME
Winter
Summer
74 pphm (3/29/71)
58 pphm (10/16/73)
56 pphm (11/9/71)
55 pphm (11/24/71)
66 pphm (5/15/70)
57 pphm (5/16/70)
53 pphm (5/14/70)
50 pphm (7/4/70)
52 pphm (10/17/73) 45 pphm (9/8/71)
NIGHTTIME
Winter
Summer
46 pphm (1/17/71) 31 pphm (5/14/70)
45 pphm (1/19/71) 21 pphm (9/28/73)
43 pphm (1/18/71) 20 pphm (4/1/70)
41 pphm (1/31/71) 20 pphm (7/8/70)
40 pphm (2/11/71) 20 pphm (5/15/70)
It also appears that that summer daytime maxima are slightly smaller
than the winter daytime maxima. An examination of all the data for Downtown
Los Angeles (a larger data set than our processed data base), indicates that
the typical winter maximum for the 1969-1974 period is around 60 pphm, while
the typical summer maximum is. around 50 pphm. Since the summer maximum is
not small compared with the winter maximum, our analysis for yearly maximum
one-hour N02 should consider both the summer and winter daytime periods.
In formulating predictive models for yearly maximum one-hour N02 concen-
trations, procedures were followed analogous to those used for annual mean
concentrations. Since the analysis for the yearly maximum is restricted to
the daytime period, the appropriate regression formulas are given by Equation
(15) and Table 10.6. One new problem arose in the analysis of yearly maxima.
The regression formulas for daytime peak N02 are actually applicable only to
average conditions; the formulas are based on all days in the data base. In-
sufficient data were available to develop separate regression formulas for
the few days with extreme N02 concentrations. The most realistic use of the
regression formulas would be to predict seasonal averages of daily maxima, not
-------
230
yearly one-hour maxima. When applied to days of extreme N02 levels, the re-
gression equations tended to under predict the one-hour maxima by as much as 40X.
To circumvent this problem, the predictive model was formulated with the
regression formulas by entering the precursor levels (N025, INTNO, and
NMHCPR) associated with the most extreme days of N02 concentrations. The
percentage change indicated by this model was then applied to the actual one-
hour maximum (60 pphm in winter and 50 pphm in summer).
The results of the predictive models for yearly maximal N02 concentra-
tions in summer and winter are presented in Tables 10.18 and 10.19. These
tables show that the effect of hydrocarbon control on maximal N02 levels
in Downtown Los Angeles is almost as great as the effect of NOV control.
A
Hydrocarbon control has a slightly greater impact in winter than in summer.
Applying the percentage changes in Tables 10.18 and 10.19 to the typical winter
and summer maxima (60 pphm and 50 pphm, respectively) indicates that the
yearly maximum would tend to occur in winter for all degrees of control
listed in the tables. Thus, Table 10.18 (the winter case) can be considered
as representative of the one-hour maximum for the entire year.*
Table 10.20 lists predicted values for yearly one-hour N02 maxima as
a function of NOX and hydrocarbon control This table has been derived by
applying the percentage changes in Table 10.18 to the typical winter maximum
In years with unusual meteorology, the yearly maximum may actually occur
in summer. Rather than complicate the predictive model, we will neglect this
possibility and deal with the winter maximum only.
-------
239
Table 10.18 Percentage Changes In Winter Yearly Peak One-Hour N02 as
Function of NO and Hydrocarbon Control
A.
0)
i 2
c
^* ^«
"* £
•* ^3
3 £
J) (/)
E g>
s c
10
0
7 +20%
O
^O
O^
§ 0%
i-
S -20%
£Z
1C
^ -40%
fjt
(O
§ -60%
n\j uuit i rvuu •• — w~
>ercentage Changes from 1969-1974 Level)
+20% 0% -20% -40% -60%
+24% +8% - 9% -25% -42%
+16% 0% -16% -32% -47%
7% . 8% -23% -38% -53%
. n -15% -30% -44% -59%
. 9% -23% -37% -50% -64%
Ol
Q.
-------
240
Table 10.20 Yearly One-Hour Maximum NOp Levels in Downtown Los Angeles
as a Function of Hydrocarbon and NO Control (All Values pphm)
rt
NO.. CONTROL
gj (Percentage Changes frosi 1969-1974 Level)
_l
Oi
1
cr>
• 1Q
3 CTl
i i
C ">
r OJ
? 0)
C
ro
O
(U
C7/
fO
-M
0)
0
r ^5
r\
+20%
0%
-20%
-40%
-60%
+20%
76
69
62
55
49
0%
66
60
54
48
42
-20%
57
51
46
40
35
-40%
47
42
37
32
27
-60%
38
33
29
25*
20*
*Attains the California one-hour N02 standard of 25 pphm
(60 pphm) for the 1969-1974 period. Table 10.20 indicates that hydrocarbon
control would be nearly as important as NO control for attaining the one-
J\
hour California N02 standard in Downtown Los Angeles.
The significance of hydrocarbon control for yearly maximum one-hour
N02 in Downtown Los Angeles is somewhat surprising. However, it is actually not
implausible from a physical viewpoint. Reducing the hydrocarbon- to-NO
X
ratio should delay the formation of maximal NOo. This delay is particularly
important in Downtown Los Angeles because dispersive conditions become
much stronger late in the morning as the sea breeze establishes and the
mixing height elevates.
-------
241
Because of the uncertainties in our analysis, it will be important to
check the empirical models quantitatively against smog-chamber results and
against historical air quality trends. These checks will be conducted in
Chapter 12 of this report. Qualitatively, the air quality trends discussed
in Part I of this study provide reason for encouragement. Part I demonstrates
that maximal N02 levels in central/coastal Los Angeles decreased slightly in
the past decade even though NOX emissions and ambient NOX levels increased
significantly. This could be the result of the hydrocarbon control that has
been achieved in the central/coastal parts of Los Angeles.
One further remark should be made in regard to the predictive models
summarized by Tables 10.12 through 10.20. These tables list the changes in
ambient N02 that should result from NOX and hydrocarbon control, but they do
not quantify the errors in the predictions. Based on the statistical errors
in the regression coefficients which underly the models, it would be possible
to compute error bounds. However, these statistical error bounds would have lit-
tle meaning because they would not be representative of the conceptual limita-
tions inherent in the models. As noted in Section 7.2.4, these limitations
include the neglect of transport, the omission of meteorology, and the as-
sumption that precursor changes produced mostly by meteorology can be used to
model the effect of control strategies. It is not possible to quantify the
potential errors that arise because of these fundamental limitations in the
models.
-------
242
10.5 REFERENCES
1. W. S. Cleveland, B. Kleiner, and J. L. Warner, "Robust Statistical Methods
and Photochemical Air Pollution Data," Journal of the Air Pollution Control
Association. Vol. 26, p. 36, 1976.
2. L. R. Reckner, "Survey of Users of the EPA Reference Method for Measurement
of Nonmethane Hydrocarbons," EPA-650/4-75-008, December 1974.
3. F. Bonamassa and H. Mayrsohn, "Distribution of Hydrocarbons in the Los
Angeles Atmosphere, Aug.-Oct. 1971," California Air Resources Board,
November 1971.
4. H. Jeffries, D. Fox, and R. Kamens, "Outdoor Smog Chamber Studies: Effect
of Hydrocarbon Reduction on Nitrogen Dioxide," prepared for EPA Office of
Research and Development by University of North Carolina, EPA-650/3-75-011,
June 1975.
5. E. R. Stephens, "Proceedings of the Conference on Health Effects of Air
Pollution," U.S. Senate Committee on Public Works, U.S. Government
Printing Office Stock No. 5270-02105, 1973.
6. E. L. Meyer, Jr., EPA Office of A1r Quality Planning and Standards,
personal communication, August 1976.
-------
243
11.0 EMPIRICAL MODELS APPLIED TO VARIOUS CITIES
Chapter 10 of this report formulates statistical models of the
N02/precursor relationship at Downtown Los Angeles. The investigation
results in an empirical control model based on a series of linear regression
equations and on certain simple physical assumptions. This chapter uses
the same procedure to derive empirical control models for 7 other lo-
cations: Lennox (CA), Azusa (CA), Pomona (CA), Denver, Chicago,
Houston/Mae, and Houston/Aldine.
11.1 GENERAL METHODOLOGY
The empirical N02 control model for Downtown Los Angeles was based
on regression equations for daytime N02 (Equations (12), (14), and (15)),
regression equations for nighttime N02 (Equations (18) and (19))> and simple
physical assumptions which transformed these equations into predictive con-
trol models (Section 10.4). The exact procedure is used here to derive
empirical control models for 7 other cities. This section provides a brief
summary of that procedure. The reader is referred to Chapters 9 and 10 for
more detailed descriptions of the procedures and for exact definitions of
\
the variables used.
The empirical control models for all 10 cities are based on regres-
sion equations which do not explicitly include wpather variants. As
noted in Chapter 10, the inclusion of weather variables raises questions
as to whether the observed dependence of N02 hydrocarbons is real or
whether it is partially an artifact produced by unaccounted for weather
-------
244
variables. This stresses the need to check the results of the empirical
models against historical air quality trends and smog-chamber experiments
These checks will be performed in subsequent chapters.
11.1.1 Regression Equations for Daytime NO,,
The dependent variables for the daytime analysis are daytime peak
one-hour N02 (DPKN02) and daytime average N02 (DAVN02). Regressions are
run (separately for DPKN02 and DAVN02, separately for winter and summer)
of the form:
DPKN02 = A + B^NOgS + Bg-INTNO . (26)
(or DAVN02)
The B-j term represents the contribution of early -morning N02 carried over from
the previous night. The Bg term represents the contribution from the con-
version of NO (both carry-over NO and early morning NO emissions). The
constant (usually small) represents the contribution from other factors,
such as late-morning NO emissions.
It is assumed that hydrocarbons (NMHCPR) affect daytime N02 by
governing the amount of INTNO converted to N02< The effect of hydrocarbons
is estimated by performing a stepwise regression,
y = CQ + C1X1 + C2X2 , (27)
where
y = DPKN02 - A - B] • N025 - B2« INTNO,
(or DAVN02)
X] = [RATIO - RATH)"] • INTNO,
X2 = [NMHCPR - NMHCPR] . INTNO,
RATIO = NMHCPR/NOX69,
and "_ _ " = average values.
-------
245
This results in a final equation of the form
DPKN02 = (A + CQ)+ BrN025 + INTNO-[B2 + CrRATIO + C2-NMHCPR] , (28)
(or DAVN02)
where BZ = B2 - Cr RATIO - C^NMHCPR.
11.1.2 Regression Equations for Nighttime NO,,
The basic equations for nighttime N02 are obtained by stepwise regres-
sions (separately for NPKN02 and NAVN02, separately for winter and summer)
of the form
NPKN02 = A + B1-N0216 + B2-NITENO + B3-NITENO-03AFT . (29)
(or NAVN02)
The second term indicates the contribution of N02 carried over from the
afternoon. The third and fourth terms represent the contribution of NO
(carried over from the afternoon or emitted during the early evening). The
conversion of NO to N02 is allowed to depend on the afternoon oxidant level
(fourth term). The constant, A, represents contributions from other factors,
such as nighttime NO emissions.
Since the afternoon N02 level (N0216) may depend on early-morning
hydrocarbons, a regression (separately for winter and summer) is also run
of the form
N0216 = DQ + D] • RATIO, (30)
where RATIO is the morning NMHCPR/NOX69 level.
11.1.3 Empirical Control Models
The regression equations are transformed into empirical control models
by adding certain physical assumptions. For the daytime models, based on
-------
246
Equation (28), the assumptions are listed in Table 11.1
Table 11.1 Assumptions to Convert Equation (28) into a
Control Model for Daytime NO/,
Term in
Equation (28)
Control Assumption
Remarks
N025
• INTNO
C1 • RATIO • INTNO
NMHCPR . INTNO
Proportional to NOX
Independent of HC
Proportional to NOX
Independent of HC
Proportional to NOX
Independent of HC
Independent of NOX
Proportional to HC
Proportional to the
product of NOY and HC
J\
This is the simplest as-
sumption to make. Fortun-
ately, this assumption is
usually not critical.
This assumption is supported
by the models for nighttime
average N02.
The effect of NO/ control is
cancelled by proportionality
between INTNO and NOX69 (de-
nominator of RATIO).
The nighttime model is formed by inserting Equation (30) into
Equation (29). The assumptions which transform the equation into a
control model are listed in Table 11.2.
-------
247
Table 11.2 Assumptions to Convert Equation (29) into a
Control Model for Nighttime N(L
Term in
Equation (29)
Control Assumption
Remarks
VD0
B1 • D1 • RATIO
NITENO
NITENO
' °3AFT
Proportional to NO,
Independent of HC
Proportional to NOX
Independent of HC
Independent of NOX
Proportional to HC
Proportional to NOX
Independent of HC
Independent of NOX
Proportional to HC
This is the simplest assump-
tion to make; it is usually
not critical
This term is obtained by sub-
stituting Equation (30)
into Equation (29). The
parameter Dn should be
directly proportional to NO .
/\
This term is obtained by sub-
stituting Equation (30) into
Equation (29). The parameter
DI should be directly propor-
tional to NOX.
It is implicitly assumed that
OsAFT is porportional to the
HC/NOx ratio. This assumption
should be approximately true
for many central-city locations.
However, it may not hold for
Houston or for downwind sites
in Los Angeles (e.g., Azusa
or Pomona). Fortunately, this
assumption is usually not critical
11.2 CONTROL MODELS FOR VARIOUS CITIES
Using the procedures outlined in the previous section and in Chapter 10,
this section formulates empirical control models for 7 cities. For each
location, a model is developed for annual mean N02 and yearly peak one-hour
N02. The model for annual mean N02 involves synthesis of four submodels,
for daytime average N02 and nighttime average N02 in both winter and summer.
-------
248
The synthesis is based on Equation (25), page 234. The model for yearly
peak one-hour N02 is developed by using the regression equation for DPKN02
corresponding to the season and time of day when the yearly peak occurs. "Worst-
case" conditions are used in the regression equation for DPKN02.
The following discussions deal only with the resultant control models
for each city. The regression equations which serve as the foundation of the
control models are presented in Appendix D.
11.2.1 Lennox, California
The Lennox monitoring site is located about eleven miles southeast of
Downtown Los Angeles and three miles from the coastline. Like Downtown
Los Angeles, Lennox is within the area of high emission density that spreads
over the central/coastal parts of the Los Angeles basin. However, Lennox
is in the upwind part of the source-intensive area, while Downtown Los
Angeles is in the center of the area.
The empirical control model for annual mean N02 at Lennox is summarized
in Table 11.3. Percentage changes in annual average N02 at Lennox are listed
for various changes in NO and NMHC concentrations at Lennox. Also presented
J\
are predicted annual average N02 concentrations (Table 11.3b). As was the
case for Downtown Los Angeles (Section 10.4.1), annual mean N02 at Lennox
is essentially directly proportional to NOV, with minute benefits accrued
A
from NMHC control. Attainment of the federal annual mean standard for N02
must be accomplished through NOV control.
A
The submodels for annual mean N02 at Lennox indicate that nearly all
the benefit from NMHC control occurs in the daytime N02 average rather than
the nighttime N02 average. Also, the effect of hydrocarbons is greater
for winter daytime N02 than summer daytime N02< These patterns are totally
consistent with the corresponding submodels for Downtown Los Angeles.
-------
249
Table 11.3 The Effect of NO and Hydrocarbon Control on
Annual Mean N02 at Lennox
Table 11.3a Percentage Change in Annual Mean NCL
DC
O
0
a:
3S.
Z
§
?
1
L>
£
0)
01
CO
o>
VO
2
1
V)
2
N0x
-20%
-18%
-19%
-20%
-21%
-22%
Levels,
CONTROL
(Percentage Changes from
+20%
0%
-20%
-40%
-60%
+20%
7.7
7.7
7.6
7.5
7.6
•
0%
6.5
6.4
6.4
6.3
6.2
-20%
5.3*
jl
5.2
5.1*
5.1*
it
5.0
-40%
-37%
-38%
-39%
-40%
-41%
pphm
1969-1974
-40%
4.1*
^
4.0
*
3.9
3.9*
it
3.8
-60%
-56%
-57%
-58%
-59%
-60%
.
Level)
-60%
2.8*
*
2.8
2.7*
2.6*
*
2.6
*Atta1ns federal standard of 5.3 pphm (100 yg/m3)
-------
250
The yearly maximum one-hour N02 concentration at Lennox occurs almost
invariably in the daytime during the winter. The highest N02 concentra-
tions in the winter are approximately 20% greater than summer peaks, and
the highest daytime concentrations are about 30% greater than the night-
time peaks. Thus, the control model for peak winter daytime N02 represents
the control model for yearly maximum N02.
The control model for yearly maximum N02 is summarized in Table 11.4.
This model has been derived by a procedure entirely parallel to the analysis
for Downtown Los Angeles (Section 10.4.2). The control model indicates
that maximal N0? concentrations are slightly less than proportional to NO
C. A
concentrations at Lennox. Moderate improvement in maximal N02 can be
gained from hydrocarbon control. The benefit of hydrocarbon control on
yearly maximum NO,, appears to be considerably less at Lennox than at Down-
town Los Angeles (Tables 10.18 and 10.20).
11.2.2 Azusa. California
The Azusa monitoring site is located about 21 miles ENE of Downtown
Los Angeles. Azusa is on the northeast fringe of the area of high emission
density which spreads over the central and coastal parts of the Los Angeles
region. As such, Azusa can be regarded as a downwind receptor site in
the Los Angeles basin.
Table 11.5 summarizes the empirical control model for annual mean NO,,
at Azusa. As was the case with Downtown Los Angeles and Lennox, annual
mean N02 is essentially proportional to NO , with very small benefits
resulting from hydrocarbon control. Attaining the federal annual mean
standard for N02 must depend on NO control.
-------
251
Table 11.4 The Effects of NOv and Hydrocarbon Control on Yearly
Maximum One-Hour N02 at Lennox
Table 11.4a Percentage Changes in Yearly Maximum N02
O
0
QJ
—1
•a-
r*.
Ol
i
1
M-
1
-C
O
OJ
en
I
»
01
r—
•o
OH
1
01
en
10
St.
(Percentat
0%
+4%
0%
-4%
-8%
-12%
-20%
-12%
-16%
-20%
-24%
-28%
Maximum N0£
NOY CONTRO
^v
(Percentage Changes
+20%
0%
-20%
-40%
-60%
+20%
49
47
46
44
43
0%
42
41
39
38
36
from
-20%
36
34
33
31
29
fL
m 1969-1974 Level)
-40%
-29%
-32%
-36%
-40%
-44%
-60%
-45%
-49%
-52%
-56%
-60%
Concentrations, pphm
L-
1969-1974
-40%
29
28
26
*
24
23*
Level )
-60%
23*
21*
*
19
*
18
16*
*Atta1ns the California one-hour standard (25 pphm)
-------
252
Table 11.5 The Effect of NOX and Hydrocarbon Control on Annual
Mean N02 at Azusa
Table 11.5a Percentage Changes in Annual Mean N02
n
-v (Percent Changes from 1969-1974 Level)
C*
o
11
!
3
«>
01
o>
CTl
1
VO
CTl
0)
C
<0
O
C
u
t-
OJ
Q.
"ol
o>
s
CTl
s-
0)
en
C
.C
(Percentage C
+20%
0%
-20%
-40%
-60%
+20%
+20%
+19%
+19%
+18%
+17%
0%
+1%
0%
-1%
-2%
-3%
Table 11. 5b Annual Mean
N0x
-20%
-19%
-19%
-20%
-21%
-22%
N02 Le
CONTROL
(Percentage Changes from
+20%
0%
-20%
-40%
-60%
+20%
7.3
7.3
7.2
7.2
7.1
0%
6.2
6.1
6.1
6fO
5.9
-20%
5.0*
4.9*
4.9*
4.8*
4.8*
-40%
-38%
-39%
-40%
-40%
-41%
vels, pphm
1969-1974 Level
-40%
3.8*
3.7*
3.7*
3.6*
3.6*
-60%
-57%
-58%
-59%
-60%
-60%
....
)
-60%
2.6*
2.6*
2.5*
2.5*
2.4*
Attains federal standard of 5.3 pphm (100 yg/m3}
-------
253
The submodels for annual mean N02 at Azusa indicate that the maximum
benefit from hydrocarbon control is attained in daytime average N02 during
the winter. This is consistent with the results for Downtown Los Angeles
and Lennox. Nighttime average N02 in the summer is also somewhat sensitive
to hydrocarbon control. This could mean that oxidant is especially signifi-
cant to nighttime N02 in the case of Azusa; oxidant affects the amount of
evening NO converted to NOp.
Yearly maximum one-hour N02 at Azusa invariably occurs during the
winter season. The yearly peak is slightly more likely to occur in the
nighttime period than in the daytime period. Thus, empirical control models
of yearly maximum N02 at Azusa were completed for both winter daytime
conditions and winter nighttime conditions. These results are presented
in Table 11.6.
Table 11.6 indicates that yearly maximum N02 in both the daytime and
nighttime periods is essentially proportional to NOX, with moderate effects
occurring from hydrocarbon control. The benefit of hydrocarbon control is
greater for the daytime peak than for the nighttime peak. For virtually all
degrees of control listed in Table 11.6, the yearly maximum will be more
likely to occur in the nighttime period than in the daytime period. Thus,
the nighttime case (Table 11.6b) is used as the control model for yearly
maximum N02. Predicted yearly maxima as a function of hydrocarbon and NOX
control are listed in Table 11.7.
11.2.3 Pomona. California
Pomona is located approximately 30 miles east of Downtown Los Angeles.
Under the prevailing daytime wind flow, Pomona is downwind of the source-
intensive, central/coastal parts of the basin.
-------
254
Table 11.6 The Effect of NOX and Hydrocarbon Control
on Yearly Maximum NC^ at Azusa
Table 11.6a Percentage Changes in Winter Daytime Peak
1
o
o
o
o
i
u
L>
i
"ol
1
CT>
VO
CTi
§
Il-
l/I
C
(O
f""
• V
o
g,
(Percent*
r™
0)
_l
CTl
1
U3
O>
14-
tn
0)
c
fmm
O
Ol
C7)
K)
C
U
s.
0]
Q.
(Percentage Changes from 1969-1974 Level)
+20%
0%
-20%
-40%
-60%
+20%
+26%
+19%
+11%
+4%
-3%
0%
+6%
0%
-6%
-12%
-18%
-20%
-13%
-19%
-24%
-29%
-34%
Table 11. 6b Percentage Changes
NOX CONTROL
(Percentage Chanaes from
+20%
0%
-20%
-40%
-60%
+20%
+20%
+17%
+16%
+14%
+11%
0%
+2%
0%
-2%
-4%
-7%
-20%
-16%
-18%
-20%
-22%
-24%
-40%
-33%
-37%
-41%
-46%
-50%
in Winter Ni
1969-1974 Level
-40%
-34%
-36%
-38%
-40%
-42%
-60%
-52%
-56%
-59%
-62%
-66%
ghttime Peak
,
-60%
-51%
-54%
-56%
-58%
-60%
-------
255
r
Table 11.7. Predicted Yearly Maximum NO? Concentrations (pphm)
at Azusa as a Function of NOX and
Hydrocarbon Control
o
ce.
o
o
iw
^J (Percentage Changes
O!
i
i—
i
^
<£ +20%
E
| 0%
| -20%
(O
JC
s, "40%
(rt
S -60%
+20%
47
46
45
45
44
o
S-
-------
256
Table 11.8 presents the empirical control model for annual mean N02
at Pomona. Following the pattern at the other Los Angeles sites, annual
mean N02 at Pomona is almost directly proportional to NOX control, with
slight benefits provided by hydrocarbon control. Again, consistent with
the other sites, the submodels indicate that the greatest benefit from
hydrocarbon control is accrued in the daytime period during the winter.
Yearly maximum one-hour N02 concentrations at Pomona occur almost
invariably in the nighttime period during the winter. Thus, the appro-
prite submodel for yearly maximum N02 is the nighttime peak model for the
winter. Table 11.9 presents the resulting empirical control model for
yearly maximum N02 at Pomona. Hydrocarbon control apparently yields sig-
nificant reductions in the winter nighttime maximum at Pomona and is about
two-thirds as important as NOX control. The regression model indicates that
oxidant is an important determinant of the nighttime N02 maximum at Pomona.
The benefit from hydrocarbon control occurs because hydrocarbon reductions
serve to decrease oxidant.
11.2.4 Denver, Colorado
The Denver CAMP site is a "center-city" monitoring site located in
downtown Denver. Table 11.10 presents the empirical control model for
annual mean N02 at the Denver CAMP site. As was the case with the 4
Los Angeles sites, annual mean N02 at Denver is approximately proportional
to NO concentrations. However, contrary to the results for Los Angeles,
7\
hydrocarbon control tends to produce slight increases in annual mean N02 at
Denver.
The submodels for average N02 at Denver indicate that the main dis-
advantages from hydrocarbon control occur during the winter (in both the
daytime and nighttime periods). At all sites which have been examined,
-------
257
Table 11.8 The Effect of NOX and Hydrocarbon Control on Annual
Mean N02 at Pomona
Table 11.8a Percentage Changes in Annual Mean N0?
g£
oc
o
o
o
jg
^y
8
E
B
£
i
r— •
>
3
en
o>
e
o
M-
!/)
01
ID
x:
o
0)
en
CO
c
Ol
o
i-
<1)
o>
c
ro
O
Ol
CD
o>
0
s.
d)
CL
(Percentage Changes from 1969-1974
+20%
0%
-20%
-40%
-60%
+20%
+21%
+17%
+12%
+8%
+3%
Table 11. Sb
0%
+4%
0%
-4%
-9%
-13%
Annual-
-20%
-13%
-17%
-21%
-25%
-30%
Mean 'N02
NOX CONTROL
-40%
-30%
-34%
-38%
-41%
-45%
Levels,
(Percentage Changes from 1969-1974
+20%
0%
-20%
-40%
-60%
+20%
9.2
8.8
8.5
8.2
7.8
Attaii
0%
7.9
7.6
7.2
6.9
6.6
-20%
6.6
6.3
6.0
5.7
5.4
-40%
"A*
5.3
*
5.0
#
4.7
*
4.4
*
4.1
Level )
-60%
-47%
-51%
-54%
-58%
-61%
•
pphm
Level )
-60%
*
4.0
*
3.7
*
3.5
*
3.2
*
2.9
is federal annual mean standard of 5.3 pp
-------
258
Table 11.9 The Effect of N0« and Hydrocarbon Control on
Yearly Maximum N02 at Pomona
Table 11.9a Percentage Changes in Yearly Maximum N02
qj" (Percentage Changes from 1969-1974 Level)
|
o
O
Sp»
^pl
U
1
| +20%
| 0%
| -20%
ia
.c
a -40%
CD
£ -60%
+20%
+20%
+12%
+5%
-3%
-11%
u
s-
O)
o.
Table 11. 9b
'fi-
0%
+8%
0%
-8%
-15%
-23%
-20%
-5%
-12%
-20%
-28%
-35%
Yearly Maximum NO
N0x
JS (Percentage Changes
Ef
E
B
L>
E
P
i
at
en
°? +20%
c
i: 0%
tn
0)
g> -20%
IO
j*j
<„ -40%
0)
ID
4J
g -60%
U
+20%
49
46
42
39
36
S-
01
Q.
*
0%
44
41
37
34
31
CONTROL -
-40%
-17%
-25%
-32%
-40%
-48%
-60%
-30%
-37%
-45%
-52%
-60%
2 Concentrations, pphm
from 1969-1974 Level)
-20%
39
35
32
29
26
-40%
34
30
27
*
24
*
21
-60%
29
25*
-------
259
Table 11.10 The Effect of NOv and Hydrocarbon Control
on Annual Mean NO? at Denver
Table ll.lOa Percentage Changes in Annual Mean N02 Levels
o
1—
o
o
z
1
D
L>
L>
1
>
0)
CTl
1
CTl
2
|
-------
260
it was found that hydrocarbon reductions tend to increase afternoon N02.
This effect appears to be especially important at Denver during the winter.
The increased afternoon N02 leads to increases in both daytime average N02
and nighttime average N02 at Denver.
As was the case with Downtown Los Angeles and Lennox, yearly one-
hour maximum N02 at Denver invariably occurs in the daytime period during
the winter. Table 11.11 presents the empirical control model for yearly
maximum N02 at Denver. It is evident that yearly maximum N02 is nearly
proportional to NOV and that hydrocarbon control yields slight to moderate
J\
benefits.
It is interesting to note that, at Denver in the winter, hydrocarbon
control reduces daytime peak N02 levels but increases daytime average N02
levels. The increase in afternoon N02 from hydrocarbon control evidently
more than compensates for the reduction in peak morning concentrations. On
the contrary, at the Los Angeles sites, the reduction in daytime peak N02
affects the daytime average more than the increase in afternoon N02 concen-
trations.
11.2.5 Chicago, Illinois
The Chicago CAMP site is a "center-city" monitoring location located
in the southeast part of Chicago. Table 11.12 presents the empirical con-
trol model for annual mean N02 at Chicago. The model indicates that annual
mean N02 in Chicago is directly proportional to NO control and independent
of hydrocarbon control.
-------
261
Table 11.11 The Effect of NCL and Hydrocarbon Control on
Yearly Maximum NOo at Denver
Table 11.1 la Percentage Changes in Yearly Maximum NCU
HVJ UUli 1 UVL, ~ ~ "" — •••"•»• ' • -..—I....— "••*»
(Percentage Changes from 1969-1974 Level)
; t
O
(—
O
CJ
CJ
5r^
yr
o>
Q)
Oi
T +20%
o>
(£1
CTl
g 0%
o
t-
-20%
OJ
OT
| -40%
0
O)
J -60%
+20%
+20%
+17%
+14%
+11%
+8%
c
0>
u
i-
d)
o.
Table 11. lib
0%
+3%
0%
-3%
-6%
-9%
Yearly
-20%
-14%
,
-17%
-20%
-23%
-26%
Maximum
— NO.. COMTROl
JU t
us
a! (Percentage Changes
_i
^
z
%
J
£
5
£
O\
S +20%
| 0%
a>
% -20%
-40%
CT
§ -60%
o
+20%
44
43
42
41
40
0)
Q.
0%
38
37
36
34
33
X
-40%
-31%
-34%
-37%
-40%
-43%
-60%
-48%
-51%
-54%
-57%
-60%
N0? Concentrations, pphm
from 1969- 1974 'Level)
-20%
31
30
29
28
27
-40%
25
24
• 23
22'
21
-60%
19
18
17
16
.
15
-------
262
Table 11.12 The Effect of NOX and Hydrocarbon Control on Annual
Mean N02 at Chicago
Table 11.12a Percentage Changes in Annual Mean NC^,
o
t—
2;
O
O
O
s:
1
1
_>
j>
1
r—
01
_J
1
Ol
IO
E
s
VI
_)
•0
at
£
8
0)
en
c
5
s
a.
liU l,UM I r.Ul- ~ '
(Percentage Changes from 1969-1974
+20%
0%
-20%
-40%
-60%
+20% 0%
+20% 0%
+20% 0%
+20% 0%
+20% 0%
+20% 0%
Table 11.12b Annual
-20%
-20%
-20%
-20%
-20%
-20% .
Mean N02
rnuToni
-40%
-40%
-40%
-40%
-40%
-40%
Level )
-60%
-60%
-60%
-60%
-60%
-60%
Concentrations, pphm
(Percentage Changes from 1969-1974
+20%
0%
-20%
-40%
-60%
+20% 0%
6.9 5.8
6.9 5.8
6.9 5.8
6.9 5.8
6.9 5.8
-20%
*
4.6
4.6*
*
4.6
*
4.6
*
4.6
-40%
*
3.4
3.4*
*
3.4
3.4*
*
3.4
Level )
-60%
2.3*
2.3*
*
2.3
*
2.3
2.3*
3
Attains federal annual mean standard of 5.3 pphm (100 yg/m )
-------
263
The submodels for annual mean N02 indicate that hydrocarbon control
does yield a modest benefit in daytime N02 averages during the winter.
However, this benefit is almost exactly cancelled by an increase in
nighttime N02 averages (in both summer and winter). N02 levels at night
are increased by hydrocarbon control because hydrocarbon reductions lead
to greater levels of N02 in the afternoon.
Unlike Denver and the Los Angeles sites, yearly maximum N02 in Chicago
will almost always occur in the summer during the daytime period. The
statistical models for peak N02 in the summer at Chicago indicated that
there was no statistically significant effect from hydrocarbons. Thus,
yearly maximum N0? at Chicago should be directly proportional to MOY control
*— f\
and independent of hydrocarbon control (see Table 11.13).
11.2.6 Houston/Mae. Texas
The Mae Drive site is located about two miles north of the Houston
Ship Channel, immediately downwind of the large, heavily industrialized
area that surrounds the channel. The Mae Drive station can be considered
representative of air quality near a source-intensive area.
As indicated by the regression results in Appendix D, the Houston
sites (Aldine as well as Mae) were unique among all the sites studied
in the sense that a significant dependence between daytime NQg and
hydrocarbons was never found, neither for peak N02 nor for average N02,
neither during winter nor during summer. One reason for this result
might be the sparsity of available data for the Houston sites, typically
about 60 to 90 days for each season as compared to 300 to 700 days for
each season at CAMP sites and Los Angeles sites. There may have been
-------
264
Table 11.13 The Effects of Hydrocarbon Control on
Yearly Maximum One-Hour N02 at Chicago
Table 11.13a Percentage Changes in Yearly Maximum N02
liU lUiilKUI. ' •
A
CIT (Percentage Changes from 1969-1974 Level)
0)
ni
o
o
o
o
:c
g
1
en
en +20%
r~
| o%
| -20%
ra
* -40%
O)
ia
g -60%
+20%
+20%
+20%
+20%
+20%
+20%
u
L.
£
0%
0%
0%
0%
0%
0%
Table 11.13b Yearly
? NOX (
> x
01
-1 (Percentage Changes
_i
£
|
J
C
?
a>
i
10
en
^ +20%
1
0%
V)
| -20%
o
S. -40%
ja
§ -60%
+20%
30
30
30
30
30
0%
25
25
25
25
25
-20%
-20%
-20%
-20%
-20%
-20%
Maximum
•fiMTRni
-40%
-40%
-40%
-40%
-40%
-40%
One-Hour
-60%
-60%
-60%
-60%
-60%
-60%
Concentrations, pphir
from 1969-1974 Level)
-20%
20
20
20
20
20
-40%
15
15
15
15
15
-60%
10
10
10
10
.10
0)
-------
265
Insufficient data for the regressions to arrive at statistically sig-
nificant hydrocarbon coefficients. The other reason would be that no
hydrocarbon effect actually exists for daytime N02 in Houston. This
possibility is reasonable because the NMHC/NO ratio at Houston is
A
quite high (around 15 to 20), and because photochemical systems tend
to be less sensitive to hydrocarbon control at high NMHC/NO ratios.
A
The nighttime regressions for Houston/Mae revealed a significant
relationship between afternoon oxidant and nighttime NO-- However, we
were hesitant to translate the nighttime N02/oxidant dependence into a
control model. The reason for caution is that the nighttime model
requires an assumed_ relationship between oxidant and the primary pollutants,
NMHC and NO . For the Los Angeles and CAMP sites, we had assumed that
X
oxidant would be proportional to the NMHC/NO ratio. This assumption
X
would be more dubious for Houston because investigations have shown
little relationship between NMHC and oxidant at Houston[l]. The high
ambient NMHC/NO ratio at Houston also lends doubt concerning the
X
effectiveness of small-to-moderate hydrocarbon reductions on oxidant
in Houston.
Fortunately, our calculations demonstrated that the control model
for annual mean N02 is insensitive to the assumed relationship between
oxidant and precursors. Regardless of what assumption is adopted, the
control model indicates that annual mean N02 is essentially proportional
to NO control, with very slight changes produced by hydrocarbon control.
X
For instance, if we assume that afternoon oxidant is proportional to
-------
266
the NMHC/NO ratio, the control model would indicate that a 50% hydro-
/v
carbon reduction produces only a 6% decrease in annual mean NO,,. If
we made a very different assumption, that afternoon oxidant is proportional
to NO and independent of hydrocarbons, the control model would indicate
j\
that a 50% hydrocarbon reduction produces a 1% increase in annual mean
From the above considerations, we conclude that a control model such
as Table 11.12a (in the Chicago discussion), where annual mean N02 is
proportional to NO and independent of hydrocarbons, is a good approxi-
X
mation for Houston/Mae. The present (1975-1976) level of annual mean
N02 at Houston/Mae is 2.5 pphm. Managing annual mean N02 air quality at
Houston/Mae should depend on strategies for NO emissions only.
/\
The yearly maximum one-hour N02 concentration at Houston/Mae is
approximately 13 pphm. The yearly N02 maximum is most likely to occur
in the winter season, but is equally likely to occur during the daytime
and nighttime periods. The daytime regression equations (Appendix D)
indicate that the winter daytime N02 peak at Houston/Mae will be pro-
portional to NO control and independent of hydrocarbon control. Thus
A
a control model such as Table 11.13a is appropriate for the daytime yearly
maximum at Houston/Mae.
Calculations based on the nighttime regressions reveal that the
winter nighttime peak N02 at Houston/Mae will be as sensitive to oxidant
control as to NOX control. Since we are unsure of the relationship
between oxidant and primary precursors in Houston, we have not constructed
-------
267
an empirical control model relating the winter nighttime N02 peak to the
primary precursors. It suffices to note that a strategy for reducing
oxidant at Houston/Mae should also yield substantial benefits in terms
of nighttime yearly peak N02. Our calculations show that a 50% reduction
in oxidant (with constant N0v) would produce a 30% reduction in the
A
nighttime yearly maximum N02 concentration at Houston/Mae.
11.2.7 Houston/Aldine. Texas
The Houston/Aldine monitoring site is located about 12 miles north
of downtown Houston and about 13 miles northeast of the Houston Ship
Channel. Since the dominant wind direction is from the southeast, the
Aldine site can be regarded as a receptor location, about 12 to 13 miles
downwind of the main source areas in Houston.
As was the case with the Mae Drive site, the daytime regressions for
Houston/Aldine revealed no significant relationships between daytime N02
(peak or average) and NMHC concentrations. The lack of a statistically
significant hydrocarbon effect could be due to the sparsity of data at
the Houston sites. The other possibility is that no hydrocarbon effect
actually exists for daytime N02 in Houston.
The nighttime regression models for Houston/Aldine did not provide
good statistical fits to the data. The winter nighttime regressions
achieved a correlation coefficient of less than 0.6, and the summer night-
time regressions failed to produce any statistically significant relation-
ships between nighttime N02 and the "independent" variables: afternoon
N02 (N0216), evening NO (NITENO), and afternoon ozone (03AFT).
The failure of the nighttime regression models at Aldine most likely
results because of the neglect of transport. Contacts with personnel
-------
268
of the Texas Air Control Board [2] indicate that the elevated nighttime
N02 concentrations at Aldine most likely result from pollution transport
from the upwind source areas. Evidence of transport is demonstrated in
Figure 9.12, which shows that N02> NOX, and NMHC concentrations simultaneously
jump upwards at about 7:00 PM. This could be due to the arrival of after-
noon industrial emissions and evening traffic emissions transported to
the Aldine site. The persistence of high oxidant levels as late as 6:00 P.M.
(see Figure 9.12) is also evidence of transport. Since the empirical
models are based on the assumption that transport is not a dominant factor,
the models may be inappropriate for the Houston/Aldine location.
The failure of the statistical approach in the case of nighttime
N0? at Aldine precludes our formulating an empirical control model for
annual mean N0~ at Aldine; a control model for annual mean N02 would re-
quire submodels for both the daytime and nighttime periods. Also, since
yearly maximal N02 concentrations at Aldine invariably occur during the
nighttime period, we cannot formulate an empirical control model for peak
i
one-hour N02 at Aldine.
**
*
Note that the results for Aldine also place doubt on our models
for Azusa and Pomona in Los Angeles. Azusa and Pomona are downwind
receptor locations, and they exhibit diurnal patterns similar to Aldine
(although not as extreme). In the next chapter, it will be shown that
the control models for Azusa and Pomona are not verified by historical
air quality trends. The neglect of transport may be inappropriate for
Azusa and Pomona as well as for Aldine.
**
For reference, the reader may wish to note that the present annual
mean N02 concentration at Aldine is 1.70 pphm. The yearly one-hour
maximum is 11 pphm and occurs during the winter nighttime period.
-------
269
11.3 REFERENCES
1. G. K. Tannahill, "The Hydrocarbon/Ozone Relationship in Texas,"
presented at the Air Pollution Control Association Conference
on Ozone/Oxidants, Texas Air Control Board, Dallas, March 1976.
2. 0. Price and T. Echols, Air Quality Evaluation Division of the
Texas Air Control Board, personal comnunication, May 1977.
-------
270
12.0 VALIDATION OF EMPIRICAL MODELS AGAINST
HISTORICAL AIR QUALITY TRENDS
The empirical N0? control models developed in this report are subject
to several limitations: the omission of meteorological variables, the
neglect of transport phenomena, and the assumption that precursor changes
produced by variance in meteorology can be used to model the effect of
control strategies. The uncertainties in the models were highlighted in
Chapter 10, where analyses with weather variables indicated that the
observed effect of hydrocarbons on N02 might partially be due to unaccounted
for meteorological factors. These uncertainties stress the need to conduct
independent checks of the empirical control models. Accordingly, this
chapter checks the predictions of the models against historical air quality
trends.
Although the empirical models and the historical trends are both
based on ambient data, the trend studies do provide an independent valida-
tion of the models. For one, the trend studies employ several more years
of data than the empirical models. Also, the trend studies are based on
year-to-year changes in precursors and N02, while the empirical models are
based on day-to-day changes in precursors and N02.
The procedure for validating the empirical models is quite simple.
First, best estimates of historical precursor changes are derived based
on emission trend data and ambient trend data for NOV and NMHC. Next,
A
these historical precursor changes are entered into the control models
to predict historical N02 trends. Finally, the predicted trends for N02
are compared with actual trends for N02<
-------
271
The validation studies will be conducted for 5 locations: the
central Los Angeles area, coastal Los Angeles area, inland Los Angeles
area, Denver, and Chicago. The empirical control models for Houston cannot
be checked against historical trends because of the lack of long-term data
for the Houston sites.
12.1 CENTRAL LOS ANGELES AREA
This section tests the empirical control model for Downtown Los Angeles
against historical air quality trends. To provide generality in the test,
the verification is performed for 3 locations in the central part of
the Los Angeles basin: Downtown Los Angeles (DOLA), Burbank, and Reseda.
The verification proceeds in two steps. First, net changes in precursor (NOX
and NMHC) levels are estimated over the nine years, 1965 to 1974. Second,
the precursor trends are entered into the control model, and the resulting
predictions of N02 changes are compared with actual N02 trends.
12.1.1 Precursor Trends. 1965-1974
Two types of data can be used to estimate trends in photochemical pre-
cursors: emission data and ambient precursor data. Both are examined
below to arrive at "best estimates" of precursor trends at DOLA, Burbank,
and Reseda.
Emission Trends
A recent report of the Caltech Environmental Quality Laboratory pro-
vides emission trend data for the Los Angeles region [1]. Figures 12.1 and 12.2
summarize the EQL estimates of basin-wide emission trends for NOX and RHC,
respectively. Basin-wide NOX emissions increased by 35% from 1965 to 1974,
while basin-wide RHC emissions decreased by 18%. Nearly all of the NOX
-------
272
1600
1400
1200
YEARLY 1000
AVERAGE
TONS/DAY
(CUMULATIVE)
800
600
400
200
^FLIGHT-DUTY VEHICLES
OTHER STATIONARY SOURCES
1965 1966 1967 1968 1969 1970 1971 1972 1973 1974
Figure 12.1 Total NO Emission Trends in the
Los Angefes Basin[l]
-------
273
2100
1800
1500
YEARLY
AVERAGE
TONS/DAY
(CUMULATIVE)
1200
900
600
300
LIGHT-DUTY VEHICLE,,
EVAPORATIVE AND CRANKCASE
'LIGHT-DUTY VEHICLE EXHAUS
GASOLINE HEAVY"-DUTY VEHICLE
OTHER MOVING SOURCES
ORGANIC CHEMICAL
ORGANIC FUEL 6 COMBUSTION
^(GEOGENIC)
///////////s/s/
1965
19661967 1968
1969
1970 1971 1972 1973
1974
Fiaure 12.2 Total Reactive Hydrocarbon Emission Trends
9 in the Los Angeles Basin[l]
-------
274
increase and the RHC decrease resulted from changes In emissions from
gasoline-powered motor vehicles.
The EQL report also documents emission trends on a county-by-county
basis. Because of low growth rates in Los Angeles County (see Figure 12.3),
Los Angeles County emissions decreased relative to the basin-wide total emissions.
Los Angeles County emission changes were +25% for NOX and -24% for RHC from
1965 to 1974[1].
Trends in emissions affecting DOLA, Burbank, and Reseda differ from
countywide emission trends because of variations in the spatial distribution
of growth and in the specific sources affecting those 3 locales. As
shown in Figure 12.3, DOLA is in (and downwind of) an area that has exhibited
particularly low growth rates. Burbank is in a low-growth area but is near
moderate-growth areas. Reseda lies in a region of moderate growth. Esti-
mating trends in the emissions that affect these specific sites requires
educated guesswork. Judging from the results of the EQL report, we estimate
that emissions affecting these 3 sites changed as follows from 1965 to
1974:
Estimated NOX Estimated RHC
Emission Increase Emission Decrease
DOLA
Burbank
Reseda
10%-20%
15%- 25%
25%-35%
30%-40%
25%- 35%
15%- 25%
Ambient NOY Trends
A
An alternative method of estimating precursor trends is to examine
ambient data. To minimize statistical fluctuations in the trend estimates,
-------
Los Angeles County
1. DOLA
2. Burbank
3. Reseda
4. Lennox
5. West L.A.
6. Long Beach
7. Azusa
8. Pomona
Figure 12.3
Geographical Distribution of Percentage Change in Population
in the Los Angeles Basin, 1965 to 1975 [2]
-------
276
a large sample of air quality data should be used. The net changes in
ambient NO listed below are based on changes in three-year averages of
annual mean NOX from 1964-1966 to 1973-1975[3]:
Net Nine-Year Change
in Annual Mean NO *
A
DOLA
Burbank
Reseda
+ 1%
+ 7%
+31%
The nine-year change in ambient NO at Reseda agrees quite well with
y\
the estimated NOX emission change for Reseda. However, the ambient NO
/v
increases at DOLA and Burbank are less than the estimated emission increases
for those sites. Part of the discrepancy between emission trends and ambient
trends might be due to low air pollution potential in 1973-1975[2]. Some
of the discrepancy might also arise from the potential errors in the emission
trend estimates for DOLA and Burbank.
Ambient NMHC Trends
Ambient trend data for total hydrocarbons (THC) are available at DOLA
and Burbank. Estimating long-term changes in NMHC concentrations with this
data, however, is a tenuous procedure. Ambient hydrocarbon measurements
are considerably more error-prone than are other monitoring data[4]. Also,
conceptual difficulties arise in translating THC trends into NMHC trends.
*
Similar results are obtained if one examines trends in the annual
average of daily one-hour maximum NOY.
/\
-------
277
Using a very simple procedure to calculate NMHC levels from THC levels,*
approximate estimates of ambient NMHC trends can be derived. The resulting
estimates of nine-year changes in ambient NMHC concentrations are as follows^]:
Net Nine-Year Change
in Annual Mean NMHC
DOLA -.42%
Burbank - 8%
The ambient NMHC trends at DOLA agree with the estimates of RHC emission
trends, but the ambient NMHC reductions at Burbank are significantly less
than the estimated RHC emission reductions. The discrepancy at Burbank
most likely arises from errors in the ambient trends. In particular, the
reader should note that hydrocarbon monitoring at Burbank was~discontinued
from 1966 to 1969[3].
Best Estimates of Precursor Trends
By considering both emission trend data and ambient trend data, one
can arrive at reasonable estimates of precursor changes at DOLA, Burbank,
and Reseda. In deriving best estimates of NOX trends, emphasis should be
placed on ambient data, because the ambient trends best represent overall
changes in emissions affecting each location. Because of uncertainties in
ambient hydrocarbon trends, emission data should be given greater weight
in the case of hydrocarbons.
NMHC trends are estimated from THC trends, using the relation NMHC =
(THC-lppm)/2 (see Chapter 10). The accuracy of this formula changes as
relative THC and NMHC levels alter with time. This leads to a basic con-
ceptual difficulty in estimating NMHC trends from THC trends.
-------
278
Table 12.1 presents our best estimates of NO and NMHC trends from
y\
1965 to 1974 at the 3 central Los Angeles basin locations. The estimates
are rounded to the nearest 5%. Also presented are approximate error bounds;
these are based on subjective analysis of the uncertainties.
Table 12.1 Best Estimates of Nine-Year NOX and NMHC Trends
at DOLA, Burbank, and Reseda
Station
DOLA
Burbank
Reseda
12.1.2 Test of the
NOX Change
1965-1974
+ 5% - 5%
+10% - 5%
+30% - 5%
Empirical Control Model
NMHC Change
1965-1974
-40% - 10%
-25% - 10%
-20% - 10%
The empirical N02 control models for DOLA can be tested against
historical air quality trends at DOLA, Burbank, and Reseda. The procedure
is very simple. The NO and NMHC trends in Table 12.1 are entered into the
A
control models, Table 10.16a for annual mean N02 and Table 10.18 for yearly
maximum N02. The resulting predictions are then compared with actual changes
in N02 concentrations from 1965 to 1974.
Table 12.2 presents the verification test for annual mean N02. The
actual and predicted changes in annual mean N02 are almost exactly equal
at DOLA and Reseda and are off 5percentage points at Burbank.
*
Tables l().16a and 10.18 present only values up to a +20% NOX change.
The tables were extended to greater NOX changes 1n order to test tho model
at Reseda.
-------
279
Table 12.2 Test of DOLA Empirical Control
Model for Annual Mean N00
Station
DOLA
Burbank
Reseda
Average
Precursor Changes,
1965-1974
NOV RHC
A
+ 5% -405!
+10% -25%
+30% -20%
+15% -282
Predicted Nine-Year
Change in Annual
Mean N02 Cone.
0%
+ 7%
+25%
+11X
Actual Nine-Year
Change in Annual
Mean N07 Cone.
+ n
+12%
+23%
+12%
Table 12.3 presents the verification test for yearly one-hour maximal
N02 concentrations. The 99th percent!le of daily one-hour maximum N02
is also used in the test, because this air quality index is subject to
less statistical noise than the single yearly maximum value. The agree-
ment at DOLA and Reseda is again very good. The discrepancy between
actual and predicted changes at Burbank is 9 percentage points.
Table 12.3 Test of DOLA Empirical Control Model
for Yearly Maximum One-Hour NO?
C. +
Station
«>LA
Burbank
Reseda
Average
Precursor
1965-1974
NOX
+ 5%
+10%
+30%
+15%
Changes
RHC
-40%
-25X
-20%
-28%
Predicted Nine-
Year Change in
Yearly One-Hour
Kax. NO,
4.
-17%
- 6%
+11%
- 4%
Actual Nine-Year N02 Cone.
Tearly One-
Hour Max.
-19%
+ 3%
+19%
+ 1%
Changes
99th Percent! le
of Daily Max.
- 7%
+ 3%
+ 9%
+ 2X
*Change in three-year average W64-1966 to 1973-1975
-------
280
In a qualitative sense, the test of the DOLA empirical control model
is extremely encouraging. The control model predicts that hydrocarbon re-
ductions should decrease yearly maximum N02 relative to yearly average N02;
this effect has occurred at all 3 monitoring sites (see actual trends
in Tables 12.2 and 12.3). It is also encouraging that the models for both
annual mean N02 and yearly maximum N02 exhibit good quantitative accuracy
at DOLA and Reseda.
12.2 COASTAL LOS ANGELES AREA
This section tests the empirical control model for Lennox against histori-
cal air quality trends. The test is performed for 3 coastal locations:
Lennox, Long Beach, and West Los Angeles.
12.2.1 Precursor Trends, 1965-1974
The first part of the verification study is to determine historical
precursor trends from 1965-1974. Below, both emission data and ambient
precursor data are used to arrive at "best estimates" of precursor trends
at the 3 coastal locations.
Emission Trends
Trends in emissions which affect Lennox, Long Beach, and West Los Angeles
can be estimated by considering the results of the EQL trend study [1], the
source mix near the areas [5], and the growth patterns within the Los Angeles
region (Figure 2.3). Our estimates of emission changes from 1965 to 1974
are as follows:
Estimated NOX Estimated RHC
Emission Increase Emission Decrease
Lennox 5% - 15% 20%- 30%
Long Beach 0% - 10% 25% _ 35%
West Los Angeles 20% - 30% 20%- 30%
-------
281
Ambient NOV Trends
A
Trends in ambient NOX are determined by examining changes in three-year
averages of annual mean NOX, from 1964-1966 to 1973-1975. These changes
are as follows:
Net Nine-Year Change
in Annual Mean NOX
Lennox +5%
Long Beach -16%
West Los Angeles +9%
For all 3 sites, ambient NOX increased less than the estimated
change in NOX emissions. Some of this discrepancy may be due to more
favorable meteorology in 1973-1975[2]. Also, it should be noted that,
for these 3 sites, slightly more positive trends for ambient NOX
are obtained using three-year averages of daily peak NOX rather than three-year
averages of annual mean NO .
X
Ambient NMHC Trends
Data on ambient hydrocarbon trends at coastal sites in Los Angeles
are available only at Lennox, and only for the years 1970-1975 [3], For
*
those years, the decrease in NMHC concentrations at Lennox appears to be
about one-half of the decrease in NMHC at DOLA. Using the nine-year trends
at DOLA, extrapolation indicates that the net nine-year change at Lennox
(1965 to 1974) was a decrease of 20%. This estimate of ambient NMHC trends
at Lennox agrees fairly well with the estimated RHC emission change.
*NMHC concentrations are estimated from THC concentrations as explained
previously.
-------
282
Best Estimates of Precursor Trends
Table 12.4 presents our best estimates of NO and NMHC trends from
/\
1965 to 1974 at the 3 coastal locations. In obtaining these best
estimates, ambient data were given the greatest weight for NO , and emission
/\
data were given the greatest weight for NMHC. Again, approximate error
bounds are specified based on a subjective analysis of the uncertainties.
Table 12.4 Best Estimates of -Nine-Year NOX and NMHC
Trends at Lennox, Long Beach, and West LA
Station
Lennox
Long Beach
West LA
NOX Change
1965-1974
+5% ± 5%
-10% + 10%
+15% + 5%
NMHC Change
1965-1974
-25% ± 10%
-30% ± 10%
-25% ± 10%
12.2.2 Test of the Empirical Control Model
To test the empirical control model for Lennox, the NO and NMHC trends
A
in Table 12.4 are entered into Tables 11.3a and 11.4a. The resulting pre-
dictions are then compared with actual trends in HOp concentrations" from
1965 to 1974.
Table 12.5 presents the test for annual mean NO^. The agreement between
actual and predicted is good at Lennox, fair at West LA, and poor at Long
Beach. The discrepancies at West LA and Long Beach could be due to errors
in the precursor trend estimates for those sites.
-------
283
Table 12.5 Test of Lennox Empirical Control Model
for Annual Mean N02
Station
Lennox
Long Beach
West LA
Average
Precursor Changes
1965-1974
NOX RHC
+5% -25%
-W% -30%
+15% -25%
+3% .27%
Predicted Nine-Year
Change In Annual
Mean NOg Cone.
+3%
-11%
+13%
+2%
Actual Nine-Year
Change 1n Annual
Mean N0£ Cone.
-1%
+10%
+22%
+10%
Table 12.6 presents the test for yearly one-hour maximal NO,,. The
agreement at Lennox is good. The agreement at Long Beach and West LA is
very sensitive to which air quality index is used to measure actual trends
in maximal NC^ concentrations. The statistical noise in the actual trends
is quite large for maximal concentrations because they are based on few
observations.
Table 12.6 Test of Lennox Control Model for
Yearly Maximum One-Hour U02
Station
Lennox
Long Beach
West LA
Average
Precursor Changes
1965-1974
NOX RHC
+5% -25%
-10% -30%
+15% -25%
+3% -27%
Predicted Nine-Year
Change in Yearly
One-Hour Max. N02
-1%
-14%
+7%
-3%
Actual Nine- Year
Yearly One-Hour
Max.
-3%
-20%
+31%
+3%
Cone. Changes
99th Percent! le
of Daily Max.
+1%
+1%
+6%
+3%
Again, in a qualitative sense, the verification study is encouraoing.
The control model predicts that hydrocarbon control should reduce maximum
-------
284
N02 relative to yearly average N02. This effect is apparent in the actual
N0? trends, especially if the actual trends are averaged over the 3
locations.
12.3 INLAND LOS ANGELES AREA
Empirical control models have been formulated for two eastern/inland
sites in the Los Angeles basin—Azusa and Pomona. This section tests those
models against historical air quality trends.
12.3.1 Precursor Trends. 1965-1974
Estimates of historical precursor trends are required to test the
control models. Both emission data and ambient data are used to arrive at
"best estimates" of precursor trends at Azusa and Pomona.
Emission Trends
Azusa and Pomona are located in areas of moderate-to-high growth rates
(see Figure 2.3). Considering the growth rate of sources near those areas
and the results of the EQL study [1], we estimate that emissions affecting
Azusa and Pomona changed as follows from 1965 to 1974:
Estimated NOX Estimated RHC
Emission Increase Emission Decrease
Azusa 25%- 35% 15%- 25%
Pomona 25%- 35% 15%- 25%
Ambient NOX Trends
Trends in ambient NO are determined by examining changes in three-year
/\
averages of annual mean NOV from 1964-1965 to 1973-1974. These results are
X
as follows:
-------
285
Net Nine-Year Change
in Annual Mean NOX
Azusa +46%
Pomona +25%
Similar results would be obtained if yearly averages of daily maximum one-
hour NOX were used instead of annual mean NOX concentrations.
Ambient NMHC Trends
Ambient hydrocarbon data are available at Azusa for the entire ten-
year period. Estimated NMHC trends at Azusa are a 41% increase or an
11% increase, using annual mean concentrations and yearly average of daily
maximum concentrations, respectively. The increase in ambient hydro-
carbons directly contradicts the estimated decrease in RHC emissions.
Most of the discrepancy probably arises from potential errors in deter-
mining ambient NMHC trends.
Best Estimates of Precursor Trends
Table 12.7 presents our best estimates of nine-year trends in precursors
affecting Azusa and Pomona. Ambient data were again given the greatest
weight for NO , while emission estimates were given greatest emphasis for
A
NMHC. There is a large error bound on the hydrocarbon trend estimates
because of the discrepancy between RHC emission trend estimates and
ambient NMHC changes at Azusa.
-------
286
Table 12.7 Best Estimates of Nine-Year NOX
and NMHC Trends at Azusa and Pomona
Station
Azusa
Pomona
NOX Change
1965-1974
+40% + 5%
+25% ± 5%
NMHC Change
1965-1974
-10% + 15%
-15% ± 15%
12.3.2 Test of the Empirical Control Models
The empirical control models for Azusa and Pomona are tested by
entering the precursor trends in Table 12.7 into the control models
(Tables 11.6 through 11.9 ). The resulting predictions of ambient N02
trends are then compared with actual N02 changes from 1965 to 1974.
Table 12.8 presents the test for annual mean NO^. The agreement is
good at both Azusa and Pomona. It is interesting to note that the empir-
ical model for Pomona indicates a larger hydrocarbon effect than the
empirical model for Azusa, and that this agrees with the relative long-
term trends in annual mean NCL.
Table 12.8 Test of Azusa and Pomona Control
Models for Annual Mean N02
Station
Azusa
Pomona
Average
Precursor Changes
1965-1974
NOX RHC
+40% -10%
+25% -15%
+33% -13%
Predicted Nine-Year
Change in Annual
Mean N0« Cone.
+39%
+17%
+28%
Actual Nine- Year
Change in Annual
Mean N02 Cone.
+37%
+11%
+24%
Tables 11.6 and 11.9 were extended to account for the large NO,
increases at Azusa and Pomona.
-------
287
Table 12.9 presents the verification test for yearly one-hour maximum
N02. The agreement between the model predictions and the actual trends in
the 99th percentile of daily maxima is very good. However, the agreement
with actual trends in yearly one-hour maxima is fair to poor.
Table 12.9 Test of Azusa and Pomona Control
Models for Yearly One-Hour Maximum N02
Station
Azusa
Pomona
Average
Precursor
Changes
NOY RHC
A
+40% -10%
+25% -15%
+33% -13%
Predicted Nine- Year
C*V\Sknrt& *i-n Va-av*1 w .
unange in Tcari y
One-Hour Max. N0£
+34%
+11%
+23%
Actual Nine-Year
Yearly One-Hour
Max.
+39%
+28%
+34%
N02 Cone. Changes
99th Percentile
of Daily Max.
+35%
+11%
+23%
In a qualitative sense, the Azusa and Pomona models perform signifi-
cantly poorer than the models for central and coastal Los Angeles. The
Azusa and Pomona models predict that maximum N02 should have been reduced
relative to annual mean N02 because of hydrocarbon control. This effect
is not apparent in the historical air quality trends at Azusa and Pomona.
There are several possible reasons for this discrepancy. First, statistical
air quality fluctuations may be masking a real decline in maximal N02 rela-
tive to annual mean N02. Second, our estimates of RHC changes for Azusa
and Pomona may be in error; it is possible that these high-growth sites
have not experienced a decrease in hydrocarbons. Third, the neglect of
transport, a potential error in all the models, may be a more significant
-------
288
error at Azusa and Pomona. Maximal NC^ concentrations at Azusa and Pomona
occur during the evening, and transport phenomena may be an essential part
of the evening maxima. Fourth, the Azusa and Pomona nighttime models may
contain an invalid assumption. The benefit of hydrocarbon control on the
evening maximum occurs because oxidant concentrations are a significant
factor to the evening NOp maximum. We have assumed, in all cases, that
oxidant concentrations are proportional to the RHC/NOX ratio. This assump-
tion may be less appropriate for Azusa and Pomona than for the central and
coastal sites. Oxidant at Azusa and Pomona would probably depend more on
overall precursor levels than on the RHC/NOV ratio.
/\
12.4 DENVER
This section tests the empirical control model for the Denver CAMP
site against historical air quality trends at that location. Unlike
the Los Angeles cases, where the tests could be performed against nine-
year trends, the data for Denver permit a check only against five-year
trends.
12.4.1 Precursor Trends. 1967-1972
Estimates of precursor trends are required to test the empirical
control model. Best estimates of precursor trends at Denver are derived
below by examining both emission data and ambient data.
Emission Trends
Historical emission trends for Denver have apparently never been
documented[6,7]. It is possible to derive a very rough estimate of emis-
sion trends by combining a 1974 emission inventory for Denver[8] with
-------
289
national emission trends[9,10] and with data on growth rates in Denver[ll].
Table 12.10 summarizes these results.
Table 12.10 indicates that hydrocarbon emissions in the Denver re-
gion remained essentially unchanged from 1967 to 1974, while NO emis-
/\
sions increased 35%. Since we are interested in the period from 1967 to
1972, five years instead of seven, these emission changes should be mul-
tiplied by five-sevenths. Accordingly, for the period of interest, hydro-
carbon emissions remained constant and NOV emissions increased about 25%.
/\
The emission changes derived above can only be regarded as very
crude estimates. It has been implicitly assumed that control strategies,
fuel switches, etc., in Denver have paralleled nationwide developments.
It has also been assumed that source growth near the CAMP monitor has
paralled growth throughout the Denver AQCR. The potential error in our
estimates of emission changes affecting the Denver CAMP site may be as
high as + 10% or 20%.
Ambient NOX Trends
Five-year trends in ambient NO are determined by examining the
y\
change in four-year averages of annual mean NOX, from 1965-1968 to 1970-
1973. A longer averaging period is chosen for Denver than for Los Angeles
because the Denver data are less complete and appear to contain more noise.
Annual mean NO at Denver changed from 6.95 pphm in 1965-1968 to
X
8.83 pphm in 1970-1973, an increase of 27%. This agrees almost exactly
Our estimates of annual mean NOX are based on averages of quarterly
means for NO and N02. In some cases, the SAROAD output did not list the
quarterly mean concentration. When this was the case, we estimated mean
NO (or N02J by taking an average of the 50th and 70th percentile concen-
trations.
-------
290
Table 12.10 Estimates of Hydrocarbon and NOX Emission
Trends for the Denver Region
Source Category
HYDROCARBONS
Motor Vehicles
Aircraft
Gasoline Marketing
Other Stationary Sources
Total
NITROGEN OXIDES
Motor Vehicles
Aircraft
Stationary Sources
Total
1974 Emissions
(Tons/Day) [8]
199
6
12
21
238
88
4
108
200
Nationwide
Emission
Change3
1974 * 1967
0.85
1.0
0.98
1.06
1.24
1.1
1.13
Estimated
Denver Emis-
sion Changeb
1974 * 1967
0.98
1.15
1.13
1.22
Estimated
Emissions
in 1967
(Tons/Day)
203
5
11
17
236
1.43
1.27
1.30
62
3
83
148
(a) Based on EPA documents[9,10]. Note that these two EPA documents
do not agree in the common year, 1970. Our estimates for the
change from 1967 to 1974 are based on relative changes from
1967 to 1970[9] and 1970 to 1974[10].
(b) Nationwide change has been factored by 1.15, the ratio of the
seven-year population increase in the Denver Metropolitan Area
(1.25) to the seven-year population increase nationwide (1.09),
[113.
-------
291
with the estimated five-year increase in N0x emissions, 25%. Although the
agreement is, no doubt, partly fortuitous, it does provide us with some con-
fidence in air estimates of five-year NOV trends at the Denver CAMP site.
/\
Ambient NMHC Trends
Ambient hydrocarbon trend data at Denver are available only for total
hydrocarbons. Using an empirical formula relating NMHC to THC at Denver,*
the THC trends can be transformed into NMHC trends. We estimate that an-
nual average NMHC in Denver changed from 72 pphm in 1965-1968 to 80 pphm
in 1970-1973, an increase of 11%. This disagrees somewhat with the esti-
mate that hydrocarbon emissions remained constant over the five-year period.
Best Estimates of Precursor Trends
Best estimates of precursor trends at Denver can be derived by con-
sidering both the emission trend data and the ambient trend data. In
arriving at these estimates, greatest emphasis should be placed on the
ambient data because of the crude nature of the emission calculations.
Table 12.11 presents our best estimates of NO and NMHC trends from 1967
A
to 1972 along with approximate error bounds based on a subjective anal-
ysis of the uncertainties.
Table 12.11 Best Estimates of Five-Year NOX and NMHC
Trends at Denver
HQ Change NMHC Change
Station 1987-1972 1967-1972
Denver +25% + 5% +5% + 10%
"Based on data for 1969-1973, this formula is NMHC =0.6[THC - 135 pphm].
-------
292
12.4.2 Test of the Empirical Control Model
The empirical control model can be tested by entering the precursor
trends in Table 12.11 into the control models for Denver (Tables ll.lOa
and ll.lla). The resulting predictions of N02 trends are then compared
with actual five-year trends in ambient N02 concentrations.
Table 12.12 presents the test for annual mean N02. The actual in-
crease in annual mean N02 concentrations (8%) is significantly less than
the predicted increase (26%). This disagreement probably has little to do
with hydrocarbon trends since hydrocabon changes were very small. The rea-
son annual mean N02 trends did not follow NOX trends is not obvious, but
the discrepancy may be due to errors in the ambient data or undocumented
changes in monitoring procedures for NO, or NOV.
£ A
Table 12.12 Test of Denver Control Model for Annual Mean N00
Station
Denver
Precursor Changes
1967-1972
NOX RHC
+25% +5%
Predicted Five-Year
Change in Annual
Mean N02 Cone.
+26%
Actual Five-Year
Change in Annual
Mean N02 Cone.
+8%
Table 12.13 presents the test for yearly one-hour maximum NO/,. Again,
the actual increase in ambient N02 are less than the increases predicted by
the control model. This disagreement would appear to have little to do
with the hydrocarbon effect since hydrocarbons changed very little over the
five-year period.
Based on change in four-year averages from 1965-1968 to 1970-1973
-------
293
Table 12.13 Tests of Denver Control Model for Yearly
Maximum One-Hour N00
Station
Denver
Precursor Changes
1967-1972
N0¥ RHC
A
+25% +5%
Predicted Five-
Year Change in Year
ly One-Hour Max,NOx
+22%
Actual Five- Year NO? Cone. Changes
fc Yearly One- 99th Percent! 1e
Hour Max. of Daily Max.
+m +m
Testing of the empirical control model for Denver against historical
air quality trends has not been very fruitful. We cannot really test the
hydrocarbon effect predicted by the models,since hydrocarbon changes have
been relatively small over the five-year test period. The control models
indicate that the historical decrease in the RHC/NOX ratio should have
produced a slight reduction in the ratio of maximal N02 concentrations
to mean N02 concentrations. This effect is not apparent in the actual
trends. However, the predicted effect is so small that we could not really
expect to discern it in the ambient trends.
12.5 CHICAGO
In this section, the empirical control model for the Chicago CAMP
site is checked against historical trends at that location. The verifi-
cation study is conducted for an eight-year period, 1964 to 1972.
"Based on changes in four-year averages from 1965-1968 to 1970-1973.
Note thaTthe original data for yearly maxima have been revised according
to the results of our data quality check.
-------
294
12.5.1 Precursor Trends. 1964-1972
Best estimates of historical precursor trends in Chicago are derived
below by considering both emission data and ambient precursor data.
Emissions
Contacts with air pollution control agencies in Chicago[12,13] re-
veal that a study of historical emission trends has never been conducted
for that region. Following the procedures used for Denver, crude esti-
mates of historical emission trends can be derived for Chicago. These
results are summarized in Table 12.14. Table 12.14 indicates that hydro-
carbon emissions in Chicago remained nearly constant from 1964 to 1972,
while NO emissions increased by 33%.
A
As was the case with Denver, the emission trend estimates for Chicago
must be regarded as very approximate. The potential errors in the esti-
mates of emission changes affecting the Chicago CAMP site may be as great
as + 10% to 20%.
Ambient NOX Trends
Eight-year trends in ambient NO at Chicago are determined from the
X
change in four-year averages of annual mean NOX, from 1962-1965 to 1970-1973.:
Annual mean NOY at Chicago changed from 7.03 pphm in 1962-1965 to 9.93 pphm
yv
in 1970-1973, a net increase of 41%. This is in fair agreement with the
estimated emission increase of 33%.
Ambient NMHC Trends
Ambient hydrocarbon trend data at Chicago are available only for
total hydrocarbons. The THC trends can be transformed into NMHC trends
*
Data for 1971 have been omitted for NO and N02 because of problems
with the Chicago NO? monitor during that year. The poor quality of
data during parts of 1971 is obvious from a scan of the hourly data.
-------
295
Table 12.14 Estimates of Hydrocarbon and NOX Emission Trends for Chicago
Source Category
HYDROCARBONS
Motor Vehicles
Aircraft
Gasoline Marketing
Other Stationary Sources
Total
NITROGEN OXIDES
Motor Vehicles
Aircraft
Stationary Sources
Total
1972 Emissions3
(Tons/Day)
192
12b
35
124
363
132
8b
179
319
Nationwide
Emission
Change0
1972 T 1964
0.95
1.15
0.96
1.14
1.29
1.25
1.43
Estimated
Chicago
Emission
Changed
1972 * 1964
0.93
1.13
0.94
1.12
1.26
1.23
1.40
Estimated
Emissions
in 1964
(Tons/Day)
206
11
37
111
365
105
7
128
240
(a) A 1973 inventory was obtained from reference [13]. We made some
slight adjustments to make this inventory representative of 1972.
(b) The aircraft emission data available from reference [13] seemed
highly dubious, especially in the ratio of hydrocarbon to NOX
emissions. We have adjusted the aircraft emissions somewhat,
decreasing them for hydrocarbons and increasing them for NOX.
Since this is a minor source category, these adjustments are not
of great consequence.
(c) Based on EPA documents[9,10]. Note that these two EPA documents
do not agree in the common year, 1970. Our estimates for the
change from 1964 to 1972 are based on relative changes from 1964
to 1970[9] and 1970 to 1972[10].
(d) Nationwide change has been factored byO.98, the ratio of eight-
year population increase in the Chicago region (1.08) to the
eight-year population increase nationwide (1.10)[11].
-------
296
using an empirical formula relating THC to NMHC.* In this way, we esti-
mate that annual average NMHC concentrations at Chicago changed from
136 pphm in 1962-1965 to 90 pphm in 1970-1973. This decrease in ambient
NMHC, 36%, differs greatly from our estimate that hydrocarbon emissions
did not change from 1964 to 1972.
One possible reason for the discrepancy between estimated hydro-
carbon emissions trends and ambient trends could be errors in the ambient
data. As noted previously, ambient hydrocarbon data tend to be of poorer
quality than other types of aerometric data[4].
A second reason for the discrepancy could be that the estimate of
hydrocarbon emission trends in Chicago is overly conservative. Contacts
with the City of Chicago Department of Environmental Control reveal that
their control program did not start to focus on hydrocarbons until 1975.
However, from 1964 to 1974 an exodus of some large emission sources from
Chicago apparently did occur for economic reasons. This exodus of emis-
sion sources may have reduced hydrocarbon emissions in Chicago relative
to our estimates based on population growth patterns and nationwide con-
trol strategies.
Best Estimates of Precursor Trends
Table 12.15 presents our best estimates of precursor trends in
Chicago from 1964 to 1972. In arriving at these estimates, greatest em-
phasis has been placed on ambient trend data because of the crude nature
of the emission calculations.
Based on data for 1969-1973, this formula is NMHC = .6[THC - 78 pphm].
-------
297
Table 12.15 Best Estimates of Eight-Year NOX
Trends at Chicago
md NMHC
Station
Chicago
NOX Change
1964-1972
+40% +
NMHC Change
1964-1972
-25% + 20%
12.5.2 Test of the Empirical Control Model
Table 12.16 presents the test of the empirical control model for an-
nual mean N02 at Chicago. The predicted eight-year change in annual mean N02
is based on Table 11.12a,which indicates annual mean N02 should be directly
proportional to NO control with no effect from hydrocarbon reductions.
A
The agreement between predicted trends and actual trends for annual mean
N02 is quite good.
Table 12.16 Test of the Chicago Control Model for
Annual Mean N02
Station
Chicago
Precursor Changes
1967-1972
NOV RHC
+40% -20%
Predicted Eight- Year
Change in Annual
Mean N02 Cone.
+40%
Actual Eight- Year
Change in Annual
Mean N02 Cone.
+38%
Table 12.17 presents the test of the empirical control model for
yearly maximum N02- The empirical control model (Table 11.13a) indi-
cates that yearly maximum one-hour N02 should be directly proportional
to NO control and independent of hydrocarbons. However, the historical
Based on change in four-year average from 1962-1965 to 1970-1973.
-------
298
NOp trends indicate that maximal N02 concentrations have decreased
relative to NO concentrations.
/\
Table 12.17 Test of the Chicago Control Model for Yearly
Maximum One-Hour NO,,
Station
Chicago
Precursor Changes
1964-1972
NOX RHC
+40% -20%
Predicted
Eight-Year
Change in Yearly
One-Hour Max. N02
+40%
Actual Eight-Year N02 Cone. Changes
Yearly One- 99th Percent! 1e
Hour Max. of Daily Max.
+15%
+26%
The historical trends for annual mean NQg confirm the predictions of
the empirical control models for Chicago, but the historical trends for
maximal NOg do not. The historical trends for maximal NOg appear to be
more consistent with the empirical control models for other cities, which
indicated that hydrocarbon reductions would yield a benefit in terms of
maximal N02 concentrations. It is possible that this hydrocarbon effect
really does occur in Chicago during the summer daytime period (the season
and time when the yearly maximum occurs), but that the statistical model
for the summer daytime period in Chicago somehow failed to discern the ef-
fect. In this regard, it should be noted that the statistical model for
the winter daytime period in Chicago did indicate a significant hydrocar-
bon effect.
Based on change in four-year average from 1962-1965 to 1970-1973.
Note that the original data for yearly maxima have been revised according
to the results of our data quality check.
-------
299
12.6 SUMMARY OF VALIDATION STUDIES
Validation studies for the empirical N02 control models have been
conducted for 5 areas: the central Los Angeles area, coastal Los Angeles
area, inland Los Angeles area, Denver, and Chicago. In the central and
coastal Los Angeles areas, the model predictions agreed quite well with his-
torical N02 trends. The historical air quality trends in these 2 areas
confirmed the conclusion that hydrocarbon reductions would have little im-
pact on annual mean N02 concentrations but would bring moderate benefits
in terms of maximal N02 concentrations.
The test for the inland Los Angeles area was less successful. The
historical air quality trends did not confirm the model predictions that
hydrocarbon control would reduce maximal N02 concentrations relative to
mean N02 concentrations. Several reasons for the disagreement between
predicted and actual trends at the inland Los Angeles sites have been
discussed in Section 12.3.2. In particular, we noted that the neglect
of transport and the assumed relationship of oxidant to the NMHC/NOX ratio
may be least appropriate for the inland Los Angeles sites.
Historical trends at Denver did not provide a proper test for
the empirical control model. The existence of a hydrocarbon effect on
N02 concentrations could not be checked with trend data because hydrocar-
bon levels remained essentially unchanged at Denver during the period of
interest.
At Chicago, the empirical models indicated that hydrocarbons would
affect neither annual mean nor yearly maximal N02 concentrations. The
-------
300
historical trends confirmed this conclusion for annual mean N02> but
seemed to indicate that yearly maximum N02 had been reduced by hydrocarbon
control.
In a general way, the studies of historical air quality trends
do seem to confirm the qualitative conclusions of the empirical control
models. Although the empirical control models vary with location, season,
and time of day, three general conclusions are apparent:
1. With other factors held constant, annual mean and yearly
maximum N02 concentrations are directly proportional to
NOX control.
2. Hydrocarbon control provides very slight, essentially negli-
gible, benefits in terms of annual mean NO,, concentrations.
3. Hydrocarbon control provides moderate (less than proportional)
reductions in yearly maximal N02 concentrations.
In an aggregated sense, these conclusions are supported by historical
trends at the 4 study locations that have experienced hydrocarbon
reductions. Table 12.18 summarizes this agreement. It is evident that
hydrocarbon control has generally been associated with little effect on
annual mean N02 concentrations and with moderate benefits in terms of
yearly maximal N02 concentrations.
-------
301
Table 12.18 Summary of Historical Precursor Trends
and Ambient NOg Trends for the 4 Study Areas
Experiencing Significant Hydrocarbon
Control
Locati on
CENTRAL LOS ANGELES AREA
(DOLA, Burbank, Reseda)
COASTAL LOS ANGELES AREA
(Lennox, Lonq Beach,
West LA)
INLAND LOS ANGELES AREA
(Azusa, Pomona)
CHICAGO CAMP SITE
AVERAGE OF 4 AREAS
Trend
Period
(Years)
9
9
9
8
Precursor
NOX
+15%
+3%
+33%
+40%
+23%
Changes
RHC
•28%
-27%
-13%
-25%
-23%
Ambient
Annual
Mean
+12%
+10%
+24%
+38%
+21%
N02 Changes
99th Per-
cent! le of
Daily Maxima
+2%
+3%
+23%
+26%
+14%
-------
302
12.7 REFERENCES
1. J. Trijonis, T. Peng, G. McRae, and L. Lees, "Emissions and Air
Quality Trends in the South Coast Air Basin," EQL Memorandum No. 16,
Environmental Quality Laboratory, California Institute of Technology,
January 1976.
2. Y. Horie and J. Trijonis, "Analysis and Interpretation of Trends in
Air Quality and Population Exposure in the Los Angeles Basin," pre-
pared for EPA Monitoring and Data Analysis Division by Technology
Service Corporation under Contract No. 68-02-2318, March 1977.
3. California Air Resources Board, Air Analysis Section, "Ten-Year
Summary of California Air Quality Data, 1963-1972" and Supplements
for 1973-1975, January 1974.
4. J. Eldon and J. Trijonis, "Statistical Oxidant/Precursor Relationships
for the Los Angeles Region," Interim Report No. 1, "Data Quality
Review and Evaluation," prepared for California ARB under Contract
No. AS-020-87, January 1977.
5. J. Trijonis, G. Richard, K. Crawford, R. Tan, and R. Wada, "An
Implementation Plan for Suspended Particulate Matter in the Los Angeles
Region," prepared for EPA by TRW Environmental Services under Contract
No. 68-02-1384, March 1975.
6. D. Wells, EPA Region VIII, Denver, personal communication, March 1977.
7. W. Rieser, Air Pollution Control Division, Colorado Department of
Public Health, personal communication, March 1977.
8. Colorado Department of Public Health, Air Pollution Control Division,
"Analysis of Air Quality in the Denver Air Quality Maintenance Area,"
March 1977.
9. J. Cavender, D. Kischer, and A. Hoffman, "Nationwide Air Pollutant
Emission Trends, 1940-1970," EPA Office of Air Quality Planning and
Standards, January 1973.
10. EPA Monitoring and Data Analysis Division, "National Air Quality and
Emissions Trends Report, 1975," Office of Air Quality Planning and
Standards, November 1976.
11. U. S. Department of Commerce, "Statistical Abstract of the United
States, 1975," Bureau of the Census, 1975.
12. R. LaMorte, Cook County Bureau of Air Pollution Control, personal
communication, March 1977.
13. James Masterson, City of Chicago Department of Environmental Control,
personal communication, March 1977.
-------
303
13.0 COMPARISON OF EMPIRICAL MODELS AGAINST SMOG-CHAMBER RESULTS
Section 7.1 of this report reviewed smog-chamber results concerning the
dependence of N02 on the photochemical precursors, NOY and NMHC. The review
y\
indicated that the various experimental studies agreed with respect to the
dependence of N02 on NOX input; both annual mean and yearly maximum N02 con-
centrations should be approximately proportional to NOV input. The various
A
chamber studies disagreed somewhat concerning the dependence of N02 on hydro-
carbon input. However, we were able to arrive at the following consensus
based on the chamber studies: Fifty-percent hydrocarbon control should
have little effect—a change of +10 on mean N02 concentrations—but should
yield moderate benefits—a reduction of 10% to 20%—in terms of maximal N02-
The purpose of this chapter is to check the empirical control models
against the conclusions based on smog-chamber experiments. In order to provide
an appropriate basis for the comparison, we will consider only the daytime
empirical models. The durations of the various smog-chamber tests ranged
from six to twelve hours; thus, the chamber experiments basically represent
daytime conditions.
The empirical control models for all 8 cities concur with the smog-
chamber results concerning the dependence of N02 on NOX control. The empirical
models indicate that, with other factors held constant, mean and maximum N02
concentrations are approximately proportional to NOX input. The slight
deviations from proportionality that sometimes occur in the empirical models
-------
304
are all in the direction of a less-than-proportional relationship. Some of
the smog-chamber experiments indicate similar slight deviations away from
*
proportionality (see Figures 7.1, 7.2, 7.3, and 7.7).
A more crucial test of the empirical control models involves the hydro-
carbon effect. Tables 13.1 and 13.2 summarize the hydrocarbon effect
predicted by the winter/daytime and summer/daytime models, respectively.
Table 13.1 indicates that, at the 6 non-Houston sites, the predictions
for winter/daytime maximum N02 agree extremely well with the conclusions
based on smog-chamber results. For a 50% hydrocarbon reduction, the
predicted changes in winter/daytime maxima range from an 8% decrease to
a 25% decrease and average a 15% decrease over the 6 non-Houston sites.
These results compare very well with the 10% to 20% decrease in maximal
N02 indicated by our review of smog-chamber studies.
The empirical models indicate that 50% hydrocarbon control should produce
anywhere from a 19% decrease to an 8% increase in winter/daytime mean N02.
The average of the predicted changes for the 6 non-Houston sites is an
8% decrease. These results are fairly consistent with the conclusion based
on the smog-chamber studies, that a 50% reduction in hydrocarbons could change
mean N02 by about * 10%.
The empirical control models generally indicate smaller hydrocarbon
effects in summer than in winter. For 50% hydrocarbon control, pre-
dicted effects on summer daytime maximal N02 range from no change to
a 19% decrease. The average predicted reduction in the summer maximum is
10% for the 6 non-Houston sites. The empirical models indicate that 50%
*
The reason for these deviations is discussed in Chapter 14.
-------
305
Table 13.1 Predicted Impact of a 50% Hydrocarbon
Reduction on Daytime N02 in the Winter
Empirical Model
Downtown Los Angeles
Lennox
Azusa
Pomona
Denver
Chicago
Houston/Mae
Houston/Aldine
Effect on Winter /Day time
Maximum* N02
-25%
-10%
-15%
-20%
- 8%
-14%
0%
0%
Effect on Winter/Daytime
Mean N02
-14%
- 5%
- 8%
-19%
+ 8%
- 8%
0%
0%
Maximum one-hour N02 during the entire season
Table 13.2 Predicted Impact of a 50% Hydrocarbon
Reduction on Daytime N02 in the Summer
Empirical Model
Downtown Los Angeles
Lennox
Azusa
Pomona
Denver
Chicago
Houston/Mae
Houston/Aldine
Effect on Summer/Daytime
Maximum* N02
-19%
-16%
0%
-17%
- 5%
0%
0%
0%
Effect on Summer/ Day time
Mean N02
- 6%
- 4%
+ 1%
- 9%
- 7%
0%
0%
0%
*Maximum one-hour N02 during the entire season
-------
306
hydrocarbon control would produce anywhere from a ]% increase to a 9%
decrease in summer/daytime mean NC^. The average predicted change over the
6 non - Houston sites is a 4% decrease. These summer results are also very
consistent with the conclusions based on smog-chamber tests.
The reader may be concerned about the differences in the hydrocarbon
effect predicted for different locations. One possibility is that the
hydrocarbon effect is universal and that the differences between cities
are a product of the errors, or limitations, in the empirical models. In
this case, the aggregate conclusions, that a 50% hydrocarbon reduction would
decrease maximal N02 by 10% to 20% and would yield very slight (if any)
benefits in mean Nt^* is most useful. The other possibility is that the
NOg/hydrocarbon relationship varies with location, depending on clima-
tology, the existing NMHC/NOV ratio, and other factors. That the hydro-
A
carbon effect may depend on local conditions is supported by the variance
in the observed l^/hydrocarbon relationship under different smog-chamber
conditions. All considered, the variability in the hydrocarbon effect
observed at the 8 locations is probably due to both factors, errors in
the empirical models and dependence of the hydrocarbon effect on local
conditions.
-------
307
14.0 CONCLUSIONS OF THE EMPIRICAL MODELING STUDY
The objective for Part II of this project was to characterize the
relationship between N02 and precursors by statistical analysis of air
monitoring data. In line with this objective, we have formulated empiri-
cal control models for nitrogen dioxide, applied these models to 8 ;
cities, and tested them against smog-chamber results and historical air
quality trends. This chapter summarizes the main conclusions resulting
from the investigation.
14.1 SUMMARY OF THE 8-CITY STUDY
The empirical control models for nitrogen dioxide are based on re-
gression equations between N02 and precursors, and on certain simple
physical assumptions. The control models for annual mean N02 involve
synthesis of submodels for daytime average N02 and nighttime average N02,
for both summer and winter. The control models for yearly one-hour
maximal N02 are derived from submodels for peak N02 under the conditions
(e.g., season and time of day) when the yearly maximum is likely to occur.
The formulations of empirical models for the 8 selected locations
proceeded smoothly with the exception of nighttime models for the 2
Houston locations. Lack of nighttime models for Houston/Mae and
Houston/Aldine precluded development of annual mean or yearly maximum
control models for those 2 sites. Accordingly, this summary is restricted
to the 6 other sites studied.
-------
308
14.1.1 Dependence of N02 on NOX
The empirical models for all 6 locations (as well as the daytime
models for the 2 Houston sites) point to the basic conclusion that
both annual mean N02 and yearly maximal N02 are essentially proportional
to NOX input. With other factors held constant, reducing NOX by 50%
should halve both mean and maximal NOo concentrations.
The slight deviations away from proportionality that sometimes occur
in the empirical models are all in the direction of a less-than-proportional
relationship. As noted in Chapter 13, similar slight deviations away from
proportionality are often observed in smog-chamber simulations. The empiri-
cal models exhibit these deviations only when a significant hydrocarbon
effect exists (e.g., as in most of the models for yearly maximum N02). The
slight deviations from proportionality result because reducing NOV has the
X
side effect of raising the NMHC/NOV ratio; this increase in the ratio may
A
produce an increase in NO? relative to NO .
t A
The conclusion that, with other factors held constant, N02 concentrations
are essentially proportional to NO input is supported by smog-chamber results
A
and historical trends. This conclusion is also considered reasonable on
basic physical and chemical grounds.
14.1.2 Dependence of N02 on Hydrocarbons
Table 14.1 summarizes the effect of hydrocarbon control on yearly
one-hour maximum N02 and on annual mean N02- Although the results vary
from site to site, the aggregate conclusion is that 50% hydrocarbon control
should yield slight-to-moderate reductions (about 10% to 15%) in yearly
-------
309
maximum N02 and essentially negligible benefits (about 0%to 5%) in annual
mean N02. As shown in previous chapters, this general conclusion is sup-
ported quite well by smog-chamber results and historical air quality trends.
Table 14.1 Predicted Impact of a 50% Hydrocarbon
On* Sn M°n.Annual Mean N0? and Yearly
One-Hour Maximum N0? 2
1
Empirical Model
Downtown Los Angeles
Lennox
Azusa
Pomona
Denver
Chicago
Effect on Yearly One-
Hour Maximum N0£
*
-25%
*
-10%
**
-6%
**
-19%
it
-8%
***
0%
Average for 6 Locations -11.32
Effect on Annual
Mean N02
-6%
-2%
-2%
-n%
+5%
0%
-2.7%
Effect on the
Maximum/ Mean
Ratio for N02
-20%
-8%
-4%
-9%
-12%
0%
-8.8%
Maximum occurs in winter/daytime period.
**
Maximum occurs in winter/nighttime period.
***
Maximum occurs in summer/daytime period.
-------
310
Table 14.1 seems to indicate that the model predictions for the
maximum/mean N02 ratio are more consistent from city to city than are
the predictions for the yearly maximum N02 or annual mean N02. Where
hydrocarbon control yields relatively high (or low) benefits in terms
of maximal N02, hydrocarbon control also yields relatively high (or low)
benefits in terms of mean N02. As remarked in Chapter 7, the various smog-
chamber studies agreed that hydrocarbon control should reduce the maximal/
mean N02 ratio but disagreed as to how this decrease would be produced (i.e.,
decreasing the maximum with no change in the mean vs. increasing the mean
with no change in the maximum). Thus, there appears to be consistency
between the types of variations observed in different smog-chamber studies
and the types of variations observed in empirical models for different
cities.
The variations in the empirical models among cities can be due either
to errors in the individual models or to real variations in the N02/precursor
relationship from one location to the next. The differences in the N02/pre-
cursor relationship found in different smog-chamber studies indicate the latter
cause is certainly a possibility. However, considering the potential errors
in the empirical models, we are more sure of the general conclusions con-
cerning the N02/precursor relationship than we are of the specific models
for individual cities.
14.2 CONFIDENCE IN THE RESULTS
The empirical control models developed here are subject to several
limitations: the omission of meteorological variables, the neglect of
transport, and the assumption that precursor changes produced by variance
-------
311
in meteorology can be used to model the effect of control strageties. The
potential importance of these limitations was stressed in Chapter 10, where
analyses with weather variables indicated that the observed effect of hydro-
carbons on N02 might be partially an artifact produced by unaccounted for
meteorological factors. Because of the uncertainties in the simplified em-
pirical models we have employed, we could not place great confidence in our
understanding of the N02/precursor relationship if it were based solely on
the empirical models.
We become much more confident of our understanding of the N02/precursor
relationship when we consider the empirical models in conjunction with
smog-chamber studies and historical trend analysis. All three approaches
yield results that are consistent with the same general conclusions:
• With other factors held constant, yearly maximal and annual
mean N02 concentrations are essentially proportional to NOX
input.
• Hydrocarbon control yields slight-to-moderate benefits in
yearly maximal one-hour N0£; reducing hydrocarbons by 50%
should decrease yearly maximal N0£ by about 10% to 20%.
• Hydrocarbon control yields very slight, essentially negligible,
benefits in annual average N02>
• The exact form of the N02/precursor relationship may vary some-
what from one location to the next,depending on climatic conditions,
reaction times, and the existing hydrocarbon/NOx ratio.
Although empirical models, smog-chamber simulations, and historical trend
studies all involve uncertainties, the overall agreement between the three
>
types of analyses indicates that we do have a basic understanding of the
dependence of ambient H09 concentrations on precursor control.
-------
312
APPENDIX A
STATION-YEARS WITH 75% COMPLETE DATA ON
SAROAD AS OF 3-6-76 (INCLUDES
CORRECTIONS DISCOVERED IN DATA QUALITY CHECK)
-------
C1TT STATt
Vt*(<
10JH 30TH bOfH 70TH 90TH 9STH 991H MAX AK1TH GtOM GtOH
H1CROGKANS PER CUBIC MtTtR Mt*N MtAN S10 OtV
P*Gt
M! 000,?
?30POl
?.*?I>4
5 S200.1J
b
AMA
A "A
As A
A-A
A* A
A*! 4
ft^'A
Ai.fe
A»I
CA|.
CAL
CAL
CAL
CAL
CAL
CAL
CAL
CAL
CAL
SliOfc 02
A/U
SuOOfl?
A/0
A/U
3&K
;ts*
bin.
<>•„•';(,«,>
b
5
5
S
•»
b
5
5
5
b
5 I2«>0001
b 1?f-CO"l
b Mt.Ocoi
s i6.>n.oi
CAL
CAL
CAL
CAL
C»L
CAL
CAL
CAL
CAL
CAL
CAL
CAL
CAL
CAL
CAL
CAL
CAL
CAL
CAL
M 'K
C A rt
LAL
CAL
CAL
CAL
CAL
C-J
cm
C1"!
klC
tic
£.301002
CAL
CAL
CAL
CAL
CAt
CAL
CAL
CAL
CAL
CAL
CAL
1973
1967
1966
19h9
1970
1971
1972
I97i
197«
1965
19A6
1967
196«»
19ft9
1970
1971
1972
1973
1974*
1966
1»70
1967
1969
1972
1973
1974
1974
1963
19669
I97'J
1971
1974
197*
1973
1971
1972
1973
1974
1974
1974
197«
1972
1973
197S
1963
7905
6749
73!>6
7985
7429
7875
0004
7340
7993
7012
773*
7690
7622
f«53
7311
7998
8001
6096
*?69
7416 -
7732
692?
7219
7729
7614
7641 ^
t>9»B
7143
7367
7674 ,
7*36
- 760* =
7583
7753
7603
7833
8200
;?9t
777«
7835
/MI
7S»2
7*26
«»I57
7772
7970
7570
7864
7295
7042
9
9
9
38
9
38
38
38
38
9
9
9
9
9
9
9
19
38
38
-.<,
3*
9
38
9
19
19
9
38
38
34
38
*6
75
75
75
56
56
9
19
9
9
19
19
19
19
9
9
9
19
9
9
38
56
56
56
56
56 .
56
56
30
3d
38
9
56
56
75
rs
'S
75
—•se- -
5*
30
56
W
36
38
30
56
56
75
75
- 9«
132
113
113
113
75
34*
30
9
9
19
19
38
30
IV
30
19
19
9
9
56
94
75
75
75
75
75
94
56
75
75
5«
75
94
113
94
90
113
56
7*
56
75
56
30
3ft
5*
94
94
113
113
132
169
150
150
ISO
113
b6
38
9
38
19
19
56
38
3*
56
38
38
38
30
94
113
113
113
113
94
113
113
94
113
113
113
113
150
132
132
ISO
150
94
113
f5
113
75
96
96
94
132
132
11«
150
188
296
207
207
207
150
75
56
30
30
30
30
75
56
75
75
56
36
30
94
169
207
169
169
108
ISO
169
18*
150
169
188
188
180
«6
207
207
226
207
132
150
132
150
113
94
113
150
22*
207
244
?4«
320
376
320
320
320
244
113
94
75
75
75
75
132
94
132
113
94
56
75
132
207
2«2
207
207
244
207
226
263
180
207
226
244
226
2*2
263
263
263
263
16*
1*8
ISO
180
112
132
132
180
242
244
WO
399
395
470
414
395
395
301
150
113
94
75
75
94
150
113
169
150
132
56
ts
169
320
395
320
!20
414
320
395
395
?02
301
320
414
33«
414
395
414
376
395
226
??6
?26
?26
188
1*9
169
?44
451
357
409
470
602
658
583
5*4
564
414
fit,
150
113
113
113
132
244
169
20?
?26
188
75
113
413 1S.O
500 04.0
.639 103.0
658 95.0
639 92,0
1015 90.0
620 06,0
921 90,0
696 106,0
677 74,0
5«3 "6.0
602 90.0
639 8b.O
650 9«,0
800 117.0
696 H6.0
027 115.0
602 117,8
046 119,0
414 71.0
4t4 «1,9
357 *!».<>
451 09.0
2«2 59,0
244 55*0
882 54,9
046 75. 0
731 Ht. 9
$64 106.0
08* 128.0
752 IHUO
1093 1*7,0
1015 201.0
940 101.0
902 172.0
996 I7i.0
650 134.0
414 *c-,0
330 51.0
1*8 31.0
207 33.0
180 3b,0
180 36,0
696 64,0
376 51,0
564 5f,0
451 59.0
357 52.0
188 32.0
180 3S, 0
19.2
56.1
74.3
74.9
69.7
73.2
65.2
78.5
85.1
49.4
58.0
60.1
46.0
6S.6
•1.*
04.9
§6.1
«*.*
96.4
53,7
76.*
47.9
n.0
•3.7
«5.l
44,1
52.0
•A, 6
03.0
95.6
«0,6
1*1.3
1*4.2
149,7
141.0
130.2
111.0
42. /
«1.7
21. U
23.2
29,0
30.0
47.6
42.0
36.0
41. J
39.2
26.1
25.1
2.0
2.5
2.5
2.2
2.3
2.3
2.3
2.0
2.1
2.7
2.7
2.0
3,3
2,6
2.7
2.S
2,4
2.1
2.0
2.3
l.»
2,«
1.9
2.4
1.9
1.9
2.5
2.2
2.2
««*
2.3
4*1
2.0
1.9
2.0
2.1
1.9
2.6
2.0
2.4
*,4
1.7
1.0
2.3
1.9
2.6
2.5
2.2
1.7
2,3
S3
(*)
-------
Ml IN'
MfcTHOU
NUHHfcR IflTH 30TH SOTH 70TH 90|H 9%TH 9«HH
HOURS HlCRUGKAiiS PtR CUBIC
MAX A«lTh
PAUL 2
MEAN STO OtV
S 2'' 0.i)<)3
5 3 '.i <* i'.' o n i
5 i'U '00 i
!j i/4'»;il)0 I
5 i '"• (i n ^
.- < > 0 0 >
. HuV
a 1 e o o o i
'II J ^ 0 0 1
u 1 :<.' o 0 i
<4 1 .-5 '- 0 0 1
4 1 "." 0 0 1
4 I o ''• 0 0 i
41««030 i
<41 o '• 0 C 1
C*L
CAL
r ')
LAri
CM.
CAL
CAL
CAL
CAL
CAL
.57«!«001 LAr<>
CAL
CAL
CH
LF •
(.to
CAL
CAL
CAL
CAL
CAL
CAL
CAL
CAL
CAL
CAI.
LO-
Lie
CAL
CAL
L'1-
10.
Lf'J
LOS
CAL
CAS.
C«L
CAL
CM.
CAL
C'.L
(-'•I
en
CAL
CAL
)96<4
1 ?bS
1969
1972
1973 -
1974
1 97?
197$
1974
1963
1964
1970
1971
1972
1973
197-4
1972
1973
1974
1966 "
1 96 J
1968
196V
1970
1971
1972
1973
1974
197(i
1973
1 t97<4
I 1963
1 1968
t96b
I9ho
19*7
196b
1909
1970
1971
1974
1972
1963
19*»u
1965
l <*be>
1967
1966
1 19*>9
1 1970
7f<51
7U46
7197
79b5
7898
7B98
8011
8fl20
82H1
7?16
6927
enao
756$
7«29
7932
7827
8?t>0
6013
8? 38
7397
7521
7151
7ttK>
6965
6030
8338
6294
8254
8021
6036
6154
7140
7061
7«14
73U6
/505
7492
7827
8073
8075
7890
6769
7044
7120
7*16
7817
7781
7623
7399
7908
9
9
9
9
19
19
9
9
9
b«>
56
9
16
3R
36
i8
9
19
1«
-.. 33 - -
3*
36
56
46
5*
" 48
bt>
56
38
19
19
3«
38^
38
58
56
56
««
b6
56
b6
9
38
9
3ft
56
38
38
38
38
9
18
3*
38
30
38
9
19
19
7b
75
9
56
56
56
56
9
19
19,
56.
75
7S>
75
75
75
75
75
75
5*
JS
38
56
56
56
75
75
94
94
94
75
75
56
5o
56
7b
9<4
7b
7b
75
75
38
38
b6
b6
b6
38
9
19
19
94
94
i8
94
94
94
94
9
1»
94
in
tu
ii*
113
113
94
111
94
56
56
" S£.-
9*
94
75
9ft
111
113
113
132
94
113
75
94
75
113
113
94
113
US
113
38
56
75
56
56
56
38
38
38
150
150
56
113
113
112
132
9
19
38
lie
159
150
132
158
ISO
1S2
112
132
7S
f 5 -
r $6
ts?
112
MI
132
ISO
169
150
169
138
150
ill
132
113
150
169
132
150
150
150
75
94
113
94
94
94
b&
75
56
263
226
111
106
169
184
188
b6
jfl
5*
?26
2&2
263
?26
224
226
207
107
?tf7
Ui
Hi
94
?fc3
226
SO?
226
?63
301
?63
263
226
226
169
244
?67
?63
?6i
226
244
?07
263
It3
132
132
94
113
113
75
75
75
338
301
1S2
226
226
244
244
Ti
56
T5
2*2'
395 -,
•jj8
282
301
301
263
244
2h$
132
132
1.1*.
2!»2
263
262
3)6
395
338
338
301
263
2?6
320
2A.3
338
33U
1?0
301
263
338
150 3?0
1MB 320
18ft 301
IbO 244
IbO 282
169 bB3
113 263
113 166
94 169
583 1090
432 752
207 414
376 639
*5»7 806
395 959
357 602
113 166
75 150
US 180
41« 752
73* 1260
S45 1«72
395 752
4*1 752
489 600
414 7*3
395 7 IS
176 898
1*9 "jdj
T»8 329
lt>9 338
489 1072
414 U09
«S1 665
395 790
«69 8*5
620 1316
50* 99fr
50ft 970
489 B65
395 696
357 639
«89 10341
432 1316
SOS 1316
U89 1504
508 1015
489 084
414 1S04
508 1335
39.0
52,0
58,0
48.0
55,0
50,0
28. 0
36. 0
31.0
132.0
l
-------
S I 4T TlIM l».-u*ill K
STATt
MtTHOO
Y£A« NUMHE* 10IH 30TH 50rH 70TM 90T* 95TH 99IH
HUURS MICRUGKAMS PS* CUBIC
MAX
GtUM
510 OfcV
PAGt 3
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
\5
5
5
5
5
5
-5
5
5
S
5
5
•j
5
5
5
5
L''<3
CAt
CAt
«l«OCB<»
u 1 '.«()(> ii £
CAt
CH
CAt
IMS
LI'S
C*L
tPS
CAt
CAt
CAL
Lf'8
tns
'it -jCu! t
IDS
t'1S
(.'IS
L IS
t'lS
CAL
CAt
c*t
CAt
CAt
CAt
CAt
C*t
CAt
C*t
C»t
Mf'j
•* 720001
CAt
CAt
CAt
CAt
CAt
CAt
H. -i r o r
ii « " o ti y
ftF A
5 ^
«i •;
«> •»! bPO
CAL
CAL
CAL
CAt
CAL
c*t
CAt
CAL
CAL
1
I
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
I
I
I
1
1
1 •
1
1
1
1 :
1
' 1
I
I
1
1
1
1
1
1
1
t
1
1
t
1
1
t
1
1
1971
1972
1973
1974
1963
1964
1965
1966
1967
1V6&
1969
1970
1971
1972
19/3
1974
1963
1964
19Ji5
19*7
4*966
1967
1966
1969
1970
t97l
1972
1973
1*74
1964
1965
19*6
1974
1972
1973
I97fl
1969
19/2
1973
1974
1974
1972
1973
19/4
19'/«l
19-oi
1967
1968
1969
0017
8215
TT17
7959
7364
7576
7S82
7779
7632
7798
8103
BOOS
7918
8042
8307
7ti9
M99
7««0
7SJ9
7565
7T64
78*1
8*18
8596
6910
7918
7692
T866
6J01
7691
8006
82.81
6180
8060
6)18
75 113
56 94
56 7b
94
56
9
38
J8
38
38
56
56
56
56
38
56
56
S6
'" .- 9
9
38
••' -*t'
38
38
m
is
56
-'••*-$*--
56
38
9
19
19
9
19
19
19
9
19
19
»9
9
9
9
9
38
56
56
56
56
75
7*
75
941
7*
94
75
TV
JS
150
132
113
113
56
75
94
94
94
113
94
113
US
9«
its
UV
US
*v its
50
75
75
75
75
75
'**.-'
75
4n
75
18
3*
3*
V
9
19
19
18
38
3d
38
38
5b
38
38
3d
94
94
#»"
i i-i
All
113
ivr
98
38
38
56
38
9
19
19
38
5*
Sft
5*
38
56
56
56
56
169
132
AK
-.»**;
169
113
56
56
56
38
38
38
19
56
75
75
56
75
75
75
188 30!
169 244
150 207
169 244
94 188
113 188
132 244
132 207
132 226
ISO 244
132 207
132 226
SSO 244
ISO 226
!*• 244
ztrf
113 MB
IJZ 226
112 22*
112 20T
207
*ftT
»«
169
75
75
94
•75
56
56
3S
75
150
132
94
113
150
113
113
395
320
263
320
244
226
320
263
301
320
263
301
282
301
320
26*
6U2 1391
489 733
39S 1090
489 1203
414 771
376 752
508 996
43? 884
432 940
508 884
395 902
451 978
508 1138
432 790
435 884
451 t»?*
9*0
8t7
173.0
145.0
125.0
142.0
86,0
91.0
121.0
109,0
114.0
136.0
119,0
126,0
141,0
IIB.O
132.0
2S2
202
263
432
126
fir
470
982
962
96, §
865 117,0
656 169.0
361
7t4
244
101
33ft
120
226
94
9«
113
94
75
75
56
94
169
150
«st
3f6
37*
696 11&«0
150
150
169
470
338
t 1 3
150
150
IS?
9«i
94
94
132
282
207
150
207
226
262
226
902
696
207
204
338
207
207
169
863
583
J76
414
620
470
utft
105,0
*7,0
34.0
53.0
37.0
26,0
32.0
26,0
49,0
7o.O
71.0
57.0
53,0
58.0
73,0
64,0
fcP ft
144.1
120.5
106,2
117.1
59,9
68,7
88,8
83.2
89.3
110,4
99,0
103,2
IU.9
92.7
110.8
. 7
n.9
83.6
9*. 9
96.1
»*,5
16.8
85.7
39.7
46>,4
46.6
25.6
ltt.2
27.«
23.6
43.1
55.7
59.0
48. V
41.9
42.7
53.9
47.2
it tl 1
1.9
1.9
1.8
1.9
2.6
2.3
2.2
2,1
2.0
1.9
2,0
*.»
2*1
1.8
.**«:•
a.o
•=»••
1.9
1.9
1,6
1.7
2.4
2.2
1.6
1.5
1.7
2.4
1.9
1.8
2.0
2.3
2.4
(ft
-------
91 *f ll-
"fit-! TOR
MtTHOU
HCIUH5
lOTH 30Th 50IH 70TH
PtN CUBIC
!
«J
b 6 -a-H'Tl
b 6 " « • u 0 1
b
b &x
b
s
b
b *2<.''1i.>e i
b
b to
b 6\i."»jOi
S 6V, n\)ft^
b b < vi "if'' .4
I'll
Pi r
PU-I
pn-i
pfiM
Rf U
wto
pro
»t u
wf
CH
CAL
CAL.
CAL
CAL
CAL
CAL
CAL
CAL
CAL
CAL
H t V
Rl V
til V
tAL
CAL
CAL
CAL
CAL
C*L
CAL
CAL
CAL
CAL
(970
197a
1972
1V7
56
169
169
150
150
132
75
7S
75
56
2?6
244
244
261
9«
56
301
!*>
75
1S0
150
13*'
90
1(18
tw
169
113
94
Ul
113
75
132
94
150
169
169
132
*1S
112
132
Ho
94
150
113
188
207
?44
?26
113
94
94
75
357
35>7
176
376
414
SOt)
470
376
414
470
357
188
169
112
620
75.0
7b,0
6/.0
70,0
6U.O
28,0
2o.O
32,0
24.0
93,0
677 109,0
639 lOb.O
620 106,0
696 121.0
«8« 138.0
1S2
IS*
94
m-
338
395
470
4|«
451
4*1
112 «**0
2?6 «0.0
207 39,0
S4b 116,0
696 111,0
790 104,0
046
752 lOfe.O
113
h*
169
169
2«7
226 157
257
2&2
J9S
307
169
169
244
188
T8R
132
282
282
?63
4A9
395
263
301
6?«
395
176
357
Sflb
263
470
S«3
470
58.0
60.2
53.8
59.0
49,«>
19.1
17.7
26.2
21.6
67.4
86.7
81.1
81.4
99,4
117.8
27,«*
31.0
26, «J
ie.6
"7.S
111.0
126.0
124.1
79,0
74.0
82,0
72,0
50,0
64.0
54.0
56.0
46,0
51,0
62.0
70,0
36,0
64.0
5*.0
70.0
82.0
94.0
25.6
29^0
5«,7
7Q*.*
59.*
14.8
45.0
15^0
«5.9
53.0
25.2
05.5
44, 't
45.2
80^5
2.2
2.0
l!*
1.9
2.3
2.3
1.9
1.7
2.0
2.1
a!i
2,0
1.9
elo
2*0
U7
.8
.8
.8
,8
.8
.1
.2
.7
.7
1.6
1.9
2«*
2*. 3
1.6
2.4
1.8
2.8
*.7
1.8
CO
•^J
en.
-------
SMI JON
6TATF:
NUMBER
IOTH JOTH bOTH 70TH 901" 95TM 99TH
MJCRU6HAMS PER CUBIC
MAX
GfcO*
GfcOM
810 DtV
b
5 oS80(
5
5
b
5
b
5
5
5
5
5
5
5
Iti
C*L
on*) 0 Jo
b
b
b
5
S
SIC
SAC
S'-C
s/c
5&C
S'-t.
S'U
5 A '4
SAN
SAM
6;-of ! 0?
CAL
CAL
CAt
C»L
CAL
UAL
CAL
CAL
CAL
Cfl.
C*L
CAL
C*L
C»L
CAL
CAL
f»4>l CAL
1>*N CAL
CAL
CAL
CAL
r.AL
C*L
CAL
b
S
0 '*«>'• '• '>({
>tn C»L
.-"I CAl
)/• . C*l
jft< CAL
,••••, CAt
SAM CAL
-A • CAI.
i.. - CAL
1969
1971
1972
1974
1971
I
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
i
1
J 1974
1 I9?.l
1 1*72
1 19t3
1 1974
1
1
1
1
1
1
1
»
1
1
1
3
3
1
1
1
1
i
1
1
1
1
1
1
1
f 1974
19f>7
1968
T97Q
1964
1P74
!<«»>/
1969
1996
6232
7487
7173
7600
7336
6930
dl!6
7925
7524
7817
7912
77S1
8117
8143
6114
6847
8^59
6575
7573
7120
7169
7895
7333
75b»
7?87
6691
6836
8140
7518
77«9
/7bi
6836
7543
7BV2
78i>9
762«
9
9
38
38
9
9
9
'9
9
9
19
19
9
- '-!? Afc
9
9
9
J*.
9
9
9
9
9
19
9
9
9
9
9
36
38
38
38
3»
19
3P
38
9
9
9
38
9
38
19
19
38
38
56
38
38
38
38
38
38
56
38
9
f
9
9
It
,4-t~
56
56
sa
36
9
9
9
9
9
W
9
S6
36
S6
56
36
38
36
56
7*
b6
38
56
5.6
38
38
38
S«
38
9
16
38
3*
3*
.*«"••
JW
fi
•**..
75
7*
7*
If
9
9
38
9
36
9
Ifr,
b*
38
9* **
38 bfe
56 75
75 94
S6 75
S> 56
33 b6
51.
75
94
b6
56
75
56
'75
94
7b
56
75
7*
S6
56
5e
56
56
38
•3S
56
50
*€
'' W
--si'
111
••:*€:
56
94
3*
56
56
94
113
113
' 94
7b
T5
7b
94
113
56
7b
75
94
94
9«
150
150
113
94
1S8
94
94
75
7b
94
94
*<•
-75
,75
'75
75
TS
75
• 7§
169
132
94
132
75
113
S*
9«
113
9«
tbO
IbO
169
132
94
94
94
56 75
113
13?
IbO
IbO
132
113
150
113
ISO
113
H3
113
113
113
75
•?s
94
94
94
94
"94
"94
488
ISO
112
•8
113
150
H3
2?6
169
1U
132
113
132
ISO
!*.<»
1R8
207
1*9
150
168
263
202
263
207
244
169
150
150
188
It?
113
113
m
113
113
94 a
207
226
13?
320
263
1 ti8
207
IbO
301
VM
?07
2^4
320
301
2*?
376
505
602
b64
SOB
S64
376
282
470
263
414
320
188
186
244
1*8
207
207
2*3
282
508
5**
357
846
620
658
978
752
*39
470
32ft
S«>5
639
33H
752
620
771
639
6?0
39b
301
602
771
451
658
39b
508
500
54b
54b
50.0
•68.0
77.0
73.0
50.0
66,0
5b,0
53,0
<*6,0
• 49,0
5«,0
52.0
27.0
32.0
41,0
41, 0
42.8
42.0
43,0
39,9
93,0
8'J.O
7o,0
Bi.O
bl.O
35.0
41,0
61,0
32.0
50,0
21.0
«2,0
6b.O
4b,0
74,0
92.0
106.0
80,0
M.O
64.0
56,0
80.0
99.0
b6.0
63,0
67,0
81.0
80,0
71,0
67,0
36,2
45,2
*t.»
S9.8
3*.l
51.1
40.9
41.4
35.4
39.2
46,b
43,6
19,0
22.1
30.2
31,0
30.0
32.V
71.2
66.8
61.0
62.2
>»..*
19,b
23.b
34.8
19.3
37.b
14.0
28.6
47.3
30.0
46.4
74.7
86.0
66. t*
S5.0
56.2
49.1
6H.1
77.8
41.4
47.0
51.2
64.9
59.8
55.9
2.3
2,7
2.1
2.0
2.4
2.2
2.3
2.2
2.2
2.1
1.7
1.8
2.2
2,4
2.3
2.3
I.J
l.e
1.8
2.3
2.2
2.0
i.J
2.a
2.7
2,6
3,1
2,5
2.2
2.2
2,5
2.4
2,6
2.7
2.0
2.0
1.9
1.8
1.6
1.7
2.1
2,2
2.3
2,3
2.2
2.1
2.3
2.1
2,0
CO
-------
S i Al
CMf t>IATt .-in.iITriK
'tfcTHOLi
YLAK
JOTH StifH iO'lH 70TH 90IH 95TH 991H MAX AHITti GEO* GtOM
M1CRUGKAMS PtR CUBIC MtUR MtAN MtA« SID
PAGt
o r» f) «,> o 'i a
b
b
b /fl«0i'i!
s nooo"
5 /16^'J
b 7l6".iP
b M 6 10
b /UO'.ci
'b 7 160*11 J
b 7;di.'i/«2
b
*.
5 /20'>uf-4
5 /?0t>cr>«
b ??0'
b fSJQuOi
5 7300'tOi
5 / $00..0<>2
S fc04Co02
5
b
5
S
S
5
5
b
b
5
5
5
5
5
b
S
5
. .•> 2
M.'2
S'.N CAL
,>*:•. CAL
O-M: CAL
CAL
CAL
CAL
CAL
CAU
CAL
CAL
34.'.
CAL
si>. CAL
S.'-iM CAL
iAo CAL
CAt
t*L
CAL
C»t
CAL
CAL
CAL
CAL
CAL
S10 C*t
•jtJ CAL
STU C*L
i>U» CAL
siu
C*L
CAt
CAL
CAL
c»i
CAL
CAL
VIC CAL
vIS CAL
<4
1971
1V64
1965
1972
1972
1973
1974
19*7
1966
1969
197^
1971
1972
l*7i
197'*
i<»70
*9?«
1965
1966
1967
1966
»9*>9
1970
l'?71
1972
1773
1974
1974
1V74
1*74
1974
l«»7t
1971
1972
1973
1974
J972
1973
1974
7791
7597
7ft » 7
7813
7893
7714
7560
7999
8?«fl
7846
«2'>0
75»6
7105
7165
7744
6922
76b6
7661
7«y»
ft|82
74SO
7*99
6040
8065
8103
7963
7762
79,16
7113
7496
7022
696«
6963
7506
7160
7029
745 '4
7570
7819
7164
7784
8046
6597
7606
7936
7632
7604
.8305
806«
6096
36
9
19
19
9
9
9
9
9
19
19
9
9
9
3»
38
19
19
9
9
-;* .1-
•'-f -;'
' 9 "
9
•W...:
"If '
9
»9
9
•9, -;
"9
- . rf-.'Si"
9
9
9
IT
19
39
19
36
36
19
19
9
*
19
19
56
56
38 .
56
9
19
IV
38
36
56
36
38
36
38
9
9
36
56
18
38
38
9
9
9
9
9
9
if
19
9
If
9
f
9
".SB:
5«
9
38
5S
38
18
«56
75
56
36
36
38
9
19
19
94
TT5
56
56
38
3ft
3A
56
56
56
36
38
56
56
38
38
56
56
56
S6
56
9
" ,*
••-'*
9
"; 9
38
r9
19
36
36
36
3«
36
&>
56
36
38
38
36
bfc
7$
94
75
38
56
38
36
38
3ft
115
115
94
75
36
3«
36
7b
75
56
56
56
75
56
56
56
•?6
75
Tb
T5
75
38
•71*
- 36
••; 9
16
36
38
19
SO
,».*
is
J»
- 16
W
75
38
56
b6
56
TS
94
132
113
56
94
56
56
56
56
ISO
132
113
132
7*
75
75
113
94
94
75
75
94
94
132
75
'94
113
94
94
94
56
S6
b6
38
56
56
38
3*
S6
75
75
7%
75
113
169
75
75
75
75
94
13?
267
169
75
IbO
94
75
75
75
244
207
168
1*9
94
75
75
132
113
113
94
«»4
115
94
169
94
113
152
113
113
111
75
75
?s
%6
56
75
56
«?6
75
75
7V
94
94
150
en 7
75
94
94
94
115
J69
244
1P6
94
169
113
94
94
94
320
263
244
2»?
115
113
113
226
169
1bO
113
13?
169
132
262
132
169
169
IbO
IbO
»69
113
113
113
75
75
113
75
7«»
113
115
U2
132
150
I**
2»2
IS?
132
113
ISO
169
?62
320
?65
132
262
169
132
132
U2
470
39S
376
564
226
207
169
451
338
520
226
265
557
320
658
338
583
301
395
262
451
169
226
282
113
1*9
263
169
ISO
263
262
263
Jot
338
414
414.
320
2*2
207
301
3«57
583
714
526
263
414
282
226
226
244
1166
902
752
76.0
57.0
41,0
36,0
62,0
5b.O
53.0
42.0
46.0
56,0
55,0
54.0
36,0
53,0
67.0
59,0
57,0
56.0
2H.O
27,0
26, ft
20.0
23,0
32,0
27,0
25.0
34,0
37,0
3b,0
34,0
3*,0
5«,0
71.0
36,0
«1.0
43,0
50.0
60.0
77,0
112,0
92.0
49,0
70,0
4b.O
40,0
4b.O
43.0
140.0
123.0
102,0
63.2
27. «!
3b.3
32.7
46,2
40,9
42,0
32. v
36.7
50. t
49.1
30.9
27.8
40.7
56.2
52.4
«9.7
49.2
19.9
19.2
»7,6
l«.8
17.3
23. S
2«,2
•22.8
26.2
31,0
24.6
23,2
26.6
42.0
44,6
26.1
31.4
3
-------
5 I A?
C1IY STATfc
rt AK
HUUhS
IdTH 30Trt 50TH 70TH 90TH 95TM 991H MAX AR1TM
PtR CUBIC METfcR "tAN
MEAN
GtOM
STD OEV
f»*Qt 7
If -K'O,*
6 S>f;oiic'
(, 5 it 0 (i i\ ^
6
7
7
9
9
9
9
1 t
<
M i
1(t
I'l l*"'}e?
1 u J-'eiPo'V
Jo < ,>,•(>»•?.>
If « 0 j> 1M
1 o ' v Ov'l 1
i« t <.-. ( i >
ir« s-',-:"".-t
1 t, )\:i\M,:-' |
I » i •'' c.' 1 (.'" 1
1 * i">.-iO ;)i t>
•,, -i •. -. i.,i^
I -J
I ,
0 I: H
1i -.1 n o
.'"••lu-V
V!.rJ
VM
»' il
L':.4
U» i*
Lif M
yf N
K'S
Alt.
C.'f
C-l
C i
I- -J
CAL
C*L
C*L
COL
cni
COL
COL
COL
tOL
CO..
DTS
OT!>
Ola
015
01$
Ulij
ILL
lit
RL
111
UL
ILL
ILL
ILL
RL
RL
Rl
ILL
RL
Kf i\
Kff.
t.f N
r.Et.
KEN
1
1
1
3
3
3
1
1
3
3
3
i
3
3
A
i
3
3
3
3
i
i-m
1970
1972
1974
19*4
1973
19*3
19*5
1971
•i97«
1963
19611
H72
]97tt
197}
J97U
197(4
1973
1973
1'ir
197(1
1974
7908
6986
mi
7321
7370
6646
7164
7528
6861
7176
7260
8999
.7167
7599
6*79
6680
6970
7640
71**?
66QC»
8tb2
698?
6974
t>790
77S«
7519
9
19
19
38
as-
9
38
9
9
9
i»
*6 '
38
38:
56
56
19
38
38
19
9
18
38
9
38
19
31
V
19
19
So
36
50
53
2S
19
-18
S«
75
%66
S6
16
•5*.
7S
75
75
19
39
Sft
41
66
9
7'i
36
53
38
Jfl
3d
56
56
56
56
73
61
34
3*
75
'«*'-
sV
85
56
94
'75-
7S
75
T> ' 9ft.
7i /-»*
75 9S
75
.56
60
9
73
7*1
143
90
9U
56
68
36
36
75
75
75
75
9(1
too
Id
7*
75
7S
7S
"WJ
113
tor
Hi
111
in iso
9«
103
75
79
10>
.9
19}
75
i so"
132
56
06
56
7S
113
94
113
132
132
'«*'
. I IS
til
13?
169
J4t
-11?
150
112
•til
150
113
169
150
132
150
ifc*
169-
132
113
105
75
75
94
113
113
113
ISO
16V
169
1R2
60
113
U2
150
152
1*2
150
152
169
ISO
:»«'..
263
til
171
ISO
ISO
135
19 58
160 197
113 ISO
2o7 2'. "»«,0
«S£ 70,0
470-- .''t^O
**,0
. '6*0
/V.O
SS5S 78, ft
JJf So *'*-
S83T 10?«0
476 -9J.O-;
.2*3,
263
376
265
19?
197
102
«»
414.
4X2
676 12^,6
263
756
517
529
320
40? 5-»;0
312 6i.O
71,0
7», o
64.0
66,0
>6
IHft
120
320
336
150
•24 a
'414
13,0
67.0
3i.o
I2g.o
9/.0
526 111.0
282 53,0
330 7i.O
740
31.1
59.7
53.6
47./
56.1
71*9
73.6
30.7
24. *
40.0
6JT.8
70.2
77.4
7«.«*
97.3
8«.«
38, «!
56.0
75.'2
56.0
44,7
78|2
10,9
70.0
63.2
9(»*7
«2.9
2.3
1,7
t.6
1.7
2^2
2.2
2.3
t.9
2.1
2,0
1,6
i.e
1,7
1.6
1.6
1,6
It*
'I.*
i.s
U7
1*5
J.5
1.6
rf.O
2l«
1.7
I.a
1.0
2,3
2.1
U5
2.0
1.9
2.3
2.3
2,8
2,0
2.0
i.a
(O
-------
.J I 1 , ,
cn» stAU win.
Yt Art
NUMBER
HiHJHS
luTrt 50TH 70Th 90TH
PtR CUBIC Mfc.ttR
99JH H*X AHITH
Gt HH
SID OtV
P*Gfc «
') It f> M 0 ,<
i 0 ,1:) I
I 0 4 0 0 ()
?6
?6 «<
26
? !) <4 i U 0 V P
i>TA
STL
ST:,
.'.IS
11 IS
11S
MIS
/•IS
Mli
MIS
MIS
MIS
S1L
ML
HIS
HIS
«tH HEV
1 »'>!.<•>
sr
31
3i isoanj
31 t80005
51
31
31
31 720 r,0 3
31 '72'o.cai
31
31
31
31
31
n
ME*
NEh
NC*
UClk
tU
l.f W
?if «s
31 4 ? 4 flc i". £ r '• i
31 <«M .'><» i'" I
Si <>
Vfn
1
1
1
1
1
1
1
3
3
3
i
1
2
i
1
i
i
4
U
3
I
5
3
1
3
1
3
1
1
3
2
1
3
1
1
1
1973
1975
1974
1973
1971
705?
7904
1965
I960
1968
19/1
1974
1971
197-t
1972
1971
1967
1968
197«
1971
197i
1971
1970
1971
1972
1971
1974
19/0
19/1
1972
1971
1974
1V7?
1973
1974*
1974
1974
1974
802?
7812
736S
7947
7261
7701
7311
70»4
7071
8067
81*8
6.019
7735
WO 7
763»
7903
7«77
*»*?«»
6055
7845
7410
7«17
816*
6311
798?
7769
/37-J
733?
701'i
7544
9
24
9
9
9
9
21
9
38
9
9
38
19
9
iB
30
9
36
55
64
19
at
53
S6
38
41
2fr
26
12
9
1.9
9
3?
56
30
36
41
3V
36
38
36
21
:*•
9
71
53
41 9
fab
56
'J«
60
56
62
05
9
71
92
90
66
6«
79
86
75
*>6
ui
03
36
9
55
79
S6
55
60
56
68
38
56
36
36
56
56
38
S3
2*
3?
*«
81
'
ft!
79
9$
n
71
21
90
US
ill
92
111
94
86
5«
58
71
53
51
75
103
85
83
86
75
9fl
56
56
T5
94
58
?9
56
6*'
108
152
141
135
133
111
119
75
111
75
75
111
113
96
100
98
100
107
79
103
103
92
98
94
36
115
135
115
HI
Ul
113
115
75
75
90
77
Tl
34
9V
90
137
lal
147
11 /
143
147
147
tii
71
160
175
1S4
169
197
169
10S
107
124
113
55
126
177
169
171
162
132
171
94
132
94
94
111
150
169
233
U5
»18
ise
»77
184
196
164
171
175
1R2
139
165
167
171
158
171
167
160
68
ISO
199
209
160
2?2
199
233
ins
mo
122
141
150
135
68
312
201
166
243
112
169
113
132
132
207
175
293
322
457
5U
444
464
472
451
244
376
101
336
263
414
560
656
186
286
1*8 630
545
254
241
246
219 568
196
?27 506
2IS 1002
24ft
21*
124
259
286
2/1
320
28«
316
244
252
Io5
167
197
438
613
41 <4
188
575
681
*53
637
6R8
530
617
526
632
336
357
344
3201
60,0
69|o
69,0
70,0
61,0
76.0
47.0
64,0
62.0
7J.O
46,0
65.0
41.0
feilo
s/.o
94,0
115 235
106.0
121.0
100,0
94,0
61.0
64,0
76, U
62,0
61.0
27,0
47,7
67.0
50.4
49.1
51.7
46,6
59.2
36.8
51.5
34./
32.6
59|l
32.4
*4,4
26.2
30.6
«»,«
80,0
85.0
69,0
61.0
87,0
86.0
74^0
84..0
81,0
3o|o
99,0
115,0
118,0
94,0
85.8
89,9
66.1
69*8
71.3
46.0
75.0
71.1
76.1
S7.4
70.6
?l|l
87.4
99. a
106.2
81.2
87,6
95.9
107.9
87.0
81.8
51.6
66^7
44.tt
20^0
2,1
2,2
2,«
2.4
2,1
2.2
2.2
2.2
2,0
2.2
2.1
1.7
1.9
2.«
2.5
2.6
2,0
U8
t.e
1.7
2.0
2.2
2.0
1.8
l.V
2,2
2.1
.8
.9
2.1
.7
.9
.5
.«
.9
. 7
.6
, 8
.8
1.8
1.8
1.7
2.1
2.1
c*
IN)
O
-------
j-JT» ilAfe
10TH 3UTH bOTH 70TM 90TH 95TH 991H WAX *«ITH GtO"
MTCBUGWAMS PtR CUBIC «t fe» "tAN MEAN
sto otv
PAGE 9
NE"
1974
33 '• •• 0 0 ',' H ii t. V „
3i l *Frt
33 .4/4»t..-iti TA
33
i" 3
wE*
fvE-v
Nfe,.
Nfc *.
NErt
3b
Sb
• JTl
C T '4
tlN
GIN
k'lt
l4i.Ov.P2
l':7 I
39
39
39 i14i-.-.V
19 /1KO,''V
a i 3 o 0 >j« -3
-a
OMl
UHi
OHJ
OH1
UH!
OKI
0-1
ORt
PP.r-f
Pf-
PE'i
HHU
••) J.
41 4 {.• 0 • '7 »* *''
'j| *C-r '? <'ui; "i»i.'
4J \v/fli^7 i'"Vj »ri.J
'14 • •> 4 •.' ' • > '. 1 f t'.; >
-* 6U JtM
,.-s ->n fft
."I 'v^L l'T *
I' ••» . 11 V I r.
19/a
3
3
3 190?
1974
1971
1974
197J
197U
1 1970
1 1971
1 1972
1
£
5 19*5
5
3
3
a
2
1
i.
3
3
4
i
a
1
I
i.
1
19/1,
19'0
1V /«
1<»74
1^73
I97b
7576
21
9
19
9
19
36
9
9
9
9
9
9
12
9
24
9
9
9
18
38
58
19
19
19
19
19
4
9
38
38
38
19
36
38
38
36
38
38
38
9
3H
33
9
38
32
9
9
9
19
26
43
9
39
30
49
75
21
9
g%
19
9
34
4V
32
30
so-
so
56
56
30
30
30
10
30
36
19
19
56
56
56
56
56
56
56
75
56
56
56
9
66
47
47
S3
53
9
V
9
Mi
07
62
2ft
60
75
75
94
39
28
45
30
26
51
62
43
75
56
56
56
5*
56
56
36
S6
36
38
' '96
38
30
75
56
56
56
75
56
56
75
56
75
75
26
05
66
66
66
71
9
9
9
56
66
66
45
61
94
107
113
62
43
62
45
47
7V
77
§6
94
75
56
56
75
?*
56
5b
5b
56
56
~ T*
. 56
58
75
7f
TS
75
. 94
'9*
T5
113
75
94
94
41
115
99
79
79
86
9
9
19
75
75
130
70
122
IbO
169
Jb9
100
73
68
71
6*
too
103
83
13*2
94.
94
113
94
1 13
yS
94
75
75
, 75
94
94
56
113
113
113
H?
152
132
113
150
132
132
132
fe4
160
122
109
11)9
|<>«
3fl
9
19
us
113
152
fll
145
1*8
205
180
1?6
86
103
63
107
115
1»7
94
160
113
113
1)2
113
112
99
113
94
«?4
, 94
\ 94
112
64
150
150
132
150
150
132
169
150
150
ISO
71
188
150
126
1?8
101
56
19
36
150
132
?01
109
?07
2H?
2»4
2*3
182
115
145
113
IbO
150
150
122
?07
169
ISO
{88
150
tb#
1 i2
1 32
150
132
131
150
150
79
226
207
169
207
907
?«fb
169
?44
226
168
207
94
254
id*
171
167
168
V4
7S
56
226
1*9
357
169,
470
451
641
470
320
186
•gSJj
14M
372
321
Mr
2:1$
316
470
376
111
320
,«*!
451
3?0
451
244
357
391
357
15*
602
489
37t>
AY2
376
263
39«>
545
432
470
150
4*9
329
301
244
JHb
330
2?6
207
395
282
70.0
35,0
66.0
80,0
8/.0
95,0
50.0
35.0
«9,0
35,0
38,0
55,0
66,0
47,0
76, C
57,0
57,0
60,0
65,0
69,0
5.i,0
50,0
54.0
47,0
•7.0
55,0
47,0
31,0
73.0
71,0
69,0
73,0
flw.O
74,0
69,0
91.0
72.0
74.0
8S.O
3J.O
95,0
75.0
66.0
70,0
71.0
16.0
12,0
15,0
66,0
66.0
56.2
25,9
51.5
57,0
64.6
79.5
33.0
26.0
57,"
26.6
24,9
"3,1
59.0
59.0
62.9
*6, t
«7.2
48.7
59,0
St.*
46,3
01.6
46.7
«0.0
«0.9
*7«2
37.4
59*0
59.5
59.7
*».*3
61,7
59.0
80,5
60.3
62.9
72.6
24.0
81.0
6^.4
52.5
63, /
66.2
lo|i
12.0
53.4
56.6
2.1
2.2
2.2
2.5
2.4
2.0
2.5
2,2
2.*2
1.6
2,0
2.0
2.1
2.0
2,0
1.6
i.'e
il?
1.7
1.7
1.0
2.0
1.9
2.0
2.0
1.6
1,0
i.a
1.9
1.9
1.7
1.9
1.9
I./
2,1
1.0
1.8
lib
1,7
1.0
ill
2.1
1.0
-------
322
APPENDIX B
DERIVATION OF FORMULAS FOR DISTRIBUTION OF MAXIMA
General
Let an individual hourly concentration, X, have a cumulative
frequency distribution
P(C) = probability that X] (B-3)
Gamma Distribution
For the Gamma distribution, P(C) equals G(C), where
-------
. _L /"
rta) J
323
C/3
I \ J - dv
(a) JQ
Changing variables to t - C/e,
"'IT-) /" v"1
dv. (B-5)
t
Since we are examining maximal values, we can assume t is large.
For large t,
G(C) ~ 1 - Y^T t06'1 e't .
Using Equation (B-2), the distribution of the maximum is
-N o-l e-t
) = e 1W (B-6)
where t = C /3-
To make the distribution of the maxima independent of both e and a, let
S = t - A = Cm/e - A (B-7)
where
M A01'1 e"A = 1 • (B-8)
ra
-------
324
Then, substituting (B-7) and (B-8) into Equation (B-6) yields
M(Cm) =exp [-(f + I)"-1 e's] (B-9)
For the data base in question, a tends to be near to one, and s tends to
be small compared with A. Thus, we use the approximation
With this approximation, the formula for the distribution of maxima from
Gamma distribution is
M(C ) = e~e (Ml)
where
s = c / g - A
and A is the solution to
e'A - 1
e
-------
325
APPENDIX C
DATA FOR CHARACTERIZING PRESENT N0g AIR QUALITY
STATION
1. Phoenix, AZ
(002A01 )
2. Anaheim, CA
(001 101)
3. Azusa, CA
(002101 )
4. Bakersfield, CA
(003F01)
5. Barstow. CA
(001 101 ]
6. Burbank, CA
(002101 )
7. Camarillo, CA
(001 101)
8. Chico, CA
(001 F01)
9. Chi no, CA
(001 101)
10.' Concord, CA
(001 101)
11. Costa Mesa, CA
(001 101)
12. El Cajon, CA
(001 101)
13. Eureka, CA
(002F01 )
YEARS
OF DATA
73
72,73,74
72,73,74
72,73,74
74
74
72,73
72,73,74
74 .
74
74
72,73
73
ARITHMETIC
MEAN
"in"
(ppnm)
1.9
5.1
6.2
3.0
4.0
7.1
3.0
1.8
3.4
2.7
3.0
3.0
I-7
90TH
PERCENTILE
11 f* 11
^90
(pphm)
5.6
9.0
11.4
5.7
8.0
13.0
5.5
4.0
7.0
5.0
7.0
5.5
3.0
YEARLY MAX
ONE-HOUR CONC.
11 y ii
*m
(pphm)
22.7
40.7
41.0
14.7
47.7
35.4
20.8
10.6
37.7
20.5
30.8
' 22.2
10.3
MAX- TO-
MEAN
RATIO
V
12.2
7.9
6.6
5.0
12.0
5.0
6.9 -
5.8
11. 1
7.5
10.2
7.5
6.0
t
-------
326
STATION
14. Fresno, CA
(002F01 )
15. Indio, CA
(001 101)
16. La Habra, CA
(001 1 01)
17. Lancaster, CA
(001 1 01)
18. Lennox, CA
(001 101)
19. Livermore, CA
(002101)
20. Long Beach, CA
(002101)
21. Los Alamitos, CA
(001 101)
22. Los Angeles (Down-
town), CA
. (001 I 01)
23. Los Angeles (West-
wood), CA
(002101)
24. Los Angeles (Reseda]
(001101)
25. Lynwood, CA
(001 101)
YEARS
OF DATA
72,73,74
72,73,74
72,73,74
72,73,74
72,73,74
72,73,74
74
72
72,73,74
72,73,74
72,73,74
74
ARITHMETIC
MEAN
"m"
(pphm)
2.7
1.7
5.6
1.4
6.4
3.3
6.7
4.7
7.3
6.8
6.3
5.5
90TH
PERCENT I LE
UP n
L90
(pphm)
5.0
3.3
9.7
2.7
11.0
5.7
12.0
9.0
12.3
12.3
11.7
9.0
YEARLY MAX
ONE- HOUR CONC.
My i»
Am
(pphrn)
20.0
11.2
42.9
9.4
40.7
17.2
37.7
36.3
54.6
55.8
• 36.7
37.7
MAX -TO -
MEAN
RATIO
Xm*m
7.4
6.8
7.7
6.8
6.4
5.2
5.6
7.8
7.6
8.1
5.8
6.9
-------
327
STATION
26. Modesto, CA
(001 101)
27. Monterey, CA
(001 101)
28. Napa, CA
(003101 )
29. Newhall, CA
(001 101)
30. Norco, CA
(001 101)
31. Oakland, CA
(003601 )
32. Ojai, CA
(001 101)
33. Palm Springs, CA
(001 101)
34. Pasadena, CA
(004101 ) ,
35. Pittsburg, CA
(001 I 01)
36. Pomona, CA
(001 101)
37. Redding, CA
(002F01 )
38. Redlands, CA
(001101)
YEARS
OF DATA
72,73,74
72,73,74
74
72,73,74
74
72,73,74
72
72,73,74
74
f2,73,74
74
r2,73,74
'2,73,74
ARITHMETIC
MEAN
"m"
(pphm)
2.7
1.5
2.6
3.6
2.8
3.5
1.5
1.5
7.3
1.9
6.9
1.8
4.0
90TH
PERCENTILE
HP ii
U90
(pphm)
4.3
2.7
4.0
6.7
5.0
6.0
4.0
3.0
12.0
3.7
11.0
3.0
7.7
YEARLY MAX
ONE- HOUR CONC.
V
(pphm)
14,2
10.8
14.1
23.0
22.4
22.7
19.9
. 8.8
47.7
10.2
34.3
9.9
24.7
MAX- TO -
MEAN
RATIO
V"
5.2
7.3
5.4
6.3
7.9
6.6
13.4
6.1
6.5
5.5
5.0
5.4
6.2
-------
328
,1
STATION
39. Redwood City, CA
(001 1 01)
40. Richmond, CA
(003101)
41 . Riverside, CA
(003F01 )
42. Rubidoux, CA
(001101)
43. Sacramento, CA
(003F01 )
44. Salinas, CA
(001 101)
45. San Bernardino, CA
(001 1 01)
46. San Diego, CA
(004101 )
47. San Francisco, CA
(003101 )
48. San Jose, CA
(004A05)
49. San Luis Obispo, C/
(001 F01)
50. San Rafael, CA
(001 101)
£1. Santa Barbara, CA
(002F01 )
52. Santa Barbara, CA
(004F01 )
YEARS
OF DATA
72,73,74
74
74
74
72,73,74
72,73,74
72,73,74
74
72,73,74
73,74
72,73,74
72,73,74
72
72,73,74
/
ARITHMETIC
MEAN
"m"
(pphm)
2.7
2.8
5.0
2.7
2.8
2.2
4.3
2.7
3.3
3.8
2.1
2.8
3.6
3.1
90TH
PERCENTILE
up M
U90
(pphm)
5.0'
5.0
9.0
5.0
4.7
4.0
7.7
6.0
5.0
6.5
4.0
4.7
6.0
5.0
YEARLY MAX
ONE- HOUR CONC.
11 v I'
*m
(pphm)
21.4
14.2
25.5
20.3
18.0
13.5
32.0
25.6
23.9
30.2
10.9
16.9
16.4
20.5
MAX -TO.
MEAN
RATIO
X^m
8.0
5.1
5.1
7.6
6.5
6.2
7.3 '
9.6
7.2
8.0
5.3
6.0
4.6
6.6
-------
329
STATION
53. Santa Cruz, CA
(0011 01)
54. Santa Rosa, CA
(002101)
55. Stockton, CA
(002F01 )
56. Sunnyvale, CA
(001 1 01)
57. Upland, CA
(003101)
58. Upland, CA
(004F01 )
59. Vail e jo, CA
(003101)
60. Victorvilie, CA
(001 1 01)
61. Visalia, CA
(001 F01)
62. Whittler, CA
(001 101)
63. Yuba City, CA
(001F01)
64. Denver, CO
(002A05)
65. New Britain, CT
(002F01)
YEARS
OF DATA
72,73,74
74
72,73,74
74
74
74
74
74
72,73,74
72,73,74
72,73,74
74,72,74
73,74
ARITHMETIC
MEAN
"m"
(pphm)
1.5
2.0
2.7
4.1
6.0
4.9
2.6
3.7
2.3
6.5
1.9
4.4
1.8
90TH
PERCENT I LE
tlr» II
C90
(pphm)
2.3
4.0
4.3
7.0
11.0
9.0
4.0
8.0
4.0
11.3
3.7
7.3
3.2
YEARLY MAX
ONE-HOUR CONG.
II V 'I
*m
(pphm)
10.5
15.2
15.8
31.7
39.7
28.6
14.2
23.5
12.6
50.6
14.6
40.2
10.1
MAX- TO-
MEAN
RATIO
Xm*m
6.9
7.8
5.8
7.7
6.7
5.8
5.4
6.3
5.6
7.8
7.6
9.2
5.6
-------
330
STATION
66. Washington, DC
(003A05)
67. Atlanta, GA
(001A05)
68. Chicago, IL
(002A05)
69. Chicago, IL
(023A05)
70. Ashland, KY
(008F01 )
71. Louisville. KY
(011G01)
72. Louisville, KY
(017A05)
73. Newport, KY
(001 F01)
74. Ohio, KY
(006N02)
75. Owens boro, KY
(008F01 )
76. Baltimore, MD
(018F01 )
77. Silver Spring, MD
(006F01 )
YEARS
OF DATA
74
74
74,72,73
74
74
73,74
74
72,73,74
73
73
73
73
ARITHMETIC
MEAN
"m"
(pphm)
3.6
4.8
5.7
2.6
3.8
4.3
3.6
3.7
0.7
4.6
6.4
5.2
90TH
PERCENTILE
up it
(pphm)
6.0
7.5
9.7
5.0
7.0
6.8
6.0
6.2
1.0
8.5
11.0
10.0
YEARLY MAX
ONE-HOUR CONC.
'V
(pphm)
17.1
25.1
27.7
14.2
41.6
22.6
17.1
19.7
13.1
35.5
51.9
,
45.1
MAX- TO -
MEAN
RATIO
Xm ' m
4.7
5.2
4.8
5.6
11.0
5.3
4.7
5.5
18.9
7.7
8.1
8.8
-------
331
STATION
78. Springfield, MA
(005A05)
79. Detroit, MI
(020A05)
80. Lansing, MI
(002F01)
81. Saginaw, MI
(002F01)
82. Afton, MO
(001601)
83. BelleFontaine
Neighbors, MO
(002601)
84. Clayton, MO
(001601)
85. St. Ann, MO
(001 601)
86. St. Louis, MO
(002A10)
87. St. Louis, MO
(006601 )
88. Las Vegas, NV
(009601 )
89. Reno, NV
(005101 )
YEARS
OF DATA
74
74
74
74
73
73,74
73
73,74
74
73,74
72, 73
73,74
ARITHMETIC
MEAN
"m"
(pphm)
5.9
2.8
3.8
3.2
4.5
3.7
3.7
3.6
3.8
3.0
2.3
3.2
90TH
PERCENTILE
"Cgg"
(pphm)
11.0
5.0
6.3
5.7
8.1
7.3
7.1
6.7
6.0
6.0
4.9
5.5
YEARLY MAX
ONE-HOUR CONC.
(pphm)
28.6
15.4
18.0
17.9
24.9
26.6
25.2
25.4
22.9
33.6
18.8
26.6
MAX-TO-
MEAN
RATIO
Xm * m
4.9
5.5
4.7
5.6
5.5
7.3
.
6.8
7.1
6.1
11.5
8.4
8.5
-------
332
STATION
90. Bayonne, NJ
(003F01 )
91. Camden, NJ
(003F01)
92. Elizabeth, NJ
(004F01 )
93. Newark, NJ
(002F01 )
94. Phillipsburg, NJ
(002F01)
95. Buffalo, NY
(005F01)
96. Buffalo, NY
(007F01 )
97. Glens Falls, NY
(003F01)
98. Hemps tead, NY
(005F01 )
99. Kingston, NY
(002F01 )
100. Mamaroneck, NY
(002F01 )
101. New York City, NY
(006A05)
102. New York City, NY
(050F01)
YEARS
OF DATA
72,73,74
72,73,74
74
72,73,74
72,73,74
74
74
74
74
74
74
74
74
ARITHMETIC
MEAN
"m"
(pphm)
4.2
4.3
5.3
5.6
3.6
3.3
3.2
1.4
3.7
1.9
3.5
4.3
4.6
90TH
PERCENTILE
Mr. II
C90
(pphm)
7.4
7.4
8.5
9.2
5.9
6.6
6.0
2.9
6.9
3.7
6.5
8.0
9.0
YEARLY MAX
ONE-HOUR CONC.
"V
(pphm)
23.9
26.4
31.1
31.3
18.8
17.8
13.6
12.9
19.5
9.2
25.6
25.6
34.8
MAX- TO-
MEAN
RATIO
V
5.7
6.1
5.9
5.6
5.3
5.4
4.3
9.0
5.3
4.8
7.3
6.0
7.5
-------
333
STATION
103. New York City, NY
(061A05)
104. Niagara Falls, NY
(006F01 )
105. Rensselaer, NY
(001 F01)
106. Rochester, NY
(004F01 )
107. Schenectady. NY
(003F01 )
108. Syracuse, NY
(005F01 )
109. Syracuse, NY
(011F01)
110. Utica, NY
(004F01 )
111. Akron, OH
(013H01)
112. Cincinnati, OH
(019A05)
113. Portland, OR
(002F01 )
114. Lancaster City, PA
(007F01 }
115. Philadelphia, PA
(002A05)
YEARS
OF DATA
74
74
74
74
74
74
74
74
73
74
72,73,74
74
73,74
ARITHMETIC
MEAN
"m"
(pphm)
5.1
2.7
1.9
2.6
1.9
2.9
3.5
2.5
4.0
2.7
2.6
1.7
3.9
90TH
PERCENTILE
n r ii
L90
(pphm)
9.0
5.3
3.9
4.7
3.8
5.3
5.5
4.4
7.0
5.0
4.7
3.0
7.0
. YEARLY MAX
ONE- HOUR CONC.
II V II
*m
(pphm)
25.4
17.8
10.2
11.5
9.4
17.6
13.3
13.3
18.9
17.3
18.3
8.8
27.2
MAX- TO-
MEAN
RATIO
Xm"m
5.0
6.6
5.5
4.4
5.0
6.0
3.8
5.3
4.7
6.5
7.0
5.3
7.0
-------
334
STATION
116. Philadelphia, PA
(004H01 )
117. Scranton, PA
(006F01 )
118. Providence, RI
(005F01 )
119. Providence, RI
(007A05)
120. Memphis, TN
(027N02)
121. Stewart, TN
(005N02)
122. Salt Lake City, UT
(001A05)
123. Alexandria, VA
(009H01)
YEARS
OF DATA
72
74
72,73
72,73,74
74
73,74
74
74
ARITHMETIC
MEAN
"m"
(pphm)
4.5
1.7
4.5
3.7
0.9
0.7
3.6
3.6
90TH
PERCENTILE
»P ii
L90
(pphrn)
!•
7.0
3.4
7.5
6.1
2.0
0.7
6.0
6.0
YEARLY MAX
ONE-HOUR CONC.
ii y M
*m
(pphm)
25.1
8.4
22.3
17.1
18.2
11.9
21.6
15.4
MAX- TO-
MEAN
RATIO
Xm*m
5.6
5.1
4". 9
4.6
21.4
16.8
6.0
4.3
-------
335
APPENDIX D
SUMMARY OF DAYTIME AND NIGHTTIME REGRESSION
MODELS FOR LENNOX, AZUSA, POMONA. DENVER,
CHICAGO, HOUSTON/MAE, AND HOUSTON/ALDINE
-------
336
Table D-l Summary of Daytime Regressions for Lennox
1. Regression of Daytime N02 vs. N025 and INTNO
DPKN0
or
DAVN02
TOTAL
B?- INTNO
WINTER
DPKN02
DAVN02
SUMMER
DPKN02
DAVM02
CORR. % VARIAflCE
COEF. EXPLAINED
.76 57%
.78 61%
.74 55%
.78 60%
A
2.4
2.2
3.4
2.7
B!
.70
.46
.70
.47
B2
.14
.08
.24
.13 -
2. Estimation of the Hydrocarbon Effect
DPKN02
or
DAVN02
(A + CQ)
TOTAL
CORR. % VARIANCE
+ C^RATIO + c2-NKHCPR)
WINTER
DPKN02
DAVM02
SUMMER
DPKH02
DAVN02
COEF. EXPLAINED
.77 59%
.78 62%
.
.77 59%
.78 61%
2.6
2.3
3.5
2.7
Bl
.70
.46
.70
.47
B2
.08
.06
.09
.10
C1
'.on
.004
.021
.005
C2
*
*
.00011
*
Not significant from zero at 95% confidence level.
Note: NMHCPR = (HC69 - 100 pphm)/2
Units of all variables are in pphm.
-------
337
Table D-2 Summary of Nighttime Regressions for Lennox
1- Regression of Nighttime N02 vs. N0216, NITENO, and 0,AFT-NITENO
NPKN02
or » A + B,.f
NAVNOo
+ B2-NITtNO + Bg-NITENO-Oj AFT
TOTAL
WINTER
NPKN02
NAVN02
SUMMER
NPKN02
NAVN02
CORR. 2 VARIANCE
COEF. EXPLAINED
.81 66%
.75 56%
.80 65%
.74 55%
A
2.4
2.4
2.0
1.5
B1
.77
.48
.77
.48
B? 1
.04
*
*
.06
B3
.026
.020
.048
.021
Not significant from zero at 95% confidence level.
2. Dependence of Afternoon N02 (N0216) on MMHC/NOX Ratio
WINTER; N0216 = 8.3 pphm (1 - .028
SUMMER: N0216 = 6.4 pphm (1 -
Note: Units of all variables are in pphm.
-------
338
Table D-3 Summary of Daytime Regressions for Azusa
1. Regression of Daytime N02 vs. N025 and INTNO
DPKN00
or
DAVNO,
A +
+ B2- INTNO
TOTAL
WINTER
DPKN02
DAVN02
SUMMER
DPKN02
DAVM02
CORR. % VARIANCE
COEF. EXPLAINED
.87 75%
.88 78%
.88 78%
.85 73%
A
1.6
1.0
1.4
1.7
Bl
1.05
.78
.92
.56
B2
.52
.33
.57
.30 '
2. Estimation of the Hydrocarbon Effect
DPKN02
or
DAVN02
(A + C0)
TOTAL
CORR. % VARIANCE
INTNO-
C^RATIO
WINTER
DPKN02
DAVN02
SUMMER
DPKN02
DAVNO£
COEF. EXPLAINED
.88 77%
.89 79%
.88 78%
.86 73%
A + co
1.7
1.1
1.4
1.7
Bl
1.05
.78
.92
.56
B2
.21
.14
.57
.32
C1
.010
.006
*
*
C2
.00099
.00060
*
-.00016
Not significant from zero at 95% confidence level.
Note: NMHCPR - (HC69-100 pphm)/2
Units of all variables are in pphm.
-------
339
Table D-4 Summary of Nighttime Regressions for Azusa
1. Regression of Nighttime N0£ vs. N0216, NITENO, and 0,AFT-NITENO
NPKN02
or - A + BrN0216 + Bg-NITENO + B^NITENO-Oj AFT
NAVNO,
TOTAL
WINTER
NPKN02
NAVN02
SUMMER
NPKN02
NAVN02
CORR. % VARIANCE
COEF. EXPLAINED
.91 84%
.90 80%
•
.74 54%
.76 58Z
A
1.5
0.9
2.7
1.7
BJ
.82
.48
.82
.61
B,
.44
.16
.34
.12
B3
.042-
.035
.042
.033
2. Dependence of Afternoon N02 (N0216) on NHHC/MOX Ratio
WINTER; N0216 = 9.9 pphm (1 - .017
NMHCPR
SUMMER: N0216 = 5.8 pphm (1 - .011 NQX69)
Note: Units of all variables are in pphm.
-------
340
Table D-5 Summary of Daytime Regressions for Pomona
1. Regression of Daytime N02 vs. N025 and INTNO
DPKM0
or
DAVNOj
TOTAL
A +
INTNO
WINTER
DPKNO?
DAVN02
SUMMER
DPKN02
DAVN02
CORR. 55 VARIANCE
COEF. EXPLAINED
.83 69%
.84 70%
.85 73%
.85 73%
A
.5
.7
2.0
2.1
Bl
1.14
.87
.90
.58
B2
.12
.07
.30
.19
2. Estimation of the Hydrocarbon Effect
DPKN02
or
DAVN02
(A + CQ) + B^NOgB + INTNO-(B2 + C^RATIO +
TOTAL
CORR. % VARIANCE
WINTER
DPKN02
DAVN02
SUMMER
DPKN02
DAVN02
COEF. EXPLAINED
.86 74X
.86 74%
.87 75%
.86 74%
A*C0
.5
.8
2.2
2.2
Bl
1.14
.87
.90
.58
B2
-.14
-.09
.03
.05
Cl
.020
.012
.018
.010
C2
.00100
.00065
.00119
.00056
Note: NMHCPR = (HC69 - 100 pphm)/2
Units of all variables are in pphm.
-------
341
Table D-6 Summary of Nighttime Regressions for Pomona
1. Regression of Nighttime N02 vs. N0216, NITENO, and 03AFT-
NITEND
NPKN02
or = A + B,.l
NAVNOo
+ B2-NITENO + B3-NITENO-63 AFT
WINTER
NPKN02
NAVN02
SUMMER
NPKN02
NAVN02
TOTAL
CORR. % VARIANCE
COEF. EXPLAINED
.87 75%
.84 71%
.71 50%
.71 51%
A
1.6
2.4
3.1
2.4
B3
.85
,50
.81
.57
B?
*
-.05
*
*
B3
.067'
.043
.058
.041
Not significant from zero at 95% confidence level.
2. Dependence of Afternoon N02 (N0216) on NMHC/NO Ratio
WINTER; N0216 = 9.9 pphm (1 - .027
SUMMER: N0216 = 7.3 pphm (1 - .018
Note: Units of all variables are in pphm.
-------
342
Table D-7 Summary of Daytime Regressions for Denver
1. Regression of Daytime N02 vs. N025 and INTNO
DPKN0
or = A +
DAVNO,
+ Bg-INTNO
TOTAL
WINTER
DPKN02
DAVN02
SUMMER
DPKN02
DAVN02
CORR. % VARIAHCE
COEF. EXPLAINED
.65 42%
.70 49%
.70 50%
.72 51%
A
1.5
1.3
1.5
1.1.
B]
.71
.41
.58
.33
B2
.32
.17
,
.37
.18
2. Estimation of the Hydrocarbon Effect
DPKN02
or
DAVNCfc
(A + CQ) + Bj-NO^ + INTNO-
TOTAL
CORR. % VARIANCE
-RATIO +
HINTER
DPKN02
DAVH02
SUMMER
DPKN02
DAVN02
COEF. EXPLAINED
.66 44%
.73 53%
.71 51%
.73 53%
A + co
1.6
1.6
1.6
1.3
Bl
.71
.41
.58
.33
B2
.23
.23
.28
.12
Cl
' .010
-.005
.007
.004
C2
*
-.00016
*
*
*
Not significant from zero at 95% confidence level.
Note: For Denver, NMHCPR is defined as .6(HC69-135 pphm). This formula results
from regressing the Denver HMHC69 measurements against the Denver HC69
measurements.
Units of all variables are in pphm.
-------
343
Table D-8 Summary of Nighttime Regressions for Denver
1. Regression of Nighttime N02 vs. N0216, NITENO, and O^FT-NITENO
NPKNOg
or -A
NAVNO,
lS + B2-NITENO + B3.NITENO-63 AFT
TOTAL
WINTER
NPKN02
NAVN02
SUMMER
NPKN02
NAVN02
CORR. % VARIANCE
COEF. EXPLAINED
.82 67%
.78 60%
•
.50 2555
.55 30*
A
2.0
2.0
•
3.2
1.7
Bl
.73
.35
.66
.39
B,
.16
.10
*
.09
B3
*
*
*
*
Not significant from zero at 95% confidence level.
2. Dependence of Afternoon N02 (NQ216) on NMHC/NOX Ratio
WINTER: N0jl6 = 7.56 pphm (1 - .044 j
SUMMER: N0216 = 2.70 pphm (1 - .007
Note: Units of all variables are in pphm.
-------
344
Table D-9 Summary of Daytime Regressions for Chicago
1. Regression of Daytime N02 vs. N025 and INTNO
DPKN02
or « A + B.j-N025 + BZ> INTNO
DAVNO,
TOTAL
WINTER
DPKN02
DAVH02
SUMMER
DPKN02
DAVN02
CORK. % VARIANCE
COEF. EXPLAINED
.65 43*
.74 54%
.70 49%
.78 60%
A
1.7
1.5
2.0
1.3
Bl
.86
.66
.98
.82
B2
.07
.05
.14
.11 -
2. Estimation of the Hydrocarbon Effect
DPKNO-
or
DAVNOo
(A + CQ) + Bj'HOgS + INTNO-(Bg + C,-RATIO + Cg-NKHCPR)
TOTAL
CORR. X VARIANCE
WINTER
DPKN02
DAVN02
SUMMER
DPKN02
DAVN02
COEF. EXPLAINED
.70 49«
.76 58%
.70 49%
.78 60%
A + co
1.6
1.5
2.0
1.3
Bl
.86
.66
.98
.82
B2
-.01
.01
.14
.11
Cl
.029
.015
*
*
C2
*
*
*
*
Not significant from zero at 95% confidence level.
Note: NMHCPR = .57(HC69-144 pphm)
Units of all variables are in pphm.
-------
345
Table D-1Q Summary of Nighttime Regressions for Chicago
1- Regression of Nighttime N02 vs. N0216, NITENO, and O/FT-NITENO
i
NPKN02
or = A + BrN0216 + Bg-NITENO + Bg-NITENO-Og AFT
NAVNOo
WINTER
NPKN02
NAVN02
SUITER
NPKN02
NAVN02
TOTAL
CORR. % VARIANCE
COEF. EXPLAINED
.86 75%
.80 63%
.90 80%
.84 70%
A
1.0
1.3
1.3
1.4
!i
.78
.49
.89
.58
B?
.03
.04
*
.02
B3
-.007
-.006
.014
.008
Not significant from zero at 95% confidence level.
2. Dependence of Afternoon N02 (N0216) on NMHC/NOX Ratio
WINTER; NQ2i6 independent of
NMHCPR
SUMMER: N0216 = 9.6 pphm (1 -
NMHCPR.
NOX79;
Note: Units of all variables are in pphm.
-------
346
Table D-ll Summary of Daytime Regressions for Houston/Mae
1. Regression of Daytime N02 vs. N025 and INTNO
DPKN0
or = A +
DAVN02
TOTAL
+ B- INTNO
WINTER
DPKN02
DAVN02
SUMMER
DPKN02
DAVN02
CORR. 2 VARIANCE
COEF. EXPLAINED
.77 60%
.80 65%
.79 62%
.75 57%
A
.93
.55
1.23
.81
Bl
.86
.48
1.00
.52
B2
.15
.074
.092
.033-
2. Estimation of the Hydrocarbon Effect
DPKN02
or
DAVN02
= (A + CQ) + Bj-NOgS + INTNO- (B^ + C^ RATIO + Cg-NMHCPR)
TOTAL
CORR. % VARIANCE
WINTER
DPKN02
DAVN02
SUMMER
DPKN02
DAVN02
COEF. EXPLAINED
.77 60%
.80 65%
.79 62%
.75 57%
A + C0
.93
.55
1.23
.81
Bl
.86
.48
1.00
.52
B2
.15
.074
.092
.033
Cl
•
*
*
*
*
C2
*
*
*
*
Not significant from zero at 95% confidence level.
Note: NMHCPR = .38CHC69 - 70 pphm]
Units of all variables are in pphm.
-------
347
Table D-12 Summary of Nighttime Regressions for Houston/Mac
1. Regression of Nighttime N02 vs. N0216, NITENO, and 03AFT-NITENO
NPKNO,
or
= A + B.j.N0216 + B2-NITENO + B3'NITENO-63 AFT
NAVNO
WINTER
NPKN02
NAVN02
SUMMER
NPKN02
NAVN02
TOTAL
CORR. % VARIANCE
COEF. EXPLAINED
.77 60%
.76 57%
.41 16%
.37 14%
A
2.57
1.68
3.00
1.74
Bi
.43
.27
.53
.30
B? 1
*
-.11
*
*
B3
.096
.074
.039
*
Not significant from zero at 95% confidence level.
2. Dependence of Afternoon N02 (N0216) on NMHC/NOX Ratio
WINTER; N0216 = 2.7 pphm (1 - .010
SUMMER; N0216 = 2.4 pphm (1 -
,, NMHCPR)
•UM NOX69;
Note: Units of all variables are in pphm.
-------
348
Table D-13 Summary of Daytime Regressions for Houston/Aldine
1. Regression of Daytime N02 vs. N0g5 and INTNO
OPKN0
or
DAVN02
TOTAL
INTNO
WINTER
DPKN02
DAVN02
SUMMER
DPKN02
DAVM02
IUKK. » VAKl/UIlt:
COEF. EXPLAINED
.75 57*
.70 49*
.67 45*
.67 45%
A
.96
.52
1.02
.41
Bl
1
.68
.40
.50
.21
By
Z
.19
.050
.12
.080-
2. Estimation of the Hydrocarbon Effect
DPKN02
or
DAVN02
(A + C) +
+ INTNO- (B^ + C^RATIO + Cg-NKHCPR)
TOTAL
CORR. % VARIANCE
WINTER
DPKN02
DAVH02
SUMMER
DPKM02
DAVNOg
COEF. EXPLAINED
.75 57*
.70 492
.67 45X
.67 45X
A + C0
.96
.52
1.02
.41
Bl
.68
.40
.50
.21
B2 1
.19
.050
.12
.080
Cl 1
' *
*
*
*
C2
*
*
*
*
Not significant from zero at 95% confidence level.
Note: NMHCPR • .5[HC - 133 pphm]
Units of all variables are in pphm.
-------
349
Table D-14 Summary of Nighttime Regressions for Houston/Aldine
1. Regression of Nighttime N02 vs. N0216, NITENO, and O/FT-NITENO
NPKNO,
or •« A + B.,.N0216 + B2-NITENO + B3«NITENO-63 AFT
WINTER
NPKN02
NAVN02
SUMMER
NPKN02
NAVN02
TOTAL
CORR. % VARIANCE
COEF. EXPLAINED
.59 34*
.56 322
* *
* *
A
2.82
1.78
*
*
BI
*
*
*
*
B? ,
*
-.28
*
*
B3
.119 '
.090
*
*
Not significant from zero at 95X confidence level.
2. Dependence of Afternoon N02 (N0216) on NMHC/NOX Ratio
; No significant relationship between N0,16
and NMHCPR/NOX69
: No significant relationship between N0216
and NMHCPR/NOX69
Note: Units of all variables are 1n pphm.
-------
350
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
1. REPORT NO.
EPA-600/3-78-018
3. RECIPIENT'S ACCESSION NO.
4. TTB1E AND SUBTITLE
EMPIRICAL RELATIONSHIPS BETWEEN ATMOSPHERIC NITROGEN
DIOXIDE AND ITS PRECURSORS
6, REPORT DATE
February 1978
6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
J. Trijonis
8. PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING ORG -VNIZATION NAME AND ADDRESS
Technology Service Corporation
2811 Wilshire Boulevard
Santa Monica, CA 90403
10. PROGRAM ELEMENT NO.
1AA603 AC-09 (FY-77)
11. CONTRACT/GRANT NO.
68-02-2299
12. SPONSORING AGENCY NAME AND ADDRESS
Environmental Sciences Research Laboratory - RTP, NC
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
13. TYPE OF REPORT AND PERIOD COVERED
Final _
14. SPONSORING AGENCY CODE
EPA/600/09
15. SUPPLEMENTARY NOTES
16. ABSTRACT
Aerometric data were examined to define relationships between atmospheric H
and its precursors.. A descriptive and critical analysis of the nationwide data
base of N0? was carried out, followed by the formulation application and testing
of empirical models relating ambient N02 changes to NO .and hydrocarbon (HC)
emission controls.
The examination showed that (1) other factors being constant, annual mean
and yearly maximum N0? are proportional to NO input; (2) HC control yields
slight-to-moderate reauctions in yearly maximOm N0«; (3) HC control yields
essentially negligible benefits for annual mean N02; and (4) the exact form
of the N0?/precursor relationship may vary somewhat from one location to the
next, depending on local conditions.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.lDENTIFIERS/OPEN ENDED TERMS
c. COSATI Field/Group
*Air pollution
*Nitrogen oxides
*Nitrogen dioxide
*Hydrocarbons
*Empirical equations
*Atmospheric models
13B
07B
07C
12A
04A
8. DISTRIBUTION STATEMENT
RELEASE TO PUBLIC
19. SECURITY CLASS (This Report)
UNCLASSIFIED
21. NO. OF PAGES
368
20. SECURITY CLASS (Thispage)
UNCLASSIFIED
22. PRICE
EPA Form 2220-1 (9-73)
------- |