JULY 1986
NUMERICAL SIMULATIONS OF PHOTOCHEMICAL AIR POLLUTION
IN THE NORTHEASTERN UNITED STATES: ROMl APPLICATIONS
ATMOSPHERIC SCIENCES RESEARCH LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
RESEARCH TRIANGLE PARK, NC 27711
-------
NUMERICAL SIMULATIONS OF PHOTOCHEMICAL AIR POLLUTION
IN THE NORTHEASTERN UNITED STATES: ROM1 APPLICATIONS
Robert G. Lamb
Meteorology and Assessment Division
Atmospheric Sciences Research Laboratory
Research Triangle Park, North Carolina 27711
ATMOSPHERIC SCIENCES RESEARCH LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
RESEARCH TRIANGLE PARK, NC 27711
-------
NOTICE
The information in this document has been funded by
the United States Environmental Protection Agency.
It has been subject to the Agency's peer and admin-
istrative review, and it has been approved for
publication as an EPA document. Mention of trade
names or commercial products does not constitute
endorsement or recommendation for use.
The author, Robert G. Lamb, is on assignment to the Atmospheric
Sciences Research Laboratory, U.S. Environmental Protection Agency, from
the National Oceanic and Atmospheric Administration, U.S. Department
of Commerce.
ii
-------
ABSTRACT
The first-generation Regional Oxidant Model (ROM1) was used to
simulate pollutant concentrations during the nine-day period 23-31 July
1980. Two simulations were performed. The first, which is considered to
be the base case, used the 1980 NAPAP 4.2 inventory for all hydrocarbon
and NOX emission rates. The second simulation, or control case, was
identical in all respects except that the county-by-county hydrocarbon
and NOX emissions rates were modified in accordance with baseline
projections for 1987 contained in SIPs. The one-hour and daily daylight
(0900-1600 LST) averaged ozone concentrations produced in each simulation
were compared to assess the effectiveness of the proposed emissions
changes on air quality.
It was found that ozone concentrations in the control case were
everywhere lower than those in the base case, but the percentage reduction
was not uniform in space. In areas near the major VOC and NOX sources,
the maximum one-hour averaged ozone levels were reduced by about 25%
while in areas farther than 100 km from these sources peak values were
only about 10% lower. Slightly smaller percentage reductions were found
in the daily daylight averaged ozone concentrations. It was also found
that the emissions reductions lowered peak ozone concentrations by con-
siderably larger percentages than they reduced the median or mean
concentration values.
The analyses of the model results are prefaced by discussions of a
number of basic issues on regional scale modeling, including model
-------
initialization, selection of meteorological data, effects of grid size on
model performance, estimating long-term concentration statistics from
short-period simulations, probabilistic vs quasi-deterministic modes of
model operation, uncertainty in emissions estimates, the characteristics
of VOC and NOX sources in the Northeast, and other topics. Preliminary
results of analyses of the SAROAD monitoring data, which reveal the
characteristics of the ozone problem in the northeastern United States,
set the stage for the model simulations.
-------
CONTENTS
Abstract ill
Figures vi
Tables x
Acknowledgments xi
1. Introduction 1
2. Overview of the Regional Oxidant Model 2
3. Characteristics of Observed Ozone Concentrations
in the Northeastern United States 11
4. The Sources of Ozone Precursor Species in the
Northeastern United States 26
The spatial scales of the VOC and NOX sources 35
Uncertainty in gridded NOX and VOC emissions 39
5. Definitions of Receptor Classes for Use in
Analyzing Simulated Concentrations 47
6. Meteorology of the Test Period: 23 - 31 July 1980 58
7. The Control Strategy Emissions Inventory . . 72
8. Results of Model Simulations 81
Specification of input parameters for the model
simulations 81
Initial and boundary conditions 82
Deposition velocities 84
Miscellaneous Fields 86
Comparison of predicted and observed ozone
concentrations 86
Results of base emissions simulations 91
Simulated effects of emissions controls on ozone
concentrations in the Northeast 100
9. Conclusions 119
References 125
Appendices
A 127
8 137
C 146
0 152
-------
FIGURES
Number Page
2-1 Modeling domain treated in the present study. Each dot
represents a grid cell 5
2-2 Illustration of the daytime (A) and nighttime (B) meteorolo-
gical phenomena treated by the ROM and the vertical structure
(C) used to simulate these processes 6
2-3 Schematic illustration of the network of processors and
core module that compose the Regional Oxidant Model (ROM) ... 7
3-1 Maximum hourly averaged ozone concentration (ppb)
observed during July - September 1980 20
3-2 Locations and record lengths (in days) of hourly averaged
ozone measurements during July - September 1980 21
3-3 Maximum hourly averaged ozone observed during an episode
on 21 July 1980 22
3-4 Difference (ppb) between the maximum hourly averaged ozone
concentration observed during the daytime (1000-1759 LSI)
at each monitoring station during June-August 1980 and
the maximum value observed at night (2100-0959 LSI) at
the same station over the same 3-month period 23
3-5 Maximum hourly averaged ozone (ppb) observed during the
nighttime hours (2100-0959 LSI) during June - August 1980 ... 24
3-6 Number of days in 1980 in which the maximum hourly averaged
ozone exceeded 120 ppb . 25
4-1 Locations of the 70 major source counties listed in
Tables 4-1, 2. See Eq. 4.7 for definitions of "high VOC",
"high NOX", and "high VOC and NOX" 43
4-2 Cumulative VOC and NOX emissions densities vs. cumulative
source area and characteristic length scale. See text, pages
36-37, for details 44
4-3 Comparison of NAPAP and NECRMP estimates of gridded VOC
emissions for the 200 strongest source cells in the ROM grid
network 45
4-4 Comparison of NAPAP and NECRMP estimates of gridded NOX
emissions for the 200 strongest source cells in the ROM grid
network 46
-------
Number Page
5-1 Locations of the 70 cells selected in each of the four
receptor classes urban (U), suburban (S), rural (R) and
wilderness (W) 53
5-2 Locations of the 70 major source counties, defined as
"urban" type receptors 54
5-3 Non-major source Standard Metropolitan Statistical Areas
(SMSAs) defined here as "suburban" type receptors 55
5-4 Non-major source, non-SMSA counties in which the predominant
land use is in agricultural activities, defined here as
"agricultural" type receptors 56
5-5 Counties in which forest and/or wetlands are the predominant
land use, defined here as "natural" type receptors 57
6-1 Quartile plot of the maximum hourly averaged ozone concentra-
tions observed at some 200 SAROAD monitoring stations each day
during the period June - August 1980. See text, page 60, for
explanation of symbols 67
6-2a Quantile - quantile comparisons of the cumulative frequency
distributions of hourly ozone concentrations at a rural
site in Connecticut (070570007). The ordinate (y-axis)
represents the distribution observed during the meteorological
scenario period 23-31 July 1980 while the abscissa (x-axis)
represents the distribution observed over the entire summer,
June - August 68
6-2b Same as 6-2a except data are from Whiteface Mountain,
New York (33020002) 69
6-2c Same as 6-2a except data are from a New Jersey site
(312750001) about 50 km east of Philadelphia 70
6-3 Trajectories calculated from wind fields used in Layer 1 of
the ROM for the 9-day meteorological scenario 23-31 July 1980
(Julian dates 205-213). Numbers along the trajectories refer
to Julian date 71
7-la VOC emissions in the control strategy expressed as a
percentage reduction in the base emissions. (See
Appendix B for details.) 75
7-lb NOX emissions in the control strategy expressed as a
percentage reduction in the base emissions. (See
Appendix B for details.) 76
vii
-------
Number Page
7-2 Comparison of base and controlled emissions of VOC (moles/day)
of the 200 strongest VOC emitting cells in the ROM domain,
Canada excluded. (All point and area sources within each
cell are combined.) 77
7-3 Comparison of base and controlled emissions of NOX (moles/day)
of the 200 strongest NOX emitting cells in the ROM domain,
Canada excluded. (All point and area sources within each
cell are combined.) 78
7-4 Base vs. controlled VOC emissions (moles per krn^ per day) of
the 70 major source counties listed in Tables 4-1,2 79
7-5 Base vs. controlled NOX emissions (moles per km per day) of
the 70 major source counties listed in Tables 4-1,2 80
8-1 Maximum ozone concentration during the 9-day scenario
simulated using the base emissions inventory. Isopleths
are drawn at intervals of 40 ppb. Numbers beside isopleths
are ppb/100 104
8-2 Maximum hourly averaged ozone concentration observed at SAROAD
monitoring sites during the scenario period 23-31 July 1980 . . 105
8-3 Simulated ozone in Layer 1 at hour 1600 EST on the first day
of the 9-day scenario, using the base emissions inventory . . . 106
8-4 Same as 8-3 except hour 1600 of day 2 of the scenario 107
8-5 Same as 8-3 except hour 1600 of day 3 of the scenario 108
8-6 Same as 8-3 except hour 1600 of day 4 of the scenario 109
8-7 Same as 8-3 except hour 1600 of day 5 of the scenario 110
8-8 Same as 8-3 except hour 1600 of day 6 of the scenario Ill
8-9 Same as 8-3 except hour 1600 of day 7 of the scenario 112
8-10 Same as 8-3 except hour 1600 of day 8 of the scenario 113
8-11 Same as 8-3 except hour 1600 of day 9 of the scenario 114
8-12 Quartile plots of simulated base case daily daylight averaged
(0900-1600 1ST) ozone concentrations in each of the four
county receptor classes defined in Section 5. See page 96
for explanation of symbols 115
8-13 Quartile plots of simulated base case hourly ozone concentration
in each of the four receptor cell classes defined in Section 5.
See page 96 for an explanation of the graphs 116
viii
-------
Number
8-14 Simulated effects of emissions controls on daily daylight
(0900-1600 LSI) averaged ozone concentrations in each of the
four county receptor classes. See page 101 for an explanation
of the symbols 117
8-15 Simulated effect of emission controls on hourly averaged ozone
concentrations in each of the four receptor cell classes.
See page 101 for an explanation of the symbols 118
9-1 Counties in which simulated one-hour averaged ozone concentra-
tions exceeded 120 ppb during the 9-day scenario. Light shading
denotes counties in which exceedance occurred only in the base
case simulation. Dark shading indicates areas where exceedance
occurred in both the base and control runs 124
-------
TABLES
Number
2-1 Chemical reactions included in the Demerjian/Schere
mechanism and the rate constants assumed for each .
2-2 Comparison of the principal differences between the first
generation regional oxidant model used in the present study
and the second generation model currently in development. ... 10
4-1 Major county sources of VOC ranked in order of source
strength Q (= mass per area per time). Counties marked
with an asterisk (*) are also major sources of NOX (listed
in Table 4-2) 41
4-2 Major county sources of NOX ranked in order of source
strength Q (= mass per area per time). Counties marked
with an asterisk (*) are also major sources of VOC (listed
in Table 4-1) 42
8-1 Values used for the initial and boundary concentrations
in both the base and control ROM applications 85
-------
ACKNOWLEDGEMENT
The work reported here is the product of the efforts of many people
to whom the author is sincerely grateful. Joan Novak coordinated data
acquisition, programming and computer operations with unprecedented
thoroughness and efficiency. James Reagan provided timely, masterful
analyses of the ozone monitoring data. Ken Schere managed the chemistry
aspects of the model with skill and precision, and Barbara Hinton exhibited
her usual patience and proficiency in typing the manuscript. I am also
indebted to the following members of Computer Science Corporation for his
or her special contribution to this work: Lucille Bender, Larry Bergman,
Russ Bullock, Tony Griffin, Don Pelles, Ted Smith, Joe Vaughan, and Jeff Young.
-------
SECTION 1
INTRODUCTION
The development of the Environmental Protection Agency's Regional
Oxidant Model (ROM) began in the late 1970's as a part of the Northeast
Corridor Regional Modeling Project (NECRMP). The NECRMP was initiated
out of the recognition that the adverse ozone concentrations observed in
the Northeastern United States are due in large part to the regional trans-
port of ozone and its precursor species. The principal role envisioned
for the ROM in this project was to assist the states in developing emis-
sions control plans that would effect compliance with the Federal ozone
air quality standards in the most equitable and cost effective way. This
report describes the second of a series of applications of the Regional
Oxidant Model in this role. It considers projected 1987 emissions, based
on 1982 State Implementation Plans (SIPs), and compares the ozone
concentrations simulated using these emissions with the corresponding
concentrations predicted using 1980 emissions data. This study also
examines the character of the ozone problem in the Northeast; the strengths
and spatial distributions of the sources of the precursor species, namely
hydrocarbons and NOX; meteorology; and other factors important in
interpreting model results and in developing a basic understanding of the
relationship between observed ozone levels and source emissions. These
analyses begin in Section 3 following a brief summary of the structure of
the ROM in Section 2. The projected 1987 emissions are described in
Section 7 and results of the model simulations are presented in Section 8.
-------
SECTION 2
OVERVIEW OF THE REGIONAL OXIDANT MODEL
The ROM is a three-dimensional Eulerian model that simulates hourly
averaged species concentrations over the geographical area shown in Figure
2-1 (page 5) with a spatial resolution, or grid size, of approximately 18
km. (The actual grid dimensions are 1/4 degree longitude by 1/6 degree
latitude.) The model has 3 layers in the vertical in which species
concentrations are treated prognostically and it has a shallow layer
adjacent to the ground in which concentrations are handled diagnostically.
The latter is used primarily to simulate dry deposition processes and
subgrid scale chemistry phenomena associated with line and point source
plumes.
Each of the three prognostic layers has a thickness that varies in
space and time in a manner designed to keep track of the dominant
meteorological processes within that layer. Both the processes and
the model's layer structure are illustrated in Figure 2-2 (page 6).
Generally speaking, the lowest layer, Layer 1 (see Figure 2-2c),
handles wind shear and marine layers by day and the radiation inversion
layer by night. Layer 2 simulates the free convective subcloud mixed layer
by day and the reservoir of aged species above the radiation inversion by
night. And Layer 3 represents the convective cloud layer by day and the
zone of cloud residue by night. At any given location, the elevation of
the top of Layer 3 is equivalent to the elevation of the tops of convective
clouds at that location, regardless of how high the clouds might be. The
-------
entrainment of stratospheric ozone into the lower atmosphere can be simulated
by prescribing the ozone flux across the top surface of Layer 3. Complete
details on the structure of the ROM can be found in Lamb (1983a).
All of the chemical and physical processes that occur within the
four layers that we have discussed are described by a system of differential
equations that constitute the basis of the model's predictions. Computer
solvable analogues of these equations comprise what we refer to as the
core model. All of the independent variables that enter these equations,
such as wind velocities, layer thicknesses, emission rates, etc, are
prepared by a network of peripheral processors or modules for input to
the core model. The relationships among the processors and CORE are
depicted in Figure 2-3 (page 7). By structuring the ROM in this modular
fashion, we can change the method used to generate individual variables
without having to overhaul the entire model. Note that even the chemical
kinetics mechanism, which is a major component of any photochemical model,
is among the interchangeable components. Thus, when we refer to a
particular "generation" of the ROM, we are referring to the network of
modules, including the kinetics scheme, that drive the CORE model. The
CORE, as described briefly above and in detail in Lamb (1984), will remain
fixed for all foreseeable generations of the ROM.
The model simulations that we have conducted to date, including the
ones to be presented in this report, have used the Demerjian and Schere
(1979) chemical mechanism described in Table 2-1 (page 8). This scheme
treats 36 reactions among 23 species. The second-generation ROM will
employ a condensed version of the CBM-IV chemical mechanism. Table 2-2
-------
(page 10) summarizes the principal differences between the first generation
ROM used in the present study and the second-generation model, which is
presently undergoing tests.
Concerning the present status of our model verification studies, we
have completed what we regard as tests of the necessary conditions for
model validity. This work, described in Lamb and Laniak (1985), demonstrates
that the solutions of the mathematical equations that constitute the CORE
model (see Figure 2-3) are close facsimiles of the solutions of the
corresponding differential equations that describe the chemistry and
physics that the ROM is intended to simulate. It remains to be shown
that the model satisfies the sufficient conditions for validity, namely
that its predictions of species concentration statistics are acceptably
close to the values one would measure under the meteorological and
emissions conditions simulated. This final step of the verification
process will be applied in a rigorous way only to the second-generation
model. A partial verification of the first generation model, used in the
present study, has been performed by Schere (1986). Some additional
comparisons of the model's predictions with observations are presented in
Section 8.
-------
^45.00°
If 44.00
H
43.00°
tf 42.00°
'. U 41.00"
'. N 40.00°
tf 39.00
:: y aa.oo0
Figure 2-1. Modeling domain treated in the present study.
represents a grid cell.
Each dot
-------
^ » - ' 7 ' -
. - • ,< •.,-.. ...•/.. . ••-
B
Figure 2-2. Illustration of the daytime (A) and nighttime (B) meteorological
phenomena treated by the ROM and the vertical structure (C)
used to simulate these processes.
-------
lai'OtiHAHW
DATA
1 {MISSIONS
DATA
5
SUIUACt
Mtl
DATA
SAieilllt
CtQUII COVtll
DATA
IANOUSE
OAIA
SIJIIFACC AIH
MONITORING
DATA
Figure 2-3. Schematic illustration of the network of processors and
core module that compose the Regional Oxidant Model (ROM),
-------
Table 2-1. Chemical reactions included in the Demerjian/Schere
mechanism and the rate constants assumed for each.
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
N02
0 + 02 + M
03 + NO
03 + N02
N03 + NO
N03 + N02 + H20
HONO
HO + CO
H02 + N02
H02 + NO
H02 + N02 + M
HOONO 2
. HO + HONO
HO + N02 + M
HO + NO + M
HO 2 + 03
HO + 03
H02 + H02
OLEF + 0
OLEF + 03
OLEF + HO
PARAF + HO
Reaction
hv
> NO + 0
+ 03 +M
> N02 + 02
> NOo + Oo
O L.
-v 2N02
+ 2HONO
hv
* HO + NO
(02)
> H02 + C02
> HONO + 02
->• HO + N02
> HOON02 + M
* H02 + N02
* N02 + H20
+ HON02 + M
+ HONO + M
> HO + 202
> H02 + 02
> H202 + 02
> R02 + ALD + H02
> R02 + ALD + H02
> R02 + ALD
> ROo
Rate Constant*
(units3)
variable
2.3 x 10'5 c
2.7 x 10+1
4.8 x 10"2
3.0 x IO4
3.4 x IO-3 c
variable^
4.1 x IO2
4.4
1.2 x IO4
1.5 x 10'3 c
3.3
9.8 x IO3
1.5 x IO-2 c
7.4 x 1U-3 c
3.0
1.0 x IO2
3.7 x IO3
5.1 x IO3
1.4 x 1Q-2
3.1 x IO4
5.0 x IO3
23.
ALD
hv
0.5R02 + 1.5H02 + l.OCO variabled
-------
Table 2-1, continued
Reaction
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
ALD +
R02 +
RO +
R102 +
RO +
R02 +
R102 +
AROM +
R202 +
R20 +
R202 +
R102 +
HO
NO
°2
N02
N02
03
NO
PAN
HO
NO
°2
03
03
Rate Constant
(units3)
->• 0.3R102 + 0.7H02 + 0.7CO 1.4 x 104
+ RO +
+ ALD +
-> PAN
> RON02
-» RO +
(02)
> R02 +
->• RlOo
(02)
-> R202
* R20 +
> ALD +
> R20 +
-»• R02 +
N02
H02
202
N02
+ N02
+ 2ALD + CO
N02
H02 + 2CO
202
202
1.1 x 104
9.0 x 10'1
8.9 x 103
1.0 x 102
2.0
t.O A ^
1.4 x 10'1 b
i\
2.3 x 104
1.1 x 104
8.9 x 10"1
2.0
2.0
* Values of rate constants that vary by temperature are evaluated here for
298°K and 1 atm pressure.
a Rate constant units are ppm~l min'l unless otherwise noted.
b Units of rate constant are min'1.
c Units of rate constant are ppm"2 min~^-.
d Photolysis rate constants are based on data compiled by Demerjian, Schere
and Peterson (1980) and vary as a function of solar zenith angle.
Species definitions:
Alkyl Nitrate
H02 Hydroperoxyl Radical
H04N Pernitric Acid
RO Alkoxyl Radical
R02 Alkyl peroxy Radical
R20 Alkoxy Radical
R102 Peroxyacyl Radical
R202 Peroxy Radical
-------
Table 2-2. Comparison of the principal differences between the first
generation regional oxidant model used in the present study
and the second generation model currently in development.
First Generation Second Generation
Layers Constant Thickness Thicknesses variable in
space and time
Winds Nondivergent Divergent
Chemistry Demerjian-Schere CBM-IV
Emissions 1980 NAPAP 4.2 1980 NAPAP 5. + Biogenic
10
-------
SECTION 3
CHARACTERISTICS OF OBSERVED OZONE CONCENTRATIONS
IN THE NORTHEASTERN UNITED STATES
Within the geographical area shown in Figure 2-1 that is presently
simulated by the Regional Oxidant Model, i.e., the region that we refer
to in this report as the Northeast, the highest ozone concentrations
attributable to anthropogenic activities occur during the summer
months. Isolated occurrences of extreme ozone levels are occasionally
reported in other seasons, particularly the winter and spring, but these
are more likely to be due to intrusions of stratospheric ozone than to
photochemical reactions among anthropogenic emissions (see Lamb, 1977;
Viezee et al, 1983).
The spatial distribution of maximum 1-hour averaged ozone concentrations
observed during June through September 1980 is shown in Figure 3-1 (page 20).
(We have excluded data from October through May to minimize the influence
of stratospheric ozone.) The information presented in Figure 3-1 is
based on data reported by some 200 SAROAD stations whose locations and
record lengths are given in Figure 3-2 (page 21). Despite the relatively
large number of monitoring sites, the spatial coverage is not adequate to
resolve the full detail of the ozone distribution. Moreover, sizable
areas of West Virginia and Pennsylvania contain no monitoring sites at
all. For these reasons, the concentration isopleths shown in Figure 3-1
should be viewed qualitatively.
11
-------
The data reveal five general subregions where ozone concentration in
excess of the current primary standard, 120 ppb, occur. One is a vast
area roughly parallel to the East Coast, extending from Virginia into
southern Maine and extending inland from the coastline some 200 kilometers.
Within this area is a zone roughly 100 km wide lying along the Eastern
Corridor cities from Washington to Boston in which ozone concentrations
greater than 160 ppb are observed. And in northern Long Island and
southern Connecticut, it is not uncommon to observe ozone levels in excess
of 240 ppb.
In the four other regions shown in Figure 3-1 where ozone concentrations
exceed 120 ppb, only the region around Pittsburgh contains an area where
values exceed 160 ppb, and even here the area is apparently rather small.
The fact that ozone levels in the Detroit airshed are relatively low is
surprising since the modeling studies we have conducted to date indicate
that the VOC and NOX emissions in that area are capable of producing
ozone concentrations comparable to those produced by East Coast sources.
The same is true of the simulated emissions from the Toronto area. We
will return to this matter in Section 8 when we discuss the results of
the model runs.
From the standpoint of model design, an important characteristic of
the ozone distribution is the spatial scale of concentration variations.
Figure 3-1 suggests that in the areas of highest concentration, the
characteristic scale is less than 100 km. This is supported by the
distribution of maximum hourly ozone values shown in Figure 3-3 (page 22)
which were recorded during one of the major episodes observed in the
Northeast in 1980. The concentrations vary markedly from station to
12
-------
station along the Eastern Corridor. For example, in southwestern Connecticut,
maximum ozone values of 155 and 302 ppb were measured at sites within
about 40 km of each other. Similar gradients can be seen in New Jersey.
It has been known since aircraft observations of ozone in the Northeast
were begun in the late 1970's that the highest ozone values occur within
*
urban plumes several tens of kilometers wide that are often discernable
up to 200 km downwind of the major sources of NOX and VOC (see, for
example, Alkezweeny, et al 1981; Clark and Clarke, 1984; and Westberg,
1985). The fact that emissions from urban areas are concentrated in
relatively narrow plumes during the first day of their travel can be a
significant factor in determining the fate of emissions. Because the
higher the concentrations of chemical species, the faster are the reactions
among them; and the longer the concentrations remain high, due to ineffi-
cient atmospheric mixing, the larger are the quantities of product species
produced (such as ozone). Models with grid meshes larger than the lateral
dimensions of urban plumes, i.e., several tens of kilometers, cannot
simulate plume chemistry unless a scheme to approximate subgrid scale
processes is incorporated. We will return to this point in the next
section when we discuss the character of the hydrocarbon and NOX sources
in the Northeast.
One of the primary objectives of regional modeling is to identify
those areas that are primarily sources of ozone and those that are
primarily receptors. Information of this kind is essential in formulating
the most equitable, cost effective emissions control plans. We can begin
to form a picture of the source and receptor areas by analyzing the
observed hourly ozone concentrations in light of the following basic
fact: ozone is effectively generated only during daylight hours.
13
-------
If one examines the chemical mechanisms that govern photochemical
air pollution, one finds that the only source of ozone is the reaction
0 + 02 + M -> 03 + M , (A)
with the principal source of atomic oxygen (0) being the photolysis of
NOg, namely
N02 + hv -» NO + 0 (B)
where hv represents a photon of sunlight. Reaction A is several orders
of magnitude faster than B, and the vast majority of oxygen atoms produced
by reaction B are removed by reaction A. Therefore, for all practical
purposes reactions A and B can be combined in the form
NO 2 + 02 + hv > NO + 03 . (C)
The fact that this reaction is the only source of ozone implies the
following premise:
Ozone observed at any given location at night (when hv =0) is
ozone that was generated somewhere else. (PI)
Here we use the terms "night" and "somewhere else" in a general sense
which we will make more specific shortly. Essentially, premise PI provides
a means of using ozone values measured during the hours of darkness to
assess the extent to which distant sources affect the ozone burden at the
measurement site. Premise PI has the following corollary:
The difference between the maximum ozone concentration observed
at a given site during the day and the maximum ozone observed (Cl)
at the same location at night is a measure of the quantity of
ozone generated in the vicinity of that site.
To implement PI and Cl we adopt the following definitions of day and
night:
day = 1000 - 1759 LST (3.la)
night = 2100 - 0959 LST (3.1b)
The definition of "day" is intended to span the period in which reaction C
14
-------
is most effective in generating ozone. The definition of night, (3.1b),
is intended to achieve two ends. First, by placing a gap between the end
of the daytime period and the beginning of the period defined as night we
minimize the extent to which locally generated ozone influences the values
to which we ultimately apply premise PI. For example, under our definition
(3.1), in which the gap between day and night is "3 hours, ozone generated
within 30 km of a monitor would have no effect on the measured nighttime
ozone levels under conditions of a 10 km/hr mean wind. Second, by extending
the period defined as night to include several hours after sunrise, we
capture the effects of ozone transported long distances that fumigates
into the surface layer following the erosion of the nighttime radiation
inversion. Let us now apply PI and Cl to the SAROAD ozone data, using
(3.1) as the definitions of day and night.
Figure 3-4 (page 23) shows isopleths of the difference between the
maximum hourly daytime ozone measured throughout June-August 1980 and the
corresponding maximum hourly nighttime value observed in the same three
month period. According to Cl, the areas of largest concentration
differences are the areas of largest ozone production. We will refer to
these areas as "virtual ozone sources" with the understanding that the
ozone generated in these zones is likely due to hydrocarbon and nitrogen
oxide emissions from sources outside these zones. On comparing Figures
3-4 and 3-1 (page 20) we find that the areas of highest apparent ozone
generation coincide with the areas of highest observed ozone concentration.
A relationship of this kind between source and concentration distributions
is characteristic of species that undergo chemical transformation. The data
presented in Figure 3-4 indicate that maximum daytime ozone concentrations
15
-------
are larger than the corresponding maximum nighttime values everywhere
except portions of central Pennsylvania and northeastern New York. In
these two areas the difference between the daytime and nighttime ozone
maximum is actually negative, implying that these are predominantly areas
of ozone destruction. The fact that both these areas are remote from the
major urban source regions supports the validity of Cl. Because after a
mixture of ozone precursor species has been exposed to sunlight for a
sufficiently long time, its ozone production potential is exhausted and
the ozone that has been formed up to that time begins to disappear primarily
through uptake by vegetation and soil. These particular removal mechanisms
are most effective during the daytime when turbulence intensities are
high enough to maintain a constant flux of ozone down to the earth's
surface where removal processes are taking place. Thus, interpreting the
data given in Figure 3-4 from the perspective of premise Cl, we tentatively
conclude that the bulk of the ozone attributable to anthropogenic sources
is generated within about 100 km of the major urban areas; and that
beyond distances of about 200 km, ozone is primarily in a state of decay.
Let us look now at the areas where ozone transported from source
regions has the largest effect. Figure 3-5 (page 24) shows isopleths of
the maximum nighttime ozone concentrations observed during June - August
1980, the same period represented by Figure 3-4 (page 23). Under premise
PI the concentration values shown in Figure 3-5 are a measure of the
maximum ozone burden imposed on a site by distant sources, namely sources
farther than about 30 km from the site in question. On comparing the
locations of the zones of highest nighttime ozone shown in Figure 3-5
with the locations of the areas of highest ozone generation indicated in
16
-------
Figure 3-4, we find that the former are displaced some 50 to 100 km from
the latter. For example, a zone of nighttime ozone concentrations in
excess of 120 ppb is evident in Southeastern Pennsylvania some 50 km north
of the source areas associated with the Washington, DC to Philadelphia
corridor. Similarly high nighttime ozone concentrations also occur in an
area east and northwest of Boston. A check of the ozone data showed th5t
the high concentrations in this area were observed around midnight on
days when afternoon concentrations in the Connecticut virtual-source
area, shown in Figure 3-4, were high.
The nighttime data suggest that the maximum ozone burden due to
imported ozone exceeds 100 ppb within an elongated zone stretching along
the eastern seaboard from Washington, DC to just northeast of Portland,
Maine. Relatively small areas of similarly high burdens appear near
Detroit, Pittsburgh and Buffalo. The data also indicate that nighttime
ozone levels ranging from 100 ppb to just over 120 ppb occur throughout
upstate New York and a large part of Vermont. The model simulations that
we present later suggest that the high ozone concentrations observed in
these areas are attributable to east coast sources with occasional
contributions from ozone generated from Toronto emissions. If the
influence zones delineated by the model are correct, we can deduce from
the observations made at Whiteface Mountain in upstate New York that
ozone concentrations slightly in excess of the current primary standard,
120 ppb, exist as far as 400 km downwind of the major source areas along
the East Coast. This is not meant to imply, however, that values this
high would be observed at all ground-level stations within 400 km of the
major eastern cities. At sites near sources of NOX, observed ozone values
17
-------
would be smaller due to scavenging by NO. At remote sites where the ozone
plume arrives overhead during the predawn hours, surface ozone readings
would remain well below 120 ppb if a strong enough radiation inversion
were present to prevent turbulent transport of the ozone to the ground.
It is reasonable to conclude, however, that within 400 km of the Eastern
Corridor cities, the New York City area in particular, hourly averaged
ozone concentrations as high as 120 ppb can occur at all elevations
higher than the depth of the radiation inversion, i.e., elevations higher
than 300-500 meters above the lowest local relief. This characteristic
of the ozone distribution has a major bearing on the ozone exposures of
forests in the Northeastern United States.
As we stated at the outset, the data analyses that we just presented
provide only a qualitative description of the ozone problem in the
Northeast, because the available data cover only a limited geographical
area and the accuracies of some of the data themselves are questionable.
Moreover, some of our results were based on the speculations embodied in
premises PI and C-l. We presented this analysis in an effort to gain a
general view of the areas in which ozone is generated and the areas into
which it spreads. We believe that this knowledge is useful in formulating
model experiments, in interpreting the results, and in designing the most
effective emissions controls for ozone reduction. In the next section we
use the information derived here to determine the locations of the ozone
precursor species that have the largest effect on ozone concentrations.
As a final piece of information that illustrates the current ozone
problem in the Northeast, we show in Figure 3-6 (page 25) the number of
days during 1980 in which the maximum hourly ozone concentration exceeded
18
-------
120 ppb. The areas in which the primary ozone standard was not attained,
i.e., the area in which two or more exceedances were observed, correlates
well with the areas of apparent ozone generation shown in Figure 3-4
(page 23). The close spacing of the frequency isopleths shown in Figure
3-6 also gives an indication of the difficulty that will be faced in
reducing ozone concentrations below the current standard. That is, the
data suggest that if emissions changes were adopted that reduced the
number of exceedances of 120 ppb ozone by one-half everywhere, the
geographical area in which the standard was not attained would be reduced
by only 10 to 20 percent.
19
-------
Figure 3-1. Maximum hourly averaged ozone concentration (ppb)
observed during July - September 1980.
20
-------
. ~ r^S^SSL-
Figure 3-2. Locations and record lengths (in days) of hourly averaged
ozone measurements during July - September 1980.
21
-------
Figure 3-3. Maximum hourly averaged ozone observed during an
episode on 21 July 1980.
22
-------
Figure 3-4. Difference (ppb) between the maximum hourly averaged ozone
concentration observed during the daytime (1000-1759 LSI)
at each monitoring station during June-August 1980 and the
maximum value observed at night (2100-0959 LSI) at the
same station over the same 3-month period.
23
-------
Figure 3-5. Maximum hourly averaged ozone (ppb) observed during the
nighttime hours (2100-0959 LSI) during June - August 1980.
24
-------
Figure 3-6. Number of days in 1980 in which the maximum hourly
averaged ozone exceeded 120 ppb.
25
-------
SECTION 4
THE SOURCES OF OZONE PRECURSOR SPECIES IN THE
NORTHEASTERN UNITED STATES
As we pointed out in the previous section the only atmospheric
source of ozone is the composite reaction
kl
N02 + 02 + hv •»• NO + 03 (4.1)
(see reactions 1 and 2 in the Demerjian/Schere mechanism described in
Table 2-1, page 9). Hydrocarbons enter the picture in a rather subtle
way that is linked to the chemical sinks of ozone, in particular its reaction
with NO:
k2
NO + 03 > N02 + 02 (4.2)
In full sun conditions reactions (4.1) and (4.2) are so fast that the two
achieve a quasi-photostationary state in which the concentration of ozone
is given approximately by
ki N02
03 » — . — . (4.3)
J k2 NO
The ratio (k]_/k2) is determined solely by the intensity of sunlight. It
has a value of zero at night and a maximum of order 10 ppb at ground
level at noon on a clear summer day along the 35° latitude circle.
By contrast the ratio (N02'.NO) is sensitive to the presence of
hydrocarbons, because these species are sources of free radicals that are
capable of oxidizing NO to N02. In this action hydrocarbons are doubly
effective in amplifying ozone levels because they remove an ozone sink, NO,
at the same time that they create a source, N02. Reactions of NO with
alkylperoxy R02 and peroxy radicals R202 (reactions 25 and 33 in Table 2-1,
page 9) are two of the pathways in which NO is converted to NO 2*
26
-------
Not all of the radicals produced by hydrocarbons act to accentuate
the (N02:NO) ratio. The alkoxyl radical, RO, for example, depletes NO 2
concentrations by transforming this species into a stable end product,
alkyl nitrate (see reaction 28, Table 2-1). Nitrogen dioxide is also
removed from the ozone production cycle by reactions with hydroxyl radical,
which produce nitric acid. Thus, according to the the Demerjian-Schere
chemical mechanism, all nitrogen oxide emissions are eventually converted
into nitrate compounds and in the process ozone is produced.
Likewise, all hydrocarbon emissions are converted ultimately into
C02. An important aspect of the fate of the hydrocarbons is that three
of the four lumped species, namely olefin (OLEF), paraffin (PARAF) and
aromatics (AROM) disappear monotonically because there are no chemical
sources of any of these species. The fourth lumped species, aldehydes
(ALD), is produced by olefins and aromatics (see reactions 19-21 and 32
in Table 2-1, page 8), by the oxidation of certain organic free radicals
(reactions 26 and 34), and it is emitted by some anthropogenic sources.
Aldehydes are destroyed by photolysis and by reactions with OH. Both
processes generate free radicals that are capable of converting NO to N02.
These aspects of hydrocarbon behavior suggest the following general
scenario. In the daytime urban plume, concentrations of all the hydro-
carbons decrease downwind of the city as a result of chemical reactions
that produce free radicals. These, in turn, cause ozone concentrations
to rise by converting the ozone scavenger NO into an ozone source, NO 2.
(Model simulations indicate that over 90 percent of the olefins and aromatics
initially present are consumed within 8 hours exposure to sunlight.)
27
-------
After sundown aldehyde concentrations in the plume may begin to rise due
to the oxidation of excess radicals left over from the daytime and to
reactions of residual olefins with ozone. (The chief mechanisms of
aldehyde decay, namely photolysis and reaction with OH, are absent at
night.) Thus, by the beginning of the second day, olefins and aromatics
will have virtually disappeared from the plume but aldehyde and paraffin
levels might still be high enough to cause enhanced ozone formation from
fresh emissions of nitrogen oxides that the plume might encounter during
the day.
In regulatory studies the current practice is to lump all hydrocarbons
except methane together and treat the sum as a single entity referred to
as non-methane hydrocarbon (NMHC), or volatile organic compounds (VOC).
This practice simplifies the treatment of emissions but it fails to
distinguish between the ozone production potentials of the individual
hydrocarbon species. Although the model simulations that we describe
later in this report are based on emissions data in which hydrocarbons
are split into the four lumped species OLEF, PARAF, AROM and ALD, it is
convenient in the present section, where we are concerned only with
qualitative features of ozone precursor sources, to look only at VOC
emissions. In the context of the Demerjian-Schere chemical mechanism
that is presently used in the model (see Table 2-1, page 8), VOC is defined
as follows:
VOC = 3.57 x OLEF + 4.56 x PARAF + 1.50 x ALD
+ 7.56 x AROM + 1.02 x (Non reactives) (4.4)
where the units of VOC, OLEF, and all remaining species are moles or ppb.
Similarly, we consider nitrogen oxides emissions in the combined form
NOX = NO + N02. (4.5)
28
-------
Our main objective in this Section is to determine the strengths and
locations of the major sources of ozone precursor, i.e., VOC and NO x. We
are especially interested in seeing where these sources lie in relation
to the virtual ozone sources that we mapped out earlier in Figure 3-4
(page 23). Our motivation is the belief that once the major sources of
precursor species have been identified, one can utilize controls on those
particular sources as a means of achieving maximal changes in ozone
concentrations.
The U.S. emissions data used in this study are from version 4.2 of
the NAPAP inventory, documented by Toothman et al., 1984. Canadian emis-
sions were estimated from data provided by Environment Canada. Due to
differences between the source classification schemes used by the U.S.
and Canada and to the availability of only annual Canadian emission, it
was necessary to apply U.S. temporal and species allocation factors to
the Canadian data to acquire the necessary model inputs. Also, Canadian
population, housing and land use data were used as surrogate indicators
for spatial distribution of emissions into the ROM grid network.
In this section we will adopt the practice commonly used in regulatory
analyses of dealing with emissions on a county basis. We wish to emphasize,
however, that in the model simulations the emissions have been apportioned
to the model's grid cells, whose dimensions are 1/4° longitude x 1/6° latitude,
or roughly 18 x 18 km (see Figure 2-1, page 5). Most of the counties within
the model domain contain several grid cells.
To identify the major sources of NOX and VOC we are guided by the
simple, basic relationship
29
-------
C = j. (4.6)
uh
which expresses the concentration C (mass/volume) of material from an area
source of strength Q [mass/(area.time)] emitting into a mixing layer of
depth h and wind speed u". In other words, the ability of a source to
cause high concentrations is measured by its emission density Q rather
than its mass release rate m (mass/time). Thus, we determined the VOC and
NOX emissions densities Qvoc anc' QNOx °f eacn °f tne 439 counties within
the model domain, and then ordered each set with the largest Q value
first and the smallest last. Tables 4-1 and 4-2 (pages 41 and 42) list
the counties with the top 50 values of Qvoc anc' QNOx» respectively.
On examining these lists, one finds that 30 counties, designated by
asterisks in each table, are common to both sets. As a result, the top
50 NOX and VOC sources consist of only 70 counties rather than 100. For
purposes of discussion we shall divide these 70 counties into three groups
designated as follows:
(1) "High VOC" = listed in Table 4-1 but not Table 4-2;
(2) "High NOX" = listed in Table 4-2 but not Table 4-1;
(3) "High NOX and high VOC" = listed in both Tables (and indicated
by asterisks in each).
Figure 4-1 (page 43) shows the locations of these sources. An important
property of this set is that together these sources generate about 50% of
all VOC and NOX emitted in the Northeastern United States (excluding Canada).
We will discuss other properties of these and the entire set of sources
later in this section.
30
-------
For purposes of discussion, we will refer to the 70 counties listed
in Tables 4-1,2, and shown in Figure 4-1, as the major source counties.
Comparing the locations of these 70 counties with the virtual ozone source
distribution plotted earlier in Figure 3-4 (page 23), we see a distinct
correlation. The area of high ozone generation along the East Coast
apparent in Figure 3-4 surrounds the major source counties of the Eastern
Corridor shown in Figure 4-1. The smaller zones of ozone generation
apparent in Figure 3-4 in the Pittsburgh, Cleveland and Detroit areas lie
in the immediate vicinity of major source counties, but they do not
engulf the counties as is the case along the Eastern seaboard. We also
see that the areas of ozone destruction lying along the Appalachian
Mountains of Pennsylvania and the Adirondack Mountains of New York fall
within a gap separating the major sources of the Eastern seaboard from
the major sources of the Ohio Valley and Michigan areas.
The two chief discrepancies between the areas of ozone generation
and the major ozone precursor species sources are first, the absence of
an area of ozone production around the major source counties of southern
Ohio and western West Virginia; and second, the absence of major sources
in the vicinity of the high ozone production areas apparent in Figure 3-4
in southwestern Ohio. The latter discrepancy is attributable to emissions
from the Cincinatti and Dayton areas which are just outside the western
boundary of the model domain. The fact that high ozone production is not
indicated in Figure 3-4 in the vicinity of the major sources in southern
Ohio and West Virginia is due to the paucity of ozone data in this area
rather than to a known absence of ozone production.
31
-------
Our primary purpose In comparing our so-called major source counties
with the areas of apparent ozone production is to substantiate the
significance of the role that these particular sources play. One of the
things that the comparison has revealed is that the major sources within
the Eastern Corridor are significantly more effective in producing ozone
than those west of the Appalachians. We believe that there are four main
reasons for this.
First, the sources along the East Coast are considerably stronger
than those west of the Appalachians. Table 4-1 indicates that all but
one of the top 20 VOC sources is in the East and Table 4-2 shows that all
but 4 of the top 20 NOX sources are east of the Appalachians.
Second, the VOC:NOX ratios of the eastern sources are generally
higher than those of the west. This is evident in Figure 4-1 (page 43),
where we see that the Eastern Corridor consists almost entirely of "high
VOC" and "high VOC and NOX" sources. Our modeling studies have shown
that the ozone production potential of a given mixture of precursor
species is realized fastest when the VOC:NOX ratio is high. This is
important in the production of ozone in the atmosphere; because if emissions
are mixed with relatively clean ambient air at a rate faster than ozone
is produced, the total amount of ozone generated will be diminished. The
time scale of mixing of an urban plume with its environment is of the
order of 1 day, because the principal mechanism of mixing at spatial
scales of the order of an urban area is the vertical wind shear present
in the nocturnal boundary layer. Thus, if the emissions of VOC and NOX
of a given source were in a ratio such that the maximum ozone concentration
32
-------
would not occur until the second day, the ozone production potential of
that source would generally not be realized. Our studies with the
Demerjian-Schere chemical mechanism described in Table 2-1 indicates that
the critical VOC:NOX ratio of the emissions is around 5. That is, if the
ratio QVQC : QNOx < $, maximum ozone levels do not occur until the second day,
Otherwise, peak levels occur during the first day of travel. This is
perhaps a good point to comment on why the VOC:NOX ratios of the major
sources listed in Table 4-1,2 (pages 41, 42) are so much lower than the
ratios observed in air samples collected in and around major cities.
Measured VOC:NOX ratios reported by Allwine and Westberg (1977) are
generally larger than 10 whereas the ratios of the emissions of the major
sources listed in Tables 4-1, 2 are for the most part smaller than 5 and
in many cases are smaller than 1. The higher ratios observed in the
atmosphere are probably due to the fact that NOX is transformed to product
species such as HN03, HONO, PAN and others more rapidly than VOC is
converted into products. Hence, the VOC:NOX ratio increases with travel
distance from the source. Another contributing factor is the apparently
high degree of uncertainty that currently exists in gridded emissions
data. We will discuss this point in some detail later in this section.
A third reason for the higher ozone generation around the eastern
sources is the arrangement of these sources in a line parallel to the
prevailing wind direction in the months of highest ozone potential. As a
result of this arrangement air arriving in Connecticut during southwesterly
winds contains not only emissions from the strong New York City sources
but also ozone and precursor species arising from all the Eastern Corridor
sources between New York and Washington, DC. It is not surprising,
33
-------
therefore, that the highest ozone concentrations lie in Connecticut (see
Figures 3-1, 4, pages 20 and 23). A dramatic example of air flow moving
in virtually a straight line along the Eastern Corridor cities was reported
by Clark, et al (1984), who tracked the air motion with a marker balloon
and conducted aircraft sampling of air at various points along the balloon's
path.
A final contributing factor to the higher ozone production of the
eastern sources is their proximity to the ocean. The model has predicted,
and field observations have confirmed (see Siple et al., 1977; Lyons and
Cole, 1976), that when air carrying fresh VOC and NOX emissions moves
from the warm land over colder water the mixture becomes trapped in a
relatively shallow layer next to the water due to the loss of the surface
sensible heat flux that drives convective mixing. If the sky is clear
over the water — and this is often the case since convective clouds are
absent — high concentrations of ozone can be generated. Often this
ozone-rich air moves back on shore causing high ozone levels along the
shoreline and progressively lesser concentrations inland as the shallow
layer of air is mixed upward by convection into cleaner air. Later in
this report we discuss a meteorological scenario observed in the Northeast
in which air picks up emissions from the Eastern Corridor sources, moves
out over the Atlantic Ocean, and after a period of one to two days returns
to land crossing the same line of sources.
We have presented evidence that the 70 major sources shown in Figure
4-1, and detailed in Tables 4-1, 2, play a dominant part in causing the
high ozone concentrations observed in the Northeastern United States.
Consequently, these sources will form a focal point of the analysis of
34
-------
the model simulations that we present later in this report.
Before leaving this discussion of the sources, we want to examine two
additional important points: the spatial scale of the sources vis-a-vis
model grid size requirements; and the current level of uncertainty in the
gridded VOC and NOX emissions inventories.
The spatial scales of the VOC and NOX sources
As we noted earlier the major sources shown in Figure 4-1 (page 43)
are responsible for about one-half of all the VOC and NOx emissions in the
Northeast. However, it is evident in Figure 4-1 that all these sources
combined cover only a small fraction of the land area contained within
the model domain. Together, these facts suggest that the characteristic
length scale of the emissions is small. In order to model ozone accurately,
it is necessary for the model's grid size to be small enough to resolve
the "granularity" of the VOC and NOX source distributions. If the grid
size is so large that major sources that are actually separated from each
other in space are mixed together within a single grid cell, essential
nonlinear characteristics of the chemical processes that act upon the
emissions will be lost and systematic errors in predicted ozone levels
will result. (See Lamb, 1986 for further details). There are two basic
ways of handling the source resolution problem. First, one can use a
coarse grid network and parameterize the nonlinear, subgrid scale chemistry
phenomena that otherwise would not be simulated. This approach is attractive
because the number of grid cells used in the model can be kept small
enough to hold the computer time and memory requirements to acceptable
levels. Its major drawback is that there does not exist a parameterization
35
-------
scheme of the kind required that is both accurate and universally valid.
The second method of mitigating the source resolution problem is to make
the model grid cells small enough to resolve the dominant features of the
source granularity. This method is computationally expensive; but it
imposes no assumptions on the model and therefore it adds no additional
sources of error. This is the approach to model design that we have
adopted in the ROM. Here we would like to quantify the scale of the VOC
and NOX source granularity in order to obtain an idea of how much subgrid
scale variation in concentration exists in a model of given grid cell
dimensions.
For this purpose we have compiled in Figure 4-2 (page 44) plots of
the cumulative emissions densities Q of both VOC and NOxas a function of
cumulative area. The plots were created as follows.
Consider, for example, VOC; and let QvocO) denote the emissions
density (moles km'2 day"1) of the i-th source. Let A(i) be the area (km2)
of the i-th source. And let the sources be numbered such that i = 1
represents the source (county) with the largest value of QVOC5 "* = 2
represents the second strongest source; and so on, with i = 439 representing
the weakest source. (The numbering assignments are different when
considering NOX emissions.) In reference now to the plot of cumulative
QVOC given in Figure 4-2, i.e., the upper curve, the ordinate Q and abscissa
A of any point on that curve are given by
I I
Q = Z QVQC(i)A(i)/ £ A(i) (4.8a)
i=l i=l
where I is the largest integer such that
36
-------
I
S A(1) < A (4.8b)
1=1
For example, the left edge of the upper curve in Figure 4-2 has
p i
ordinate Q^ = 53457 moles km day , which one can see in the first
entry in Table 4-1 (page 41) is the emissions density of the strongest
VOC source. The abscissa of this point is A = 58.1 km2. This is the
area of the strongest VOC source. Thus, our model would need a mesh size
of order A = /A" = 7.6 km to resolve the strongest VOC source in the
anissions inventory. Using the terminology we introduced earlier, we say
that the smallest "grain" size of the VOC sources is of the order of 7 km.
Obviously, the actual sources of hydrocarbons have dimensions much
smaller than this, but in compiling the emission inventory it is impossible
to describe the full microstructure of the emissions distribution. However,
sources that emit more than 100 tons per year of pollutant are considered
to be significant enough to warrant detailed specification of their
location, effective height, and so on. In the emissions inventory used
in this study, over 50,000 sources fall into this category. In the
model these sources are treated as point sources and their emissions are
placed in the proper layer based upon plume rise calculations performed
using meteorological conditions prevailing at the source site each hour.
However, within each layer the emissions of all point sources that fall
within a single grid cell are mixed together and treated as an area source.
Thus, the spatial resolution of the source distribution is limited in
the model by the grid size, which is about 18 km in the case of the ROM.
We have just seen that the grain size of the strongest area source of VOC
37
-------
in the inventory is of the order of 7 km, so we wonder how large the sub-
grid scale variations in source strength would be in a model with an 18 km
grid. The answer can be found from the curves plotted in Figure 4-2.
Reading the ordinate of the VOC curve at the abscissa A = 18 km, we find
that at best, an 18 km grid can capture only 83% of the actual peak source
strength; and, from the NOX curve, we see that only 78% of the peak NOX
emissions density is resolved. In other words, if the highest gridded
VOC source strength Q were 83 units, the largest actual value in the
field would be at least 100 units. The percentage value represents an
upper bound, because the gridded emissions would achieve this level only
if all the top sources that together fill an area A = (18)2 km^ happened
to lie within a single grid. For example, the two top VOC sources listed
in Table 4-1 together cover 317 km^, which is just under the area (18^) =
324 km^ of a single grid cell, and both happen to fall within the same cell.
So in the case of the ROM, 83% of the peak VOC source strength is resolved.
The same is not true of the NOX sources, however, because as we can see
in Table 4-2 (page 42), one of the top sources does not lie in the same grid
cell as the others. Therefore, the ROM's resolution of the NOxsource
strengths is somewhat less than the maximum achievable, 78%.
Figure 4-2 indicates that a model with 80 km resolution captures
less than 33% of the strengths of both NOX and VOC sources, and that a
model with 450 km grid cells is practically no better than one in which
the entire domain is treated as a single cell. As we pointed out earlier,
when a model is unable to resolve the strengths Q of the actual sources,
errors can occur in simulated concentrations of photochemical species.
Just how much resolution is necessary is not clear. It has been shown
38
-------
(Lamb, 1986) that compared to a model with 18 km resolution, models with
80 km grid cells or larger underpredict ozone concentrations by significant
margins. One would expect that the 18 km grid model also underpredicts
because it, too, is unable to resolve details of the VOC and NOxemissions.
We surmise that the grid resolution necessary to obliterate subgrid scale
effects altogether is determined jointly by the rate of turbulent mixing,
the time scales of the dominant chemical reactions, and the sizes of the
clusters in which individual emitters are grouped. Until this question
has been resolved, we must rely on comparisons of model predictions and
observed ozone concentrations to assess the significance of subgrid scale
phenomena.
Uncertainty in gridded NOX and VOC emissions
In the course of changing from the 1979 NECRMP emissions inventory
to the more recent 1980 NAPAP data, version 4.2, we discovered considerable
differences between the gridded VOC and NOX emissions values. Figures 4-3
and 4-4 (pages 45 and 46) show scatter plots of the 200 largest VOC and NOX
grid values, respectively, comparing the NAPAP and NECRMP data. The
regression lines in both cases show that the NAPAP inventory has effectively
reduced the estimated strengths of sources that were strongest in the
NECRMP data set and raised the estimated strengths of weaker sources.
For the most part the NAPAP values are within a factor of 3 of the corre-
sponding NECRMP estimates. The uncertainties are largest in the NOX data
where discrepancies as large as a factor of 100 occur. As we shall see
later, this "noise" in the emissions data is over 10 times larger than
the changes in emissions that the control strategies would cause. This
39
-------
brings into question the significance of the model calculations themselves,
Because the response of ozone concentrations to changes in VOC and NOX
emissions is a function of the base emission levels. Therefore, if the
estimated base emissions are in error, the simulated response of ozone
concentrations to emissions controls will also be in error. We are
pursuing the cause of these discrepancies in the emissions inventory and
hopefully it will lead to refined estimates of the true source strengths.
40
-------
r
Q t" ?"N c« •_' o *•' o r* -'jr. *-. .x ^* r i C1" c; o :. —< r** •*- ' - TJ *^ *?• "'? co • j c- *v f - *• • ;, . * *T **~ =;, \
> -r *r — tj <5 »- — ca -c r- -„• —. —• i. -^ o cj "• r*-. --. r- •" - - PI ^— f* -"? L~ c- -. re -. -; •- r • — ^- >
O1 n Pi r-n *r M -^ —. o o o cj r. -.1 r" ci a ~* 1-. o > c» c- r^ i*-- PV >u -.1 -,1 L* i." •.* . i.~ L"t i" i." t~: i
JJ x ** M ci o r^. n f j r** r*. M o *^ T r. PI «r *r c- UT -* T CJ cs *-• r«i -- r-s ""• *o r *r c »* c j -a T-N C3 —• 3* r i cs -o o c.~ ui *-" Q-- i~j *- c
tu o o •^ ~J c; PJ *r CD 'O T P. *T c-j •-* ro o -~: c- 3- o 3 LT r*^ c^ ^-» ',1 —* LI ri -o ' :o o- -o - 3 rx -- "* c; p; r-1 a --j -o & o M '^ '.- p^ c^
T PJ O O r*. — C3 -* U1 M CO T T "*5 -C <^ C^ C^ ^" LI >J 3^ «=* •>-* P) Pi « f- CO C3 Pi "". CS •** LT C' -^ -3 N. T -O -j i^ *T C- -J fv -H i,- p-
" *-• n o — r-i o
^ « o* o
33 CC O O LI !"1 *O CJ L") C- CD ft T f*S r*1 L^ M 0- C- Ci C^ -O S3 '-T -O "*» CO "^ « TJ O* O T-. — i (^ C" -O U"j n t,") O
— ca rv f*3 «r "j LT o o o M o o rv „ ;x u-j m o CK *r -o n 03 o r*^ rj *r cc c- -c ta LT c^ n n rj > •*> rt r,
M •«-» -o ca M n pi « -o ri m « •— o o^ i- u~ -^ > •« «»• T rj -• r% Li i-i
§2 ^>«or^iPicQ«oroooiiTTOncioo;'-t-ro>ooa3-ooops — afi-cr-'
_, LJ 3: G 2
tn
- >• uj >- ;- »c ;- ~z. — i—
1 UJ L. 0" LJ L— _. L. T _ i-1
k: Q ^ ^: ui ci ;; cn ^; _i tn cn — .£ a —
«i:5rju£u;issii:a:(K«:*-f-ttr«ju=
™ |OO>OLUUjUUiOauj O O Z Lu >~ tJ
"
2
en = a r a: =
* * •«
Table 4-1. Major county sources of VOC ranked in order of source strength
Q (= mass per area per time). Counties marked with an asterisk (*;
are also major sources of NOX (listed in Table 4-2).
41
-------
r.
t» L~ o r-- -* -j •' :~, c* r-- ." ••• t~~. v*. << — r, t: o* •-• f- c: o •-. r- •-* t> c^ r*. *?• ••" c^ ;~ r; CM ;;-*-•>•
^" L~ r. *r ci ~: c i -« r. c. ^ i-» r« —. ^ — T-. —. —
-* rt r*s M o o LT o rx «- rj rs *r f
I '
L
ca r>. > n x3 r- ^ ** n o o CJ
— o — 03 o 1.1 sj o» c* ^ >. r^
•o ca co ^ o* o x ~ C3 »- co o
T cs rj -* rs o M cs « -^ m rv
** v a- o* o fN. o o co -* N !?•
f c- n LT « o C3 -i o
•
•> « -o c, ri -o ^'• "*. o
— -3 o -3 • . ix r-
t-^ c-j o «r -r- -3 —
c: -. ri :-a i,~ -r TJ
i^ c^ r . CN c: -.3 r.
&» rv o* > ui 1^ si »* rj ri cc «r ^3 «r
« o n n a n o rj — o rj *-» "^ c-j c
^- - S3 LT ^^ f" ~* O T '."; r. i-"5 O ^x. r" ^. ~ fv L." r-» C*1 •>_.*"' ~
• •** -r* C"> r< T^ i-t -• rj ri rj n 5- ^* -~ " **•
ri rv t/i CN v ^ fx o (x rj s.
n n -a o ci ci M co c^ a
•^ m -a ri ri f-i -o >o co ss C
f4 -Q * ~l rv 1,1 *f O fj C3 -
^rirj«ca^uT^>rjr
o IN- o o* M rj ^ -«r IN. ui U
> — • ^ ri C3 > >• T c f
o «o — rs. o ^ -3 o -:
» Li O T 03 *» t."! Ci C3 LT
>^>r^o««MO
^o ui T LI o in co m
»* ^ n r^ -o T LT C" PV
01 CD rx o* o rs « co n
O Cs. TJ -* > *T TH T >.
'
h~ ^^ o to — * f! «* r^ ^ r; i-« t^
m c; T -o o w «*s -- s T > -,3
"*- '-1 fS *w i/1 O CO rO > -r '.T CC
''
- ~t 23
L* ca
. f> O
ca c-i -
ui rj •
rn m r-i M m m rn u
ri rn rn i-s m
m ri n m ;-i n
5
• 3
-
O 3 W 3 —
-------
X
o
CJ
o
r
ID
X
o
z
I
(3
111
cn
Figure 4-1. Locations of the 70 major source counties listed in
See Eq. 4.7 for definitions of "high VOC", "high NO,
"high VOC and NOX".
Tables 4-1,2.
", and
43
-------
NOX AND YOG
(CUMULATIVE MASS / CUM AREA)VS(OIM AREA)
10
A (km
I 1 ' I I »
10 20 30 50 80 100
A (km)
10°
200 300 500 800 1000
Figure 4-2. Cumulative VOC and NOX emissions densities vs. cumulative source
area and characteristic length scale. See text, pages 36 - 37,
for details.
44
-------
NAPAP BASE VS NECRMP BASE
LOG VOC
7.5-
5.0
5.5
6.0 6.5
NECRMP BASE LOG VOC
7.0
7.5
Figure 4-3. Comparison of NAPAP and NECRMP estimates of gridded VOC emissions
for the 200 strongest source cells in the ROM grid network.
45
-------
NAPAP BASE VS NECRMP BASE
LOG NOX
7.0-
N
A
P 6.5-t
A
P
*
*
B 6.0-
A
S
E
x
5.5-
L
0 5.0^
G
N
0 4.54
X
4.0-
* *
3X
/'*
/ *
* ,' * *
,* ** **
* &?!&£?*,''
• xJ^bdBdP3!'^
* ^^ ^fl*
«ii« 9" IK 5K
-yiiUS^ .*'
X
/
/
I I
4.0
I I
4.5
] 1 I I ! I |
5.0 5.5 6.0
NECRMP BASE LOG NOX
6.5
7.0
Figure 4-4. Comparison of NAPAP and NECRMP estimates of gridded NOX emissions
for the 200 strongest source cells in the ROM grid network.
46
-------
SECTION 5
DEFINITIONS OF RECEPTOR CLASSES FOR USE IN ANALYZING
SIMULATED CONCENTRATIONS
In several earlier reports (Lamb, 1983a, b; 1984) the author pointed
out that the concentrations predicted by a regional scale model using a
discrete set of meteorological data and a given source inventory are
stochastic rather than deterministic variables. In the following para-
graphs we will review the basic concepts behind this premise. They are
important to the manner in which models should be employed in decision
making roles and they form the basis of the technique we use in this
report to analyze model results.
The root of the premise is the mathematical fact that a discrete set
of meteorological data do not uniquely define the atmospheric flow field.
Rather, they delineate an infinite set, or ensemble, of flow fields any
member of which is a possible description of the flow that actually
existed at the time the meteorological measurements were made (see Lamb
and Hati, 1986). Hence, for a given source distribution specification,
the equations governing species concentration in the atmosphere define a
one-to-one mapping of members of the ensemble of flow fields to members
of an ensemble of concentrations.
The network of processors that compose the ROM framework (see Figure
2-3, page 7) was designed to produce the information necessary to simulate the
ensemble of concentration fields associated with a given set of meteorological
and emissions conditions. From the simulated concentration ensemble one
can derive probabilities, expectations or any desired statistical property
47
-------
of concentration at a single site or group of sites. We refer to this as
the "probabilistic mode" of ROM operation. Strictly speaking this is the
operating mode that should be employed both in model verification studies
and in any control strategy assessments in which one is interested in
effects at one or more specific receptor sites.
Complementing the probabilistic mode is the "quasi-deterministic mode"
in which the ROM generates a single concentration field for each species.
This single field, or realization, is computed by the ROM using a procedure
designed to select from the concentration ensemble the member that has
the highest probability of occurrence. Despite the fact that this is the
most likely concentration field, the significance of values at any given
point in space and time is limited in the same sense that the most probable
outcome of a random process is not necessarily the outcome that one would
observe on any given occasion. In this case one might wonder what point
there is in running the model in the quasi-deterministic mode.
One reason is that useful information may exist in certain aggregate
properties of individual realizations of the concentration field, as
opposed to the value of concentration at a specific site. Based largely
on theoretical arguments, Lamb (1983b) hypothesized that receptor sites
can be grouped into classes, determined by the distances to the nearest
meteorological station and major emissions sources, such that concentration
averaged over all sites in a given receptor class at a given hour is a
quantity that is approximately the same for all realizations in the
concentration ensemble. If this is true, then with respect to the receptor
class average (RCA), the concentration field is a quasi-deterministic
48
-------
variable, and we can interpret model simulations of single realizations
in the conventional deterministic manner. Although this hypothesis has
not yet been tested, all applications of the ROM in the foreseeable
future will proceed under the assumption that it is correct. Because if
the hypothesis is false, that is, if there does not exist any useful
aggregate property of the concentration field that is relatively invariant
across the ensemble, then meaningful results can be derived only from
probabilistic mode operations of the model. And at present this mode of
operation is prohibitively expensive. As an example, analysis of a n-day
period requires only a single run of the ROM in the quasi-deterministic
mode but five to ten runs in the probabilistic mode. Thus, the latter
mode is five to ten times more expensive. In the case of the ROM in its
current form the difference in costs in the two modes of operation for a
5-day simulation period could be of the order of $100K.
Our digression on the model's operating modes was intended to give
the reader the proper basis for interpreting the results of the control
strategy simulations that we are reporting here. In this study we applied
the ROM in the quasi-deterministic mode, once to the 23-31 July 1980
time period using the base case source emissions as inputs, and then a
second time to the same period using the modified emissions. According
to the points we raised above, the only aspect of the model results that
we can use to assess the effects of the control strategy is statistical
properties of concentration over receptors of the same class. Comparing
the concentrations predicted at a single receptor using the base emissions
with the corresponding value computed with the control emissions is not
meaningful because we are examining only a single realization of the
49
-------
concentration ensemble and the relationship between the two concentration
values could vary radically from one member of the ensemble to another.
By contrast, differences in the receptor class average (RCA) concentration
obtained for the base and control emissions do not vary appreciably within
the ensemble, according to our untested hypothesis, and therefore we can
use differences in the RCA's as a measure of the impact of the emissions
control strategy.
Since the ROM is a grid model that simulates concentration averaged
over the volume of each grid cell, rather than concentration at individual
points in space, our analyses of the model results must be done in terms
of receptor cells rather than receptor points. Thus, in analogy with the
receptor classes mentioned earlier, we will define cell classes. For the
purposes of the present report, we define four cell classes — urban,
suburban, rural and wilderness — as follows:
urban cell => 70 strongest VOC and NOX cells in the
model domain, excluding Canada;
suburban cell => "urban" land useage U > 10% and cell is
within 50 km of any cell of the "urban" class;
rural cell => 0 < U < 10%, distance d to the nearest
cell of the "urban" class is in the range
50 < d < 100 km, and "agricultural" land
useage > 10%;
wilderness cell => U = 0 and distance d to the nearest cell of
the "urban" class is greater than 100 km.
These definitions are largely heuristic and are meant mainly to illustrate
applications of the ROM in the quasi-deterministic mode. In a future
study we will test the hypothesis concerning the determinism of receptor
class average concentrations and at that point rigorous specifications of
class definitions will be developed, assuming that the hypothesis is
validated.
50
-------
In selecting specific cells from the ROM's domain for inclusion in
each of the four classes, we began by finding all cells that satisfied
the "urban class" criterion. A total of 70 cells in the 60 x 42 cell
domain were found to meet these conditions. (Most of the 70 major source
counties listed in Table 4-1, 2 (pages 41, 42) are represented by at
least one of the urban cells.) Next, we searched for cells fulfilling
the "suburban" class criteria. More than 70 cells met these requirements
so we selected 70 at random from the candidate set. This procedure was
repeated for the "rural" and "wilderness" classes, resulting finally in
sets of 70 cells of each of the four categories. Having equal numbers of
cells in each group helps minimize sampling aberrations. In the selection
process only the land area within the United States was considered.
Figure 5-1 (page 53) shows the locations of each of the cells in all four
receptor classes. We emphasize that some grid cells do not fall within
any of the four receptor types.
The criteria that we used to define the receptor classes were designed
to satisfy what we believe are necessary conditions for the validity of
the ensemble invariance hypothesis that we discussed earlier. However,
the characteristics of the receptors that meet these criteria are not
necessarily best suited to the needs of regulatory analyses where one is
concerned with exposures of population, crops, and forest. Furthermore,
regulatory studies must consider all locations rather than just subsets
of randomly chosen sites. To serve these needs we have defined four
county types analogous to the four receptor classes defined above. These
are as follows:
51
-------
major source county = any one of the 70 major source counties listed
in Tables 4-1, 2 (pages 41, 42);
non-major source SMSA = Standard Metropolitan Statistical Area (SMSA)
other than major source counties;
agricultural county = not a county of either of the above types
agricultural land useage exceeds the sum of
forested land plus unforested wetlands;
natural county = all counties not falling into any of the
above categories.
The counties in each of these four groups are listed in Appendix C and
illustrated in Figures 5-2, 3, 4 and 5 (pages 54-57). In our analyses of
the model results in Section 8, we will make use of both the receptor
cell and receptor county classes defined above.
52
-------
fi € * *
R. I ?4S.O°X
• v
R J 3
38.0°.\r
Figure 5-1. Locations of the 70 cells selected in each of the four
receptor classes urban (U), suburban (S), rural (R) and
wilderness (W).
53
-------
CLASSCD
AGRIC
NATURAL
SUBURBAN
URBAN
Figure 5-2. Locations of the 70 major source counties, defined as
"urban" type receptors.
54
-------
CLASSCD
AGRIC
NATURAL
SUBURBAN
URBAN
Figure 5-3. Non-major source Standard Metropolitan Statistical Areas
(SMSAs) defined here as "suburban" type receptors.
55
-------
CLASSCD
AGHIC
NATURAL
SUBURBAN
URBAN
Figure 5-4. Non-major source, non-SMSA counties in which the predominant
land use is in agricultural activities, defined here as
"agricultural" type receptors.
56
-------
CLASSCD
AGRIC
NATURAL
SUBURBAN
URBAN
Figure 5-5. Counties in which forest and/or wetlands are the predominant
land use, defined here as "natural" type receptors.
57
-------
SECTION 6
METEOROLOGY OF THE TEST PERIOD: 23 - 31 JULY 1980
It is customary in air pollution modeling studies to specify the
meteorological conditions used in performing the simulations, i.e., wind
speed and direction, mixing depth, cloud cover, and the like. In urban
modeling this task is rather straightforward because one is concerned
with the limited spatial domain of a single city and with a time period
of only one or two days, usually days in which meteorological conditions
were stagnant. Therefore, in these particular applications meteorological
conditions are quite well defined.
However, in regional scale studies that extend over 1000 km areas
and periods of a week or more, "meteorology" in the sense in which it is
used in urban studies is not well defined; because conditions vary from
site to site within the spatial domain at any instant and from time to
time within the simulated period at any site. One can discuss the passage
of fronts through the simulated area or meteorological conditions at
selected times and locations, but for the most part these discussions are
of little or no value because they do not describe the cumulative effects
of transport and mixing over time that affect concentrations over a regional
area, particularly concentrations in rural and remote areas. Moreover,
discussions of the meteorological variables themselves provide no perspec-
tive from which to judge the significance of the results of the model
simulations. For example, if the model is being used to assess the
effectiveness of emissions control strategies, one needs to know whether
the predicted effects can be regarded as applicable to other seasons
58
-------
or other years; or whether, due to the use of meteorological data from an
anomalous period, the results represent conditions that rarely occur.
These considerations have led us to the conclusion that in regional
scale, multi-day modeling studies the focus of the meteorological
discussions should not be on the variables themselves but rather on the
effect the meteorology had on species concentrations observed during the
period of interest. In particular, we believe that the most relevant
indicator of the air pollution potential of the meteorological scenario
selected, and the best index for judging how applicable the model results
are likely to be to other times within a season or to different years, is
the frequency distribution of observed concentration values, principally
ozone, during the scenario period and the extent to which this distribution
differs from that observed over the entire season or over the same season
in a number of years combined. Detailed analyses of the meteorological
variables can be useful in understanding why particular concentration
phenomena occur. We will present an example of an analysis of this type
later in this section.
Figure 6-1 (page 67) shows one of the concentration frequency distri-
butions that provides information about the effects of the meteorology
used in the model simulation. The figure shows a quartile plot of the
maximum hourly ozone concentrations observed each day at all SAROAD
monitoring sites in the model domain, some 200 sites altogether, during
the three summer months June, July and August of 1980. (June through
September is the period in which the highest ozone levels are generally
observed each year in the Northeastern U.S.) Below is an explanation of
59
-------
the symbols used in the quartile plot shown in Figure 6-1.
largest maximum hourly ozone observed at all SAROAD
sites on the day indicated by the abscissa
75-percentile of maximum hourly ozone observed
at all SAROAD sites this day
median value
25-percentile value
minimum of all maximum hourly ozone values
observed at all SAROAD sites this day.
The data presented in Figure 6-1 indicate that the meteorological
conditions that are to be used in the model simulations, namely data
from the period 23-31 July 1980, are not capable of producing ozone
concentrations as high as those that were observed on other occasions in
1980. However, this particular meteorological scenario did manage to
drive the median value of the daily maximum ozone values to a level almost
as high as it achieved at any other time during the summer.
Note that at the beginning of the chosen period ozone concentrations
throughout the model area were as low as at any other time during the
season. This aspect of the concentrations was the principal reason for
selecting 23 July as the beginning of the simulation period. Because as
we have already discovered in our earlier model applications, attempts to
initialize a simulation at a time when significant amounts of ozone or
precursor species are present in the domain introduce considerable error
60
-------
into the calculations because monitoring data are not available in
sufficient density to specify the concentrations of all the simulated
chemical species accurately. The simplest way to minimize the initialization
errors is to begin the calculations at a time shortly after a "clean" air
mass has moved across the model domain and concentrations are near
tropospheric background levels at a majority of the monitoring sites.
Consequently, in our model applications we have adopted a policy of using
meteorological scenarios that begin at times when the median value of the
observed hourly maximum ozone concentrations over the entire model domain
is minimal.
One might argue that if the only purpose of the model simulations
is to compare concentrations that would result from different source
fields, it should make no difference when the meteorological scenario
begins. But this is not true because the concentrations that occur at
certain locations, particularly those remote from the major source areas,
may be the result of the cumulative effects of transport, mixing and
emissions over a period of several days. For example, if we started the
meteorological scenario at an arbitrary hour and assumed, for the purposes
of initializing the model that "clean" conditions existed at that hour,
we might easily underpredict the peak concentrations that a given set of
emissions would cause if the starting hour happened to be in the midst of
a developing air pollution episode. However, the chances of such an error
are greatly reduced by starting the meteorology at a time when concentrations
were observed to be near minimum values.
The data presented in Figure 6-1 provide a measure of how well the
chosen meteorological scenario represents the air pollution modeling
61
-------
region as a whole. Similar information pertaining to individual
locations within the region is provided by quantile-quantile plots such
as those shown in Figure 6-2a,b,c (pages 68-70) for selected monitoring
sites. These graphs compare the cumulative frequency distribution of
hourly averaged ozone concentrations observed during the scenario period
23-31 July at a given monitoring station with the corresponding distribu-
tion observed over the whole season, June - August, at the same site.
The graphs are read as follows. Let (xy, yj) be the coordinates (abscissa,
ordinate) of the point marked "7" in Figure 6-2a (or 6-2b or c). Then
70% of the hourly averaged ozone concentrations observed during the
season at the given site were below xy(ppb) while 70% of those observed
during the scenario period were below y7(ppb). The same interpretation
applies to the point marked "5" except in this case reference is to the
50% decile rather than 70%. Points marked by asterisks refer to quantiles
other than the decile points explicitly noted.
If the concentration distribution observed during the scenario period
were identical to that observed over the entire season, the points in the
quantile-quantile plots would all lie on the "45-degree" line (drawn in
each figure). Looking at each of the graphs in Figure 6-2 we see that
below about the 90 percentile, the scenario and seasonal concentration
distributions are in surprisingly close agreement at each of the three
stations represented (named below), while above the 90 percentile point
significant differences occur.
In the case of Figure 6-2a (page 68), which represents ozone data
collected at a rural site in Connecticut about 150 km northeast of New
York City, the top 10% of the concentrations observed during the 9-day
62
-------
scenario period ranged from about 100 ppb to a high of Ib9 ppb, whereas
the top 10% of the values measured during the season span an interval
nearly 3 times as wide — 100 ppb to 262 ppb. Since this particular site
is in the heart of the area where the highest ozone concentrations observed
in the Northeastern U.S. occur (see Figure 3-1, page 20), we might conclude
that the meteorological scenario that we have chosen for the model simula-
tions is not the best for testing emissions control strategies in this
region. (As we have stated before, formal applications of the Regional
Oxidant Model for regulatory studies will not be undertaken until the
final version of the model is complete, and we have resolved issues such
as the one in question here regarding the relevance of short-period simu-
lations to the ozone problem in general.)
Figure 6-2b (page 69) compares the seasonal and scenario ozone frequency
distributions measured at Whiteface Mountain, a remote, high-altitude
site in upstate New York. Again the distributions are in good agreement
up to the 90-th percentile. Beyond that point the extreme concentrations
observed during the scenario are smaller than those seen over the entire
summer.
Finally, in Figure 6-2c (page 70) we have an instance in which concen-
trations higher than the 90-th percentile value occurred more frequently
during the scenario period than during the season — just the reverse of
the situation at the other two monitoring stations. The data represented
in Figure 6-2c were collected at a New Jersey site about 50 km east of
Philadelphia.
63
-------
In summary, the ozone observations indicate that:
(1) over the modeling area as a whole meteorological conditions that
occurred during the scenario period 23-31 July 1980 were not
adequate to cause extreme concentrations, although at isolated
locations values equal to the seasonal peak were observed;
(2) At seven representative receptor locations, three of which we
discussed (see Figure 6-2), concentration values below the 90-th
percentile level were observed at each station with about the same
frequency during the 9-day scenario period as during the whole
season, June-August. However, concentrations higher than the
90-th percentile value at any station occur with significantly
different frequencies during the scenario than during the season.
We infer from the latter finding that 90 percent of the time, concen-
trations at any location are dominated by processes whose time scales are
much shorter than 9 days. We have already pointed out that there is a
strong diurnal variation in ozone at surface sites with a net production
of ozone occurring during daylight hours and net destruction occurring at
night. So we suspect that it is diurnal variations in wind speed, turbulent
mixing and sunlight that determine concentrations 90 percent of the time.
The highest concentrations observed at any site, namely those seen less
than 10 percent of the time, appear to be caused by processes whose
periods are longer than 9 days. This conjecture is based on the following
reasoning. Suppose the period of the events in question is less than 9 days.
In this case the 9-day sample window will capture several events which would
tend to cause the scenario frequency of high concentrations to approach the
frequency observed over a long period. However, if the event cycle is
64
-------
longer than 9 days, the scenario frequency of high concentration will
either exceed or fall below the corresponding long-term frequency depending
on whether the 9-day sample window captured or missed, respectively, one
of the events in questions. It is the latter behavior that we observe in
the ozone frequency data.
The fact that so much of the seasonal frequency distribution of ozone
concentrations is emulated by a distribution based on only 9 days of
observations suggest that quite reliable estimates of seasonal mean
concentrations could be derived from 9 day model simulations. The same
is not true, however, of peak concentrations. We are presently studying
the ozone data in detail in an effort to determine criteria for selecting
meteorological scenarios that will yield model outputs that can be used
to infer estimates of all seasonal concentration statistics.
At the beginning of this section we stated that analyses of meteorolo-
gical variables performed in the conventional way could be useful in
understanding why particular concentration phenomena occur. Air parcel
trajectory analyses in particular are often helpful in diagnosing causes
of air quality events because it is the Lagrangian characteristics of the
wind field that govern concentration distributions. Figure 6-3 (page 71)
shows trajectories computed from the 9-day meteorological scenario
wind fields. The trajectories begin at hour 0000 of day 206, i.e., the
second day (July 24) of the 9-day period, at a number of major urban
centers in the Northeast. The trajectories terminate on day 211, with
the exception of those that pass outside the model boundaries before this
day. Day 206 is significant because a cold front had just crossed the
65
-------
East Coast by the beginning of this day (see Appendix A), and a new,
clean air mass had established itself over the region.
The trajectories from the Eastern Corridor cities are especially
interesting. They indicate that air that crossed the Eastern Corridor
cities on day 206 returned to the same general area after 2 to 3 days of
travel and crossed the cities a second time traveling in roughly the
opposite direction. According to the simulations, air that originated
over cities between Philadelphia and Boston spent a large part of its
travel over the ocean where, as we have explained before, sunlight condi-
tions and weak vertical mixing are highly conducive to ozone formation
from VOC and NOX emissions. The recirculatory nature of the trajectories
is a consequence of the latitude of the cities in question relative to
that of the path of the anticyclone that moved across the region following
the frontal passages. Thus, we are led to wonder whether the combination
of latitude and proximity to the ocean makes the Eastern Corridor cities
more susceptible to air quality problems than other geographical areas of
the country. We are investigating this question in our present studies
because it has important bearing on the formulation of effective emissions
control plans for this area.
66
-------
MAX
HOURLY
OEONE
20CH
ppb
100
(300
200
100
JUNE 1380 | JULY J
23 JULY
AUGUST
31 JULY
Figure 6-1. Quartile plot of the maximum hourly averaged ozone concentrations
observed at some 200 SAROAD monitoring stations each day during
the period June - August 1980. See text, page 60, for explanation
of symbols.
67
-------
Figure 6-2a.
Quantile - quantile comparisons of the cumulative frequency
distributions of hourly ozone concentrations at a rural
site in Connecticut (070570007). The ordinate (y-axis)
represents the distribution observed during the meteorological
scenario period 23-31 July 1980 while the abscissa (x-axis)
represents the distribution observed over the entire summer,
June - August.
200-
-Q
Q.
Q.
o
IM
O
03
-------
Figure 6-2b.
Same as 6-2a except data are from Whiteface Mountain,
New York (33020002).
100
a.
a.
o
N
O
0)
oo
50
1OO
Seasonal Ozone (ppb)
69
-------
Figure 6-2c. Same as 6-2a except data are from a New Jersey site
(312750001) about 50 km east of Philadelphia.
300
150-
O.
CL
c
o
o
o
OJ
O
oo
loo-
*-<, PKOBT .-ill
50
loo
ISO
200
Seasonal Ozone (ppb)
70
-------
Figure 6-3. Trajectories calculated from wind fields used in Layer 1 of the
ROM for the 9-day meteorological scenario 23-31 July 1980
(Julian dates 205-213). Numbers along the trajectories refer
to Julian date.
71
-------
SECTION 7
THE CONTROL STRATEGY EMISSIONS INVENTORY
The control strategy inventory was specified by the Strategies and
Air Standards Division of OAQPS in the form of county-by-county changes
in total hydrocarbon and NOX emissions. The control inventory represents
projected 1987 emissions based on controls contained in the 1982 SIPs.
No control measures proposed after 1982 are included. The specified
change in hydrocarbon was applied equally to each of the simulated hydro-
carbon species — paraffin, olefin, aldehyde and aromatic — and the change
in NOX was applied to both NO and NOg. Figure 7-1 (pages 75 and 76)
shows isopleths of the hydrocarbon and NOX emissions changes. Details
are provided in Appendix B. The most significant aspect of the control
emissions is that hydrocarbons are reduced everywhere from the base case
levels whereas NOX is decreased in some counties and increased in others.
For the most part hydrocarbon emissions are reduced by 32% and NOX is
reduced by 8%.
A different perspective on the character of the emissions reductions
is provided by the scatter plots in Figures 7-2,3 (pages 77 and 78) which
compare the base and control VOC and NOX emissions, respectively, of the
200 strongest source cells in the ROM domain. The regression line (dashed)
in each figure shows that in the case of both the VOC and NOX controls,
there is a slight tendency to reduce the emissions of the strongest sources
by smaller percentages than the weaker ones. For example, a source (in
this case a grid cell roughly 18x18 km) emitting 10? moles VOC per day is
reduced 22% on the average whereas one emitting 10^ moles/day is reduced
35% on the average.
72
-------
Earlier, in Section 4, we alluded to the high level of error that
apparently exists at present in the base emissions inventories. Comparing
Figure 7-2 with 4-3 (page 45) and 7-3 with 4-4 (page 46) one gets the
impression that the effects of the emissions reductions that we are
trying to assess are likely to be overshadowed by the system "noise"
in the base emissions inventory. Until the causes of the uncertainties in
the emissions data have been diagnosed and corrected, we must interpret
the results of the model simulations with caution.
We argued in Section 3 that the emissions reductions imposed on the
major sources of VOC and NOX would have a proportionately greater impact
on ozone levels than the changes made in lesser sources. Thus, it is
instructive to examine the effects of the proposed control strategy on
the 70 major source counties identified earlier in Tables 4-1,2 (pages 41
and 42). Figures 7-4 and 7-5 (pages 79 and 80) show scatter plots of the
base and controlled VOC and NOX emissions, respectively, of these 70
counties. The data presented in Figure 7-4 indicate that about 40% of
the major sources have little or no VOC reductions at all — the remainder
have 30% to 40% reductions. Figure 7-5 indicates that on the average the
NOX emissions of these sources remain virtually unchanged, as is evident
in the regression line.
We should emphasize that the emissions that we refer to here as the
"controlled" emissions actually represent the projected 1987 baseline
emissions. Thus, the model simulations that we have performed comparing
the ozone concentrations simulated with the 1980 "base" emissions with
the ozone levels simulated using the "controlled" inventory are intended
73
-------
to assess the magnitude of the ozone problem that one could expect in 1987
if no further controls beyond those in the 1982 SIPs are placed in effect.
As we shall see in the next section, the model predicts that 1987 ozone
levels will still exceed the primary standard in many parts of the North-
east. Consequently, one of our future modeling tasks will be to test
various control strategies in an attempt to find the most cost effective
one.
74
-------
Figure 7-la.
VOC emissions in the control strategy expressed as a
percentage reduction in the base emissions. (See
Appendix B for details.)
75
-------
Figure 7-lb.
NOX emissions in the control strategy expressed as a
percentage reduction in the base emissions. (See
Appendix B for details.)
76
-------
NAPAP BASE VS NAPAP CONTROL
LOG VOC
5.0-
5.0
5.5
6.0 6.5
NAPAP CONTROL LOG VOC
7.0
7.5
Figure 7-2. Comparison of base and controlled emissions of VOC (moles/day)
of the 200 strongest VOC emitting cells in the ROM domain,
Canada excluded. (All point and area sources within each
cell are combined.)
77
-------
NAPAP BASE VS NAPAP CONTROL
LOG NOX
4.5-
4.5
5.5 6.0 6.5
NAPAP CONTROL LOG NOX
7.0
7.5
Figure 7-3. Comparison of base and controlled emissions of NOX (moles/day)
of the 200 strongest NOX emitting cells in the ROM domain,
Canada excluded. (All point and area sources within each
cell are combined.)
78
-------
NAPAP BASE VS NAPAP CONTROL
LOG VOC
3.5
NAPAP BASE LOG VOC
I
4.0
4.5
T
5.0
Figure 7-4. Base vs. controlled VOC emissions (moles per km^ per day) of
the 70 major source counties listed in Tables 4-1,2.
79
-------
NAPAP BASE VS NAPAP CONTROL
LOG NOX
2.0
3.5 4.0
NAPAP BASE LOG NOX
4.5
5.0
o
Figure 7-5. Base vs. controlled NOX emissions (moles per knr per day) of
the 70 major source counties listed in Tables 4-1,2.
80
-------
SECTION 8
RESULTS OF MODEL SIMULATIONS
In this section we present the results of model simulations that
were performed to assess the effects of projected VOC and NOX emissions
reductions on ozone levels in the Northeast. The presentation is divided
into three parts. In the first we describe the input parameters used in
each simulation. Following that we provide a limited comparison of the
ozone predicted in the "base" case simulation with measured ozone data to
give at least a qualitative overview of the model's accuracy. In the
final subsection we discuss the predicted ozone concentrations from both
the base and control runs.
SPECIFICATION OF INPUT PARAMETERS FOR THE MODEL SIMULATIONS
Two, 9-day model runs were made in this study. One, referred to as
the "base" case, used the 1980 NAPAP version 4.2 emissions data discussed
earlier in Section 4. The second run, referred to as the "control" case,
used emissions fields derived by multiplying the base emissions by the
projected 1987 VOC and NOX changes described in Section 7 and specified
in Appendix B. Both the base and the control runs used identical values
of all other model inputs including the meteorological fields, which we
discussed in Section 6 (see also Appendix A). Following we elaborate on
our choice of initial and boundary conditions for the simulations, and
other parameter fields that we have not yet mentioned.
81
-------
Initial and Boundary Conditions
In every model application, regardless of its purpose, one must
specify the 3-D spatial distribution of the concentrations of each of the
simulated species at the initial instant of the simulated period. In the
case of the ROM in its current form, this means specifying values of the
•
concentration of each of 23 chemical compounds in each of some 7500 grid
cells. Unfortunately, this requirement vastly exceeds the information
content of presently available air monitoring data bases. Within the
present ROM domain, measured hourly ozone concentrations are available at
about 150 surface sites; lower quality data are available for NOX at
fewer sites; and no measurements are available for any of the remaining
23 species. Moreover, no data are available on the concentrations of any
species above ground-level.
In preliminary studies, we attempted to devise a scheme for generating
estimates of the concentrations of all the required species from the few
ozone and NOX data available. In order to appreciate the difficulty, and
indeed futility, of this task, keep in mind that the objective of this
scheme is to reconstruct the 3-D spatial structure of the concentration
fields of 23 compounds given the concentration of only two of the species
at a few surface locations. If the nature of the pollutant chemistry
were such that the concentrations of all species were unique functions of
the concentrations of the two given species, then initialization would
not be a problem. But this is not the case. In fact, it appears that
the ozone concentration that evolves from a given initial mix of species
is quite sensitive to the initial levels assumed for hydrocarbon and NOX,
and as a consequence anomalous ozone concentrations may arise in the
82
-------
course of the simulation that are merely artifacts of the initialization.
This problem is made all the more acute in the regional model by the
prolonged residence time, 4-5 days or more, of the initially present
species within the model domain. We have concluded from our studies that
there is presently no initialization technique available that will allow
consistently accurate simulations of historical air pollution episodes
unless the period is selected so that it begins at a time when species
concentrations are at minimum levels throughout the modeling region.
A related modeling problem is that of specifying the concentrations
of all species at inflow boundaries. Most of our model simulations run
long enough that species that enter the domain through the inflow boundaries
eventually permeate the model area to nearly the same extent that the initial
concentration field does. When we defined the ROM domain originally we
attempted to mitigate the effects of uncertainty in the lateral fluxes of
species by making the domain large enough to encompass the majority of
the pollutant sources that affect air quality in the Northeast. The
rationale was that if we treat the majority of the sources explicitly,
the influence of transboundary fluxes on species concentrations at receptors
in the interior of the domain will be at most a second-order effect and
well below the level of tolerable error in the predictions.
Specifying the initial and boundary conditions accurately is necessary
only if one plans to compare the predicted concentrations with observed
values for model verification purposes. If the model application is for
emissions control strategy assessment, meaningful results can be obtained
by using hypothetical initial and boundary values with meteorological data
associated with historical pollution episodes. In fact, one could argue
83
-------
that a polluted state of air quality observed in the past is irrelevant
in control strategy development because if emissions are changed, that
state of air quality might not be seen again in the future. The same
argument does not apply, however, to the meteorology associated with
historical episodes.
«
The hypothetical initial and boundary conditions that are most
appropriate in emissions strategy studies are those regarded as representing
the clean troposphere. Simulating the effects of emissions changes in
this context makes the task of interpreting the model results particularly
easy because all the computed ozone and product species in excess of the
initial values can be attributed to emissions rather than the initial
and/or boundary conditions. The ROM applications that we will report
here were conducted using the "clean" species values given in Table 8-1
(next page) for both the initial and boundary conditions.
Deposition Velocities
In the first-generation ROM, deposition of each chemical species is
treated using the same state-of-the-art technique that the second generation
model will employ. In this method each land use type is assigned an
uptake resistance for each chemical species simulated by the model. The
value of the resistance is a function of atmospheric stability which is
in part a surrogate of relative humidity and sunlight, both of which
influence the rate at which plants absorb gasous material. The deposition
resistances are converted in turn to deposition velocities using the
local friction velocity, Obukhov length, and surface roughness. The
resulting distributions of deposition velocities have highly complicated
84
-------
Table 8-1. Values used for the initial and boundary concentrations
in both the base and control ROM applications.
Species
Concentration (ppm) value used for initial
and boundary conditions in all model layers
NO
N02
03
Olefin
Paraffin
Aldehyde
Aromatic
CO
HN02
HN03
PAN
RN03
H202
0
N03
HO
H02
H04N
RO
R02
R20
R102
R202
6.58
1.82
.037
3.09
3.92
4.32
1.22
0.100
1.94
1.53
4.93
1.80
1.80
2.18
1.34
7.95
3.54
2.11
1.25
1.72
1.19
7.09
1.52
10-6
10-3
io-4
io-4
io-5
io-4
10'6
io-4
10-8
io-11
io-11
10-12
IO-5
10-9
IO-6
IO-6
10-12
IO-6
10-13
10-10 •
10-7
85
-------
spatial patterns and they undergo radical variations in magnitude between
night and day. In the case of ozone, the largest daytime deposition
velocity anywhere in the model domain is of the order of 1 cm/sec. At
night the largest value is less than .1 cm/sec and in most locations it
is two orders of magnitude below the daytime levels.
»
The principal weakness in the parameterization of species deposition
is the paucity of resistence data. In applications of the second-generation
ROM we may be able to update the empirical data base that we presently use.
Miscellaneous Fields
In the first-generation ROM that we used in the present study:
0 the terrain is flat;
0 the thicknesses of each of the three layers is constant in space
and time with Layer 1 = 400 meters, Layer 2 = 800 m and
Layer 3 = 800 m (these values were selected on the basis of
vertical sounding data);
0 horizontal diffusivity is everywhere zero;
0 vertical fluxes between layers are based on turbulence estimates
and vary in space and time;
0 vertical exchange between layers 2 and 3 occurs only where cumulus
clouds are present;
0 the mean vertical air speed is everywhere zero.
COMPARISON OF PREDICTED AND OBSERVED OZONE CONCENTRATIONS
In order to give the reader a basis for judging the credibility of
the emissions control simulations that we will present later in this
86
-------
section, we provide here a very limited comparison of the ozone concentrations
predicted during the 9-day base case simulation with ozone levels observed
during the same period. More detailed analyses of the model's performance
can be found in Schere (1986). Our model verification studies have been
limited thus far because the first-generaion verison of the ROM that we
are currently using will soon be replaced by the more advanced second-
generaton model.
Figure 8-1 (page 104) shows contours of the maximum, hourly averaged
base case ozone concentrations predicted within the model domain over the
entire 9-day simulation. Figure 8-2 (page 105) shows the corresponding
concentrations measured at the SAROAD sites during the 9-day period from
which the meteorological parameters used in the model were taken, namely
23-31 July 1980. Comparing the patterns and magnitudes of the concentra-
tion distributions shown in the two figures reveals the following.
First, the simulated concentration field exhibits distinct maxima
in four areas. Three of these lie over the Great Lakes — Huron, Erie
and Ontario — while the fourth occurs over the Atlantic Ocean just south
of Long Island. In each case the high concentrations are the result of
the limited mixing and generally cloud free conditions that exist in the
summertime marine layers over the Great Lakes and the Atlantic Ocean. The
simulations indicate that the Lake Huron and Erie maxima are attributable
to emissions from the Detroit area while the Lake Ontario maximum is a part
of the Toronto plume. The high concentrations over the Atlantic are the
results of emissions from the New York City area. Confirmation of the
existence of peak ozone concentrations in these four areas is not provided
by the SAROAD data because only land based observations are available.
87
-------
However the measurements shown in Figure 8-2 that were made around the
periphery of Lake Erie tend to support the prediction that ozone
concentrations are higher over the water than inland. In this particular
area the model predicts a peak value of about 140 ppb along the Pennsylvania-
New York shoreline of Lake Erie, and in fact a measured peak of 122 ppb
was reported in that area. (One should keep in mind when comparing
predicted and observed concentrations that predicted extreme values will
generally exceed the corresponding observed value because the latter are
available at only a limited number of sites.) Measured ozone concentrations
in and around the New York City area tend to support the model predictions
in this area as well. A peak value of 184 ppb was reported in southwestern
Long Island, an area where the predicted maximum was 187 ppb. A concentra-
tion of 230 ppb was predicted just offshore. Peak ozone concentrations
of 130 to 175 ppb were measured in southeastern and central Connecticut
where the predicted values ranged from 140 to 187 ppb. There are not
enough ozone data around Lake Huron and Lake Ontario to assess the accuracy
of the high concentrations predicted over those areas.
There is general disagreement between the observed data plotted in
Figure 8-2 (page 105) and the model predictions (Figure 8-1) in four regions:
the Pittsburgh vicinity, the Washington-Baltimore corridor, the Toronto
area, and the northern portions of Vermont and New Hampshire. In the
first of these areas, measurements show ozone levels as high as 120 to
130 ppb within a radius of about 40 km of Pittsburgh but the highest
computed values in this area are of the order of 90 ppb. Underprediction
is also evident in the Washington-Baltimore area. The ozone measurements
show a narrow zone in which concentrations range between 120 and 165 ppb,
88
-------
but no similar area appears in the simulated ozone values. It is possible
that the underprediction in both this and the Pittsburgh areas is due to
errors in the emissions data that we noted earlier. Errors in the
meteorological inputs are another possible cause of the discrepancies,
particularly errors in the cloud cover values. Since the concentrations
shown in Figure 8-2 are observed maximum ozone values over a 9-day period,
failure of the model to have accurate cloud cover data at the proper
location and hour can cause the model to grossly underestimate the ozone
that would be generated in a mixture of VOC and NOX emissions. At this
point we cannot say why the model underpredicts ozone in the two areas
that we have just cited.
The opposite problem occurs in the Toronto area where peak ozone
concentrations of 200 ppb are predicted but values no larger than 100 ppb
are observed. Since our simulations have consistently shown high ozone
levels in the Toronto plume, we suspect problems with the emissions data.
In northern Vermont and New Hampshire we find on comparing Figures 8-1
and 8-2 that the model overpredicts peak ozone values by a considerable
margin — 100 ppb observed vs 140 ppb predicted. In this case we are
quite certain that the problem is due to the constant thickness of the
layers in the first-generation model. The top surface of the model in
the simulations presented here had an elevation of 2000 meters above
ground. However, at the time the simulated pollutant cloud passed through
the area in question, the meteorological data show that widespread shower
activity was present which could have transported ozone and other pollutants
to altitudes of 5000 m or higher. The second-generation ROM will eliminate
this source of error.
89
-------
In summary, the ozone observations suggest that the model overpredicts
ozone concentrations in areas remote from the major precursor pollutant
source areas. We also found that it underestimated concentrations near
the Pittsburgh, Washington, DC, and Baltimore urban areas, but overpredicted
ozone by a wide margin around Toronto.
90
-------
RESULTS OF THE BASE EMISSIONS SIMULATION
Since the primary goal of this study is to examine the effects that
projected 1987 emissions are likely to have on ozone concentrations, we
will focus our analysis of the model results exclusively on ozone and on
the differences between the predicted concentrations in the base and
control cases. Analyses of a broader scope will be performed in later
studies, primarily in connection with the second generation ROM.
We will present the results of the ozone simulations in four separate
formats:
(1) Averages over each of the 439 counties in the model domain
(Appendix D);
(2) Isopleths of concentration over the entire model area at selected
times during the 9-day simulation;
(3) Statistics of county averages within each of the four receptor
county categories defined in Section 5, (pages 51 and 52, see also
Figures 5-2, 3, 4, 5, pages 54-57);
(4) Statistics of cell averaged concentrations within each of the
receptor cell classes defined in Section 5, (page 50, see also
Figure 5-1, page 53).
We also consider a variety of averaging times: 1-hour, 3-hour, 7- hour
(the so-called daily daylight average, from 0900 - 1600 LSI), 12-hour and
24-hour averages. The 1-hour averaged ozone concentration is regulated by
the current primary National Ambient Air Quality Standard (NAAQS); and the
7-hour daily daylight average is presently being considered as a basis for
specifying a new secondary standard which would protect agriculture, forests
91
-------
and public welfare in general.
The predicted county averaged ozone concentrations for the base
emissions are tabulated in Appendix D for each of the 439 counties in the
model domain. Results are given for both the 7-hour and 1-hour averaged
concentrations. The corresponding values obtained using the controlled
emissions are given in the same table for comparison. Examples are also
given at the end of Appendix D of the type of model results that are
available for other averaging times. The entire set of results is too
lengthy to include in the report. The data contained in Appendix D are
provided for use by others in making cost-benefit analyses of the emissions
controls described in Section 7, and we will not discuss them further in
this report.
The simulated distributions of ozone over the model domain at hour
1600 EST of each day of the nine day scenario are shown in Figures 8-3
through 8-11 (pages 106-114). The concentration patterns seen in these
plots give an idea of the spatial size of the simulated urban plumes, the
areas influenced by ozone produced from the emissions of given sources,
and the temporal behavior of the ozone distribution that accompanies
changes in meteorology. One should keep in mind when examining these
data that unlike inert chemical species whose spatial distributions are
determined chiefly by transport, ozone is created and destroyed by chemical
reactions which may cause concentration patterns to change in seemingly
erratic ways. For example, a "cloud" of ozone may vanish upon crossing a
strong area source of NOX and then reappear downwind.
92
-------
Figure 8-3 (page 106) snows the simulated ozone distribution in
Layer 1 at hour 1600 EST of the first day of the scenario, i.e., 23 July
1980 (Julian date 205). (The model was initialized at hour 00 of this
day.) Concentration maxima occur over southern Lake Ontario — the result
of Toronto's VOC and NOX sources; to the southeast of Detroit; and over
the Atlantic offshore of Long Island and Massachusetts. The next figure
in the sequence shows that by 1600 of the next day, the Toronto emissions
are influencing ozone levels as far south as the New York-Pennsylvania
border. At this point concentrations are just over 60 ppb, as compared
with the background level of 37 ppb. The highest concentrations in what
appears to be the Toronto plume are between 120 and 140 ppb just east of
Buffalo. At this same hour we see that the previous day's ozone from
Detroit has merged with ozone generated from Ohio's source emissions to
elevate concentrations over northern and eastern Ohio. By the third day
of the period, shown in Figure 8-5, the areas of elevated ozone associated
with both the Toronto and Detroit emissions have reversed direction, and
moved northward. Over the next two days these areas of high ozone continue
their northward motion and by day 210, the sixth day of the simulation,
they pass through the northern boundary of the model. The trajectories
shown earlier in Figure 6-3 (page 71) are consistent with the motions of
the Detroit and Toronto ozone plumes seen in Figures 8-3 through 8-8. A
significant aspect of the simulation is that during the 9-day scenario,
ozone generated in the Detroit, Toronto and northern Ohio emissions stays
west of the Appalachians. It affects a zone stretching from the Adirondack
Mountains through western New York into northern Ohio, as well as nearly
all of the portion of Ontario that lies within the model area. The
boundaries of the impact area and the maximum concentration levels
93
-------
produced are readily apparent in Figure 8-1 (page 104).
During the first four days of the simulation, the bulk of the ozone
generated in the Eastern Corridor emissions remains just offshore of the
East Coast, but edges of the polluted air mass extend inland into all the
states bordering the Atlantic from Massachusetts to Delaware. By day
210, July 28, a strong southerly flow has developed over New England in
advance of an approaching cold front (see Figure A-6, page 133) and the
entire ozone laden air mass has been swept northward across Connecticut
and Massachusetts, into northern New York, Vermont, New Hampshire and
Maine. This is strikingly evident in Figure 8-8 (page 111). It is during
this period, as we noted earlier, that the model overpredicts ozone con-
centrations by a factor of two or more due to its failure to account for
the deep vertical mixing and precipitation that occurred in the southerly
flow. This problem will be corrected in the second-generation model.
The area influenced by the Eastern Corridor sources and the associated
maximum ozone concentrations can be seen in Figure 8-1. (When examining
this figure, keep in mind the tendency of the model to overpredict
concentrations far downwind of the major VOC and NOX sources. Remember
also that the isopleths are of maximum ozone concentration during the 9
days, not the mean concentration.) The concentration isopleths shown in
the figure outline two broad plumes, one eminating from the Eastern
Corridor sources and the other from the Detroit/Ohio area. Both plumes
appear to be oriented in roughly a southwest-northeast direction with the
Detroit/Ohio plume overrunning the Toronto plume. The Eastern Corridor
plume and the combined Detroit-Toronto plumes intersect in northeastern
New York. This indicates that on some occasions during the 9-day scenario,
94
-------
ozone in the Adirondack Mountains, for example, originated in the emissions
of the major sources around the Great Lakes and on other occasions it
arose from emissions of the major sources along the Eastern Seaboard.
It is not clear whether the broad features of these two major plumes are
indicative of the patterns one would find from longer term simulations.
We suspect that somewhat different plume shapes would be found from, say,
a 30-day simulation because as we discovered earlier in Section 6, the
meteorological processes associated with extreme ozone levels appear to
have periods longer than 9 days. We also anticipate changes in the ozone
plume characteristics when biogenic hydrocarbon emissions are added to
the source inventory. A likely result of this addition will be enhanced
ozone concentrations in the Appalachian regions of West Virginia, Maryland
and Pennsylvania due to reactions of NOX from major sources in Ohio and
West Virginia (see Figure 4-1, page 43) with hydrocarbons from the mountain
forests. In any event we expect to find relatively well defined influence
zones which should provide valuable information in the design of emissions
control plans aimed at lowering concentrations in specific areas.
Information useful in performing cost/benefit studies of emissions
controls is contained in the statistics of the concentrations in each of
the four receptor county classes defined in Section 5 (see pages 51-52
and Figures 5-2 through 5-5, pages 54-57). Here we will examine only the
predicted daily daylight (7-hour) averaged ozone values, leaving discussions
of the hourly averaged values until later when we analyze the model output
in the context of the receptor cells. The statistics of the simulated 7-
hour averaged ozone concentrations are presented in Figure 8-12 (page 115)
for each of the four county classes. A guide to the symbols used in the
95
-------
figure is given below:
maximum value in the set
75 percentile concentration
--median value of the set
25 percentile concentration
minimum value
There are different numbers of counties in each of the four groups. The
Major Source Counties contain the least number, 70; and the Natural Counties
contain the most, 169; and the Suburban and Agricultural Counties include
104 and 96, respectively.
The quartile plots in Figure 8-12 reveal no striking, systematic
differences in the simulated concentrations in each of the four county
groups. This is perhaps due in part to the definitions of the groups
which emphasize demographic and land use characteristics of counties
rather than their proximity to major source areas. An important feature
of the model output that we cited earlier in this section is the prolonged
influence of the initial concentrations on the model predictions. On the
first day of the simulation, i.e., 80205, the frequency distribution of
ozone concentrations is virtually the same in all four county classes.
It appears that it is not until the third day of the model run that
significant differences begin to appear in the concentration statistics
in each area. The data show that the extreme and median ozone values in
natural areas can be as high as those in the major source and SMSA counties.
This is probably due in part to model overprediction in remote areas and
in part to the close proximity of some of the counties classified as
"natural" to major sources. In the next subsection we will compare the
96
-------
base case quartile plots shown in Figure 8-12 with the corresponding
statistics obtained using the controlled emissions.
Better insight into the nature of the processes of ozone formation
and transport can be derived from the quartile plots of the simulated
hourly concentrations in each of the four receptor cell classes defined
in Section 5. Recall that there are 70 cells in each of the four categories
and that three of the groups, namely the Suburban, Rural and Wilderness
cells, are defined to be within given distance ranges of the Urban cells —
which represent the major VOC and NOX sources (see page 50). The quartiles
of the simulated base case ozone concentrations are plotted in Figure 8-13
(page 116) for each hour of the 9-day scenario for each of the four cell
classes. The format is the same as that used in Figure 8-12 (described
on page 96), but it appears different because the vertical boxes and
lines are so close together that they give the appearance of dark and
light shading.
On comparing the data presented in Figure 8-13 for each of the four
receptor classes, we see again that the initial concentration field
dominates the model predictions during the first three to four days of
the simulation. This implies that a regional scale (1000 km) model must
be run for at least four days before the predicted concentration levels
in areas remote from the major sources are free of artifacts of the initial
field.
We see also in Figure 8-13 a strong diurnal modulation of the 25,
50, and 75 percentile concentrations superposed on longer scale variations
that are apparently associated with changes in meteorology. Based on
97
-------
the chemistry of ozone production that we described earlier, we interpret
the diurnal variations to be evidence of ozone formation in these areas.
In the Rural and Wilderness areas, which are progressively farther from the
major VOC and NOX sources, the magnitude of the diurnal component of the
ozone variations is progressively smaller. Let us emphasize that the 0
and 100 percentile concentrations, i.e., the minimum and maximum values,
respectively, may reflect conditions at only one of the 70 cells in a
given group, whereas the 25, 50 and 75 percentile points represent more
sites and hence are more representative of conditions as a whole.
We see in Figure 8-13 (page 116) that the 75 percentile concentration
in the Urban cells peaks on the fourth day of the scenario, i.e., day
80208. It peaks the next day, 80209, in the Rural cells; and it is a
maximum still later, day 80210, in the Wilderness sites. The ozone
concentrations measured at the SAROAD sites during the scenario peaked on
the first of these days (see Figure 6-1, page 67). Thus, the predicted
concentrations are consistent with the observations inasmuch as a majority
of the monitoring stations are in the major source areas represented by
the Urban cells. The phase lag between high ozone levels in the urban
and remote areas is particularly evident in Figure 8-13 at the 150 hour
point (day 80211) where the median concentrations in the Urban and Suburban
areas are near their minimum values while the Wilderness concentrations
are near their peak.
The presence of strong NOX sources in at least some of the Urban
cells is evident in the fact that during most nighttime periods in the
9-day scenario, ozone concentrations in about 25 percent of these cells
drop to values near zero. By contrast, the 25 percentile concentrations
98
-------
at Wilderness sites seldom drop below about 30 ppb. Note also that the
extreme concentrations predicted in the Wilderness areas vary in a smoother
manner with time than in the Urban or Suburban areas, suggesting, as one
would expect, that ozone plumes broaden with distance from their source.
The highest ozone concentration predicted at any of the 70 Wilderness
sites is about 150 ppb compared to a peak of just over 200 ppb in each of
the other three receptor areas.
In summary, we see evidence in the simulated concentration statistics
presented in Figure 8-13 (page 116) that the bulk of the ozone generated
by VOC and NOX emissions is produced within about 100 km of the areas
that we have identified as major source areas. This is suggested by the
fact that the magnitude and time variation of the extreme concentrations,
and the amplitude of the diurnal modulation of the median concentrations
are all larger in the Urban, Suburban and Rural cell groups than in the
Wilderness sites. Recall that the last group are, by definition, farther
than 100 km from the nearest major source area. In a future report we
plan to compute from the model output the quantity of ozone generated in
each model cell at each time step. From this information we can map the
virtual ozone source distribution, which will show where and in what
quantities ozone is produced from VOC and NOX emissions. We believe that
this information will be especially valuable in assessing the impact of
various emissions controls on ozone formation.
99
-------
SIMULATED EFFECTS OF EMISSIONS CONTROLS ON OZONE CONCENTRATIONS
IN THE NORTHEAST
The model results from the 9-day scenario that we just discussed
used the 1980 NAPAP version 4.2 emissions inventory which we discussed in
some detail in Section 4. These data are what we have referred to as the
base case emissions. We ran the model a second time using identical
values for all input parameters except the emissions, which were replaced
by the projected 1987 inventory described in Section 7 (see also Appendix
B). These are the emissions data that we have referred to as the controlled
emissions. Here we will compare the ozone concentrations produced by
this simulation with those obtained from the base case run described
earlier. Detailed, county-by-county comparisons of the base and control
ozone values are provided in Appendix D. We will limit our discussion of
the results in this section to qualitative assessments of the general
effects of the emissions controls.
We found that the ozone concentrations from the controlled emissions
simulation were everywhere and at all times less than or equal to the
corresponding values predicted in the base case simulation. We add,
however, that biogenic hydrocarbon emissions were not treated in the
model and their presence could alter our findings, especially in the
western portion of the model region where NOX emissions of some sources
were increased in the "controlled" inventory. Due to this and other
deficiencies in the first generation ROM which we discussed earlier, one
should view the results of the emissions control simulations that we
report here as preliminary.
100
-------
Let us look first at che effect of the controls on the 7-hour, daily
daylight averaged ozone. Figure 8-14 (page 117) compares the extreme and
median statistics obtained in the base case and control simulations in
each of the four county receptor groups. An explanation of the graphical
notation used in the figure is given below.
Largest ozone concentration in the receptor cell
set in the base case simulation
Largest ozone concentration in the same set in
the controlled emissions simulation
Median value in the base simulation (solid curve)
Median value in the control simulation (dashed)
To avoid cluttering the graph, we compare only the extreme and median
values of the 7-hour average. One effect of the emissions changes that
is readily apparent is that the extreme values of the 7-hour averaged
concentration have been reduced in general by much larger percentages
than the median values. For example, on day 80207, the maximum daily
daylight averaged ozone in SMSA (suburban) counties was about 135 ppb in
the base case simulation but only 80 ppb with the controlled emissions —
a reduction of 40%. At the same time the median value fell by only 15%.
The degree of reduction of the maximum value varies widely from day to
day in a single receptor group and from receptor to receptor on the same
day. Note, too, that the controlled emissions lowered the median 7-hour
averages most in the major source counties and by progressively less
degrees in the SMSA, Agricultural, and Natural counties. A final interesting
effect of the emissions changes is alterations in the time of occurrence
of the maximum concentration. For example, in the SMSA and Natural
101
-------
counties, the largest values of the 7-hour average occurred on day 80209
in the base case simulation, but one day later with the controlled emissions,
These and other effects of the emissions changes are more clearly
evident in the comparisons of the simulated hourly averaged concentrations
in each of the four receptor cell categories shown in Figure 8-15 (page 118),
The graphical representation used here is the same as that described
above. Looking first at the beginning of the 9-day scenario we note that
it is not until the second day of the simulation that significant
differences in the median concentrations begin to appear in the Urban and
Suburban areas, and it is only after four days that differences develop
in the Wilderness cells. This result re-emphasizes the point we made
earlier, that regional scale models must be run for periods of at least a
week to produce meaningful results.
A curious feature of the early development of the concentration
fields in both the base and controlled emissions simulations is the near
equality of the extreme values in the Urban and Suburban areas during
the first two days. It may be that during this time the highest concen-
trations occur near sources whose base and controlled emissions are nearly
the same. Notice that extreme concentrations in the Wilderness sites are
unaffected by the emissions changes until the fifth day of the simulation.
This may be caused by the influence of the Toronto plume which, as we
discussed earlier, affects concentrations as far south as Northern
Pennsylvania. We assumed that Canadian emissions were the same in the base
and controlled inventories. Therefore, the Toronto ozone plume will be
identical in the two simulations up until the time that it becomes mixed
102
-------
with significant amounts of pollutants from U.S. sources.
The data presented in Figure 8-15 show rather clearly that in the
Urban, Suburban and Rural areas, the emissions changes have the largest
effects on the extreme and median concentrations during the development
of the episode that peaks about 110 hours into the simulation (day 80209).
Beyond that point the differences between the base and control simulations
decline reaching a minimum at about hour 150. In fact, at that point the
median concentrations in Urban and Suburban cells are the same for both
emissions inventories. As before, the response of concentrations in remote
areas to the emissions changes is radically different. No effect is
evident here until after 110 hours, apparently because the ozone that
built up in the urban and suburban areas does not arrive in the Wilderness
areas until this time. When it does arrive, the reductions in the peak
and median concentrations are smaller than those that were seen in the
Urban and Suburban areas as the episode developed. The hourly data shown
in Figure 8-15 also reveal the phenomenon that we saw earlier in Figure 8-14
(page 117) in connection with the 7-hour averaged concentrations, namely
that the time of occurrence of the extreme concentration may be changed
by changes in the emissions. This is dramatically evident in the Rural
data of Figure 8-15 where we see that the peak concentration occurs near
hour 110 with the base emissions but three days later, near hour 180,
with the controlled sources. This suggests that the meteorological
conditions that favor the highest ozone concentrations are different
depending on the relationship of VOC to NOX emissions within a spatially
fixed set of sources.
In the next, and last, section we summarize our findings.
103
-------
•. 3 5 2
Figure 8-1. Maximum ozone concentration during the 9-day scenario
simulated using the base emissions inventory. Isopleths
are drawn at intervals of 40 ppb. Numbers beside isopleths
are ppb/100.
104
-------
Figure 8-2. Maximum hourly averaged ozone concentration observed at SARGAD
monitoring sites during the scenario period 23 - 31 July 1980.
105
-------
o
Tl
X
Q.
Q.
cn
in
en o
a cu
en
LU
>—» ..
CJ LU
iU H-
Q. •<
cn o
LU
Figure 8-3. Simulated ozone in Layer 1 at hour 1600 EST on the first
day of the 9-ctey scenario, using the base emissions inventory.
106
-------
SK
SK
o
0_0
0.0
CO
m o
o ru
o
. . GO
cn
LU
I—» . .
O LU
UJ h-
Q. •<
cn Q
UJ
LU
Figure 8-4. Same as 8-3 except hour 1600 of day 2 of the scenario.
107
-------
o
•»•»
X
Q.
in
CO
• •
UJ
r*.
en o
o oj
o
. . CD
cn
UJ
»-* ..
O LU
UJ 1—
Q. -<
cn a
UJ
Figure 8-5. Same as 8-3 except hour 1600 of day 3 of the scenario.
108
-------
o
•rl
X
i§
CO
.. *H
en
t— ..
I—i UJ
er» o
o ru
«s
UJ
•—» ..
oil!
UJ
Figure 8-6. Same as 8-3 except hour 1600 of day 4 of the scenario.
109
-------
o
^-«
X
sS
a. o
C^B G^
GO
C^7
K— ..
1—4 UJ
O3
en o
a cu
en
UJ
H-l ..
CJ UJ
^!5
en a
13
UJ
Figure 8-7. Same as 8-3 except hour 1600 of day 5 of the scenario.
110
-------
SK
o
0:0
a. o
CO
.. •»•«
52..
I—I LU
O
CO "-I
CD ru
o
.. CD
cn
LU
t—t ,.
O LU
LU t—
c/5 a
Figure 8-8. Same as 8-3 except hour 1600 of day 6 of the scenario.
Ill
-------
o
•«
X
o
o
CO
en'
LU
o cvt
o
. . 03
cn
LU
i—i ..
O LU
UJ
LU
Figure 8-9. Same as 8-3 except hour 1600 of day 7 of the scenario.
112
-------
o
xt
X
o
rs: o
Q_ O
LU
ru
en •»•«
o.cu
o
. . 03
C/3
LU
i—« . .
O LU
UU H-
Q. •<
cn a
iq .5
• ^ ^WK.
Figure 8-10. Same as 8-3 except hour 1600 of day 8 of the scenario.
113
-------
o
•T*
X
65
t— ..
t-4 UJ
cn «M
o ru
UJ
*-* ..
O UJ
UJ H-
o. •<
UJ
Figure 8-11. Same as 8-3 except hour 1600 of day 9 of the scenario.
114
-------
0.20
0
z
o a.15 •
M
E
P 0.10
P
o.oo-L.
MAJOR SOURCE COUNTY
°"2°t NON-MAJOR SUURCE SMSA COUNTY
0 \
i
a o. is 4
N 1
E
j
P 0.101
p
M
0.05-
I
ii i
1 3 1
1
1
o.oo4
i
i
I
1
i
1
,'
i"
t
.
1
!
1 AGRICULTURAL COUNTY
0 i
a o. 15 1
N
c
P 0.10 •
p 1
"
}
* J_
0.05-
0.00-
a
z
0 0.15 J
E
P 0.10-
p
M
0.05-
0.00 •
I
1 ! t
!
T
j
j I
*i
! j
;
NATURAL COUNTY
i
t 1 I
1
4|
[
30209 90206 30207 30209 30209 30210 30211 30212 30213
Figure 8-12. Quartile plots of simulated base case daily daylight averaged
(0900-1600 LSI) ozone concentrations in each of the four
county receptor classes defined in Section 5. See page 96
for explanation of symbols.
115
-------
250
200-
158-
100-
50-
0-
t I I t f
WILDERNESS CELLS
-0123*
1 0900
0
» 7 t » I 1 t t I
000001254
o o o a a
HOURS FROM START
t I
3 i
a g
i i i
789
ooo
222
a i 2
oao
Figure 8-13.
Quartile plots of simulated base case hourly ozone concentration
in each of the four receptor cell classes defined in Section 5.
See page 96 for an explanation of the graphs.
116
-------
0.10
0.00
0.20
0.10
MAJOR SOURCE COUNTIES
0.30 {
NON-MAJOR SOURCt SMSA COUNTIES
o.is 4
0.10 -j
1
NATURAL COUNTIES
0.00
50205 8020J 30207 30208
90209
DATE
30210 30211 30212 30213
Figure 8-14.
Simulated effects of emissions controls on daily daylight
(0900-1600 LSI) averaged ozone concentrations in each of the
four county receptor classes. See page 101 for an explanation
of the symbols.
117
-------
250 -j
I
2001
URBAN CELLS
150-i
I
100-
501
250-j
200
i i i t i
i i
I i 1 I [ I 1 I I I I 1 I I j
I 1 1 I 1
I I I I I t I I I 1 1 1 I
i i i i r T i i i i i i i r
i i i i i r
Figure 8-15.
-01234567891111111111222
1 0000000000123456739012
0 0000000000000
TIME
Simulated effect of emission controls on hourly averaged ozone
concentrations in each of the four receptor cell classes.
See page 101 for an explanation of the symbols.
118
-------
SECTION 9
CONCLUSIONS
Following is a summary of the main points that we made in each of
the previous sections.
•
o The first-generation Regional Oxidant Model (ROM1) used in
this study is weak in four basic areas: (1) the chemistry
scheme employed is not considered to be the best description
currently available, (2) the emissions inventory does not
include biogenic hydrocarbons, (3) the model's 3 and one-half
layers have constant thicknesses throughout the 9-day
simulation, and (4) the wind fields are non-divergent.
(Problems may also exist in other aspects of the model, such
as the technique used to treat major point sources.) These
four weaknesses will be eliminated in the second-generation
model (ROM2) which we hope to have operational in mid-1986.
(See Table 2-2, page 10.)
o Based on the 1980 NAPAP version 4.2 emissions inventory, we
identified 70 counties in the Northeastern U.S. as major
sources of VOC and NOX (see Tables 4-1, 2, pages 41 and 42).
These counties have the highest emissions densities (moles
per area per day) of VOC and NOX of all counties in the region,
and together they produce about one-half of all VOC and NOX
emissions. The areas of highest measured ozone concentrations
are closely associated with the locations of these 70 counties
119
-------
(compare i-igure 4-1, page 43, with Figure 3-1, page 20).
o The characteristic spatial scale of VOC and NOX sources in
the Northeast is estimated to be a few tens of kilometers
(see Figure 4-2, page 44). This means that if a model with
a grid size much larger than this, say A > 50 km, is used
to simulate photochemical air pollution in this region,
significant systematic errors can occur not only in the pre-
dicted peak concentrations of secondary species, such as
ozone, but also in the predicted response of the concen-
trations of these species to changes in VOC and NOX emissions.
The latter point is of critical importance in the use of
models in regulatory studies and requires further detailed
analysis.
o On comparing the 1980 NAPAP version 4.2 emissions inventory
with the earlier 1979 NECRMP inventory, we found differences
of 300 percent and more in the gridded (~18xl8 km) VOC
and NOX emission rates (see Figures 4-3, 4, pages 45, 46).
This represents a level of uncertainty in the base emissions
that is roughly ten times the magnitude of the changes in
emissions that are contemplated in present control strategies.
The origin of this uncertainty is being investigated.
o Air quality models that treat regional scale and larger areas,
i.e., domains > 1000 km in extent, can be operated in either
of two modes, which we called the probabilistic mode and the
quasi-deterministic mode. In the former the model predicts
120
-------
the probabilities, expectations and other statistical prop-
erties of concentrations at specific sites at specific
times. In the quasi-deterministic mode, the model provides
statistics of concentrations at given times or integrated
over given periods within given receptor classes rather than
at specific sites. In the present study the model is run
only in the quasi-deterministic mode, yielding information
on the concentrations of 23 different chemical species in
four receptor classes -- Urban, Suburban, Rural, and
Wilderness — defined on page 50.
o We have tentatively proposed two criteria for selecting
historical meteorological data for use in regional scale
modeling studies of photochemical oxidant: (1) the meteoro-
logical scenario should begin on a day when the median
value of the maximum hourly ozone concentrations observed
at all measuring sites in the model domain is near the
seasonal minimum value; (2) the scenario should be long
enough that the frequency distribution of measured hourly
ozone values during the scenario period approximates the
corresponding seasonal distribution closely enough to give
the results of the model simulations broad applicability. We
showed that the scenario must be more than about 5 days long,
to minimize the effects of the initialization procedure on
predicted concentrations, and that a 9-day scenario is not
long enough to model the processes that control concentrations
above the 90-th percent!le level at any site (see Figure 6-2a,
b, c; pages 68-70).
121
-------
o We performed 9-day simulations comparing ozone concentrations
produced by 1980 emissions (base case) with those produced
by projected 1987 emissions (control case). (See Sections 4
and 7 for discussions of the emissions inventories.) The
simulations were performed using ROM1 which has the following
basic limitations: "outdated" chemical mechanism; biogenic
hydrocarbon emissions are ignored; model layers have constant
thicknesses; horizontal winds are nondivergent. On comparing
the results of the base case run with measured ozone data
we found that the model underpredicted the scenario maximum
ozone concentration in the Pittsburgh area — 90 ppb predicted
vs. 120 ppb observed — and in the Washington-Baltimore area —
90 predicted vs. 150 observed. It overpredicted near Toronto
~ 200 vs. 90 -- and in northern New England — 130 vs 100.
Better agreement was found in the Detroit and New York areas
(compare Figures 8-1 and 2, pages 104 and 105). The most
likely causes of the errors near the urban areas are believed
to be errors in the emissions data or cloud cover or both.
o Comparing the predicted ozone concentrations for the base case
with corresponding values in the control case we found the
following:
(1) In general, at any given location and hour the concentration
in the control case is less than or equal to that in the
base case;
(2) Within each receptor group, peak concentrations are reduced
by larger percentages than the median values are reduced.
122
-------
Reductions of 0 to 50% occur in the peak compared to
to 0 to 25% reductions in the median (see Figure 8-15,
page 118). This is true of all concentration averaging
times.
(3) Ozone is reduced more at sites near the major VOC and
NOX sources than at locations far away. For example,
median ozone levels at Suburban locations (defined to
be within 50 km of major source centers) were reduced
up to 25% whereas in Wilderness areas (greater than
100 km from major sources) reductions were less than 15%
(see Figure 8-15).
(4) The maximum ozone concentrations in Rural and Wilderness
areas occurred on different days in the control case
than in the base case, even though meteorological conditions
were identical in both simulations. This suggests that
the source-receptor relationship between VOC/NOX sources
and remote sites is strongly nonlinear.
(5) Overall, the 1987 emissions reductions appear to have two
basic effects on ozone: they cause a delay in ozone
formation, and they reduce the total quantity produced.
The former effect is evident in Figure 9-1 (next page)
which shows the counties in which the simulated hourly
ozone level exceeded 120 ppb in the base and control
cases. The figure shows that reduction of the peak value
to levels below 120 ppb occurred primarily in the upwind
counties only.
123
-------
BASE AND CONTROL, MAXIMUM 03 > 120 PPB
RESPONSE
BASE ONLY
BASE. CONTROL
Figure 9-1. Counties in which simulated one-hour averaged ozone concen-
trations exceeded 120 ppb during the 9-day scenario. Light
shading denotes counties in which exceedance occurred only
in the base case simulation. Dark shading indicates areas
where exceedance occurred in both the base and control runs.
124
-------
References
1. Alkezweeny, A., K. M. Busness, R. C. Easter and J. S. Wetzel,
(1981): "Northeast Corridor Regional Modeling Project: Aircraft
Measurements — New York and Vicinity. EPA-450/4-81-012. 229 pages.
2. All wine, K. J. and H. H. Westberg, (1977): "Potential Impact of
Coal-Fired Power Plants on Ground-Level Ozone Concentrations in
Wisconsin", Final Report to Wisconsin Public Service Commission,
Madison, Wisconsin.
3. Clark, T. L. and J. F. Clarke, (1984): "A Lagrangian Study of the
Boundary Layer Transport of Pollutants in the Northeastern United
States", Atmos. Envir., Vol. 18, pp 287-297.
4. Demerjian, K. L., and K. L. Schere, (1979): "Applications of a Photo-
chemical Box Model for Ozone Air Quality in Houston, Texas. Proceedings,
Ozone/Oxidants: Interactions with the Total Environment II, Houston, TX,
14-17 Oct 1979, APCA, Pittsburgh, Pa., pp. 329-352.
5. Demerjian, K. L., K. L. Schere, and J. T. Peterson, (1980): "Theoretical
estimates of actinic (spherically integrated) flux and photolytic rate
constants of atmospheric species in the lower troposphere." In Advances
in Environmental Science and Technology - Vol. 10, J. N. Pitts et al.,
eds., John Wiley and Sons, New York, pp. 369-459.
6. Lamb, R. G., (1977): "A Case Study of Stratospheric Ozone Affecting
Ground-Level Oxidant Concentrations" J. Appl. Meteor., Vol. 16, 780-794.
7. Lamb, R. G., (1986): "The Effects of Model Grid Size on Predicted
Regional Scale Ozone Concentrations". In preparation.
8. Lamb, R. G., and S. K. Hati, (1986): "The Representation of Atmospheric
Motion in Models of Regional Scale Air Pollution", submitted to
J. Climate and Appl. Meteor.
9. Lamb, R. G., (1983a): "A Regional Scale (1000 km) Model of Photochemical
Air Pollution: Part 1. Theoretical Formulation". EPA-600/3-83-035.
237 pages.
10. Lamb, R. G., (1983b): "Theoretical Issues in Long Range Transport Modeling".
Extended Abstract Volume. Sixth Symposium on Turbulence and Diffusion.
Boston, Mass. Published by the American Meteorological Society.
11. Lamb, R. G., (1984): "A Regional Scale (1000 km) Model of Photochemical
Air Pollution: Part 2. Input Processor Network Design". EPA-600/3-84-085.
310 pages.
12. Lamb, R. G., and G. F. Laniak (1985): "A Regional Scale (1000 km) Model
of Photochemical Air Pollution: Part 3. Tests of the Numerical Algorithms".
EPA/600/3-85-037. 289 pages.
125
-------
13. Lyons, W. A., and H. S. Cole, (1976): "Photochemical Oxidant Transport:
Mesoscale Lake Breeze and Synoptic-Scale Aspects. J. Appl. Meteor.,
Vol. 15, 733-743.
14. Schere, K. L., (1986): "The EPA Regional Oxidant Model: ROM1 Evaluation
for 3-4 August 1979" EPA Report.
15. Siple, G. W., G. K. Fitzsimmons, J. J. Van Ee and K. F. Zeller,
(1977): "Air Quality Data for the Northeast Oxidant Transport
Study, 1975", EPA-600/4-77-020, 92 pages.
16. Toothman, D. A., J. C. Yates and E. J. Sabo, (1984): "Status Report on
the Development of the NAPAP Emission Inventory for the 1980 Base Year
and Summary of Preliminary Data. EPA/600/7-84-01.
17. Viezee, W., W. B. Johnson and H. B. Singh, (1983): "Stratospheric Ozone
in the Lower Troposphere. II. Assessment of Downward Flux and Ground-
Level Impact. Atmos. Environ.. Vol. 17, 1979-1993.
18. Westberg, H., (1985): "Ozone Behavior in the Combined Baltimore-
Washington Plume". EPA Report.
126
-------
Appendix A
Daily weather maps for the scenario period 23-31 July 198U
127
-------
WEDNESDAY. JULY a iseo
128
-------
THURSDAY, JULY Zt. 19SO
-,>S-V74£j£_ r—?«—ss^.-t --23c ^ vr^«s' ^.'S/.-T^Jig .W
•• •^^•-t^r.^ /^£~/W '^'^^ ^'^lUv *&
^^^ff&K^^^$^ff\^ /
\ **•* '< _m to>2 !* ~"J" ~' -«—n?,MM1 ,-»l.is -i •r^oir'-el 'f*®*'*/ 1 '*^5^ ^ >-LJ-"*/^ ^'f
\ » i *>v _ / i ^C*"! I » -A',» : _-i"vi_\ ,rr,v '••s
C> x^t:j .-•=3^—- •»&( -^ •/»
.
SURFACE 'WEATHER MAP
AND STATION WEATHER -
AT 7 00 * M £ST
/ '^^4/
500-MH.UBA* HEIGHT CONTOURS _
| IT 7 GO A M £ 3 T
129
-------
FRIDAY. JULY 25. 1980
~4~o* > _'>»•*-*• ;>, *-~L' &% ^?""\. •*.*-
^-=3*^--> -r & T- ^^^
^^j^, 3&^ -^^ic:
130
-------
SATURDAY. JULY 26. 1980
'\M *«~! J^'*»r- i®.--:/r"
'-«is> * '— -'i.**" -^s1 • ~*~ /*^'
''^iS'**^" I ""v$.^*".. -~j[*
-------
SUNDAY. JULY 27, 1980
500-MI
, AT ' 00 A M ES T
-.^
132
-------
MONDAY, JULY 28, 1980
->-,
133
-------
TUESDAY. JULY 29. 1980
^^:r^m^m
134
-------
WEDNESDAY. JULY 30.1980
•»' '-^ »«• -J.O>s» Ljaiiif' ^^ u» V
' 50Q.MILLJ8AR HEtGHT CONTOURS
I AT 7 OO A.M E S T
^,
. . <*
x$S*-
•^3,—T-S—'-ff-
_.-s*/
135
-------
THUHSDAY. JULY 31. 1980
gry^iir
500-MILUBAR HEIGHT CONTOURS _
AT 7-00 A.M . E.S.T i
ii- «• ••!• .ia- :i*
jff^^_ - |7 »T •* j; >*-- - -TC^r-- ^s •*• «*
'-* **« a«M^~S"*-*^" is*"9- *^ ""•
*% ^-sl_i_ii-3-i -" ^ -r»
•-*» - „"• - -X " .S'5S-^V,-54"
s.—.
tr.r^vj!
>, 5
->>
5
136
-------
Appendix B
The county-by-county control strategy emissions inventory expressed
as precentage change in hydrocarbon and NOX emissions from the 1979 base
emissions rates. Emissions changes represent 1987 projections based on
1982 SIPs.
Notes:
(1) Unspecified changes in hydrocarbon (HC) or nitrogen oxides (NO)
indicate that the default values apply, namely -32% for HC and
-8% for NOX.
(2) Six digit numbers before county names are the NEDS county ID's.
The first two of the six digits identify the state as follows:
07 Connecticut
08 Delaware
09 District of Columbia
18 Kentucky
20 Maine
21 Maryland
22 Massachusetts
23 Michigan
30 New Hampshire
31 New Jersey
33 New York
36 Ohio
39 Pennsylvania
41 Rhode Island
47 Vermont
48 Virginia
50 West Virginia
137
-------
«>D ATA,L NEROS*CNTYEMISFORM.
I.
2.
3.
4.
5.
6.
7.
3.
9.
10.
11.
12.
13.
14.
15.
16.
17.
13.
19.
20.
21.
22.
23.
2*.
25.
26.
27.
23.
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
40.
41.
42.
43.
44.
45.
46.
47.
43.
49.
50.
51.
52.
53.
54.
55.
> 06/24/85
070265
070425
070478
070565
070705
070725
071 155
071505
080060
030130
080240
090020
130180
130300
183340
160380
130620
180720
131060
131200
131540
132140
182230
182640
182720
182860
182900
183040
183520
133560
200027
200277
200445
200547
200595
200645
200885
200907
201065
201125
201183
201325
210040
210080
210120
210140
210280
210340
210360
210420
210440
210600
210740
210800
210920
11:05:00 (1)
FAIRFIELO CO
HARTFORD CO
LITCHFIELD CO
MIDDLESEX CO
NEW HAVEN CO
NEW LOiJDON CO
TOLLAKD CO
WIKDHAM CO
KENT CO
NEW CASTLE CO
SUSSEX CO
WASHINGTON
BATH CO
BQURECM CO
BOYD CO
BRACKEN CO
CARTER CO
CLARK CO
ELLIOTT CO
FLEMING CO
GREENUP CO
LAWRENCE CO
LEWIS CO
MASON CO
MENIFEE CO
MONTGOMERY CO
MORGAN CO
NICHOLAS CO
ROBERTSON CO
ROMAN CO
ANDROSCOG6IN CO
CUMBERLAND CO
FRANKLIN CO
KENNEBEC CO
KNOX CO
LINCOLN CO
OXFORD CO
PENCSSCOT CO
SAGAOAHOC CO
SOMERSET CO
WALDO CO
YORK CO
ALLEGAHY CO
ANNE ARUNDEL CO
BALTIMORE
BALTIMORE CO
CALVEHT CO
CAROLINE CO
CARROLL CO
CECIL CO
CHARLES CO
DORCHESTER CO
FREDERICK CO
GARRETT CO
HARFORD CO
HC -23.4
HC -30.3
HC -27.7
HC -28.1
HC -30.3
HC -28.7
HC -30.1
HC -23.0
HC
HC -40.2
HC
HC -39.7
HC
HC
HC
HC
HC
HC -38.1
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC -44.2
HC -44.2
HC -44.2
HC
HC
HC -44.2
HC
HC
HC
HC
HC
NO -12.6
NO -18.9
NO -21.3
NO -3.7
NO -13.9
NO -15.5
NO -22.7
NO -15.0
NO
NO -3.3
NO
NO -29.4
NO
NO
NO
NO
NO
NO
HO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO 0.5
NO 0.5
NO 0.5
NO
NO
NO 0.5
NO
NO
NO
NO
NO
HC -44.2
NO 0.5
138
-------
56.
57.
58.
59.
60.
61.
62.
63.
64.
65.
66.
67.
68.
69.
70.
71.
72.
73.
74.
75.
76.
77.
73.
79.
80.
81.
32.
33.
34.
85.
36.
87.
33.
89.
90.
91.
92.
93.
94.
95.
96.
97.
98.
99.
100.
lOlo
102.
103.
104.
105.
106.
107.
108.
109.
110.
111.
112.
210960
211000
211160
211300
211320
211400
211500
211600
211680
211740
211800
228001
223002
228003
228004
228005
221291
223006
228007
221793
228008
228009
221274
228010
228011
228012
222121
228013
228014
220369
230060
230220
230280
230400
231700
232340
232440
232880
232960
233020
233140
233600
233640
233900
234040
234100
234620
234780
234800
234860
235160
235280
235320
300020
300060
300080
300140
HOWARD CO
KENT CO
MONTGOMERY CO
PRINCE GEORGES CO
QUEEN ANNES CO
ST MARYS CO
SOMERSET CO
TALSOT CO
WASHINGTON CO
WICOMICO CO
WORCESTER CO
EARNSTABLE
BERKSHIRE
BRISTOL
DUXES
ESSEX
ESSEX
FRANKLIN
HAMPDEN
HAMPDEN
HAMPSHIRE
MIDDLESEX
MIDDLESEX
NANTUCKET
NORFOLK
PLYMOUTH
PLYMOUTH
SUFFOLK
WOR CHESTER
WCRCHESTER
ALCONA CO
ALPENA CO
ARENAC CO
BAY CO
GENESEE CO
HURON CO
IOSCO CO
LAPEER CO
LENAUEE CO
LIVINGSTON CO
MACOMB CO
MONROE CO
MONTMORENCY CO
OAKLAND CO
OGEMAM CO
OSCOOA CO
ST CLAIR CO
SAGINAU CO
SANILAC CO
SHIAWASSEE CO
TUSCOLA CO
WASHTENAW CO
WAYNE CO
BELKNAP CO
CARROLL CO
CHESHIRE CO
COOS CO
HC -44.2
HC
HC -34.9
HC -36.6
HC
HC
HC
HC
HC
HC
HC
HC -12.4
HC -9.0
HC -12.4
HC -12.4
HC -7.8
HC -7.3
HC -3.7
HC -8.7
HC -8.7
HC -3. 7
HC -7.8
HC -7.8
HC -12.4
HC -7.8
HC -12.4
HC -12.4
HC -7. 8
HC -20.0
HC -20.0
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC -34.7
HC
HC
HC -45.2
HC
HC
HC -41.3
HC
HC
HC
HC
HC
HC -38.3
HC
HC
HC
HC
NO 0.5
NO
NO -7.5
NO 0.5
NO
NO
NO
NO
NO
NO
NO
NO 6.6
NO 4.0
NO 6.6
NO 6.6
NO 8.5
NO 8.5
NO 5.1
NO 5.1
NO 5.1
NO 5.1
NO 3.5
NO 8.5
NO 6.6
NO 8.5
NO 6.6
NO 6.6
NO 3.5
NO 6.0
NO 6.0
HO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO -13.9
NO
NO
NO -17.1
NO
NO
NO 61.1
NO
NO
NO
NO
NO
NO -5.3
NO
NO
NO
NO
139
-------
113.
114.
us.
116.
117.
118.
119.
120.
121.
122.
123.
12*.
125.
126.
127.
123.
129.
130.
131.
132.
133.
134.
135.
136.
137.
133.
139.
140.
143.
144.
145.
146.
147.
148.
149.
150.
151.
152.
153.
154.
155.
156.
157.
153.
159.
160.
161.
162.
163.
164.
165.
166.
167.
163.
169.
300240 GRAFTON CO
300300 HILLSBOROUGH CO
300440 MERRIMACK CO
300530 ROCKINGHAM CO
300640 STRAFFCRD CO
300660 SULLIVAN CO
310080 ATLANTIC CO
310300 BERGEN CO
310660 BURLINGTON CO
310740 CAMOEN CO
310780 CAPE MAY CO
311050 CUMBERLAND CO
311380 ESSEX CO
311760 GLOUCESTER CO
312240 HUDSON CO
312260 HUffTERDON CO
312980 MERCER CO
313060 MIDDLESEX CO
313180 MONMOUTH CO
313260 MORRIS CO
313900 OCEAN CO
314120 PASSAIC CO
314900 SALEM CO
315020 SOMERSET CO
315300 SUSSEX CO
315440 UNION CO
315660 WARREN CO
330060 ALBANY CO
330120 ALLEGANY CO
330600 BRONX CO
330640 BROOME CO
330840 CATTARAUGUS CO
330860 CAYUGA CO
331000 CHAUTAUQUA CO
331060 CHEMUNG CO
331080 CHENANGO CO
331120 CLINTON CO
331220 COLUMBIA CO
331400 CORTLAHO CO
331520 DELAWARE CO
331620 DUTCHESS CO
332000 ERIE CO
332020 ESSEX CO
332240 FRANKLIN CO
332340 FULTON CO
332400 GENESEE CO
332660 GREENE CO
332820 HAMILTON CO
332960 HERKIMER CO
333340 JEFFERSON CO
333440 KINGS CO
333740 LEWIS CO
333380 LIVINGSTON CO
334040 MADISON CO
334380 MONROE CO
334400 MONTGOMERY CO
334520 NASSAU CO
HC
HC
HC
HC
HC
HC
HC -29.5
HC -35.9
HC -34.9
HC -34.5
HC -24.3
HC -31.6
HC -34.7
HC -46.3
HC -28.0
HC -35.3
HC -32.2
HC -39.6
HC -37.2
HC -39.7
HC -27.5
HC -33.5
HC -49.2
HC -33.8
HC -33.4
HC -39.1
HC -41.0
HC
HC
HC -28.9
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC -26.7
HC
HC
HC
HC
HC
HC -44.2
NO
HO
NO
NO
NO
NO
HO -10.4
NO 18.4
NO 9.8
NO -10.5
HO -0.8
NO -2.0
HO -10.9
NO -8.4
NO 22.2
NO 4.3
HO -3.2
NO 10.9
NO -16.1
NO -15.3
NO -7.2
NO -15.4
HO 0.3
NO -8.1
HO -8.6
NO -10.4
NO -2.4
HO
NO
NO -15.0
NO
HO
NO
NO
NO
NO
HO
NO
NO
NO
NO
NO
HO
NO
NO
NO
NO
NO
NO
HO
NO -1.7
HO
NO
NO
NO
NO
NO -20.3
140
-------
170.
171.
172.
173.
17*.
175.
176.
177.
178.
179.
130.
181.
182.
183.
18*.
185.
186.
137.
188.
189.
190.
191.
192.
193.
19*.
195.
196.
197.
198.
199.
200.
201.
202.
203.
204.
205.
206.
207.
203.
209.
210.
211.
212.
213.
21*.
215.
216.
217.
218.
219.
220.
221.
222.
223.
22*.
225.
226.
33*660 NEW YORK CO
33*720 NIAGARA CO
335060 ONEIDA CO
335100 ONOSDAGA CO
335120 ONTARIO CO
3351*0 ORANGE CO
335180 ORLEANS CO
3352*0 OSKEGO CO
335260 OTSEGO CO
3356*0 PUTNAM CO
335660 QUEENS CO
335700 RENSSELAER CO
335720 RICHMOND CO
335780 ROCK UNO CO
335930 ST. LAWRENCE CO
335965 SARATOGA CO
3360*0 SCHENECTAOY CO
336060 SCHOHARIE CO
336080 SCHUYLER CO
336160 SENECA CO
336500 STEUBEN CO
336580 SUFFOLK CO
336600 SULLIVAN CO
336700 TICGA CO
336720 TOMFKINS CO
3368*0 ULSTER CO
3370*0 WARREN CO
337100 WASHINGTON CO
3372*0 WAYNE CO
337320 WESTCHESTER CO
337580 WYOMING CO
337600 YATES CO
3600*0 AOAMS CO
360080 ALLEN CO
360130 ASHLAND CO
360220 ASHTABULA CO
360260 ATHENS CO
360280 AUGLAIZE CO
3605*0 BELMONT CO
360820 BROUN CO
3610*0 CARROLL CO
3611*0 CHAMPAIGN CO
361260 CLARK CO
361280 CLERMOHT CO
3613*0 CLINTON CO
3614*0 COLUKBIANA CO
361520 COSHOCTON CO
3615*0 CRAWFORD CO
361600 CUYAHOGA CO
361760 DELAWARE CO
362000 ERIE CO
362080 FAIRFIELD CO
3621*u FAYETTE CO
362220 FRANKLIN CO
362260 FULTON CO
362320 GALLIA CO
362380 GEAUGA CO
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
-29.*
-36.0
-27.6
-37.6
-37.3
-*6.7
-2*. 9
-18.9
NO -*.l
NO
HO
NO
NO
NO
NO
NO
NO
NO
NO -12.2
NO
NO -1.5
NO -7.1
NO
NO
NO
NO
NO
NO
NO
NO -12.1
NO
NO
NO
NO
NO
HO
NO
NO -15.2
NO
NO
HO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO -6.7
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
HO
NO
-*.7
141
-------
227.
223.
229.
230.
231.
232.
233.
23*.
235.
236.
237.
238.
239.
2*0.
2*1.
2*2.
2*5.
2*6.
2*7.
2*8.
2*9.
250.
251.
252.
253.
25*.
255.
256.
257.
258.
259.
260.
261.
262.
263.
26*.
265.
266.
267.
268.
269.
270.
271.
272.
273.
27*.
275.
276.
277.
273.
279.
230.
231.
282.
283.
362530
362680
3627*0
362760
362800
362820
362360
3629*0
36296C
363020
363120
363160
363260
363280
363380
363**0
363580
3636*0
363720
363300
363320
363960
36*160
36*180
36**60
36*5*0
36*560
36*6*0
36*900
365260
3654*0
365*30
365500
365580
365660
365740
3658*0
365980
366020
366060
366220
366*00
366500
366700
366720
366780
366960
3670*0
367100
367160
367580
367680
390040
390100
390260
390560
390620
GREENE CO
GUERNSEY CO
HANCOCK CO
HARDIN CO
HARRISON CO
HENRY CO
HIGHLAND CO
HOCKING CO
HOLMES CO
HURON CO
JACKSON CO
JEFFERSON CO
KNOX CO
LAKE CO
LAWRENCE CO
LICKING CO
LOGAN CO
LORAIN CO
LUCAS CO
MADISON CO
MAHONING CO
MARION CO
MEDINA CO
MEIGS CO
MONROE CO
MORGAN CO
MORROW CO
MUSKINGUM CO
NOSLE CO
OTTAWA CO
PERRY CO
PICKAWAY CO
PIKE CO
PORTAGE CO
PUTNAM CO
RICHLAND CO
ROSS CO
SANOUSKY CO
SCIOTO CO
SENECA CO
SHELBY CO
STARK CO
SUMMIT CO
TRUM3ULL CO
TUSCARAUAS CO
UNION CO
VINTON CO
WARREN CO
WASHINGTON CO
WAYNE CO
WOOD CO
WYANDOT CO
ADAMS CO
ALLEGHENY CO
ARMSTRONG CO
BEAVER CO
BEDFORD CO
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC -8.6
HC
HC
HC
HC -10.7
HC
HC
HC
HC
HC -19.6
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
KC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC -35.0
HC
HC
HC
HC
HC
HC -39.3
HC -39.3
HC
HC
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
-1.9
2.3
-0.1
-4.0
-6.4
-6.4
142
-------
285.
236.
287.
289.
289.
290.
291.
292.
293.
294.
295.
296.
297.
298.
299.
300.
301.
302.
303.
30*.
305.
306.
307.
308.
309.
310.
311.
312.
313.
314.
315.
316.
317.
313.
319.
320.
321.
322.
323.
324.
325.
326.
327.
328.
329.
330.
331.
332.
333.
334.
335.
336.
337.
338.
339.
340.
390720 BERKS CO
390820 BLAIR CO
391000 BRADFORD CO
391200 BUCKS CO
391260 BUTLER CO
391300 CAMBRIA CO
391320 CAMERON CO
391330 CARBOM CO
391520 CENTRE CO
391660 CHESTER CO
391760 CLARION CO
391S20 CLEARFIELD CO
391360 CLINTON CO
391960 COLUMBIA CO
392140 CRAWFORD CO
392130 CUMBERLAND CO
392340 DAUPHIN CO
392360 DELAWARE CO
392940 ELK CO
393080 ERIE CO
393220 FAYETTE CO
393320 FOREST CO
393480 FRANKLIN CO
393540 FULTON CO
393720 GREENE CO
394200 HUNTINGDON CO
394240 INDIANA CO
394340 JEFFERSON CO
394460 JUHIATA CO
394640 LACKAWAHNA CO
394700 LANCASTER CO
394840 LAURENCE CO
394900 LEBANON CO
394940 LEHIGH CO
395220 LUZERNE CO
395240 LYCOMING CO
395360 MC KEAN CO
395660 MERCER CO
395760 MIFFLIN CO
395960 MONROE CO
396000 MONTGOMERY CO
396020 MONTOUR CO
396580 NORTHAMPTON CO
396700 NORTHUMBERLAND CO
397120 PERRY CO
397160 PHILADELPHIA CO
397220 PIKE CO
397460 POTTER CO
397960 SCHUYLKILL CO
398320 SNYDER CO
398360 SOMERSET CO
393740 SULLIVAN CO
398800 SUSQUEHANNA CO
399000 TIOGA CO
399070 UNION CO
399140 VENANGO CO
399180 WARREN CO
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
-38.5
-39.3
-38.5
-38.5
-36.1
-38.5
-36.1
-38.5
NO
NO
NO
NO -10.7
NO -6.*
NO
NO
NO
NO
MO -10.7
HO
NO
HO
NO
NO
MO
NO
NO -10.7
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO -2.4
NO
NO
NO
NO
NO
NO
NO -10.7
NO
NO -2.4
NO
NO
NO -10.7
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
143
-------
341.
342.
343.
344.
345.
346.
347.
343.
349.
350.
351.
352.
353.
354.
355.
356.
357.
353.
359.
360.
361.
362.
363.
364.
365.
366.
367.
368.
369.
370.
371.
372.
373.
374.
375.
376.
377.
373.
379.
360.
381.
332.
333.
384.
385.
386.
387.
383.
389.
390.
391.
392.
393.
394.
395.
396.
397.
399200 WASHINGTON CO
399220 WAYNE CO
399330 WESTMORELAND CO
399530 WYOMING CO
399570 YORK CO
410060 BRISTOL CO
410140 KENT CO
410200 NEWPORT CO
410320 PROVIDENCE CO
410380 WASHINGTON CO
470020 AODISON CO
470100 BENNINGTON CO
470160 CALEDONIA CO
470180 CHITTENDEN CO
470200 ESSEX CO
470240 FRANKLIN CO
470260 GRAND ISLE CO
470280 LAMOILLE CO
470360 ORANGE CO
470380 ORLEANS CO
470420 RUTLAND CO
470500 WASHINGTON CO
470580 WIMOHAM CO
470620 WINDSOR CO
480040 ACCOMACK CO
480060 ALSEMARLE CO
480200 ARLINGTON CO
480260 AUGUSTA CO
480300 BATH CO
480600 CAROLINE CO
480760 CLARKE CO
480830 CULPEPER CO
481000 ESSEX CO
481060 FAIRFAX CO
480080 FAIRFAX CO
431120 FAUQUIER CO
481220 FREDERICK CO
481380 GREENE CO
481460 HANOVER CO
481540 HIGHLAND CO
481640 KING GEORGE CO
481760 LOUDOUN CO
481780 LOUISA CO
481860 MADISON CO
482080 NELSON CO
482200 NORTHUMBERLAND CO
482300 ORANGE CO
482320 PAGE CO
482520 PRINCE WILLIAM CO
482620 RAPPAHANHOCK CO
482630 RICHMOND CO
462740 ROCKBRIDGE CO
482760 ROCKIMGHAM CO
482380 SHENANOOAH CO
483000 SPOTSYLVANIA CO
483040 STAFFORD CO
483260 WARREN CO
HC -39.3
HC
HC -39.3
HC
HC
HC -3.9
HC -3.9
HC -3.9
HC -3.9
HC -3.9
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC -42.4
HC
HC
HC
HC
HC
HC
HC -30.9
HC -30.9
HC
HC
HC
HC
HC
HC
HC -10.5
HC
HC
HC
HC
HC
HC
HC -33.5
HC
HC
HC
HC
HC
HC
HC
HC
NO -6.4
NO
NO -6.4
NO
NO
NO -17.4
NO -17.4
HO -17.4
HO -17.4
NO -17.4
NO
NO
NO
NO
NO
NO
NO
NO
HO
NO
NO
NO
HO
NO
NO
NO
NO -30.0
NO
HO
HO
NO
NO
NO
NO -9.0
NO -9.0
NO
NO
NO
NO
NO
NO
NO 10.2
NO
NO
HO
NO
NO
NO
K'O -5.9
NO
NO
NO
NO
NO
NO
NO
NO
144
-------
398.
399.
400.
401.
402.
403.
404.
405.
406.
407.
408.
409.
410.
411.
412.
413.
414.
415.
416.
417.
413.
419.
420.
421.
422.
423.
424.
425.
426.
427.
428.
429.
430.
431.
432.
433.
434.
435.
436.
437.
438.
439.
440.
441.
442.
443.
444.
445.
446.
ENO DATA.
483340
500020
500100
500140
500160
500COO
500240
500260
500380
500400
500460
500520
500560
500530
500600
500620
500640
500660
500720
500740
500760
500840
500360
500900
500980
501020
501060
501100
501140
501200
501320
501380
501440
501460
501480
501520
501560
501600
501680
501700
501840
501660
501880
501900
501960
501980
502100
502200
502220
ERRORS: HONE.
WESTMORELAND CO
BAR8CUR CO
BERKELEY CO
BOONS CO
BRAXTON CO
BROOKE CO
CASELL CO
CALHOUN CO
CLAY CO
DOODRIDGE CO
FAYETTE CO
GILHER CO
GRANT CO
GREENBRIER CO
HAMPSHIRE CO
HANCOCK CO
HARDY CO
HARRISON CO
JACKSON CO
JEFFERSON CO
KANAU'HA CO
LEWIS CO
LINCOLN CO
LOGAN CO
MARION CO
MARSHALL CO
MASON CO
MINERAL CO
MONONGALIA CO
MORGAN CO
NICHOLAS CO
OHIO CO
PEHOLETON CO
PLEASANTS CO
POCAHONTAS CO
PRESTON CO
PUTNAM CO
RANDOLPH CO
RITCHIE CO
ROANE CO
TAYLOR CO
TUCKER CO
TYLER CO
UPSHUR CO
WAYNE CO
WEBSTER CO
WET2EL CO
WIRT CO
WOOD CO
TIME: 2.101 SEC. IMAGE
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
HC
COUNT:
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
HO
HO
HO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
NO
MO
NO
NO
NO
NO
446
145
-------
Appendix C
Lists of counties in each of the four receptor county classes
146
-------
Major Source Counties
3TAFENM
CONNECTICUT
CONNECTICUT
CONNECTICUT
DELAUARE
DIST. COLUMBIA
MARYLAND
MARYLAND
MARYLAND
MARYLAND
MASSACHUSETTS
MASSACHUSETTS
MASSACHUSETTS
MASSACHUSETTS
MICHIGAN
MICHIGAN
MICHIGAN
MICHIGAN
MICHIGAN
MICHIGAN
NEU JERSEY
NEW JERSEY
NEU JERSEY
NEU JERSEY
NEU JERSEY
NEW JERSEY
NEU JERSEY
NEU JERSEY
NEU JERSEY
NEU JERSEY
NEU YORK
NEU YORK
NEU YORK
NEU YORK
NEU YORK
NEU YORK
NEU YORK
NEU YORK
NEU YORK
COUNTYNM
FAIRFIELD
HARTFORD
NEU HAVEN
NEU CASTLE
DISTRICT OF COLUMBIA
ANNE ARUNBEL
BALTIMORE CITY
BALTIMORE
PRINCE GEORGES
DUKES
MIDDLESEX
NORFOLK
SUFFOLK
BAY
GENESEE
MACOMB
MONROE
OAKLAND
UAYNE
BERGEN
CAMDEN
ESSEX
GLOUCESTER
HUDSON
MERCER
MIDDLESEX
MORRIS
PASSAIC
UNION
ALBANY
BRONX
KINGS
NASSAU
NEU YORK
QUEENS
RICHMOND
ROCKLAND
UESTCHESTER
STATENM
OHIO
OHIO
OHIO
OHIO
OHIO
OHIO
OHIO
OHIO
OHIO
OHIO
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
RHODE ISLAND
RHODE ISLAND
RHODE ISLAND
VIRGINIA
VIRGINIA
WEST VIRGINIA
WEST VIRGINIA
UEST VIRGINIA
UEST VIRGINIA
UEST VIRGINIA
UEST VIRGINIA
COUNTYNM
ADAMS
CUYAHOGA
FRANKLIN
GALLIA
JEFFERSON
L.^KE
LORAIN
LUCAS
SUMMIT
UASHINGTON
ALLEGHENY
ARMSTRONG
BEAVER
BUCKS
DELAUARE
INDIANA
MONTGOMERY
MONTOUR
NORTHAMPTON
PHILADELPHIA
VENANGO
BRISTOL
KENT
PROVIDENCE
ARLINGTON
FAIRFAX
HANCOCK
HARRISON
MARSHALL
MASON
MONONGALIA
PUTNAM
147
-------
Non-Major Source SMSA Counties
(Suburban Counties)
STATENM
COUNTYNM
CONNECTICUT
CONNECTICUT
CONNECTICUT
KENTUCKY
KENTUCKY
MAINE
MAINE
MAINE
MAINE
MAINE
MARYLAND
MARYLAND
MARYLAND •
MARYLAND
MARYLAND
MARYLAND
MARYLAND
MASSACHUSETTS
MASSACHUSETTS
MASSACHUSETTS
MASSACHUSETTS
MASSACHUSETTS
MASSACHUSETTS
MASSACHUSETTS
MICHIGAN
MICHIGAN
MICHIGAN
MICHIGAN
MICHIGAN
MICHIGAN
NEW HAMPSHIRE
NEW HAMPSHIRE
NEW HAMPSHIRE
NEW JERSEY
NEW JERSEY
NEW JERSEY
NEW JERSEY
NEW JERSEY
NEW JERSEY
NEW JERSEY
MEW YORK
NEW YORK
NEW YORK
NEW YORK
NEW YORK
NEW YORK
NEW YORK
NEW YORK
NEW YORK
NEW YORK
NEW YORK
NEW YORK
NEW YORK
NEW YORK
NEW YORK
NEW YORK
NEW YORK
NEW YORK
NEW YORK
NEW YORK
NEW YORK
MIDDLESEX
NEW LONDON
TOLLAND
BOYD
GREENUP
ANDROSCOGGIN
CUMBERLAND
PENOBSCOT
SAGADAHOC
YORK
ALLEQANY
CARROLL
CECIL
CHARLES
HARFORD
MONTGOMERY
WASHINGTON
BERKSHIRE
BRISTOL
ESSEX
HAMPDEN
HAMPSHIRE
PLYMOUTH
WORCESTER
LAPEER
LIVINGSTON
SAGINAW
ST CLAIR
SHIAWASSEE
WASHTENAW
HILLSBOROUGH
ROCKINGHAM
STRAFFORD
ATLANTIC
BURLINGTON
CUMBERLAND
MONMOUTH
SALEM
SOMERSET
WARREN
BROOME
CHEMUNG
DUTCHESS
ERIE
HERKIMER
LIVINGSTON
MADISON
MONROE
MONTGOMERY
NIAGARA
ONEIDA
ONONDAGA
ONTARIO
ORANGE
ORLEANS
OSWEGO
PUTNAM
RENSSELAER
SARATOGA
SCHENECTADY
SUFFOLK
NEW YORK
NEW YORK
NEW YORK
NEW YORK
OHIO
OHIO
OHIO
OHIO
OHIO
OHIO
OHIO
OHIO
OHIO
OHIO
OHIO
OHIO
OHIO
OHIO
OHIO
OHIO
OHIO
OHIO
OHIO
OHIO
OHIO
OHIO
OHIO
OHIO
OHIO
OHIO
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
RHODE ISLAND
VERMONT
VIRGINIA
VIRGINIA
VIRGINIA
VIRGINIA
VIRGINIA
VIRGINIA
TIOGA WEST VIRGINIA
UARREN WEST VIRGINIA
WASHINGTON WEST VIRGINIA
UAYNE "EST VIRGINIA
ALLEN "EST VIRGINIA
AUGLAIZE WEST VIRGINIA
BELMONT WEST VIRGINIA
CARROLL WEST VIRGINIA
CHAMPAIGN
CLARK
CLERMONT
DELAWARE
FAIRFIELD
FULTON
GEAUGA
GREENE
LAURENCE
LICKING
MADISON
MAHONING
MEDINA
OTTAWA
PICKAUAY
PORTAGE
PUTNAM
RICHLAND
STARK
TRUMBULL
UARREN
WOOD
ADAMS
BERKS
CAMBRIA
CARBON
CENTRE
CHESTER
CUMBERLAND
DAUPHIN
ERIE
LACKAUANNA
LANCASTER
LEHIGH
LUZERNE
LYCOMING
MONROE
PERRY
SOMERSET
SUSQUEHANNA
WASHINGTON
WESTMORELAND
YORK
WASHINGTON
CHITTENDEN
ALBEMARLE
GREENE
HANOVER
LOUDOUN
PRINCE WILLIAM
RICHMOND
BROOKE
CABELL
KANAWHA
MINERAL
OHIO
WAYNE
UIRT
WOOD
148
-------
Agricultural Counties
STATENM
DELAWARE
KENTUCKY
KENTUCKY
KENTUCKY
KENTUCKY
KENTUCKY
KENTUCKY
KENTUCKY
KENTUCKY
MARYLAND
MARYLAND
MARYLAND
MARYLAND
MARYLAND
MARYLAND
MASSACHUSETTS
MICHIGAN
MICHIGAN
MICHIGAN
MICHIGAN
NEW JERSEY
NEW YORK
NEU YORK
NEW YORK
NEU YORK
NEW YORK
NEU YORK
NEU YORK
NEW YORK
NEU YORK
OHIO
OHIO
OHIO
OHIO
OHIO
OHIO
OHIO
OHIO
OHIO
OHIO
OHIO
OHIO
OHIO
OHIO
OHIO
OHIO
OHIO
OHIO
OHIO
OHIO
OHIO
OHIO
OHIO
COUNTYNM
KENT
BATH
BOURBON
BRACKEN
FLEMING
MASON
MONTGOMERY
NICHOLAS
ROBERTSON
CAROLINE
FREDERICK
HOWARD
KENT
QUEEN ANNES
TALBOT
NANTUCKET
HURON
LENAUEE
SANILAC
TUSCOLA
HUNTERDON
CAYUGA
QENESEE
JEFFERSON
SCHUYLER
SENECA
STEUBEN
TOMPKINS
WYOMING
YATES
ASHLAND
ASHTABULA
BROUN
CLINTON
COLUMBIANA
COSHOCTON
CRAWFORD
ERIE
FAYETTE
GUERNSEY
HANCOCK
HARDIN
HARRISON
HENRY
HIGHLAND
HOLMES
HURON
KNOX
LOGAN
MARION
MORGAN
MORROW
MUSKINGUM
STATENM
OHIO
OHIO
OHIO
OHIO
OHIO
OHIO
OHIO
OHIO
PENNSYLVANIA
'PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
VIRGINIA
UEST VIRGINIA
UEST VIRGINIA
UEST VIRGINIA
UEST VIRGINIA
COUNTYNM
NOBLE
SANDUSKY
SENECA
SHELBY
TUSCARAUAS
UNION
UAYNE
UYANDOT
BRADFORD
COLUMBIA
CRAWFORD
FRANKLIN
GREENE
LAURENCE
LEBANON
MERCER
NORTHUMBERLAND
CLARKE
BARBOUR
JEFFERSON
MARION
TAYLOR
149
-------
Natural Counties
STATENM
CONNECTICUT
CONNECTICUT
DELAUARE
KENTUCKY
KENTUCKY
KENTUCKY
KENTUCKY
KENTUCKY
KENTUCKY
KENTUCKY
MAINE
MAINE
MAINE
MAINE
MAINE
MAINE
MAINE
MARYLAND
MARYLAND
MARYLAND
MARYLAND
MARYLAND
MARYLAND
MARYLAND
MASSACHUSETTS
MASSACHUSETTS
COUNTYNM
LITCHFIELB
UINDHAM
SUSSEX
CARTER
ELLIOTT
LAURENCE
LEUIS
MENIFEE
MORGAN
ROWAN
FRANKLIN
KENNEBEC
KNOX
LINCOLN
OXFORD
SOMERSET
UALDO
CALVERT
DORCHESTER
6ARRETT
ST MARYS
SOMERSET
UICOMICO
WORCESTER
BARNSTABLE
FRANKLIN
STATENM
MICHIGAN
MICHIGAN
MICHIGAN
MICHIGAN
MICHIGAN
MICHIGAN
MICHIGAN
NEW HAMPSHIRE
NEU HAMPSHIRE
NEW HAMPSHIRE
NEU HAMPSHIRE
NEW HAMPSHIRE
NEW HAMPSHIRE
NEU HAMPSHIRE
NEW JERSEY
NEU JERSEY
NEU JERSEY
NEU YORK
NEW YORK
NEU YORK
NEW YORK
NEU YORK
NEU YORK
NEU YORK
NEU YORK
NEU YORK
NEU YORK
NEU YORK
NEU YORK
NEU YORK
NEU YORK
NEU YORK
NEW YORK
NEU YORK
NEU YORK
NEU YORK
OHIO
OHIO
OHIO
OHIO
OHIO
OHIO
OHIO
OHIO
OHIO
OHIO
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
COUNTYNM
ALCONA
ALPENA
ARENAC
IOSCO
MONTMORENCY
OGEMAU
OSCODA
BELKNAP
CARROLL
CHESHIRE
COOS
GRAFTON
MERRIMACK
SULLIVAN
CAPE MAY
OCEAN
SUSSEX
ALLEGANY
CATTARAUGUS
CHAUTAUOUA
CHENANGO
CLINTON
COLUMBIA
CORTLAND
DELAUARE
ESSEX
FRANKLIN
FULTON
GREENE
HAMILTON
LEUIS
OTSEGO
ST LAURENCE
SCHOHARIE
SULLIVAN
ULSTER
ATHENS
HOCKING
JACKSON
ME I OS
MONROE
PERRY
PIKE
ROSS
SCIOTO
VINTON
BEDFORD
BLAIR
BUTLER
CAMERON
CLARION
CLEARFIELD
CLINTON
150
-------
Natural Counties, continued
STATENM
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
PENNSYLVANIA
RHODE ISLAND
VERMONT
VERMONT
VERMONT
VERMONT
VERMONT
VERMONT
VERMONT
VERMONT
VERMONT
VERMONT
VERMONT
VERMONT
VERMONT
VIRGINIA
VIRGINIA
VIRGINIA
VIRGINIA
VIRGINIA
VIRGINIA
VIRGINIA
VIRGINIA
VIRGINIA
VIRGINIA
VIRGINIA
VIRGINIA
VIRGINIA
VIRGINIA
VIRGINIA
VIRGINIA
VIRGINIA
VIRGINIA
VIRGINIA
VIRGINIA
COUNTYNM
ELK
FAYETTE
FOREST
FULTON
HUNTINGDON
JEFFERSON
JUNIATA
MC KEAN
MIFFLIN
PIKE
POTTER
SCHUYLKILL
SNYDER
SULLIVAN
TIOGA
UNION
WARREN
WAYNE
WYOMING
NEWPORT
ADDISON
BENNINGTON
CALEDONIA
ESSEX
FRANKLIN
GRAND ISLE
LAMOILLE
ORANGE
ORLEANS
RUTLAND
WASHINGTON
WINDHAM
WINDSOR
ACCOMACK
AUGUSTA
BATH
CAROLINE
CULPEPER
ESSEX
FAUQUIER
FREDERICK
HIGHLAND
KING GEORGE
LOUISA
MADISON
NELSON
NORTHUMBERLAND
ORANGE
PAGE
RAPPAHANNOCK
ROCKBRIDGE
ROCKINGHAM
SHENANDOAH
STATENM
VIRGINIA
VIRGINIA
VIRGINIA
VIRGINIA
WEST VIRGINIA
WEST VIRGINIA
WEST VIRGINIA
WEST VIRGINIA
WEST VIRGINIA
WEST VIRGINIA
WEST VIRGINIA
WEST VIRGINIA
WEST VIRGINIA
WEST VIRGINIA
WEST VIRGINIA
WEST VIRGINIA
WEST VIRGINIA
WEST VIRGINIA
WEST VIRGINIA
WEST VIRGINIA
WEST VIRGINIA
WEST VIRGINIA
WEST VIRGINIA
WEST VIRGINIA
WEST VIRGINIA
WEST VIRGINIA
WEST VIRGINIA
WEST VIRGINIA
WEST VIRGINIA
WEST VIRGINIA
WEST VIRGINIA
WEST VIRGINIA
WEST VIRGINIA
WEST VIRGINIA
COUNTYNM
SPOTSYLVANIA
STAFFORD
WARREN
WESTMORELAND
BERKELEY
BOONE
BRAXTON
CALHOUN
CLAY
DODDRIDGE
FAYETTE
GILMER
GRANT
3REENBRIER
HAMPSHIRE
HARDY
JACKSON
LEWIS
LINCOLN
LOGAN
MORGAN
NICHOLAS
PENDLETON
PLEASANTS
POCAHONTAS
PRESTON
RANDOLPH
RITCHIE
ROANE
TUCKER
TYLER
UPSHUR
WEBSTER
WETZEL
151
-------
Appendix D
County-by-county comparisons of base case vs. control case ozone
concentrations. Mean and standard deviation values are for the
9-day scenario period. All concentrations are in ppm.
152
-------
— i>-cDCKO*oo*-«o»>(S-«oa3O»oocKCoo^riaao.in o^
'
**--
c oooooooooooooooooooooooooooooooooooooooooooooooooooooo
e/i
i x
25T (x.
^-^ a
^JQQOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO
oooooooooooooooooooooooooooooooooooocooooooooooooooooo
O OOO^OOOOOOOOOOOOOOOOOOOOOOOC'OOO-COC-OOOC'OOOOOOOOOOOOOOOO
•c ooo^o — o — oooooooooooooooooooooooooooooooooooooooooooooo
a o ooooooooooooooo oooooooooooooooooooooooooooooo oooooooo
UJ
<*n
-
wac
-------
*
c ooooooooooococooooocoooc ooocoooc-o-^ooocooooo-c-ooooooooo
c
c o ooooooooooooooooooooooooooooooooooooooooooooooooooooo
o ^
y ^ ^j co>ONirvajo*'O'OfxcD^ajo-rsrjO'rs.i>X'-^o>oo»"s- C3rvu*)ui--rs.^i-;
H~" i OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO
= n
.
LU ^ oooooooooooooo-^ooooo-«ooooooooooooooooooooooooooooooooo
ig £ [[[
o
UJ ^
tO Z aon»ofvno*O(sao — rio^^(^rt^t>>r4Nibic9rvo^binasooo«»rNior4eaoai(D(oo**<«o*o*i — ^^NN^^X^. — «4 — M ^ M ^ *«„,-, — .-, — — szxxx
^uiiu — uc^u — uuu— uuu'^uuur-u
L£ LU S: OL£_JUJ LU X I— uC
uiO2:cn»-h-aujAt^ Z^UILJXUJ
Z U- ^ U i— -C -^ _J •-•S'-iCnji^H-^t-
Q LU OiCn^-3I»-«U)1— ZO JLUXli>iJ_i3_jCn
-------
— oooooooooooooooooooooooooooooooooooooooooooooooooooooo
tn
X *
c
.. c oooooooooooooooobooooooooooooooooooooooooooooooooooooo
c
!« ». o -• i *• i «• -o • < r^
^ a oooooooooooooooooooooooooooooooooooooooooooooooooooooo
H- t- - *
2T en
X
or*rj - rx
CO a oooooooooooooooooaooooooooooooooooooooooooooooooaooooo
y?
o oooooooooooooooooooooooooooooooooooooooooooooooooooooo
UJ
ui ui uj LJ
i^; ?{• ,-£ fa
Z en "• tn tf_ iy ty i.) '^MAJ ''Mi< hi ^it Ui 'A! UJ 'A* 'il Ut '^ '/Jl 'vt *'J 'iJ ^
_ ssz^cc^»
ut— ^^^ auiz •—
-* a -t_; en a •/•*- u ui
t-zzzzj: j.zj;a:u;xco) •« tnx z
z 3u;o ^
-H ui au*y) cn
155
-------
c ^oo^iiT^ortMo-OiNM-^cowrt—«tjirjr(05*^^O'O<>*ooijioo'X'Oooo5Comri^3Oi'orjo*i>CNO*^"'^o*-*«^u"ia3o-
*: C4 T4 r-j n n r* •* r*i v *•* «• p* m ** -* n r*j ci CM n -* ri ri *•• — TJ r-i m r» m -* rj rg o o o m o o o o — o o o o — -^ o n -^ «-• o o
£ COOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO
c
|5 S OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO00000300OOO
O s
o
21 c OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOGOOOOOOOOOOOOOGOOOOO
C£
-------
— o oo-^ooo-soooc ooosoo
LU <:
oooooooooooooooooooooooooooooooooooooooooooooooooooooo
«-J i OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOCOOOOOCOOOC
•J!
-nninio««»«isiniii»riin«ro-'Minpv»rio.r
-------
,£2 < OOGOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOoooooooo
o
t IN. o— o^-»»-tooo — o *•« o — o o o — o-»--«riooo-*(Norjooooo«-«oooriooo-i — rjooo--t — oooo
^- a oooooooooooooooooooooooooooooooooooooooooooooooooooooo
•£.
X
•........................,.........................*..
-i
<: o ooooooooooooooooooooooooooooooooooooooooooooo oooooooo
a:
c -il-i L«,j^jZ-iw-iZ:-*I.-*l!J — J-;-iw-jI-4l:jI-iI.J--iIjI-i — jl-il-i — -I--JZ-.1
-
c i£»-'Q-:uju.au.aaiO.i:>-i
-------
•s o OOOOOOOOOOOOOOOOOOOOOOOOOOOGOOOOOOOOOOOOOOOOOOOOOOOOO
IT
LU x
^ c o OOOOOOOOOOOOOOOOOOOOGOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO
u. •
1 S
o
oc
>_
^ r^ -» o o o *- -* *•* o PJ n o o o — o ro n n rj rj PJ rj n rj cj rj rj PJ r. rj a n r* rj « -* o ~+ -* o o o -* o o o — — — o o — o —
O a o ooooooooooooooooooooooooooooooooooooooooooooooooooooo
rrt
X
-*_; cio-rs:i^c_,'^^--r 2 LU xu.-^ i^zrTz: 3:
159
-------
C GW&frO~fr&B-~CDCQtBf'*QQ*Q+Qt33-**&*Q+Q+<>£Q
— OOOC-O-C'OCOOOC-OCOOOOO'^OOOOOOOOO-C-OOOOOOOOOOOC'OOOOOCOOOOOO
yj
o
<: oooooooooooooooooooooooooooooooooo ooo ooooooooooooooooo
cs oooooooooooooooooooooooooooooooooooooooooooooooooooooo
w
J£
. - -*'
z: *r m ci f j -< 2: tn
-------
<: a3
z -^o^o^oo
i oooooc-o
LU
C/>
< «
OQ 3:
ooooooo
OOOOOOO
ooooooo
~ ci r-j m "i r-i n n
c ooooooo
cn ;•: 01 x to v en
Ul lu UJ UJ Ul U. Ui
-J o? >• A H> C£ G •
>- u. *E Lu ai t-i O
"-333333
161
-------
Example of other types of concentration information provided
by the simulation. Here we compare the 3, 12, and 24 hour
averaged concentrations predicted at one of the 439
counties in the base and control simulations.
Base case results
---------------------------- CQUNTYNM^LITCHFIELD SI ATFKM-CdHNECT ICUT
DATE rtVtt24 MftX24J MftX12Sl HAX1US2 ftyiijSSl /V.IG3H2
80203 .021 .026 .026 .025 .019 .021 .024 .023 .024
30'.!06 ,031 .039 .039 ,038 ,028 .033 .037 ,039 .037
30207 .043 .032 .032 .056 .041 .046 .031 .030 .046
30208 ,070 .093 ,093 ,089 ,063 ,071 ,037 .093 ,097
80209 .113 ,199 .199 .096 .112 .136 .191 .145 .082
30210 ,037 ,106 .106 ,074 .097 ,104 ,097 ,077 .067
80211 .063 .081 .072 .091 .070 .035 .046 .031 .063
80.ii:.> ,088 .099 ,099 .080 .095 ,096 .098 .087 .077
S0213 .063 .073 .073 . .062 .064 .071 .066 .059
control aase results
COUNTYNM^LITCHFIELD 31ATENM«CUKNECTIGUT
DATE !y;ti24 MAX24 MrtXTJSl I1AX12S2 rtVOSSl (V,*G38;: AVG.iSJ
.t
80203 .021 .025 .025 .024 .019 .021 .025 .025 .023
30,206 .029 .036 ,036 .036 .026 .031 .034 .034 .035
80207 .039 .047 .047 .051 .039 .014 .047 .041 .040
3020U ,056 .071 ,071 ,064 .057 .062 .069 .070 ,061
80209 .069 .109 .109 .080 .070 .093 .106 .079 .062
80.MO ,077 .091 ,091 ,068 .034 .090 .085 .070 .061
30211 .055 .068 ,065 .074 .064 .032 .042 .042 .047
30212 .072 .085 .085 ,066 ,081 .034 .084 .074 .064
30213 .048 .036 .056 . .051 .052 .055 .049 .040
162
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
1. REPORT NO.
2.
3. RECIPIENT'S ACCESSION NO.
4. TITLE AND SUBTITLE
NUMERICAL SIMULATIONS OF PHOTOCHEMICAL AIR POLLUTION
IN THE NORTHEASTERN UNITED STATES: ROM1 APPLICATIONS
5. REPORT DATE
6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
Robert G. Lamb
8. PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Same as Block 12
10. PROGRAM ELEMENT NO.
CDWA1A/02-1335 (FY-86)
11. CONTRACT/GRANT NO.
12. SPONSORING AGENCY NAME AND ADDRESS
Atmospheric 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/600/9
15. SUPPLEMENTARY NOTES
16. ABSTRACT
The first-generation Regional Oxidant Model (ROM1) was used to simulate
pollutant concentrations during the nine-day period 23-31 July 1980. The one-hour
and daily daylight (0900-1600 LST) averaged ozone concentrations produced in each
of two simulations, one representing 1980 base emissions and the other projected
1987 emission, were compared to assess the effectiveness of the proposed emissions
changes on air quality.
The analyses of the model results are prefaced by discussions of a number of
basic issues on regional scale modeling, including model initialization, selection
of meteorological data, effects of grid size on model performance, estimating
long-term concentration statistics from short-period simulations, probabilistic
vs. quasi-deterministic modes of model operation, uncertainty in emissions
estimates, the characteristics of VOC and NO sources in the Northeast, and other
topics. Preliminary results of analyses of the SAROAD monitoring data, which
reveal the characteristics of the ozone problem in the northeastern United States,
set the stage for the model simulations.
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.lDENTIFIERS/OPEN ENDED TERMS
c. COSATI Field/Group
18. DISTRIBUTION STATEMENT
RELEASE TO PUBLIC
19. SECURITY CLASS (This Report!
UNCLASSIFIED
21. NO. OF PAGES
20. S
'Ms page)
22. PRICE
EPA Form 2220-1 (R«v. 4-77) PREVIOUS EDITION is OBSOLETE
------- |