EPA-450/3-74-056<
JULY 1973
HACKENSACK MEADOWLANDS
AIR POLLUTION STUDY
DEVELOPMENT
AND VALIDATION
OF A MODELING TECHNIQUE
FOR PREDICTING
AIR QUALITY LEVELS
U.S. ENVIRONMENTAL PROTECTION AGENCY
Office of Air and Waste Management
Office of Air Quality Planning and Standards
Research Triangle Park, North Carolina 27711
-------
EPA-450/3-74-OJ6-C
HACKENSACK MEADOWLANDS
AIR POLLUTION STUDY -
DEVELOPMENT
AND VALIDATION
OF A MODELING TECHNIQUE
FOR PREDICTING
AIR QUALITY LEVELS
by
James R. Mahoney, Bruce A. Egan, and
Edward C. Reifenstein, III
Environmental Research and Technology, Inc.
429 Marrett Road
Lexington, Massachusetts 02173
Contract No. EHSD 71-39
EPA Project Officer: John Robson
Prepared for
. ENVIRONMENTAL PROTECTION AGENCY
Office of Air and Waste Management
Office of Air Quality Planning and Standards
Research Triangle Park, N. C. 27711
July 1973
-------
This report is issued by the Environmental Protection Agency to report technical
data of interest to a limited number of readers. Copies are available free
of charge to Federal employees, current contractors and grantees, and nonprofit
organizations-as supplies permit-from the Air Pollution Technical Information
Center, Environmental Protection Agency, Research Triangle Park, North
Carolina 27711; or, for a fee, from the National Technical Information Service,
5285 Port Royal Road, Springfield, Virginia 22161.
This report was furnished to, the Environmental Protection Agency by the
Environmental Research and Technology, Inc., in fulfillment of Contract
No. EHSD-71-39. The contents of this report are reproduced herein as received
from the Environmental .Research and Technology, Inc. The opinions', findings,
and conclusions expressed are those of the author and not necessarily those
of the Environmental Protection Agency. Mention of company or product
names is not to be considered as an endorsement by the .Environmental Protection
Agency. . ' , ,: . :..' i'. '.; ,':" .
Publication No. EPA-450/3-74-056-C
11
-------
PREFACE
The Hackensack Meadowlands Air Pollution Study final report consists
of a summary report, five task reports, and three appendices, each bound
separately. This report is the second of the five task reports. Its
purpose is to describe the mathematical basis for predicting air quality
levels for the New Jersey Hackensack Meadowlands. The report discusses
both the development of the model and its validation and calibration. The
'- '
preparation of model emissions input data is discussed extensively in the
first task report, and the procedures for operating the software components
of the model are discussed in the fifth task report.
iii
-------
ACKNOWLEDGEMENTS
The work upon which' this report is based was performed pursuant to
contract No. EHSD-71-39 with the Environmental Protection Agency, and
Contract No. IP-290 with the New Jersey Department of Environmental Pro-
tection.
The cooperation and assistance of the many personnel from EPA and
NJDEP contributed greatly to the success of this study. The special
assistances of Mr. Roland S. Yunghans, and Dr. Edward B. Feinberg, En-
vironmental Scientists, Office of the Commissioner, NJDEP, and Mr. John
Robson, Land Use Planning Branch, Office of Air Programs, EPA, is
appreciated.
IV
-------
TABLE OF CONTENTS
Page
PREFACE
ACKNOWLEDGEMENTS . iv
LIST OF ILLUSTRATIONS vii
LIST OF TABLES
SUMMARY ix
1. INTRODUCTION ' 1
1.1 Definition of the Problem
2. DESCRIPTION OF THE DISPERSION MODEL 3
2.1 General Comments
i
2.2 Principal Modifications Incorporated in the ERT
MARTIK Model 4
2.3 Basis of the Model: The Gaussian Plume Equation 5
2.4 Geometrical Details of the ERT/MARTIK Model 17
2.5 Operation of the Model 27
3. DATA USED IN THE MODEL ANALYSES 29
3.1 Introduction 29
3.2 Meteorological Data 30
3.3 Emission Data 37
3.4 Air Quality Monitoring Data
4. MODEL VALIDATION PROCEDURES AND RESULTS 53
4.1 Introduction 53
4.2 Procedures for Validating Models
4.2.1 Selection of Data for Purposes of Validation
4.2.2 Preliminary Runs to Assess the Initial Agree-
ment Between' Predicted and Observed Values 53
-------
TABLE OF CONTENTS, contd.
Page
4.2.3 Sensitivity Analyses and Identification of
Possible Model Improvements ' 59
4.2.4 Modifications Made to Model and Final
Calibration of Model Results 67
4.2.5 Final Calibration of the Model 69
4.3 Discussion 71
REFERENCES 75
GLOSSARY 76
-------
LIST OF ILLUSTRATIONS
Figure Page
1 Coordinate System Showing Gaussian Distributions in
the Horizontal and Vertical 7
2 Geometry of the Linear Interpolation Between Adjacent
Wind Direction Sectors . 11
3 Geometry for Area Source with Dimensions g1 and g? in
the Horizontal Plane, Shown in the Receptor-Centered
Wind Oriented Coordinate System 21
4 Integration Over Source Cell Area By Summation of
Elemental Strips from X to X 23
5 Geometry for Determination of Concentration Due to
Point and Line Sources 25
6 The Four Geographical Zones Used in the Development
of the Emission Inventory 44
7 New Jersey State Bureau of Air Pollution Control
Continuous Air Monitoring Network 50
8 New Jersey High-Volume Sampler Network 51
9 Highly Correlated Regression Line Fits 54
10 Validation Sites Surrounding the Meadowlands Region 57
vn
-------
LIST OF TABLES
Table Page
1 Functions k and d of Stability Class 14
2 Glossary of Terms for Area Source Geometrical
Description . - , 19
3 Values f., r and r.. for Winter, Summer and
Annual Time Periods 62
4 Vertical Spread Statistic Constants (a = kxd) 63
5 Predicted and Observed Validation Data 70
6 Ratios of Observed to Predicted Values by Pollu-
tant and Time Period 72
Vlll
-------
SUMMARY
This report describes the development and validation of the atmospheric
diffusion model used as the tool for calculating the air quality concentra-
tion patterns expected in the Meadowlands planning area in 1990. The model
has been designed specifically to operate on the computing equipment of the
New Jersey Department of Health, and the model has been specialized to pro-
vide input data to the ERT/AQUIP program for the evaluation of total antici-
pated air quality impact. Thus the model is useful both as a stand-alone
tool and as an element within the framework of a total land use planning-
air pollution impact evaluation system.
The model was run for validation and calibration purposes for summer,
winter and annual time periods using meteorological input data from Newark
Airport and the point, line and area source emission inventory developed for
the year 1969. Initial comparison of the model results with measured data
indicated that the accuracy could be improved by some modifications to the
model.
Specific modifications made to the model included the adoption of a
half-life for SO- emissions, the inclusion of dispersive spread statistics
more characteristic of urban areas, and incorporation of wind speed
variations with height above ground. The final calibration of the model
utilized pollutant measurement data from five stations surrounding the
Meadowlands region to develop calibration factors applicable to the individual
locations and for the Meadowlands region in general.
The diffusion model development and validation has followed "standard"
procedures wherever possible. This was done both to assure the future use-
fulness of the model in the Meadowlands area, and to establish a methodology
which might be copied, and improved upon, in future related studies.
-------
1. INTRODUCTION
Definition of the Problem
The basic goal of the model development and validation effort has been
the calculation of seasonal and annual average concentrations for the pol-
lutants of interest (S02, CO, NOX, hydrocarbons and particulates) expected
within the planning region for the emission patterns associated with various
possible land use distributions which may exist in 1990.
An important secondary goal has been the documentation and reporting
of the modeling and validation schemes for two purposes: (1) so that the
model and its associated data might be used for future studies of expected
air pollution impact in the Meadowlands area; and (2) so that the experience
gained in the present applications study might be made available for similar
studies in other regions.
The model used in this study has been developed from the basic form
described by Martin and Tikvart (1968) , and documented in the report on
the Air Quality Display Model (NAPCA, 1969). To preserve standard model
notation wherever possible, the definitions of model parameters and variables
adopted in the AQDM have been retained. However, several modifications to
the Martin-Tikvart model have been incorporated into this study. The
modifications provide for controllable computational accuracy (as a function
of computation time and therefore cost), and increased flexibility in the
treatment of area sources of various geometrical shapes. These model fea-
tures are particularly desirable for planning studies, where many different
combinations of emission patterns must be evaluated. The details of the
model adopted for use in this study are described in Section 2.
-------
The model analysis incorporates several types of data: emission patterns,
meteorological data and topographical data are required as model inputs, and
air quality measurements are required for model validation and for the
development of calibration parameters where appropriate; The details of
the data requirements, and a summary of the data selected for use in this
study, are described in Section 3.
Section 4 contains a description of the model validation scheme adopted
for use in the planning projections. Various other validation techniques
which were considered but not used are also described, because they may
have application in similar studies for other regions.
Section 5 contains a summary statement on the modeling procedures
adopted for use in preparing projections of concentration levels expected
in the planning region by 1990.
-------
2. DESCRIPTION OF THE,.DISPERSION MODEL
2.1 General Comments . . .
The model used for the dispersion calculations is called MARTIK in the
ERT program library. It is a gaussian plume model which follows the physical
and mathematical basis described by Martin and Tikvart (1968). The model
has been modified to improve computational accuracy for receptors near
' ' '..' ' .. <*". - T'"..' '">'. ':.'- ' .
source areas, to permit tradeoffs between accuracy requirements and compu-
tation time, arid to permit improved flexibility .in the treatment of rectangu-
lar area sources of any size and location.
The model has also been specially modified to permit its use on the
RCA SPECTRA 70 computer operated by the New Jersey Department of Health.
l'
Specially designed data storage and data flow routines have been developed
to comply with the core limitations of the SPECTRA 70. Also, the model
input and output routines have been integrated into the AQUIP software
system (for example, the LANTRAN program which computes emissions source
distributions from given planning parameters, and the SYMAP computer graphics
program, which provides general display capabilities for land use, emissions
and air quality data).
Detailed descriptions of all programs, including the version of the
. - - i .,
MARTIK program used in the present study, are presented in the User's Manual
document. Readers interested in detailed reviews or operational use of the
model should refer to the User's Manual, as a supplement to the following
text.
The general MARTIK program provides for calculations both in an
averaging"mode and in an instantaneous mode. In the instantaneous case,
-------
fixed values of the meteorological parameters (wind speed and direction,
and stability class) are used, to determine the concentration pattern
expected at a specific time. The description of the model in this section
is restricted to the time-average mode, because the planning forecasts
required are the seasonal and annual average cases.
2.2 Principal Modifications Incorporated in the ERT/MARTIK Model
The principal modifications of the basic model reported by Martin and
Tikvart (1968), incorporated into the ERT/MARTIK model are listed here.
1. Treatment of Area-Source Emissions
A major improvement of accuracy of representation was made by replacing
the virtual point source approximation to area source emissions with a
numerical integration procedure over the area source. The use of a virtual
point source representation for area sources results in a "sawtooth" concen-
tration profile in the crosswind direction at short distances downwind from
a group of area source cells. The numerical integration scheme avoids this
difficulty. The integration is accomplished by summation of elemental strips
of the area source and may be used for any specified source height and any
wind direction. For efficiency of calculation, the computation routine uses
fewer elemental strips when the source receptor distance is large and where
finer elements would not significantly change the computed concentrations.
The maximum number of elemental strips to be used is externally specified
as part of a set of input parameters which control computational accuracy.
2. Treatment of Line-Source Emissions
Emissions from roadways may be represented as if from line-type sources.
In order to estimate the downwind concentration from these sources for any
-------
possible wind direction and elevation, another numerical integration routine
was developed. This procedure involves approximating short segments of the
line representation with upwind virtual point sources. In a manner similar
to that of the area source integration, the number of virtual points per
unit length used depends upon the distance to the receptor and the accuracy
selected.
3. Flexibility of Operating Mode and Output Format
A major feature of the MARTIK system is the ability to isolate the
contributions to the concentration at receptors by simp]e choices of input
parameters. Thus, for example, the contributions from individual sources,
wind directions or stability classes may be easily isolated.
4. Computational Efficiency Features
In addition to the controlled accuracy capabilities discussed in terms
of the integration routines for line and area type sources, a number of
other specific improvements have been made to keep computation time low.
These include: (1) input specification of argument value ranges for which
exponentials may be approximated by simple functions; (2) simple inverse
scaling of concentration contributions as a function of wind speed, for
sources with zero effective plume rise; (3) optimization of the number of
source-receptor geometry calculations for a given run.
2.3 Basis of the Model: The Gaussian Plume Equation
The model calculations are based upon multiple applications, and inte-
grated forms, of the gaussian plume equation which represents the concentra-
-------
tion pattern downwind from a point source. The general form of the equation
is
where
exp _
exp [-1/2 (i-^Ji ) ] t exp [-1/2 (i-i-S. ) ] J. (1)
° z °z
(x,y,z) are the (upwind, cross-wind and vertical)
components of a cartesian coordinate system, such
that the receptor point is located at or vertically
above the origin (expressed in units of length) and
the source at the point (x,y,H)
H is the effective height of emission and therefore
the centerline height of the plume (length)
q is source strength (mass/time)
a , a are dispersion coefficients that are measures of
cross-wind and vertical plume spread. These two
parameters are functions of downwind distance and
atmospheric stability (length)
u is average wind speed (length/time)
Figure 1 illustrates the geometry for the plume equation. The source
base is at z = 0 in the coordinate system, and the plume center-line reaches
an equilibrium height H at some distance downwind from the source.
-------
(x,-y,o)
(x,o,o)
Plume Axis
{downwind)
y
(crosswind)
X (upwind)
Figure 1 Coordinate System Showing Gaussian Distributions
in the Horizontal and Vertical
-------
The most important assumptions upon which the equation is based are the
following:
1. The wind speed and direction in the vicinity of the point source
are constant throughout the period of interest.
2. mien the effluent enters the atmosphere, the plume rises until it
reaches an equilibrium altitude; the plume er-nterline height remains constant
at all downwind distances after the equilibrium height is reached.
3. At any downwind distance, the maximum concentration occurs at the
plume centerline. The distribution of concentration values off the center-
line is given by the product of two gaussian, or bell-shaped curves.
4. The concentration profiles described by the gaussian form arc not
"instantaneous" plume profiles. Instead they represent concentrations
averaged over a short time, such as 10 minutes.
5. None of the effluent is lost from the plume. Therefore when the
plume boundary intersects the ground surface, it is assumed that all material
is reflected back above the ground.
6. The effluent emission rate is constant, and the meteorological
parameters determining plume form are constant; (i .e. , the equation represents
steady state conditions) .
Ground-level concentration estimates are obtained by setting z = 0 in
Eq. (1), resulting in
2 2
X(0,0,H) = ^-u exp [-1/2 (Jl) ] exp [-1/2 (~ ) ] . (2)
y z y z
-------
Eq. (2) can be modified to yield estimates of long-term average
concentrations (e.g., seasonal and annual concentrations), when applicable
stability wind rose data are available. A stability wind rose is a tabula-
tion of the joint frequency of occurrence of wind speed, wind direction and
atmospheric stability category at a specific location.
Although the probability of all wind directions is a continuous function
over the long time periods, for computation purposes discrete wind directions
are specified with respect to a 16-point compass, corresponding to 22.5
angular sectors. For annual periods it is often assumed that all wind
directions within a given 22.5 sector occur with equal frequency. Thus
the effluent could be assumed to be uniformly distributed in the horizontal
within the sector. However, this assumption would result in discontinuities
in calculated .concentrations at sector boundaries. A more reasonable distri-
bution is obtained by using a linear interpolation between sector center
lines. In this case, the concentration at a given receptor location is
composed of proportional contributions from both the sector containing the
receptor and from the nearest adjacent sector. The linear interpolation
term is given by
K(c-y)/c,
where
y = crosswind distance between the receptor and the sector
centerline
c = sector width at the receptor location = 2x tan (11.25 ) -
0.398x
K = a constant to be determined
-------
The geometry of the interpolation is illustrated in Figure 2. Note
that an SSW wind affects the receptor to the NNE of the source.
The linear interpolation is applied to the crosswind integrated form
of the ground level plume equation, and the constant K is determined by
the condition that, when the frequency of occurrence of wind direction is
the same in all 16 sectors, the long time average concentration at any
given downwind distance away from the source must be the same, regardless
of direction. Therefore,
and the time-averaged equation used in the computations becomes
exp [-1/2 (^ ) ]. l }
U 'c" ' ' 2 ' ' '
It should be noted that Eq. (3) is different from the analogous
equation in AQDM by a factor of '(2irx/16)/c w 0.99.
Vertical diffusion of the plume is inhibited by the existence of a
stable atmospheric layer having a base lower than the effective stack height
(the layer elevation will generally range between 100 and 3,000 feet). The
rate of vertical mixing is greatly reduced in such a layer, and the base
of the layer can thus be considered as an effective "lid" on vertical trans-
port of pollutants.. , ...,.-
When an elevated stable layer occurs locally, the estimated pollutant
concentrations can be calculated with the assumption that all the effluent
remains within a mixing layer depth D defined as the vertical distance from
the ground to the base of the stable layer. For the mode.l calculations,
10
-------
North
East
Figure 2 Geometry of the.Linear Interpolation
Between Adjacent Wind Direction Rectors
11
-------
a is considered to increase in the downwind direction until it reaches a
distance x,, at which a = 0.47 D. At this distance, pollutant concentration
at the base of the stable layer will be one-tenth that at the plume center-
line. Up to this distance, the gaussian vertical distribution is assumed,
and Eq. (3) is appropriate. At distance x the trapping effect of the
elevated stable layer begins to be effective, and uniform mixing below
the base of the stable layer is assumed to occur at downwind distance 2x_,.
For distances x >_ 2x_, the average concentration is calculated with the
assumption of full mixing in the mixing layer:
v _ Q (c-y) - ' ' . (4)
A ~~ n >
Due
For distances between XT and 2x , x is determined by a, linear inter-
polation between Eq. (3), evaluated at 'x and Eq. (4), evaluated at
For a specific source-receptor configuration, an estimate of x is
obtained by choosing a representative wind speed for each wind speed class
and solving the appropriate equation for all wind speed and stability
classes appropriate for the time period in the geographical area of interest.
The average concentration, x", is obtained by summing all concentrations
and weighting each one according to its frequency for the particular wind
direction, wind speed class, and stability class. The average concentration
is
X = 2 Z I F(J,K,L) X (J.K.L) (5)
K L J .
where
F(J,K,L)= normalized frequency during the period of interest
for wind direction interval K, wind speed class J, and
/
stability class L
12
-------
X(J,K,L) = average ground-level concentration calculated from
Eq. (3) or (4) as appropriate
The total concentration at a specific, receptor is obtained by summing
the results of Eq. (5) over all sources. For each of the 16 wind direction
intervals, wind speed is defined in 6 categories and stability class in 5
categories. Thus a three-dimensional array of 480 categories is established.
(However, only a few wind directions result in nonzero contributions for
specific source-receptor pairs. Thus the computation time is reduced signi-
ficantly.) Vertical variations in wind direction are not accounted for in
the calculation of x« (However, the ERT/MARTIK model does permit vertical
variability of wind speed in both the calculation of plume rise and of
average transport rates.)
The representative speeds adopted for the six climatological wind
speed categories (0-3, 4-6, 7-10, 11-16, 17-21 and >21 knots), are
0.67, 2.46, 4.47, 6.93, 9.61 and 12.52 meters per second. (A modification
of the 0.67 value to 0.89 was later made for purposes of model validation.
This subject is discussed later in this report.)
The five stability categories (L= 1,2,3,4 and 5, in order of increasing
atmospheric stability) are derived from surface level meteorological obser-
vations, as defined by Turner (1964). Stability in the lowest part of the
atmosphere is determined primarily by the net heating (or cooling) of the
13
-------
ground surface, and by local wind speed. Turner's classification is based
upon surface wind speed, cloud cover and ceiling, supplemented by solar ele-
vation data (latitude, time of day, and time of year); thus the stability
estimates can be obtained for any Weather Bureau station at which continuous
surface level observations have been made.
The values of a (x,L) used in the program are those of McElroy and
Pooler (1968). For computation these are represented in the power law form
xd,
where
x is the downwind distance in meters
and
k and d are functions of the Stability Class L as given in Table 1.
TABLE 1
L
k
d
1
.072
1.22
2
.072
1.22
3
.169
1.01
4
1.07
.682
5
1.01
.554
14
-------
The miriini^m v&lue of- ^ used in calculating q_ is 100 meters. When
x, < lOO/meters, the program sets x, =? 1,00 meters, prioi; to the o calculation.
1 ' ... Z
This simulates initia.} mixing, of the effluent by. perturbed wind flow in the
vicinity-of the source, and* it eliminates £he .possibility of unrealistically
large values of con9pntr:>.'it:ir.n. corresponding to npar-zcro values of o .
The mixing layer depth, I), varies greatly from season to season, day
to d\ Since it i.s impractical to account for all these
variations! a procedure reflecting only major changes is used ir) the model.
The procedure determines an effective mixing depth hy modifying the average
aftern.oon mix|n,g dep£h valuers, as t'atuijnted. hy Hpl^worth (}964), according
to the stability class be.fn.a considered,. Stability classes L - 1,2, and 3
are afternoon conditions, with L =* 1- corresponding to very unstable condi-
tions. When I- w lh the va|ue of 1,1 i,s asstped ^o be 50?n greater than the
climatologies! value, tabulated by Holzwflrj:h;. when I = 2 or 3,, the cjimatolp-
gical ya,lja.e $$, adopted1. Aecording to Turnery's critteria^ L * 5 c.an pcfur
only when nighttime ground-based, inversion conditions exist. Since a
shallow layer of neutral or weak lapse conditions has b^en found to occur
over urban areas (pv.en with strong nocturnal surface inversions in the
surrounding rural .areas), a mixing d.epth af D =s. ]00 meters is adopted for
stability cl'ass L ~ 5,. wt|en this class is indicated b,y the Jurner selection
rules. The- :1.00rrme:ter yaltje \s. based upon observations of Clarke (1969).
Stability clas.s L «= 4 is a neutral stability condition which occurs either
with high wind, speeds during the day or with high negative net radiation
during the night. The^ mixing depth for stability class 1*4 is taken
to be 80v .o.f the c.l:ini.ato.logi(;al value (class L = 3.) and 20% of the class
L = S value.
15
-------
The effective stack height> appearing in Eq. (3) is defined as the
height of the plume centerline when it becomes horizontal. Thus h = h + Ah,
. i. . . . s
. . . , ., .: ,' f " ..'. ; . ' ..'.. '' -'':'' " ... For shqrt
stacks, and for line and area, sources, .this .correction is rn.o.t--, important ,.
but it is significant for elevated point sources., .For the. rnode.1. studies
reported in this document, the value qf u,* was obtained ...from ,i; ;.. ...;/
u
where
wind speed measured at height z ,
16
-------
z, =? 6.0 meters, which is the height of the standard anemometer...
at Newark Airport
and e = 0.2 . '
^
The plume rise equation is appropriate for the neutral stability condition,
but it must be modified for application over a range of stability conditions.
The following modification is used to allow for a range of from 1.3 Ah for
very unstable conditions to Ah for neutral stability.
h = h + Ah (1.4 - 0.1LD..
The Holland plume rise equation Eq,. (-6) frequently underestimates the
effective height of 'emission; thus its .use often .provides .a slight "safety"
factor. For .some point spurees (e.g.., power plants .with ta.lj stacks), the
effective emission height will be ab,pve the mixing ;ftep'th when the mixing
depth is low. Based on the assumption .that the p}ume wil;l not disperse
downward through the stable layer, these cases are identified and eliminated
from consideration -by the .prpgram. For ,(irea sources an avenge .effective
!
height of emission is estimated, and is ,'^ntered-;as input dat;a,r
2.4 Geometrical Details pf the ERT/MARTIK;;Model
1. Treatment of Area Sources
- - - . . . ( ,
The treatment of area sources is described here firsti so that the
geometry defined for the area source case qan be used in the point and
line cases also. Two types of orientation must b^ consi4ered in the
description of the .area ;sourc^ case. These are,:
a. Geographic orientation: based upon nprth--south and
west directions.
17
-------
b. Receptor centered wind-oriented system: based on the wind
''' ! '; direction categories'reported in climatological data records.
The symbols used in the description of the area source case are defined
:\
in Table 2.
The downwind concentration ;x :(o.,:o.,z.).. at: the .receptor .point '(0,0,2)
in. the, w.ind-orie.nted coordinate system-, is! given;-for .a'paint source- located
at ,(x,,y,H). by the. Giffo:cd-.pasquill :fprmula i(Eq. 4} .<<;:;.: ;.- -.':<..".
In the case of the an area source, the contribution of each element
of area is summed in the integration over area:
r r QACx,y,H) 2 '
xCo,o,z) = / dx dy ,-£ exp \- 1/2 (-£-) 1
. ./ ./, ... ,2ir. a. a u . -^. ; .... .a-. .. -.. ,; . ....'.: .:;.
-..." ''.; .:-. V..:!1..- J ' ..;:,'.....: V 'Z'. ' ... 1-.'>./ v J
. V.. :>.'.; *l ' :;;':','.....': y
y
(8)
' . ..' -. ,:. j - '. jf.v,:. '.'.'! :'-. V :' ''<.'" ,' .
In the present application, a rectangular area source is assumed,
with a'uniform'Source distribution' Q'. See Figure 3 for the geometry
i\
;;.?:,r '. ^/r-.-i ; , ;.: '.--'ri "' ":-'i:> -:.- :.-";-- ,'-1:..' M I ;':::. : ' :''. i
applicable to this discussion. Eq. (8) becomes:
/X2 /max
d*2. a a u FZ(Z'H'X) / dy.«?,.- "V? ^ .
y z ^ ..-':'.i. -..'...''.^'. .y .....'
y
mm
where
2
18
-------
TABLE 2
GLOSSARY OF TERMS FOR AREA SOURCE GEOMETRICAL DESCRIPTION
RH, RV receptor coordinates in horizontal (east-west) and vertical
(north-south)directions, meters (system a)
SH, SV horizontal and vertical coordinates of the source center,
meters (system a)
g, dimension of rectangular area source in horizontal (east-
west) direction, meters (system a)
g~ dimensions of source in vertical (north-south) direction,
meters (system a)
x upwind direction and coordinate, meters (system b)
y cross-wind direction and coordinate, meters
(system b)
x,, x_ projection of maximum and minimum cell extent upon x-axis,
1 i.
meters (system b)
y ->-y parallel lines y = y(x) defining boundaries of cell (system b)
8 wind source direction angle, measured clockwise from vertical
(north) direction, radians
u wind speed, meters/sec
o ,a variances of plume concentration in x,y direction, meters
x y
c,d coefficients for determination of a in meters
Li
z receptor height coordinate, meters
H source height coordinate, meters
X integrated concentration due to source distribution in source,
g/meter
point source emission rate, g/sec
19
-------
TABLE 2 Contd
Q. source emission density for source (assumed uniform),
2
g/meter -sec for area source
V transfer function: concentration due to unit
source distribution (seconds per meter):
X = Q-V
Ax increment in x for numerical summation along down-
wind axis, meters
M number of x increments within source
N number of x increments in maximum .downwind
displacement
20
-------
Crosswind A
Position
Receptor
Point
X2
Upwind
Position
Figure 3 Geometry for Area Source with Dimensions g and g in the Horizontal
Plane, Shown in the Receptor-Centered Wind Oriented Coordinate System
o
r-
21
-------
and y and y . , the crosswind integration limits, are functions of the
max mm
integration variable x, and represent the intersection of the midpoint of
the elemental strip with the geometrical boundary of the area source.
Computer realization of Eq. (9) may be efficiently achieved by sum-
mation of elemental strips (see Figure 4) from the limits x to x .
Note that if Xj is negative, the limits are from 0 to x , and that if
x? is less than or equal to zero, the entire source is downwind of the
receptor point, and the integral vanishes.
For numerical evaluation the integration over x in Eq. (9) is
replaced by a summation over a total of M discrete intervals from
x.. to x«. With this substitution Eq. (9) becomes
M
x. = v =
0. u
A
U0)
where
F =
z
2ir a
1
D'
exp
2-,
[-1/2 (fi) ]*«, [-1/2 tfS
ClOa)
x > 2 x,
F is given by a linear interpolation between the two cases
above for x_ < x < 2 x.
D = mixing layer depth
and
x_ = trapping distance, such that a = 0.47 D.
22
-------
Crosswind
Direction
y
4
Receptor_
Point
Upwind
Direction
Figure 4 Integration Over Source Cell Area by Summation
of Elemental Strips from x to x_
N
1-
in
23
-------
In this area source case the y-interpolation function, F in
Eq. (10), is given by the following integral form (see Figure 5),
"iiaA
ly-y'l
F (x) =
y
max
-i f
"e J
rain
dy'
(11)
or
ycf(y ) - ycf(y . ) , (y , y . of same sign)
J v"'viiov/ J w mi r\' * wTnav*'rn-ir\
max'
mm
max
F (x) =
y
~ycf(ymax) + ycf(ym. J , (ymax, ymin of opposite sign)
mm
where
ycf (y) =
y
j
dy = y - %r,
0 < y < c
ycf(y) = 1/2, y > c
In order to avoid insignificant computations, several tests are
provided in the program to cut off the summation indicated in Eq. (10),
or to bypass terms whose contribution lies below a threshold value.
2. Treatment of Point and Line Sources
The concentration x(°,o,z) at the receptor point (0,0,z) in the
wind-oriented system is given for an elemental point source at x,y,H
by the Gifford-Pasquill formula (Eq. 1).
For a line source distribution, the concentration at the point (0,0,2)
is thus the integral along the line of elemental strips of length da
x2,y2
X(o,o,z) =
line
r
J
(12)
Xl'yl
24
-------
Crosswind Direction
Virtual Point Source
Upwind
Direction
Receptor Point
-------
1 ,z-H,2 r 1 sZ+H,2-}-,
- 2 ( } + 6XP [-2 <> ) J}
z
where Q.(x,y,H) becomes the line source emission density in g /m-sec.
lj '
The concentration for constant emission density Q may be written
LI
as :
Q X2'72
X(o,o,z) = -k / V(o,o,z; x,y,H) d£ (13)
with the transfer function V given for time average calculations by
V(o,o,z; x,y,H) = F (x,y) FZ(X,Z,H)
where
c(x) = 0.398 .x
with F2 as given in Eq. (lOa) .
The numerical integration of Eq. (13) is obtained by representing
the line integral as a summation over a set of M virtual points, and
displaced, upwind by an amount which increases the sector width c by
M
QL y
,o,z) = - A£ *-<
X(o,o,z) = - A£ *-< V(o,o,z; x,, y , H)
line k?l . - k k
The modified sector width c' is then given by
c1 = 0.398 x + |Ay| .' , .
26
-------
The geometry for determination of x(o,o,z) due to a point source
at (x,y,H) is shown in Figure 5. For point sources, y is determined
by straightforward solution of Eq. (lOa)
X(o,o,z) = J - Fy(x,y.) Fz(x,z,H)
with q the total source emission rate in grams/sec and F and F as
defined in Eqs. (lOa) and (13a).
2.5 Operation of the Model
A detailed description of the operational instructions for the model,
including specifications for input data and interfacing with the data output
routines, is contained in the User's Manual.
27
-------
3. DATA USED IN THE MODEL ANALYSES
3.1 Introduction
Three principal types of data are required for any atmospheric disper-
sion model analysis. These are:
1. Source Data
Emission data for all pollutants of interest must be available. Five
pollutants were considered in these model projections: S02, CO, nitrogen
oxides, particulates, and hydrocarbons. The emission data log may contain
a combination of point, line and area sources. Information required for
each source includes emission rate, location of the source, and engineering
data necessary to determine plume rise.
2. Transport Data
This generally includes meteorological data and information on ground
topography in the region of interest. The model requires climatological
records indicating the joint frequency of occurrence of wind speed, wind
direction, and atmospheric stability classes, appropriate for the model
region. The influence of topographic features is generally treated by
appropriate modifications to available wind speed and direction data.
In the present analysis, data from Newark Airport were judged to be rep-
resentative of the model region and were used without modification.
3. Concentration Data
Air quality measurement data must be used for validation of the model
calculations. If it were possible to model atmospheric transport and dis-
persion processes with great reliability, it would not be necessary to
incorporate ambient atmosphere monitoring data in the model program.
29
-------
However, current capabilities with models require direct evaluation of
model results to actual measurement data. For the present study an array of
air quality monitoring data gathered by the New Jersey Department of Health
at several locations near the planning region was used as the basis for
model validation.
This section of the report serves three purposes: First, the criteria
for selection of the appropriate data of each type are identified. Second,
the actual choices of data used in the present study are identified. Third,
suggestions concerning other possible uses of data to support the modeling
effort are summarized. Reasons for rejection of these analyses in the
present case are also identified. It is hoped that the discussion of
criteria for data selection, and the suggestions for use of alternative
data types may be useful to other investigators involved in similar modeling
studies dealing with air pollution impact expected in areas under planning
or development.
3.2 Meteorological Data
1. The Uses of the Meteorological Data in the Model Study
Three separate uses of meteorological data arise when a modeling project
is undertaken. These are:
a. Initial Definition of the Influence Region
The influence region for a study area is defined as the geographical
region containing the emission sources responsible for at least 90% of the
ground level concentrations (averaged throughout the study area) of all pol-
lutants considered. Emissions data from the entire influence region will be
included as input data to model calculations.
30
-------
The meteorological data required to determine the influence region is
either a stability wind rose (expressing the joint frequency of occurrence
of wind speed, wind direction and stability class) or a simple wind direc-
tion rose. The most important meteorological information entering the deter-
mination of the influence region is the frequency of occurrence of each wind
direction in the vicinity of the model region.
b. Model Validation
The full set of meteorological data required for model calculations (i.e.,
a stability wind rose) must be available for the time period used for
model validation studies. In the present case validation was carried
out for the 1969-1970 period, and the 1970 meteorological data was used
as input to these model calculations.
c. Model Projections to Future Times
The desired output from the model calculations are estimates of air
quality corresponding to land use patterns expected in 1990. The meteoro-
logical data input to the model should represent stationary values, not the
record of a single year. In this study a 10-year average record of meteoro-
logical data has been used for the projection studies.
2. Criteria for Selection of Meteorological Data
There are three principal criteria for the selection of meteorological
data for model use: representativeness, reliability and completeness. A
final choice* of data will reflect a balance among these three.
31
-------
a. Representativeness
The ERT/MARTIK model, in common with all gaussian plume dispersion
models, normally employs a constant set of meteorological parameters through- .
out the model region. Therefore the climatological data used in the calcu-
lations must be chosen for good representativeness within the region. (It
should be noted that there ore no wind trajectory data available on a clima-
tological basis which might be used in the calculations. Therefore, if the
model were modified to incorporate variable winds and stability within the
planning region, there would be no suitable input data for calculations.)
The requirement of representative meteorological data demands that an
observing site close to or in the planning region be chosen, and that the
site have exposure (relative to local topographical features) which is
typical of conditions within the planning region. When topographical features
change significantly within the planning region, model sensitivity calculations
can be used to estimate the range of error associated with this variability.
b. Reliability
The requirement of reliability in the meteorological data normally
favors the choice of records from official National Weather Service observing
stations. When data from other sources is considered, the questions of
instrument exposure and calibration in the field should be investigated
carefully.
c. Completeness
Records from first order National Weather Service stations will normally
be complete and available. Since a multiple-year average is required for
32
-------
model projections, the availability of such an extended record should be
investigated before any meteorological data from other sources is used.
In summary, the usual choice of meteorological data will be that of
the most representative, long-term record available from a local National
Weather Service observing station. Other sources of data may occasionally
be chosen because of better representativeness within the model region, but
such data must be reviewed carefully for reliability and completeness before
it is selected.
3. Data Selected for Use in the Present Study
After a review of several possibilities (discussed in the following
section) it was determined that the National Weather Service records for
Newark Airport should be used as the meteorological data input in this
study. The Newark Airport location is approximately 5 km from the south-
western corner of the Meadowlands planning region.
For determination of the influence region the Newark wind direction
rose data for 1956-1965 were examined.
For model validation studies, the stability wind rose data (available
as output from the STAR program at the National Climatic Center, Asheville,
North Carolina) for the Calendar Year 1970 were employed.
For model projections to 1990, the STAR program climatological data
for the 10-year period 1956-1965 were used. This period was selected be-
cause the National Weather Service changed its practice at the beginning
of 1966. Prior to that time climatological data for each hour was logged
at Asheville. Since that time data records have been kept on a three-hourly
basis. The 1956-1965 record represents the most recent, 10-year record of
hourly observations available for use in the model projections.
33
-------
4. Other Meteorological Data and Analysis Methods Considered for
Use in the Model Studies
Several other sources of data were investigated for possible use in the
model studies. The most important are summarized here:
a. Secaucus Data
A limited set of meteorological observations (together with air quality
observations) were taken by the U. S. Public Health Service at Secaucus,
within the planning region. These data were collected during most of a
one-year period, beginning in March 1969. While these data have benefit
because of the location of the observations, they were not selected for use
because of the lack of: (1) a complete annual record; (2) a long enough
record for the 1990 projections; and (3) information on calibration and
maintenance of the equipment...
b. New Jersey Air Monitoring System Data
New Jersey has been operating wind equipment as part of its air moni-
toring system at sites in Newark and Bayonne. These data were not selected,
because of the greater reliability of the Newark Airport data.
c. Teterboro Airport Data
Teterboro Airport is at the northern edge of the planning region.
However, it was decided that the more complete record at Newark Airport was
a better choice.
d. Other National Weather Service Stations in New York and New Jersey
The possibility of using data from a more extensive array of National
Weather Service stations was considered. However, it was decided that, to
34
-------
favor procedures readily applicable to other regions, it would be better
to use the single record from the station at Newark Airport.
In addition to the other sources of data described here, the possibility
of alternative analysis techniques was considered. These techniques included:
(1) Use of an average stability wind rose from two or more
National Weather Service stations.
The use of such average data would favor a better represen-
tation of the entire model region, and would reduce the
possible impact of unusual instrument exposures at a single
site. However, it was decided that it is preferable to
favor standard procedures, and therefore to adopt the data
record from the single, most representative station (Newark).
(2) Use of actual mixing layer distributions from JFK upper
level data or from the Environmental Meteorological Support
Unit (EMSU) station at LaGuardia Airport. Standard upper
level data are taken at JFK Airport, and twice daily vertical
soundings have been initiated at the EMSU at LaGuardia Air-
port. These data would be more representative of actual
vertical mixing conditions in the greater New York City
area than the standard distributions of mixing depth adopted
in the MARTIK model. However, for the present study these
data were not used because: (a) It would be very expensive
to,compute an appropriate data set from a long-term record.
35
-------
(b) The model does not treat the relationship between
emission rate and meteorological variables (except for
the basic seasonal differences in emission rates) and this
limitation makes it unrealistic to attempt greater accuracy
in the treatment of mixing depth.
(3) Use of "typical" mesoscale wind trajectories in the planning '
region.
There are no standard approaches available at this time which
would permit the use of wind trajectory patterns, instead of
constant wind estimates. Future model studies may incorporate
such standard trajectories, but this is a development area at
the present time.
In summary, a large number of "nonstandard" meteorological data sources
and analysis techniques were considered for use in the model program. All
were rejected in the present case, in favor of a standard technique, suited
to the limitations of the model, which can be applied readily in similar
studies for other regions.
Two ad-hoc modifications to the meteorological data were considered
during the model validation studies. These were (1) the use of lower
wind speeds to represent additional frictional drag in built-up areas; and
(2) the use of larger nighttime mixing depths, reflecting the urban heat
island effect. These are discussed in the model validation section.
36
-------
3.3 Emission Data
This section discusses the criteria for selection of emission data
used in the model, but not the methodology for development of the emission
estimates. The general description of the development of the emission data
base is contained in the Task 1 report.
1. Criteria for Selection of Emission Data
Five separate criteria for the selection of emission data to be used
in the model calculations are discussed here. The final choice of a data
set will reflect a balance among these criteria.
a. Accuracy
Generally the accuracy of the emission data should be related to
the accuracy desired in the air quality projections. The accuracy of
the model results is usually no greater than that of the input data.
However, there are three other considerations which determine the
accuracy requirements. First, because of the uncertainties in other ele-
ments of the model calculations, there is no benefit resulting from improved
accuracy in the emission data, beyond that which matches the accuracy of
the remainder of the model elements. In the present case (calculations of
seasonal and annual concentration levels in a region with total linear
dimensions of several kilometers) it is not useful to develop area source
emission data for cell sizes of less than one kilometer in linear dimension.
37
-------
In the case of point sources, all sources of smaller size (for example,
having emission rates of less than 100 tons of pollutant per year) will
have no significant influence upon the long time average model calcul-
ations, except in the case of sources which are near (within 1000 meters)
specific receptor points.
Second, the accuracy of any set of emission data is controlled by the
reliability of the basic engineering and planning data. In most cases the
specification of accuracy for the emission estimates is in fact determined
by the availability and accuracy of the basic data.
Third, more detailed emission data should be prepared, when possible,
for the region near the central planning area. Because nearby sources
always account for a major part of the average concentrations observed at
a single site, it is useful to treat the near-field emission data with
optimum accuracy.
b. Completeness
The diffusion model requires a complete set of emission data in the
model region. Therefore the emission data base must include estimates for
all defined locations. Because the basic planning and engineering data may
not be complete, a method for estimating missing data, for substituting
"default parameters" for missing information, must be available.
c. Consistency
When emission data is gathered for a large region, the estimates are
frequently based on a variety of sources. These may include direct
measurements, fuel use data, process output data, materials balance calcu-
38
-------
lations, heating demands, population density, and others. It is important
that data from such widely varied sources be examined carefully for internal
consistency, so that local anomalies in the estimated emission patterns do
not result from the variability in this initial data analysis step.
d. Geometry and Resolution of the Sources
The ERT/MARTIK model is capable of treating, at the same time, a
general array of point and line sources, and rectangular area sources of
any dimension. The basic question to be resolved in the choices among
these source geometries is "How many sources should be treated as specific
point and line sources, and how many should be aggregated with the general
area source data?"
For calculation of annual average concentration patterns the following
guidelines may aid in the choice of the number of point and line sources
selected for direct representation in the model calculations.
For a single model area the total number of point and line
sources treated individually will normally be of the order
of 100 (i.e., between 50 and 200). If more than 200 sources
are treated individually in the model, computation time becomes
large, with little benefit in improved model accuracy. (The
basic accuracy limitations of the model prevent useful reso-
lution of larger numbers of individual sources.)
It is.useful to treat point and line sources in greater detail
in the central planning region, where the most accurate concen-
tration estimates are desired.
39
-------
Smaller sized area sources are also recommended for use in
the central area.
e. Geographical Coverage
Ideally, emission data to he used in a model calculation should be
available for the entire area which contributes significantly to the observed
or estimated concentrations. (When such data arc not available, the model
must permit treatment of "background" concentrations, arising from areas
not included in the model calculation.) The definition and evaluation of an
"influence region" for emissions data is described in the following subsection,
2. Estimation of the Emissions Influence Region
The air pollution concentration observed at a point for a specific
time period is composed of contributions from all nearby upwind sources.
Sources nearest the receptor point will have a much stronger effect than
those further upwind because of wind direction variations in time, dilution
by horizontal and vertical turbulent mixing, and because of possible removal
and transformation processes. Therefore, for uniform accuracy of estimate
of the concentration field, the distribution of sources near the receptor
must be treated in more detail than the distribution further away.
For efficiency in the determination of an emission inventory,for use
in regional modeling, it is necessary to estimate the extent of the surround-
ing area which includes the significant sources of emission affecting the
receptor concentrations. The influence region is generally related to the
local meteorology (for example, skewed toward the prevailing wind direction),
the regional distribution of source strengths, and the pollutant removal
40
-------
or transformation rates. In order to form a practical measure of the
influence region, the region must be defined as that from which some high
percentage (less than 100%) of the observed concentrations arise. In this
study the influence region is defined as the 90% contribution area.
The influence region must be determined separately for each wind
direction sector defined in this study by the upwind 22-1/2 wedge-shaped
area whose emissions can contribute, to the concentration at the receptor.
Observed values of the wind direction are expected to be representative of
the average only over a mesoscale range of up to perhaps 100 km. They
would not be expected to be representative of the average over a synoptic
scale of the order of several-hundred kilometers.
For a continuous distribution of emission density in the upwind sector,
the concentration at the receptor can be expressed as a function of the
influence radius, R, as
R IT/16
X(R) = f /" Q(r,e)-V(r,6)-T(r) rdGdr (14)
J J
O -TT/16
where
2
Q(r,6) is emission rate density (g/m sec)
V(r,6) is the transfer function relating the mean transport
velocity and source strength parameters
and
T(r) is a decay term related to the source-receptor
travel time, but expressible as a function of
distance for a given transport velocity. (The
decay term is frequently expressed as a exponential
which decreases with increasing time).
41
-------
Since Q(r,9) is not known at the beginning, the equations cannot be
solved directly, and it is necessary either to solve for the influence
region boundary iteratively, or to .use equation (14) as a qualitative
guideline for evaluating the boundary. One approach which permits consider-
able insight into the problem of influence region definition is discussed
here. If we begin with an approximation of the near-field emissions (i.e.,
those within the first few kilometers) in the vicinity of the Meadowlands,
we can calculate the approximate contribution to the total concentration
from these sources. Then we can consider the influence from more distant
sources, by assuming (with uniform mixing and no decay) that these sources
will add in inverse proportion to their distance from the Meadowlands. At
larger distances (for example, greater than 50 kilometers) the effects of
decay or removal processes will become important, and will cause the impact
of contributions to fall off even more rapidly. A distance scaling system
can be developed to incorporate these cases and, for the Meadowlands region,
it is possible to make a reasonably good evaluation of the 90% emissions
influence region. We have concluded that, for the planning region itself,
over 90% of the expected concentrations arise from within the 17-county
interstate planning region, including the counties in Connecticut. The
only possible important influence from outside this region is the greater
Philadelphia area, which might contribute several percent to the background
concentrations observed in the planning region. Because any contribution
arising from the Philadelphia area would have a nearly uniform effect
throughout the planning region, it was decided that this contribution would
be treated adequately by the model validation, and that the incorporation of
a Philadelphia source in the model calculations would not improve the accu-
racy or representativeness of the results.
42
-------
3. Emission Data Selected for Use in the Present Study
The emission data selected for use in the model study is defined
^cording to four geographical zones. These zones are defined as:
Zone 1: Meadowlands plan boundary
Zone 2: Outside of Zone 1 to approximately 1 mile beyond the Meadow-
lands plan boundary; defined by town lines except to the
south (Newark and Jersey City, and includes Secaucus).
Zone 3: Outside of Zone 2 to approximately 5 miles beyond the
Meadowlands plan boundary; defined by town lines;
includes Manhattan in the New York part of the region.
Zone 4: Remaining New Jersey and New York counties in the
Abatement Region (1955/1966).
Other: Connecticut counties in the New York - New Jersey-
Connecticut Abatement Region.
Figure 6 Illustrates these four zones.
The selection rules for point sources in the model are:
Zone 1: All sources with rates greater than 100 tons/year
(for any one single pollutant); all stacks considered
to be separate point sources; where stacks have exactly
the same parameters, a stack multiplier notation is used.
Zone 2: Same as Zone 1, except all stacks at each source are
aggregated into one (or, more if more than one pre-
dominant set of stack parameters exist) source using
the major stack parameters.
43
-------
LITCHFIELD
CONN
N E W Y/O R K
JE R^SEY
7
(URLINGTON , / I
Figure 6 The Four Geographical Zones Used in the Development
of the Emission Inventory
44
-------
Zone 3: Same as Zone 2, except much of the data are estimated
according to fue] use, process output and other para-
meters .
Zone 3: Manhattan - Same, except for cutoff at 500 tons/year
The area sources for Manhattan have greater emissions
than other areas; therefore, 100 tons/year is less
significant as a point source.
Zone 4a: Remainder of Bergen, Passaic, F.ssex, and Union counties-
Same except for 500 tons/year cutoff.
Zone 4: Remainder of 17 county region - Same except for
1000 tons/year.
Other: Connecticut counties in 1969 area of Air Quality
Control Region - Same except for 1000 tons/year.
Line sources were included in the model only for major links, as
identified by the New Jersey Department of Transportation. For most of
the model case studies, a total of 50 line source elements were included.
Line sources were included explicitly only in Zones 1 and 2. In the other
regions, all line source data was aggregated into the area source distributions.
The area source data has been organized as follows:
Zones 1 through 3 fand part of Zone 4):
Area cells of 8 x 8 km dimension were used for general calculations.
Sensitivity studies, and detailed studies in the vicinity of the
Meadowlands region used 1x1, 2x2 and 4x4 kilometer cells.
45
-------
Zone 4: (The portion which is greater than 8 km from the planning
region)
Area cells of 16 x 16 kilometer dimensions were used for general
calculations.
Counties in Connecticut:
Area sources were not included. . (Only major point sources were
included from this region.)
A more detailed discussion of the choice of emission data, and a
listing of data used in the model studies, appear in the Task 1 report.
4. Other Emission Data Considered for Use in the Model Study
Two other types of emission data were considered for possible use in
the model study. These are:
a. Sources Near Receptors
In order to investigate the importance of sources very near the
measurement sites (i.e., within the first several hundred meters), a number
of special calculations were performed, incorporating near-field sources.
These investigations were directed in particular toward the question of the
influence of adjacent roads on the measurements of CO, NO,, and hydrocarbons,
It was decided that, for the long time average cases being considered, it
was not necessary (nor possible) to incorporate the influence of emissions
from adjacent, medium-duty roads.
-------
b. Sources Correlated With Meteorological Parameters
Many sources have nonzero correlations with the meteorological para-
meters treated by the model. For example, winter space heating demand is
above average on days with northerly winds. Also, motor vehicle emissions
are greater during the daytime, when mixing depths are greater.
Model accuracy would certainly be improved if the joint frequency
distribution of meteorological parameters and emissions rates were treated
specifically. However, it was considered to be outside the scope of the
present work to accumulate the necessary data base, and to make the required
model modifications, in order to include this improvement. (However, it
should be noted that gross meteorological variability related to winter-
summer differences in emission rates was included specifically in the model
studies.)
3.4 Air Quality Monitoring Data
1. Criteria for Selection of Air Quality Data
Monitoring data must be used to evaluate and refine the performance of
the diffusion model and its input data. In most cases the principal
criteria for selection of monitoring data for model validation is availability:
the total number of monitoring sites in any single region is usually very
limited. The Meadowlands planning region itself dr.es not contain any con-
tinuously operated air quality monitoring site. However, there are a large
number of monitoring stations, of various types, within 50 miles from the
planning region. Also, observations were carried out for a single location
within the planning region, in Secaucus, for slightly less than one year in
1969-70.
47
-------
Several of the criteria for selection of air quality monitoring data
are similar to those for the meterological data, and are mentioned only
briefly here.
a. Representativeness
The question of representativeness is particularly difficult to evaluate
for air quality monitoring data. Mahoney et al (1969), in a study of multiple
station SO- observations in the Boston region, showed that local variability
in concentrations (on a scale of 2 miles or less) is at least as important
as variability on a regional scale of several miles, in multiple-source
regions. Most monitoring stations are placed in areas where high concentra-
tions are expected, and data from all of these stations will not be repre-
sentative of the lower concentrations expected in the areas between observing
stations.
Because of the small amount of monitoring data available, it is usually
not possible to eliminate data from use because of lack of representativeness.
Instead, the available data are normally used, with the understanding that
they frequently reflect near-maximum concentrations in central urban locations,
b. Reliability
The types of monitoring equipment used for all pollutants except particu-
lates have changed rapidly during the past few years. Only equipment of
recent manufacture should be considered of good reliability, particularly
for the recording'of relatively low level concentrations. Instrument reli-
ability is especially important for the examination of long period average
concentrations, because of the possibility of bias errors in the lower
concentration ranges.
48
-------
c. Length of Record
For the present study a full year's record is necessary for model
validation. Longer records are desirable, to permit checking for internal
consistency in the data, anil to examine for trend:, corresponding to known
(or estimated) emission reductions over the past few years.
2. Data Selected for Use
Data sources evaluated during the present study include the following:
a. The New Jersey continuous air mcrdtoring network, including
two major trailer sites (at Newark and Bayonne) nc;:r the planning region
and other satellite monitoring stations near the region. The general loca-
tions of these sites are identified in Figure 7.
b. The New Jersey high-volume .sampler network, including a large
number of sites in the northern part of the state. Locations of these
stations are shown in Figure 8. Observations are made for one day each
week at these sites.
c. The 38 station air quality monitoring network operated by New
York City.
d. Data obtained at Secaucus by the U. S. Public Health Service
between March 1969 and February 1970.
e. Measurements during the summer of 1966 made ir, support of the
New York-New Jersey Interstate Abatement Activity.
f. Other data in southern New York and northern New Jersey avail-
able in the National Acrometric Data Bank.
Several types of data from these sources were employed during trial
calculations, and were examined extensively during initial validation studies,
49
-------
Morristown
Phillipsburg
Somervifle
Jems Grove
\ncora
Paterson
Hackensack
NEWARK TRAILER
Jersey City
BAYONNE TRAILER
Elizabeth
Perth Amboy
Camden
CAMDEN TRAILER
Paulsboro
Asbury Park
Freehold
Toms River
Atlantic City
Figure 7 New Jersey State Bureau of Air Pollution
Control Continuous Air Monitoring Network
50
-------
Orange
Dover-
Morristown
Livingston
Irvington
Newark
Roselle
Linden
Rawray
Somerville
Bound Brook
Rutgers
Edison
Sayreville
Trenton I
Trenton 2
Trenton 3
Trenton 4
Trenton A
Trenton B
Ancora
Paterson
Westwood
Passaic
Hackensack
Fort Lee
Bloomfield
East Orange
Union City
Hoboken
Jersey City
Jersey City - Hudson
Bayonne
Carteret
Woodbridge
Perth Amboy
Red Bank
South Amboy
Asbury Park
Roosevelt
Figure 8 New Jersey High-Volume Sampler Network
51
-------
For specific calibration of the model, it was decided to use data only
of the same type, (from the New Jersey network) and from locations nearest
to the planning region. Therefore, data from five stations in the New
Jersey network (Newark, Bayonne, Jersey City, Hackensack and Paterson)
were chosen. Model calibration parameters were developed for summer,
winter and annual cases/from these five locations.
3. Other data considered for Use
Several combinations of the available monitoring data were in fact
used during the model, validation studies. These are described in the next
section, which discusses the model validation procedures.
52
-------
4. MODEL VALIDATION PROCEDURES AND RESULTS
4.1 Introduction
In this section we discuss the approach adopted to validate the air
pollution dispersion model employed for the Hackensack Meadowlands project.
The primary objective of the validation effort was to assure that the
model adequately predicted concentration values over the time and space
scales of interest and over the range of expected input data values. For
purposes of definition, a distinction should be made between "calibrating"
a model with observed data and "validating" a model with observed data.
The former operation can be as simple as determining "calibration factors"
defined as the ratio of observed values to predicted values or determining a
least-square regression line relating predicted to observed values. If the
calibration factors or regression line slopes have values that are not near
unity, or if intercept values imply a negative background pollution level,
there is a strong implication .that the model is inadequately representing
some important phenomena. It is important to note in this regard that good
correlation (high positive values of the correlation coefficient) is not
necessarily evidence of realistic modeling of phenomena. High correlation
can even be obtained when highly variable model results are compared with
nearly constant observed values (see Figure 9a) or conversely, when nearly
constant model results are compared with highly variable observed values
(see Figure 9b). .
53
-------
B
T3
O)
Q>
(A
JD
O
O)
Predicted
1
./
/
/
./
/
Predicted
Figure 9 Highly Correlated Regression Line Fits
-------
The judgment of model performance thus needs to include consideration
of the intercept values and slope of the least-squares regression line.
Positive intercept values should have a physical interpretation as the
contribution of "background" levels of air quality associated with emissions
from all the other sources not incorporated in the model. Non-unity regres-
sion line slope should have a physical interpretation in terms of systematic
over or under prediction of some input parameters.
Validating a model implies a detailed investigation of the model results
and a comparison of those results with measured values in order to identify
and evaluate discrepancies. If the model results compare well with the
observed data or if, for the applications to be made, simple correction
factors are deemed appropriate, the model may then be simply calibrated with
the observed data. On the other hand, if systematic discrepancies are found,
the investigation may suggest alterations of model parameters or of the
model mechanics which would improve the representativeness of the model.
A final calibration is generally required as the'last stage of the validation
procedure to best adjust for remaining discrepancies between observed and
predicted results.
4.2 Procedures for Validating Models
The procedures for validating models will differ somewhat from applica-
tion to application depending upon the nature and purpose of the study and
depending upon the quality of the available data. The validation procedure
will normally require a thorough study of the implications of model assump-
tions and the performance of "sensitivity" studies for various input para-
meters. The following sections describe the various steps taken to validate
the dispersion model- developed for the Meadowlands study.
55
-------
4.2.1 Selection of Data for Purposes of Validation
Because the planning project is concerned with the prediction of air
quality levels within and near the Meadowlands for winter, summer, and annual
time periods, data for validation was chosen for the same time durations.
Model predictions are expected to be most representative in the near region
of the Meadowlands, where the source inventories developed are most detailed.
For this reason, only data from the five monitoring stations nearest the region
were used for validation purposes. (These are the stations at Bayonne,
Jersey City, Newark, Paterson and Hackensack). See Figure 10.
The emissions inventory for the validation study was based on estimates
for the year 1969. However, among the five stations chosen for validation,
only Newark and Bayonne have data for all pollutants of interest for 1969.
Measurements of SCL, CO and particulates are available for most of 1970 and
the entire year of 1971 at the other three stations. The trends in average
concentration levels from 1969 to 1971 observed at Newark and Bayonne were
used to extrapolate the 1971 measurements at the other three stations back to
1969 levels. This procedure permits the use of a much expanded data base in
the validation but introduces some additional uncertainty. In the final
calibration, because of the additional uncertainty, the calibration factors
from Newark and Bayonne were weighted more heavily than those of the other
stations. Table 5 includes a summary of the air quality data used.
As previously noted, meteorological data for 1970 were used in the
validation studies. The 1970 meteorological data were adopted for use
instead of the 1969 data for the following reasons:
1. Because only two stations (Newark and Bayonne) reported concen-
tration measurements in 1969, all of the initial evaluation
56
-------
HACKENSACK
MEADOWLANDS
DISTRICT
Validation Sites
Figure 10 Validation Sites Surrounding the Meadowlands Region
57
-------
studies and most of the validation studies involved comparison
of model predictions with 1970 air quality measurements, at all
five of the stations mentioned above.
2. The year-to-year variability in the seasonal and annual wind
rose data at Newark Airport is small compared to the variability
in concentrations measurements among the five stations used in
the validation studies. Therefore it was preferable to intro-
duce some uncertainity into the meteorological data, in order
to incorporate all five stations into the validation procedure.
3. Although the measured concentration data for Jersey City, Pater-
son and Hackensack were extrapolated back to 1969, these values
are directly representative of meteorological conditions in
1970 and 1971.
It should be enphasized that, whenever data, from an adequate network of
air quality monitoring stations are available for the time period of the
emission inventory, the meteorological data from the same time period
should be used. The use of differing time basis was required in this
case only because of the lack of sufficient air quality measurement
data for 1969.
4.2.2 Preliminary Runs to Assess the Initial Agreement Between
Predicted and Observed Values
Computer runs were made using the point and line source inventories and
a crude area source inventory to make an initial comparison of the values
predicted by the MARTIK model with observed data. The inventory included
192 point sources, 50 line sources and 35 area sources. The Newark wind
rose and the standard Pasquill-Gifford-Turner stability class dispersion-
58
-------
rate relations were used. These tests indicated that the model generally
overpredicted concentration values of S0_ and NOY. It predicted spatially
4 A
averaged concentrations of CO and hydrocarbons well, but the individual data
points did not correlate well with the observed values. Predictions of
particulate levels were in good agreement with the observed averages.
4.2.3 Sensitivity Analyses and Identification of Possible Model
Improvements
A number of model mechanisms which could account for discrepancies
between the calculated and observed values were investigated. These mechanisms
are listed and discussed here. Sensitivity tests were made to investigate the
effect of some of the model modifications above on the receptor points chosen
for validation. Results for these arc also presented below.
1. Inclusion of Pollutant Half Lives
The gross effects of removal processes for gaseous pollutants can
be simulated by the inclusion of an exponential time decay term in any
one of Equations 2,3, or 4. The effect of incorporating a half-life of
concentration values to simulate removal processes was investigated
specifically for the S0_ emissions. Test runs were made with a three-
hour half-life and with a one-hour half-life. The three-hour half-
life caused a 47% reduction of the calculated values while the one-hour
half-life caused an 85% reduction.
59
-------
2. Refinement of Area Source Grid and Inclusion of Nearby Roadways
Minor sources in the immediate vicinity of the sampling locations will
have a strong influence on the sampling locations. In order to improve the
representation, the area source grid was refined into smaller source grids
in the neighborhood of the receptors and roadways adjacent to the receptors
were included as line sources.
Nine 16-km area source cells in the neighborhood of the Meadowlands
were subdivided into 36 8-km squares and new emission rates were calculated
for the new cells. The average levels of concentration were not changed
by this alteration, but correlation between observed and computed was
somewhat improved. Further refinement of the area source inventory was
not thought to be justified using the county-based emission data on area-
type emissions.
The effect of adding short line-source segments in the near field of
the receptors at Bayonne, Jersey City and Newark was investigated to repre-
sent the effect on concentrations of small but nearby roadways. The local
contributions increased the concentrations by a maximum of 5%. As a result
of this relatively small influence and the inability to predict the level of
detail for the year 1990, it was decided not to incorporate the additional
roadways into the final emission inventory.
3. Incorporation of the Effects of Correlation Between Diurnal
Meteorological Variations and Pollutant Emission Rates
Emission rates of transportation and industrial related pollutants
are lower during nighttime hours. Because the atmosphere is on the average
60
-------
more stable at night, models which do not include diurnal variations in
source strength tend to overpredict seasonal and annual average pollutant
levels. The effect of diurnal variations in source strength combined with
a systematic association of stable atmospheric conditions with nighttime
was investigated by postulating a difference between day and night emis-
sion rates and analyzing the concentration contributions by stability
class. By definition, occurrences of stability class 5 are only associ-
ated with nighttime conditions.
The frequency of occurrence of stability class 4 includes the wind
speed and direction statistics of the remaining nights. If we assume that
the nighttime hours are one half of all hours, and that the nighttime
emission rate is A times the daytime rate, where A is expected to be
0 < A < 1, an estimate of the effects of the diurnal variation of
emission rate on long-term averaged concentrations is given by
where
X is the concentration calculated assuming that there is no
diurnal variability
f. are the frequencies of occurrence of stability class i
and
r and rM are the fractional times of day and night occurrence
in stability class 4.
Values of f . , r,, and r for winter, summer and annual time periods are
given below in Table 3.
61
-------
TABLE 3
VALUES OF £., TD and r FOR WINTER, SUMMER AND ANNUAL TIME PERIODS
Season
Winter
Summer
Annual
fl + f2 * f3
.041
.243
.130
f4
.703
.489
.617
fs
.256
.268
.253
rD
.655
.525
.600
rN
.345
.475
.400
Values of x were computed for values of A between 0 and 1 for S0« and CO and the
results compared with the measured values. Although overall reductions of
computed values could be as large as 40%, no improvement of correlation between
observed and computed was found. For this reason and because of the uncer-
tainty in assigning values to A a decision was made to not include this
correction mechanism in the model. Since observed correlation was not
altered, a simple scaling in the final calibration seemed to have equal
merit in accounting for the diurnal effect.
4. Incorporation of Effects of Correlation Between Wind Direction
and Emission Rates
Winter emission rates associated with space heating are expected to be
positively correlated with the cooler northerly winds. Neglect of this
mechanism causes models to overpredict space heating related pollutant levels.
On the basis of the insensitivity of the model correlation results to the
diurnal variability and recognizing the inability to specify the relation
between wind direction and emission rates, this effect was not formally
investigated and was left to be corrected by the final calibration.
62
-------
5. Modification of Dispersive Spread Statistics
Because of reduced evaporative cooling and increased capture of incoming
solar radiation in urbanized areas, thermal convection from the ground
is more vigorous than in rural areas. In addition, because of their
built-up nature, urban areas are aerodynamically rougher than rural areas.
These two factors cause enhanced mixing of pollutants and increased plume
dispersion spread rates. Unfortunately, data on dispersion rates in urban
areas are not as extensive for rural areas. The study by McElroy and
Pooler (1968) in St. Louis provides some measures of the increased disper-
sion rate expected. The table below compares the McElroy-Pooler vertical
dispersion coefficients after Koch and Thayer (1971) with the Pasquill-
Gifford-Turner values.
TABLE 4
VERTICAL SPREAD STATISTIC CONSTANTS (a = kxd)
Stability
Class (L)
1
2
3
4
5
Pasquill-Gif ford -Turner
k
.022
.064
.150
.270
.372
d
1.44
1.12
.860
.680
.580
McElroy-Pooler
k _j
-
.072
.169
1.07
1.01
d
-
1.22
1.01
.682
.554 |
The net effect on ground-level concentrations of utilizing the McElroy-Pooler
coefficients depends upon the mixing depth and the height of the source
emissions. For stability classes 4 and 5 and for low-level area and line
source emissions, the McElroy-Pooler a values will cause concentration
contributions in the near field to be nearly 1/3 the values obtained using
63
-------
the Pasquill-Gifford-Turner values. On the other hand, for upper-level
sources, ground level contributions will increase because of the increased
vertical mixing rate. At some distance downwind (of the order of 10 km for
the prevailing meteorological conditions), material is expected to be mixed
nearly uniformly from ground-level to the top of the mixing layer. Beyond
that distance the choice of the vertical spread rate statistic does not
alter concentrations. The net effect on total concentration will therefore
depend strongly on the spatial distribution of the sources. A comparison
of runs made with both sets of dispersion coefficients for annual cases
showed that the net result of utilizing the McElroy-Pooler coefficients
was to reduce the average predicted concentration for the five validation
receptors by about 40% for S0_ emissions and by about 30% for CO emissions.
6. Modification of Assumed Mixing Depths
For the same reasons as in item 5 above, the mean mixing depth for urban
areas is expected to be larger than that assumed for rural areas. The dif-
ference is expected to be largest and to influence ground-level concentration
the most under nominally stable atmospheric conditions.
Sensitivity test runs were made for assumed mixing depths of 100, 200
and 300 meters for stability Class 5. Changing the average mixing depth
associated with the most stable stability class, from 100 to 200 meters reduced
concentrations at the five receptors on the average by about 7%. This rela-
tively low sensitivity is caused by the fact that major sources contributing
to the concentrations are in the near field of the receptor and within the
distance that the finite mixing depths act as a lid to vertical diffusion of
pollutants. Increasing the height from 200 to 300 had hardly any effect on
the five receptors.
64
-------
7. Modification of Airport Velocity Measurements
Several possible modifications to the airport wind speed measurements
were investigated and are discussed below:
a. The low level measurements of wind velocity at Newark airport
are on the average lower than wind speeds existing at higher levels. This
will have two (partially compensating) effects on ground-level concentrations,
The use of the surface level winds in the plume rise equations for stack
emissions causes an overestimate of plume rise and thus contributes to an
underestimate of ground-level contributions. On the other hand, the use of
the low-level wind speed underestimates the average transport rate for pol-
lutants from elevated sources and thus contributes to an overestimate of
ground-level concentrations. Concentrations resulting from low-level emis-
sions from area sources or roadway line sources are not expected to be
altered significantly by the variation of wind speed with height.
To improve the model representation of these phenomena, power law
increases of velocity with height were postulated. The velocity calculated
for use in the computation of plume rise was
u = U
where
u was the value measured at the anemometer height z,,
and the exponent e is a function of stability class.
Velocities for use in the computation of a "ventilation" velocity were
calculated from the equation
(h/z )e
u = u i (17)
1-t-e
65
-------
where
h is the effective stack height.
This formula estimates the mean velocity averaged iron < :. und level to the
plume height.
b. The wind speed frequency records ci a?;, i i y : he .lowest speed
cases (calm) as values between 0 and 3 knots. The avenge wind speed which
really occurs during "calm" conditions (velocities too .Low to be reliably
read by the anemometer) is likely to be skewed toward the 3-l.nct limit.
Instead of utilizing the average of the windspeed class limits (i.e., 1.5
knots) to characterise the wind speed of this lowest class we chose the
slightly higher value of 1.75 knots or .89 m sec . The net effect on long-
term average concentrations is quite small because of the small difference
and the small portion of the time that calm conditions are observed.
c. The possibility of simulating an effect of the increased
"roughness" of urban areas by reducing wind speeds in the lower boundary
layer was considered. Because the model was tending to generally over-
predict, and because this type of modification would cause the model to
overpredict even further, it was not investigated in specific detail. We
note that the incorporation of the power law variation of velocity with
height in the calculation of a ventilation velocity does cause the average
velocity for low level emissions to be lower than the velocity at emission
height by the factor of l/(l+e).
66
-------
4.2.4 Modifications Made to the Model and the Final Calibration
Model Results
There were three fundamental criteria used to determine which modifica-
tions or parameter changes actually to incorporate.
1. Modifications or parameter charges must be-physically realistic
which requires that parameters chosen must be within the range of experi-
mentally measured values.
2. Modifications or parameter changes must be appropriate for use in
the planning year 1990. This criterion restricted the incorporation of
emission rate dependent modifications such as items 2, 3 or 4 in Section 4.2.3.
3. The evaluation of modifications or parameter changes must be based
primarily on their effect on the regionally averaged agreement of calculated
vs. observed values, rather than on agreement for individual stations. We
expect that measurements at the individual monitoring stations will be
strongly dependent upon the fine details of the location and operating modes
of the nearest (though perhaps small) sources. Because of the nature of the
applications of the dispersion model in this study, it is more important
that the model predict concentrations accurately over a regional scale rather
than accurately predict concentrations at specific monitoring sites. By
averaging measured values at a number of sites within a region, the effect
of local sources will be minimized, so that a meaningful'regional average
can be formed.
On the basis of these criteria, only changes in the meterological
parameters were finally adopted in the model. The emission rate dependent
modifications were not appropriate for the seasonal and annual average
calculations, and their effects were left to be corrected by the final
calibration.
67
-------
Sensitivity analyses performed on the meteorological parameters were studied
from the points of view of seasonal variations, differences of influence of
near and far sources, and effects on dispersions fro- lo'-- and high level
sources. The final decisions on specific changes to 'ie implemented were
based on our best judgment of the results of the se;;si' \-ii:y tests and
experience in other mode] studies, 've were purposeful; conservative in
incorporating model changes. The modifications made arc- listed here:
1. A half-life of 5 hours was assigned to S09 emissions. An estab-
lished procedure for simulating removal processes in large-scale dispersion
studies does not exist. The problem is complicated by the dependence of
removal processes on precipitation, atmospheric interactions with other
pollutants, and reactions with buildings and other surfaces. To simulate
average effects of removal processes for SO., a .three-hour half-life is a
reasonable number.
2. Vertical spread statistics were adopted from the McIUroy-Pooler
study in St. Louis. These values better simulate the increased turbulance
expected in urbanized areas.
3. Velocities for use in the computation of plume rise were computed
from Eq. (16) using the Newark Airport anemometer height of 2=6 meters
and e = 0.2. Velocities for use in the computation of a ventilation velocity
were calculated using Eq. (17) with the same values of z and e.
The value of e = 2 was chosen for purposes of validation as a value
representative of the most commonly occurring stability classes..4 and S.
68
-------
The value of .89 m sec was chosen as representative of the wind speed
in the lowest wind speed class and incorporated into the final model vali-
dation runs.
4.2.5 Final Calibration of the Model
With the parameter modifications implemented in tiu1 model, final vali-
dation runs were made for the five receptors for the writer, summer and
annual cases. The results were then compared with the observed or extra-
polated-observed air quality data for 1969. Table 5 summarizes the observed
and predicted data. The ratio of the predicted to ''observed" was then calcu-
lated for each station, for each season, and for each pollutant for which data
was measured. For the sites used in the validation, we recommend the use of
these ratios as simple calibration factors. For estimation of concentrations
at any other location and, specifically, in the planning region, an average
of the calibration factors for the entire region was derived. As noted in
Section 4.2.1, the validation air quality data for SO- and CO for the
Jersey City, Paterson and Ilackensack sites was extrapolated backwards to
1969 from the 1971 values. In order to discount this added uncertainty in
determining a regional calibration factor, the Bayonne and Newark data were
weighted more heavily. The weighting factors were determined simply by
requiring that the average of predicted and observed values for Bayonne
and Newark sites be weighted equally with the average for the Bayonne, New-
ark, Jersey City, Paterson and Hackensack sites. Thus, for both predicted
and observed values the regional average values foi- SO,, and CO were deter-
mined by the formula
p i A 1 /B * VN VR + VN + VJ + VP + \
Regional Average = [ -^ + p J
69
-------
TAB Lii 5
PREDICTS AND OBSERVED VALIDATION DATA
WINTER !
|
POLLUTANT i Bn>'°;;nC
;j /;._. ;,v.t,..j ! Cbv.'ivei!
SO, (ppnO i .13"! j .067
'- 1 1 '
HC (ppra) i 1.31 j 2.03
NOX (ppm) .257 j .097
"('articulates (yg/m ) \ 132. / | 77.6
'i 1
SO (iT:-0 ; .041 i .042
^ !
CO (pp-n) ; 1.7J j .97
HC (pprcl i l.rC j 1.S7
i i
NO. fpp;n) .157 i -OSS
s ; i
"Partiu.lstos ing/rr,' > . 7J :) 09.-:
i |
SO, ipi'M : .C7>/ j .032
CO (r,p.ii !i .:.': J 1.28
nc (pp--) j 1.46 I j..y::
N'O... (p;>;;0 ;' . !94 i .067
A. ' _ ' ; i
"Particuiat.es Cyg/"'-') i S<:-.i i 3.-'.8
: I
.lor-.-
rr,:0:.:tvJ
. JO7
2 , S 3
f';T
' ' *-
179.9
.0,0
4.S5
l.A.1
. 296
.13(1
2 . '" 2
,' ("sty i N'.-v.:irK !'A*!'rs;i;; j Ifackonsack
!':u-r\^J* ' . ,-.-J. ,-. i-;J ;U Hc:'t-;:..i ;"iej-.'-.; i ->hscrvf?ci* , i.rt..j ; ,. r . Obs'M'vcJ'* i
i ! : .!'".."' ':' !
.091 .n:1^ : .OMV .OS'S i .0)1 i .121 | .048
N A. ' .5^0 : ;.'.i? .SJ'O ' 'C. A. i .7:53 N. A.
; ! ! !
v. A, ,i£r- ! .;-. .:,-! ' r. . A. : .c.s-1 >;. A.
! ' ;
112.0 i UJ.O ; N. A. 101.?, i v,. A. 1 130.1 175.0
.(is: ' .c:.: ..MS .0:; ! .01.3 ! .033 i .012
i 1 '
S.f. ! i.70 3..',3 1.7.5 1 3.07 ; 2.->7, \ 1.27
i '
N. A. j .525 1 .:-.v .^9 j N. ' . ' .790 N. A.
N. A. j .iJ8 .:^ .U'J j N. A. 1 .1-7 N. A.
i i , j
.or-',, i .Oi-o .or; i-.:o .a::' 1 .075 ! ,o/s
! 1
7.s i ; .'.LI 4. i1"7 , .(-,' i ' .-' j - -f'^ -----1
.. i
11?. 0 ! b.l.O 136. ; 84.2 . > '. "' 7 : 142.9
L_ 1 i
'Extrapolated backwards fron 1970 ancl 1971 value? to I'JO1.
values (except for particulates).
**0bserved values of particulates are 1970 values.
Observed values for Sayonne and Newark are 1969 values.
-------
where V denotes the value under consideration and the subscript,
the site location, '['able 6 summarizes the calibration i'a.Mors
calculated.
4.3 Discussion
The final calibration factors for the ilackensafk : .-.;; on given in
Table 0 show that the model underpredi cts CO and hydr.v:: rbons , and over-
predicts SO., and NO for all averaging times. rart.icui.ites arc undorprcdicted
in the summer and in the annual average, hut ovorprcdicted in the winter.
The overall agreement is thought to be quite satisfactory and well within
the accuracy typically found Kith long averaging-tine period projections.
It should he pointed out that although considerable smoothing of data
is inherent in both long-time averaged predicted values and observed values,
the agreement between predicted and observed values is not. necessarily
improved by the averaging processes. Tor example, systematic associations
of emission rates with different meteorological conditions will cause syste-
matic disagreements between observed and predicted values over long averag-
ing times. On the other hand, agreement is expected to be better when
short-time period observations of air quality (such as over several hours)
are compared with model projections using observed values of meteorological
conditions (and possibly observed values of emission rates) for input. Thus,
when strong diurnal effects on concentration patterns arc observed, hourly
values are expected to he well correlated with model -results which incorpor-
ate nighttime and daytime variations in emission rate. .Since the model
developed in this study is to be used primarily for the longer averaging
time applications, it needs to be validated with corresponding long-time
averaged data sets.
71
-------
TABLE 6
RATIOS OF OBSERVED TO PREDICTED VALUES, BY POLLUTANT AND TIME PERIOD
c
JH
i ~ '' " 1 Q
J, * \.) *..»..'-' ( ~
; ... .._ _.. _ . _. i
*"0bserved" values for Jersey City, Paterson and Hackensack were extrapolated backward from 1971
measurements to provide estimates for the emission inventory year 1969.
-------
There are several reasons why, for the final calibrations, the deter-
mination of a regional-average calibration factor for each pollutant and
for each season was most appropriate. These reasons are discussed below.
1. By averaging over the five stations around the Hackensack Meadow-
lands, errors introduced by the presence or absence of numerous smal] local
emission sources tend to cancel.
2. The number of data points did not warrant a least-squares regres-
sion line fit (note that annual values aie not independent of the summer
and winter values and cannot be used as independent data points).
3. Because the emissions inventory influence region accounts for
nearly all of the concentration observed at a point, and since large changes
in emission rates are postulated for the year 1990, a forced fit of the
calibration lines through zero seemed mandatory; (i.e., a line with a
slope equal to the calibration factor and with a zero intercept value.)
Only in this way can we be assured that an ncross-thc-board reduction of
all emission rates will result in a tori' \pondinp reduction of predicted
concentrations in the same fraction.
4. The limited amount of data available did not warrant the generation
of a spatially varying calibration factor over the region. A regionally
varying calibration factor would tend to Has the impact results of alter-
native land use plans spatially throughout the region. Another level of
study effort with more air quality measurements and emission inventory
detail would be required in order to determine physical explanations for and
confidence in other than a regionally averaged calibration factor.
73
-------
REFERENCES
Clarke, T. F., 1969: "Nocturnal Urban Boundary Layer ove.'- Cincinnati, Ohio."
Mon. Wea. Rev. 97_, pp. 582-589.
Gifford, F. A., 1961: "Use of Routine Meteorological 01' nervations for
Estimating Atmospheric Dispersion, Nuclear Safety, 2(4), pp. 47-51.
Holzworth, G. C., 1964: "Estimates of Mean Maximum Mixing Depths in the
Contiguous United States," Mon. Wea. Rev. 92, pp. 235-242.
Koch, R. C., and S. D. Thayer, 1971: Validation and Sensitivity Analysis of
the Gaussian Plume Multiple-Source Urban Diffusion Model, GEOMET Report
No. EF-60, Rockville, Maryland.
McElroy, T. L., and F. Pooler, 1968: St. Louis Dispersion Study, Vol. II -
Analysis, NAPCA Publication No. AP-53.
Mahoney, J. R., W. 0. Maddaus and J. C. Goodrich, 1969: "Analysis of
Multiple-Station Urban Air Sampling Data,"Proc. of Symposium on
Multiple-Source Urban Diffusion Models, APCO Publication, No. AP-86.
Martin, D. 0. and T. A. Tikvart, 1968: A General Atmospheric Diffusion Model
for Estimating the Effects of One or More Sources on Air Quality,
presented at the 61st Annual Meeting of the APCA, St. Paul, Minn.
NAPCA, 1969: Air Quality Display Model, National Air Pollution Control
Administration, Washington, D. C,
Pasquill, F., 1961: "The Estimation of the Dispersion of Windborne Material,"
Meteorological Magazine, 90^1063, pp. 33-49.
Turner, D. B., 1969: Workbook of Atmospheric Dispersion Estimates, PHS.
publication No. AP-26.
Turner, D. B., 1964: "A Diffusion Model for an Urban Area, J. of Appl.
Meteor., 3(83).
75
-------
GLOSSARY
Activity, Activity Level - basic land use and transportat:on planning
units of intensity of use - vehicles per day on a highway, acres
of residential land use, square feet of industrial Jant space.
Activity Index - a numerical conversion factor to transform the level of
activity specif.i.ed for a land use category into demand for fuel for
heating purposes.
Air Quality Contour - a contour line in a plane (usually the horizontal
or vertical) representing points of equal concentrations for a specified
air pollutant.
Air Quality Criteria - factors used in this study that represent a basis
for decision-making, for example ambient air quality standards.
Air Quality Prediction - the calculation of current or future air pollutant
concentrations at specified receptor points resulting from the action
of meteorological conditions on source emissions.
Albedo - the fraction of solar radiation reflected from the ground surface.
Ambient Air - that portion of the atmosphere, external to buildings, to
which the general public has access.
Ambient Air Quality - concentration levels in ambient air for a specified
pollutant and a specified averaging time period within a given geographic
region.
Ambient Air Quality Standard - a level of air quality established by federal
or state agencies which is to be achieved and maintained; primary
standards are those judged necessary, with an adequate margin of
safety, to protect the public health; secondary standards are those
judged necessary to protect the public welfare from any known or
anticipated adverse effects of a pollutant.
76
-------
AQUIP - an acronym for Air Quality for Urban and Jndustri :1 planning,
a computer-based tool for incorporating air pollution considerations
into the land use and transportation planning process.
Atmospheric Boundary Layer - the lower region of the .m :<-;phere (to
altitudes of 1 to 2 km) where meteorological cor;d". I ens are strongly
influenced by the ground surface features.
Atmospheric Dispersion Model - a mathematical procedure for calculating
air pollution concentrations that result from a :<: ecif led array of
emission sources and a specified set of meteorological conditions.
Average Receptor Exposure - a measure of the average impact of air quality
levels on specific receptors; the measure is based on the integrated
receptor exposure divided by the total number of receptors in the
study region.
Background Air Quality - levels of pollutant concentrations within a study
region which are the result of emissions from all other squrces not
incorporated in the model for the study region.
Background Emissions - the emissions inventory applicable to the background
region; that is, all emission sources not explicitly included in the
inventory for the study region.
Climatology - the study of long term weather as represented by statistical
records of parameters such as winds, temperature, cloud cover, rainfall,
and humidity which determine the characteristic climate of a region;
climatology is distinguished from meteorology in that it is primarily
concerned with average, not actual, weather conditions.
Concentrations - a measure of the average density of pollutants usually
specified in terms of pollutant weight per unit (typically in units
of micrograms per cubic meter), or in terms of relative volume of pollutant
per unit volume of air (typically in units of parts per million).
-------
Default Parameters - values associated with a parameter for a category of
activities (such as heavy manufacturing) assigned to the activity para-
meter for a subcategory of activities (such as electrical machinery
production) when the actual value for the subcatogory is not known.
Degree Days - the number of degrees the average temperature is below 65°
each day; used to determine demand for fuel for heating purposes.
v
Effective Stack Height - the height of the plume center-line when it be-
comes horizontal.
Emission Factor - a numerical conversion factor applied to fuel use and
process rates to determine emissions and emission rates.
Emissions - effluents into the atmosphere, usually specified in terms of
weight per unit time for a given pollutant from a given source.
Emissions Inventory - a data set describing the location and source strength
of air pollution emissions within a geographical region.
Emissions Projection - the quantitative estimate of emissions for a specified
source and a specified future time.
Equivalent Ambient Air Quality Standards - air quality levels adopted in
this study to permit analysis of all air pollutants in terms of annual
averages; in cases where state and federal annual standards do not exist,
the adopted levels are based on the extrapolation of short period stan-
dards.
Fuel Related Sources, Fuel Emissions - fuel related sources use fuel to heat
area, or to raise a product to a certain temperature during an industrial
process, or for cooking in the house; they produce fuel emissions.
(See also Non-Fuel Related Sources.)
78
-------
Fuel Use Propensity, Fuel Demand - the total heat requirement (space
heating plus process heating) determines the fuel demand; the propensity
to use a particular fuel or fuels determines the actual amounts of various
fuels used to satisfy the heat requirement.
Heating Requirements - the demand for fuel is specified in terms of the
heating requirements:
space heating - the fuel used to heat area, such as the floor space
of a school in the winter, is that required for space heating; the
heat content or value of that fuel defines the space heating re-
quirement (BTUs, British Thermal Units of heating content).
non-space heating, process heating - the fuel used to raise a pro-
duct to a certain temperature during an industrial process or for
cooking (with gas) in the home is that required for process heating
or non-space heating. It is generally not related to outside tempera-
ture whereas space heating requirements are.
percent space heating, percent process heating - the relative pro-
portion of a fuel or its heat content that is used for space heating
or process heating defines.respectively, the percent space heating
or percent process heating.
Impact Measure (or Parameter) - a quantitative representation of the degree
of impact on air quality or specific receptors resulting from concentrations
of specified pollutants.
Influence Region - the influence region for a study area is the geographical
region containing the emission sources responsible for at least 90% of
the ground level concentrations (averaged throughout the study area) of
all pollutants considered.
79
-------
Integrated Receptor Exposure - a measure of the total impact of air quality
levels on specific receptors; the measure is based on the summation
within the study region of the number of receptors times the concentration
levels to which they are exposed.
Inventories - the aggregation of ail fuel and process c.-issions sources is
called the emissions inventory; the components for use with the model:
current inventory - all sources for 1969
background inventory - all sources for 1990 not directly related
to the meadowlands plans.
plan inventories - all sources for 1990 related to the Meadowlands
plans; this excludes any source outside the Meadowlands boundary
and also excludes existing major single sources and the highway
network.
Isopleth - the locus of points of equal value in a multidimensional space.
Land Use Intensity - the level of activity associated with a given land use
category, for example the population density of residential areas.
Land Use Mix - the percent of total study region area allocated to specific
land use categories.
Meteorology - the study of atmospheric motions and phenomena.
Microscale Air Quality - the representation of air quality in a geographical
scale characterized by distances between source and receptor ranging
from a few meters to a few tens of meters.
Mixing Depth - the vertical distance from the ground to the base of a stable
atmospheric layer (also called inversion height).
Model Calibration - the process of correlating model projections with observed
(measurements) data, usually to determine calibration factors relating
predicted to observed values for each pollutant.
80
-------
Model Validation - the detailed investigation of model results by comparison
with measured values to identify systematic discrepancies that may be
conected by alterations of model parameters or model mechanics.
Non-Fuel Related Sources, Process Emissions, Separate Process Emissions -
non-fuel related sources do not burn fuel primarily for heating purposes
or do not burn fuel at all; these include transportation sources, in-
cineration, and certain industrial processes; they produce process or
separate process emissions. (See also Fuel Related Sources.)
Ranking Index - a quantitative representation of the net impact on air
quality or specific receptors resulting from all pollutants being con-
sidered.
Receptor - a physical object which is exposed to air pollution concentrations;
objects may be animate or inanimate, and may be arbitrarily defined in
terms of size, numbers, and degree of specificity of the object.
Receptor Point - a geographical point at which air pollution concentrations
are measured or predicted.
Regional Air Quality - the representation of air quality in a geographical
scale characterized by large areas, for example, on the order of 50
square kilometers or greater.
Schedule - number of hours per year a fuel burning activity will consume fuel;
used to determine heating requirements.
Source - any stationary or mobile activity which produces air pollutant
emissions.
Source Geometry - all sources for modeling purposes are considered to exist
as a point, line, or area, defined as follows:
point source - a single major emitter located at a point.
line source - a major highway link, denoted by its end points.
81
-------
area source - a rectangular area referenced to a grid system; in-
cludes not only area-wide sources,,' such as residential emitters,
but single emitters and highway links deemed too small to be con-
sidered individual point or line sources by the model.
Stability Category - a classification of atmospheric stability conditions
based on surface wind speed, cloud cover and ceiling, supplemented by
solar elevation data (latitude, time of day, and time of year).
Stability Wind Rose - a tabulation of the joint frequency of occurrences of
wind speed and wind direction by atmospheric stability class at a
specific location.
Total Air Quality - the air quality at a receptor point resulting from back-
ground emission sources and from emission sources specifically within
the study region.
Trapping Distance - the distance downwind of a source at which vertical
mixing of a plume begins to be significantly inhibited by the base
of the stability layer, and gaussian vertical distribution can no
longer be assumed.
Wind Sector - a 22-1/2 degree wind direction range whose center-line is one
of the sixteen points of the compass.
82
-------
TECHNICAL REPORT DATA
(Please read lustructions on the reverse before co»i/>/cliii[;J
1. REPORT NO.
EPA-450/3-74-056-C
3. RECIPIENT'S ACCESSIOI*NO.
4. TITLE AND SUBTITLE
HACKENSACK MEADOWLANDS AIR POLLUTION STUDY
Development and Validation of a Modeling
Technique for Predicting Air Quality Levels
5. REPORT DATE
July 1973
6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
James R. Mahoney, Bruce A. Egan, and
Edward C. Reifenstein, III
8. PERFORMING ORGANIZATION REPORT NO.
ERT Project P-244-2
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Environmental Research and Technology, Inc.
429 Marrett Road
Lexington, Massachusetts 02173
10. PROGRAM ELEMENT NO.
11. CON TRACT/GRANT NO.
EHSD 71-39
12. SPONSORING AGENCY NAME AND ADDRESS
Environmental Protection Agency
Office of Air Quality Planning and Standards
Research Triangle Park, North Carolina 27711
13. TYPE OF REPORT AND PERIOD COVERED
Final
14. SPONSORING AGENCY CODE
15. SUPPLEMENTARY NOTES
Prepared in cooperation with the New Jersey Department of Environmental Protection,
Office of the Commissioner. Labor and Industry Building, Trenton, N. J. 08625
16. ABSTRACT
The Hackensack Meadowlands Air Pollution Study consists of a summary report and
five task reports. The summary report discusses the procedures developed for con-
sidering air pollution in the planning process and the use of these procedures to
evaluate four alternative land use plans for the New Jersey Hackensack Meadowlands for
1990. The task reports describe (1) the emission projection methodology and its
application to the Hackensack Meadowlands; (2) the model for predicting air quality
levels and its validation and calibration: (3) the evaluation and ranking of the land
use plans; (4) the planning guidelines derived from the analysis of the plans; and,
(5) the software system.
17.
I.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
Land Use
Planning and Zoning
Local Governments
County Governments
State Governments
Regional Governments
Air Pollution r.nntrnl
b. IDENTIFIERS/OPEN ENDED TERMS C. COSATI Field/Group
13. DISTRIBUTION STATEMENT
19. SECURITY CLASS (This Report/
Uncalassified
21. NO. OF PAGES
91
Unlimited
20. SECLIRIT Y. CLASS (This page!
un clas si fied
22. PRICE
EPA Form 2220-1 (9-73)
83
------- |