Final Report February 1972
EVALUATION OF THE APRAC-1A URBAN
DIFFUSION MODEL FOR CARBON MONOXIDE
By: F. L. LUDWIG and WALTER F. DABBERDT
Prepared for:
COORDINATING RESEARCH COUNCIL
30 ROCKEFELLER PLAZA
NEW YORK, N.Y. 10020
ENVIRONMENTAL PROTECTION AGENCY
DIVISION OF METEOROLOGY
RESEARCH TRIANGLE PARK, NORTH CAROLINA 27711
CONTRACT CAPA-3-680-69)
STANFORD RESEARCH INSTITUTE
Menlo Park, California 94025 • U.S.A.
-------
Final Report
February 1972
EVALUATION OF THE APRAC-1A URBAN
DIFFUSION MODEL FOR CARBON MONOXIDE
By: F. L. LUDWIG and WALTER F. DABBERDT
Prepared for:
COORDINATING RESEARCH COUNCIL
30 ROCKEFELLER PLAZA
NEW YORK, N.Y. 10020
ENVIRONMENTAL PROTECTION AGENCY
DIVISION OF METEOROLOGY
RESEARCH TRIANGLE PARK, NORTH CAROLINA 27711
CONTRACT CAPA-3-68(1-69)
SRI Project 8563
Approved by:
R. T. H. COLLIS, Director
Atmospheric Sciences Laboratory
R. L. LEADABRAND, Ex~ud~ D"~ror
Electronics and Radio Sciences Division
-------
ABSTRACT
Stanford Research Institute carried out an experimental program in
St. Louis during the period August through October 1971, for the further
development and testing of a practical, multipurpose urban diffusion
model (APRAC-1A) for carbon monoxide to determine whether earlier (1970)
SRI findings in San Jose, California could be generalized to larger
cities with highly developed urban cores. Two adjacent downtown street
canyons were instrumented to obtain measurements of CO concentration at
30 points and winds at eight. The instrumented canyons are at right
angles to each other and have aspect (height-to-width) ratios of 1.5
and 2.0; these are significantly larger than the value of 0.7 of the site
studied earlier in San Jose. Wind, temperature, and CO were also measured
to a height of 130 m above the site on a television tower. Helicopter-
and van-borne instrumentation was used to supplement data collected with
the automated street-canyon instrumentation system.
The data collected generally confirm the findings of the earlier
studies, and only minor revisions of the model were required to improve
the specification of atmospheric stability and small-scale street-canyon
effects. The distribution of CO in the street canyon indicates the
presence of a single-cell, helical circulation in the deep St. Louis
street canyons under cross-street, roof-level flow conditions; the same
pattern was found in the shallow San Jose canyon. Carbon monoxide
measurements made along the street to within 10 m of the intersection
indicate that the street-effects formulation should be applicable in
this region.
The model was applied using only routinely available meteorological
and traffic data. Concentrations were calculated for four locations in
the canyons and two at roof level. These calculations were compared with
about 600 hour-averaged observations for each location. The observed con-
centrations of CO were simulated with root-mean-square errors of 3-4 ppm.
This is half the uncertainty that had been encountered when the model was
applied to St. Louis prior to refinements of the stability and diffusion
formulations and without the street-effects model. Linear regression
(calibration) would reduce the differences by an additional 1 ppm. Median
and 90-percentile concentrations were specified within 2-3 ppm by the cur-
rent model; these errors would be halved by the use of calibrated values.
It is felt that the APRAC-IA model is now suitable for practical
applications.
iii
-------
CONTENTS
ABSTRACT. .
iii
. . . .
............
LIST OF ILLUSTRATIONS.
.....
...........
. . . . .
LIST OF TABLES. . .
. . . .
........
. . . .
PROGRAM SUMMARY. . .
.......
...............
ACKNOWLEDGMENTS. . .
.......
. . . .
. . . . .
. . . .
I
INTRODUCTION. . . .
.........
. . . .
II
EXPERIMENTAL PROGRAM.
.........
A.
General Approach. . . . . .
.........
1.
Experimental
Experimental
Model. . .
Evaluation of the Sub models. . .
Evaluation of the Composite
2.
. . . .
. . . . .
........
B.
Description of the Field Program.
. . . . .
1.
Experimental Area. . . . . .
Instrumentation and Operations.
Data from Other Sources. . . . . .
. . . .
2.
. . . .
3.
. . . . .
C.
Preparation and Analysis of Data. . .
. . . .
1.
2.
Mobile Systems. . . . . . . . . . . . . . . .
Street Canyon. . . . . . . . . . . . .
Weather Service Data. . . . . . . . . .
3.
III
EVALUATION OF THE MODEL. .
.............
A.
Mixing Height Submodel.
.............
B.
Stabi li ty . .
..................
v
vii
xi
xiii
xix
1
3
3
3
7
8
8
13
29
29
29
31
34
37
37
45
-------
CONTENTS (Concluded)
C.
Emissions Submodel.
...............
1.
2.
Traffic Volumes. . . . .
Traffic Speeds. . . . . . . . . . . .
Emissions Formula.
. . . .
. . . .
. . . .
3.
D.
Winds. . . .
..................
E.
Street Canyon
..................
F.
Freeway
.......
..............
IV
PERFORMANCE OF THE MODEL.
..............
V
CONCLUDING REMARKS. . . .
..............
Appendix A--GENERAL DESCRIPTION OF THE MODEL
Appendix B--LIST OF SYMBOLS
REFERENCES
vi
55
56
58
60
73
76
94
101
113
-------
ILLUSTRATIONS
Figure 1
9
Figure 2
Figure 3
Figure 4
Figure 5
Figure 6
Figure 7
Figure 8
Figure 9
Figure 10
Figure 11
Figure 12
Figure 13
Figure 14
Figure 15
Aerial View of St. Louis Central Business District. . .
Intersection Used for Street Canyon Experiments.
Map of the Area in the Immediate Vicinity of the
Locust-Broadway Street Canyon Experiment. . . . . . . .
Locations of Van and Helicopter Routes and Vertical
Sounding Area. . . . . . . . . . . . . . .
Freeway Measurement Site, Highway 40 at Forest Park,
st. Loui s . . . . . . . . . . . . . . . . .
Location of Street Canyon Sensors and Air Inlets.
Equipment in Boatmen's Building.
. . . .
. . . .
Organization of Data Acquisition System as Used
inS t. Loui s. . . . . . . . . . . . . . . . . . .
Schematic Diagram of Data Collection Sequence
. . . . .
Comparisons of Calculated and Observed Mixing
Height s . . . . . . . . . . . . . . . . . . .
.....
Variations of Calculated and Lidar-Observed
Mixing Heights with Time. . . . . . . . . .
. . . . . .
Normalized Concentration as a Function of
Stability and Mixing Height. . . . . . . .
Diurnal Emission Patterns for St. Louis
.....
Space and Time Variation of CO Concentration over
Downtown St. Louis, 31 August 1971. . . . . . . .
St. Louis Suburban Temperature Profiles
........
vii
10
11
12
16
18
19
23
25
39
42
44
57
63
64
-------
ILLUSTRATIONS (Continued)
Figure 16
Figure 17
Figure 18
Figure 19
Figure 20
Figure 21
Figure 22
Figure 23
Figure 24
Figure 25
Figure 26
Figure 27
Figure 28
Carbon Monoxide Mass Budget Analysis, St. Louis
0800 CDT, 31 August 1971. . . . . . . . . . . .
Carbon Monoxide Mass Budget Analysis, St. Louis
1730 CDT, 23 August 1971. . . . . . . . . . . .
Space Variation of CO over Downtown St. Louis,
and Suburban Temperature Profile. . . . . . . .
Carbon Monoxide Mass Budget Analysis, St. Louis
0730 CDT, 24 August 1971. . . . . . . . ,
Carbon Monoxide Mass Budget Analysis, St. Louis
0730 CDT, 24 August 1971. . . . . . . . . . . .
Distribution of CO Concentration in Broadway
Street Canyon. . . . . . . . . . . . . . . .
. . . .
Distribution of CO Concentration in Locuet
Street Canyon. . . . . . . . . . . . . . . . .
Locations of Along-Street CO Measurement Sites.
Comparison of Along-Street CO Concentrations
with Midblock Observations of CO and Winds. .
. . . .
Temporal Variation of CO Concentration Adjacent
to the Downwind Edge of the Six-Lane Daniel Boone
Expressway at Forest Park, St. Louis. . . . . . . . .
Distribution of CO Concentration at 3-m Height
Downwind of the Daniel Boone Expressway at
Forest Park, St. Louis. . . . . . . . . . . . . . . .
Height Variation of CO Concentration Adjacent to
the Downwind Edge of the Six-Lane Daniel Boone
Expressway at Forest Park, St. Louis. . .
. . .
Observed and Calculated CO Concentrations at 4 m. . .
viii
67
68
70
71
72
78
80
88
88
96
97
98
103
-------
ILLUSTRATIONS (Concluded)
Figure 29
Figure 30
Figure 31
Figure A-1
Figure A-2
Figure A-3
Figure A-4
Figure A-5
Figure A-6
Figure A-7
Figure A-8
Figure A-9
Figure A-10
Scatter Diagram Showing Observed CO Concentrations
Versus Those Calculated with Model. . . . . . .
Cumulative Frequency Distributions of Hourly CO
Concentrations on Broadway. .........
Cumulative Frequency Distributions of Hourly CO
Concentrations on Locust. . . . . . . . . . . .
Vertical Diffusion According to Gaussian Formulation
Vertical Diffusion as a Function of Travel Distance
and Stability Category, as Revised for Urban
Conditions. . . . .
.............
. . . .
Diagram of Segments Used for Spatial Partitioning
of Emissions. . . . . . . . . . . . . . . . . . . . .
Computer Display of Traffic Links for Chicago.
Calculated and Observed CO Concentrations for a
Midblock Location in San Jose. . . . . . . . .
Observed and Calculated CO Concentrations at the
Cincinnati CAMP Station, 14-20 December 1964. . . . .
Calculated St. Louis CAMP Station CO Concentration
Frequency Distribution for 1965 Traffic Conditions--
Weekday, Saturday, and Sunday Hours. . . . . . . . .
Calculated St. Louis CAMP Station CO Concentration
Frequency Distribution for 1990 Traffic Conditions--
1-Hour, 8-Hour, and 24-Hour Averages. . . . . .
Calculated St. Louis Concentration Patterns for
Two Grid Sizes. . . . . . . . .
........
Calculated CO Concentrations in St. Louis for
Historical and Forecast Traffic Conditions. .
. . . .
ix
107
llO
III
A-5
A-5
A-7
A-9
A-16
A-17
A-18
A-19
A-21
A-22
-------
TABLES
Tab Ie 1
Table 2
Tab Ie 3
Tab Ie 4
Table 5
Table 6
Tab Ie 7
Tab Ie 8
Tab Ie 9
Table 10
Table 11
Tab Ie 12
The Relationship of Stability Category to Standard
Deviation of Wind Direction, a . . . . . . . .
e
Helicopter and Van Operating Periods.
.......
Resolution Limitations Imposed by the
Analog-to-Digital Converter. . . . . .
.......
Approximate Number of Hours of Usable Data
Obtained from the Street Canyon Instrumentation. . .
Percent Occurrence of Various Combinations of
Wind Direction Variability and Turner Stability
Class. . . . . .
. . . . .
. . . . .
........
Joint Occurrence of Turner's Net Radiation Index
and Observed Insolation. . . . . . . . . . . .
Joint Occurrence of Insolation Classes (after
Ludwig et al., 1970) and Observed Insolation.
Joint Occurrence of Revised Insolation Classes
and Observed Insolation. . . . . . . . . . . .
Revised Stability Categories
.......
. . . . .
Percent Occurrence of Various Combinations of Wind
Direction Variability and Revised Model Stability Class 52
Percent Occurrence of Various Combinations of Wind
Direction Variability and Revised Stability Class
wi th Urban Fetch. . . . . . . . . . . . . . . . . .
Percent Occurrence of Various Combinations of Wind
Direction Variability and Revised Stability Class
with Nonurban Fetch. . . . . . . . . . . . . . . . .
xi
6
15
27
28
47
48
49
50
51
53
54
-------
TABLES ( Concluded)
Table 13
Table 14
Tab Ie 15
Table A-1
Average Vehicle Speeds in Downtown St. Louis
During the Period 24 August to 2 September 1971.
Vehicle Speeds for St. Louis.
.....
.......
Carbon Monoxide and Wind Data for Budget Analyses. . .
Values of a for Cars Produced After 1970 . . . . . . .
xii
59
61
66
A-10
-------
PROGRAM SUMMARY
During this research program, the APRAC-IA urban diffusion model
was developed to simulate carbon monoxide (CO) concentrations from
readily available meteorological and traffic data.
The model is based
on existing experimental data and previous research results.
Carbon
monoxide concentrations calculated with the model were initially com-
pared with measured data from Continuous Air Monitoring Program (CAMP)
stations; the calculated and observed values often differed signif-
icantly in magnitude, although they had similar trends.
An extensive
measurement program was undertaken in San Jose, California to determine
the causes of the discrepancies between calculated and observed concen-
trations.
It was found that local effects in street canyons and around
buildings sometimes caused CO concentrations to vary as much as a factor
of 3 from one side of the street to the other.
It was obvious that
these effects must be accounted for if the model were to avoid large
errors.
One of the principal accomplishments of the research in San
Jose was the development of a new submodel to describe street-canyon
effects.
The submodel substantially improved the agreement between
observations and calculations.
The San Jose program also uncovered and
corrected other shortcomings of the original model.
These changes
resulted in more realistic specification of atmospheric stability and
turbulent diffusion in urban areas.
San Jose is a moderate-sized city. and questions arose as to the
general applicability of the results to larger cities with taller build-
ings.
To answer this question another extensive measurement program
was undertaken in St. Louis.
The results of that program are discussed
in this report.
One of the primary concerns of the St. Louis research has been the
evaluation of the performance of the street effects submodel in street
xiii
-------
canyons deeper than those studied in San Jose.
In the earlier work it
was found that roof-level winds blowing across a street canyon will
cause a helical circulation in the canyon.
At street level the cross-
street air movement is opposite to the roof-level wind direction,
causing a downflow of relatively clean air in front of the "downwind"
buildings that face the roof-level wind, and an upflow across the street.
The resulting street-level CO concentrations in front of the downwind
buildings may be significantly less than those observed across the street.
We wished to investigate whether the helical circulation fourtd in
the relatively shallow San Jose street canyon (with a height-to-width
ratio of about 0.7) was typical of circulations in the deeper street
canyons of larger cities; wind tunnel studies had suggested that under
light wind and low turbulence conditions a double-helical circulation
*
might develop (e.g., Chang et al., 1971).
To test the generality of
the San Jose observations, two street canyons, with height-to-width
ratios of about 1.5 and 2, were instrumented in St. Louis, so that CO
concentrations could be measured on both sides of each street at five
heights, from 4 m to roof level.
Concentrations were also measured in
midstreet at about 7 m and at roof level at about 35 m.
Winds in the
street canyon were measured on either side, at roof level and at 4-1/2 m.
Larger-scale airflow and CO concentrations in the area were monitored
with instrumentation up to a height of 130 m on a television transmitting
tower on top of a building at the intersection of the two streets.
These
data are available on magnetic tape from the National Climatic Center,
Asheville, North Carolina.
The data collected show that a single-helix circulation is found in
the deep street canyons of st. Louis and that the simple model developed
from the San Jose data is fundamentally correct for these deeper canyons.
Some slight modifications were required to account for the entrainment of
*
References are given at the end of the report.
xiv
-------
recirculating polluted air in the downward-flowing part of the helical
circulation.
There had been evidence of this entrainment in the San
Jose data also.
In the revised version of the street canyon sUbmodel,
the air is presumed to begin its downward flow at roof level concen-
tration.
Carbon monoxide is then added so that the concentration
increase is linear with decreasing height.
At street level the original
and revised models give the same concentration.
That concentration is
based on "box model" reasoning, i.e., the CO concentration is inversely
proportional to street width and wind speed, and directly proportional
to the emissions in the street canyon.
For the upwind side of the street,
no changes were required in the street effects model.
It is still based
on box model reasoning; the volume into which the emissions can be mixed
is limited by the helical circulation that transports street emissions
toward the buildings. As the air moves from the source, the volume into
which the pollutants are mixed increases, so the concentration is taken
to be inversely proportional to the straight-line distance from the
receptor and the nearest traffic lane.
The data indicate that the helical circulation develops when the
roof level winds are at an angle of more than 300 to the street directions.
When the winds are more nearly parallel to the street, cross street
gradients were found to be small.
For winds parallel to the street, the
street canyon effects submodel describes the vertical gradients as an
average of the two expressions used when the winds are blowing across the
street.
The small changes that were made in the street model have
improved its ability to predict the CO gradients in the street canyons.
Observations made at street level with mobile equipment indicate that the
street-canyon submodel is applicable through most of the block, at least
to within about 10 m of the intersection.
The submodel used to calculate atmospheric stability was revised to
give results that are more consistent with the fluctuations of wind
xv
-------
direction observed on the television tower.
During the evaluation of
the stability model, it was found that for a given stability type, there
was appreciably greater fluctuation in wind direction when the air had
an urban fetch than when it had a nonurban fetch.
This fact lends
support to the revisions made during an earlier phase of the program that
effectively increased diffusion rates in the urban situation as compared
to the rural.
Model calculations of mixing depth were compared with lidar (laser
radar) observations of the aerosol layer from a concurrent program spon-
sored by the National Science Foundation, and with radiosonde measure-
ments of the temperature profile near the downtown center.
It was con-
cluded from these comparisons that the mixing depth submodel does about
as well as is possible using routinely available data.
Helicopter and van measurements of CO concentrations around the
downtown, area were combined with wind speed measurements so that a mass
budget analysis could be performed to estimate the rate of CO emissions
by traffic in the study area.
The emissions model was also applied to
the same area and the results were compared.
Uncertainties in the wind
field, possible changes in CO emission rates during the measurement
periods, and uncertainties in traffic amounts all contribute to the
difficulties in making reliable comparisons, but the results should be
good enough to uncover serious deficiencies in the model.
On the average,
the three cases analyzed in St. Louis and the four earlier cases from San
Jose show agreement within a factor of 2 between the two methods.
With
the data currently available there does not seem to be sufficient justi-
fication for changing the submodel.
The revisions and additions that have been made in the model since
it was originally formulated have substantially improved its performance.
When the revised model was applied in St. Louis and compared with observed
hour-average CO concentrations, the root-mean-square difference (RMSD)
xvi
-------
between the calculated and observed values ranged from 2.6 to 3.9 ppm
depending on the particular observation site.
This is about half the
uncertainty of the original model when it was applied to this same city.
If the calculated and observed data are fit by linear regression, the
corresponding RMSD's are reduced to values between 1.6 and 3.3 ppm.
The
correlation coefficients between calculated and observed CO have been
improved substantially.
They are now in the range 0.4 to 0.7, as opposed
to the 0.2-to-0.4 range found before revision.
The ability of the model to specify frequency distributions of con-
centration is good.
Median and 90-percentile concentrations are spec-
ified within about 2 and 3 ppm of the observed values, respectively.
Use
of regression relationships derived from the observed concentrations and
those calculated with the model reduces the error in specifying median
and 90-percentile concentrations to about 1.3 ppm.
The APRAC-1A model is now sufficiently accurate that it can be used
for planning purposes, and a users manual has been prepared (Mancuso and
Ludwig, 1972).
Some improvements anq extensions are still desirable,
including better specification of emissions and a new submodel to
describe the effects that take place in the immediate vicinity of a
freeway.
xvii
-------
ACKNOWLEDGMENTS
The authors appreciate the support provided by the Coordinating
Research Council and the Air Pollution Control Office (EPA) and the
assistance furnished by the representatives of these agencies on the
CAPA-3 monitoring committee:
J. F. Black (Chairman), A. P. Altshuller,
J. M. Colucci, R. P. Doelling, C. R. Hosler, W. B. Johnson, R. G. Larson,
F. J. Mason, J. J. Mitchell, J. S. Seward, I. Solomon, and A. E. Zengel.
The authors are grateful to a number of SRI personnel for their
assistance.
E. Uthe provided the lidar data used in the evaluation of
the computations of mixing depth.
E. L. Younker, F. H. Burch, and
W. B. Guthoerl were responsible for the design and construction of the
computerized data collection system.
B. Wheeler and T. Humphrey did the
programming for the minicomputer.
During the field operations in
St. Louis, R. Allen, A. H. Smith, W. Guthoerl, J. Kealoha, N. Nielsen,
L. Salas, A. Pijma, and S. Mueller gave assistance.
G. L. Williams,
H. Shigeishi, R. Trudeau, A. H. Smith, R. Hadfield, A. Imada, J. Kealoha,
and L. Salas assisted in various aspects of data reduction and analysis.
v. Small, M. Kucinski, E. Cox, S. Hanson, M. Taylor, and P. Monti all
aided in the preparation of this and other project reports.
The authors are also grateful to the owners and managers of the
buildings where our equipment was located:
the Boatmen's Bank BUilding,
the First National Bank Building in St. Louis, and the Federal Reserve
Bank.
Mr. Robinson, Manager of the Boatmen's Bank Building, and his
staff were particularly helpful.
The personnel of Fostaire Helicopter
Service were extremely cooperative in the help that they provided in
mounting our equipment and providing storage space.
xix
-------
The personnel of several government agencies were particularly
helpful in providing access to their data.
These included G. Brancato
and the National Weather Service station staff at Lambert Field,
D. Wuerch of the Environmental Monitoring Support Unit, C. Copley and
the staff of the St. Louis Air Pollution Control Commission, the
St. Louis Department of Streets, and the Missouri Department of Highways.
Mrs. I Doris Wallace of McDonnell Douglas Automation Company was
helpful in expediting the analysis of our data.
xx
-------
I
INTRODUCTION
The long-term objective of this research program is the development
of a methodology for predicting the concentrations of motor-vehicle-
generated air pollutants throughout an urban area as a function of local
meteorology and the distribution of traffic.
The resulting methodology
is intended to be used as a tool in planning activities for the pre-
diction of the pollution patterns in any urban region that will result
from planned traffic changes or predicted growth and also as an oper-
ational tool for short-term prediction.
The achievement of the long-term objective set forth in the pre-
ceding paragraph requires that several intermediate objectives be
satisfied.
In this program we have undertaken the development of a
mathematical model of the transport and diffusion of a relatively inert
automotive pollutant, carbon monoxide.
The nature of the model has been
governed by the desire to make it a useful planning tool for application
in a variety of cities. We have attempted to develop a model whose data
requirements are neither elaborate nor esoteric, but are readily satis-
fied by routine observations.
This program of research is envisioned as having four major steps:
(1)
Developing a practical urban diffusion model for
carbon monoxide
(2)
Testing and improving the model as much as possible
with special experimental programs
(3)
Extending the applicability of the model:
to
reactive pollutants, to smaller computational
1
-------
facilities, to direct coupling with traffic or
economic models, etc.
(4)
Further testing of the model in its extended
applications.
At this time the second step has been completed. Within the practical
constraints mentioned above, we believe that the model does as well as
is possible with present knowledge.
Furthermore, it is organized in a
modular fashion so that new developments in urban meteorology, such as
better data or better techniques for specifying mixing height and other
parameters, can be incorporated relatively easily.
This report is primarily devoted to the description of the final
phases of the testing and improvement process.
The model development
and the earlier stages of the testing and improvement have been covered
in detail in two reports already issued (Ludwig et al.,1970; Johnson
et al.,1971).
Some things are repeated here to provide background
for the descriptions of the extensive field program conducted in
st. Louis and for the treatment and interpretation of the data resulting
from that field program.
This report documents the final technical evaluation of the model
that has been developed under this program.
It has been given the name
APRAC-lA, a nomenclature intended to acknowledge both the role of the
sponsoring committee and the fact that this is the first version of
what could be a series of more comprehensive or more specialized models.
2
-------
II
EXPERIMENTAL PROGRAM
A.
General Approach
The purpose of the experimental program has been to provide the
data necessary to evaluate the performance of the diffusion model and
its components.
To do this, it has been necessary to obtain special
measurements that can be compared, either directly or indirectly, with
the values of the various parameters derived by the model from gen-
era11y available input information.
Appendix A reviews the ways in
which the model treats the available information.
The first parts of
this section describe the approaches used to evaluate the model's com-
ponents, and the necessary measurements.
These are followed by a dis-
cussion of the field program that was undertaken to provide some of the
required data.
An earlier report described a prior experimental program that we
conducted in San Jose, California (Johnson et a1., 1971).
Inasmuch as
the experimental program described here generally extends that work to
a larger city (St. Louis), it has not been necessary to repeat much of
the detailed information presented in the earlier report and its
Appendices.
The techniques and equipment used in the two cases were
similar.
1.
Experimental Evaluation of the Submode1s
a.
Emissions
Evaluation of the emission submode1 required some method of
determining the rate at which CO is generated in the urban area.
This
method should be independent of traffic measurements, because traffic
3
-------
measurements are the fundamental data source used by the model.
The
method chosen is based on the fact that the net flux of some conser-
vative property or material through a closed surface is equal to the
total rate at which that property or material is generated within the
volume enclosed by the surf~ce.
In practice we have had to make some
reasonable assumptions.
These are:
(1)
Carbon monoxide is conserved during its residence time
in the volume studied.
(2)
The mean wind varies only in the vertical through the
volume studied.
(3)
The net horizontal turbulent flux of CO is negligible
compared to its transport by the mean wind.
(4)
Vertical transport of CO through the top of the
studied volume is negligible.
The conservative nature (inertness) of CO has been estab-
lished for the relatively short periods involved in the experimental pro-
gram (Junge, 1963). The area that was studied was reasonably homogeneous
and only about 1 km on a side, so the assumption of horizontal uniformity
of the wind should be justified.
The third assumption depends on there
being no gradient of emissions or turbulence across the penetrable
boundaries of the volume.
This is generally true for the scale of the
experiments.
For measurements with a resolution of a city block or so,
the distribution of CO sources in the area studied can be considered to
be reasonably uniform, as can the turbulence field.
The final assumption is valid because the observations show
that the vertical gradients of CO are quite small at the top of the
studied volume (about 335 m). so the turbulent fluxes of the material
must be relatively small.
The vertical transport by mean motions is
also small because the mean vertical motions themselves are usually small.
4
-------
The two measurements required to determine the net flux of
CO from a selected volume under the specified assumptions are the wind
field and the CO concentration at the boundaries of that volume.
Because
of the assumed horizontal uniformity of the wind field, the U.S. Weather
Service's Environmental Meteorological Support Unit (EMSU) profile data
obtained at the Arch were sufficient.
If we choose a volume with vertical
sides, then the distribution of CO needs to be measured only on those sides,
because our assumptions allow no flux of CO through the top of the volume,
and the physical presence of the ground surface prevents fluxes through
the bottom.
In summary, the approach outlined above requires measure-
ments of winds at various levels above the surface.
It also requires
measurements of CO concentrations on the vertical surfaces of some
chosen volume, in our case, a rectangular box.
The methods used to
obtain these measurements are described later.
b.
Atmospheric Stability
The stability of the atmosphere describes the intensity
and direction of the flux of sensible and latent heat in conjunction
with the momentum flux; as such, it is related to the structure of the
turbulence and hence can be used to describe the nature of the diffusion
process.
Stability categories
can,
in turn, be determined from the
variability of the wind direction.
According to Gifford (see Slade,
1968, Chapter 3), the stability categories used in the model calcula-
tions are related to the standard deviation of the wind direction 0e as
shown in Table 1. The values of 0e shown in the table are those that
would be calculated from data collected over a period of about 10 minutes
to an hour.
Each wind direction value in the total sample should represent
the average wind direction over a very short period such as a few seconds.
Since typical wind vanes tend to have response times of a few seconds,
"instantaneous" values from such instruments will be suitable for deter-
5
-------
Tab Ie 1
THE RELATIONSHIP OF STABILITY CATEGORY
TO STANDARD DEVIATION OF WIND DIRECTION, as
ae
Degrees Stability Category
25 1, extremely unstable
20 2, moderately unstable
15 3, slight ly unstable
10 4, neutral
5 5, slight ly stable
2.5 6, moderately stable
mination of the values of ae required by Table 1. Stability values from
the submodel can be compared with measured values of a to assess the
e
submodel's performance in terms of an independent parameter.
The submodel's ability to determine the stability category
from readily available meteorological information depends in part on the
determination of daytime surface heating, using observations of opaque
cloud cover and solar elevation.
The model's estimates of insolation
can be checked against measurements of the radiative power received at
the surface.
Such measurements are helpful in determining the source
of possible inaccuracies in the procedures used to determine stability
category.
Mixing Height
c.
As used in the model, the mixing height is that height to
which surface-generated materials can reach; it is the upper boundary
of the mixing layer.
This concept of mixing height suggests one method
for its determination.
If this concept is satisfied, then the upper
boundary of the mixing layer should be marked by a sharp decrease with
6
-------
height in surface-generated pollutants.
Measurements of the vertical
distribution of pollutants, such as CO or particulates, that are
released within the mixing layer can help define the extent of that
mixing layer.
Another commonly employed approach to the determination of
mixing height makes use of the distribution of temperature in the
vertical.
The top of the mixing layer is often marked by a layer of
increased atmospheric stability, which in turn is characterized by an
increase in value of the rate of change of temperature with height,
dT/dz.
Street Canyon Effects
d.
Unlike the preceding items, the distribution of carbon
monoxide in street canyons is directly observable.
Testing the street
canyon effects requires that we observe these effects and see if they
agree with the results of the parameterization that has been devised.
The CO concentrations and airflows in and around the street canyons must
be measured with considerable detail and accuracy.
2.
Experimental Evaluation of the Composite Model
After the individual components of the model have been tested,
then the composite must be tested.
This is straightforward.
The
questions to be answered are:
1)
How accurately does the model calculate CO concentrations
for specific times and places?
2)
How accurately does the model reproduce the concentration
frequency distribution for given locations?
To answer the above questions, we must provide accurate measurements of
CO concentrations and the information necessary to make the calculations
with the model.
7
-------
Description of the Field Program
B.
On the preceding pages, we have described the general approaches
that we have used to evaluate the model and its components.
The types
of data necessary for the application of the approaches have been
discussed.
The following pages describe the program that was undertaken
in St. Louis to obtain the necessary data.
1.
Experimental Area
St. Louis was chosen as the area in which to extend the exper-
imental program that was begun in San Jose.
St. Louis has the primary
requisites that were sought.
It is significantly larger than San Jose,
has a well-developed downtown area with deep street canyons, and is
relatively free of unusual and complicating topographical features that
might mask fundamental problems in the model.
Furthermore, a significant
number of studies of urban diffusion and local meteorology had already
been conducted there (e.g., Stanford Aerosol Laboratory, 1953, Pooler,
1966; McElroy and Pooler, 1968; and Wuerch, 1970, 1971); these repre-
sented a considerable source of information that might be useful in the
interpretation of our own data.
Finally, there were several other
studies of urban meteorology planned for st. Louis at about the same
time as our field program and thus there was the possibility of obtain-
ing some additional data that might not otherwise have been available.
Within st. Louis we sought an area of deep street canyons with
relatively uniform building heights, in which we could collect the data
necessary to evaluate the street effects submodel.
For this we chose
the area near the corner of Locust Street and Broadway in the central
business district.
Figure 1
is an aerial photograph of the central
business district; Figure 2
is a closer view of the intersection used
for the street effects studies.
A map of the nearby areas is given in
Figure 3; this shows approximate building heights (in standard-height
"floors If) .
8
-------
u:>
~~~1
~Jo-
helicopter route \
.
,~,,;{J_. j,,~~--. ...;~:..,",
...._-...- -. "'. -~'!I:
La...:.: --_."~.-:WP-
"
/'
~
-...
~
"""'-----....;..
TB-8563- 114
FIGURE 1
AERIAL VIEW OF ST. LOUIS CENTRAL BUSINESS DISTRICT
-------
FIGURE 2
INTERSECTION USED FOR STREET
CANYON EXPERIMENTS
The chosen location was instrumented to study two street
canyons,
at right angles to each other.
One of the buildings at the
selected site also had the television transmission tower shown in
Figure 1; it extended to about 160 m above street level.
This allowed
us to measure winds, temperatures, and CO concentrations well away
from the street effects.
As indicated in earlier sections, evaluation of some of the
models required vertical profiles of temperature and pollutant concen-
trations.
Evaluation of the emissions submodel required measurements
of CO at various levels around the perimeter of some area with significant
CO sources.
The area chosen is shown on the map in Figure 4.
It sur-
rounds the area shown in Figure 1.
The map and photograph are marked to
show the route followed by an instrumented van and helicopter during the
experiments designed to determine emission rates.
The rectangular
10
-------
J L
ST. CHARLES
Park-
ing Courtyard Parking
0 Lot 2 0 @ Garage
@ 6 @ CD
>- @
«
0 0 s:
~ a ~
(i5 « @ :;.
o
a:
> cc
CB II> [2j@
0~ 0
UVW and CO
LOCUST . Temperature
@)
@
G) UVW
and CO
@ 0 0 Parking
@ @ Lot
OLIVE
I I
> @)
@ ~
-------
.'
",
SCALE 1 ..~<.A:(:
T A-8563-116
"
f ::=;..r--'
.~v ..H'
,tt
~.
--=,
1'"',1
--~-~
~>'"(' fH~
..
CONTOUR INTERVAL 10 FEET
[:>4$.., 0 l'~tS J:li:PlJtsr-.r ~~-)')T COHTOURS
C 4~U'" -:;;, '~C"" $£.4 lEVEl
FIGURE 4
LOCATIONS OF VAN AND HELICOPTER ROUTES AND VERTICAL
SOUNDING AREA
12
-------
route surrounds the downtown area and has the street experiment site
along the east edge, which permits the incorporation of data from that
instrumentation with those from the mobile units.
Figure 4 also shows the area over which the instrumented
helicopter measured vertical profiles of CO concentration and temperature,
and the Gateway Arch which serves as a launch site for twice-daily
balloon releases by the National Weather Service's Environmental Meteor-
ological Support Unit (EMSU).
These balloon soundings provided temper-
ature, humidity, and wind profile measurements up to 3 km.
The figure
shows that all locations are within a few kilometers of each other.
2.
Instrumentation and Operations
a.
Mobile
A helicopter and a van were instrumented to measure CO
concentration and temperature.
The equipment has already been described
in detail elsewhere (Johnson et al., 1971), and so only the general
characteristics will be described here.
For the CO measurements, filtered
air was drawn through a few meters of polyethylene tubing from the front
of the vehicle.
By having inlets at the front, interferences from the
helicopter's downwash or the van's exhaust were avoided when the vehicles
were moving forward.
The CO analyzers used in the van and in the Bell
helicopter were manufactured by the Bacharach Instrument Co.*
In these
instruments the sampled air is passed through a bed of heated mercuric
oxide; any CO present reduces the oxide and releases mercury vapor which
can be quantitatively detected through its spectral absorption of ultra-
violet radiation.
Temperatures were measured using radiation shielded ther-
mistors.
The data were recorded on strip charts, along with marks made
by the operator to identify certain locations or significant events.
*
2300 Leghorn Ave., Mountain View, California
94040.
13
-------
The CO analyzers were generally calibrated with gases of
known CO concentration (0 and 11 ppm for the helicopter, 0 and 21 ppm
for the van) before and after an experiment.
It took about 45 minutes
to an hour to complete an experiment.
During this time, the van was
driven around the rectangular route shown in Figures 1 and 4.
helicopter crossed to the east side of the Mississippi River and then
T~
climbed upward in the area shown by the spiral in Figure 4.
During
this period, the operator kept the CO analyzer flow rate constant and
marked the charts at height intervals of 50 feet, as determined from
t~ helicopter altimeter.
After completing the vertical sounding to about 610 m, the
helicopter recrossed the river and began its first circuit of the route
from the northeast corner at a height of 115 m.
At the end of that
circuit the helicopter circled and gained altitude over the area north
of the route, while the operator adjusted the instrument in preparation
for the second circuit at a height of 150 m.
This routine was repeated
at heights of 215, 275, and 335 m.
Most of the mobile measurements were made during the latter
part of August 1971,
a period during which other groups were conducting
atmospheric studies (Changnon, 1971),
e.g.,
of airflow over the city and the
vertical distribution of atmospheric aerosols.
We attempted to collect
data during the morning, evening, and midday periods of heavy traffic,
as well as the forenoon and afternoon periods in between.
Table 2 shows
the approximate time intervals during which data were collected for use
in evaluating the emissions submodel.
In addition to the mobile measurements that were made in
conjunction with the helicopter flights, the van was also used to make
special observations downtown and in the vicinity of a heavily traveled
highway.
Carbon monoxide measurements were made at a variety of locations
14
-------
Table 2
HELICOPTER AND VAN OPERATING PERIODS
(Central Daylight Time)
Date Morning Forenoon Midday Afternoon Evening
23 Aug 71 0905-0955 1730-1820
24 Aug 71 0730-0820 1145-1235 1430-1520 1630-1720
28 Aug 71 0730-0820 0930-1020 1145-1235 1445-1535 1630-1720
31 Aug 71 0730-1820 0930-1020 1145-1235 1630-1720
1 Sept 71 1100-1150 1330-1420 1630-1720
2 Sept 71 0730-0820 0930-1020 1145-1235 1430-1520 1630-1720
on Locust Street and Broadway to examine along-street variations of CO
with respect to the continuous observations at the fixed sampling sites.
Because of local traffic regulations, the van could not be parked during
the rush periods and so the observations were usually made between 0900
and 1600 CDT.
As an additional check, the van was also operated at the
Continuous Air Monitoring Project (CAMP) station at the intersection of
12th and Clark Streets (see Figure 4 and the discussions in Ludwig et a1.,
1970).
Local contamination by the van itself for these stationary
aPPlications was avoided because the instrumentation is powered by
storage batteries and it is not necessary to run the engine; a more
detailed description of the van system is given by Johnson et a1. (1971).
The van borne instrumentation system was also used to make
measurements for basic studies of the diffusion of CO in the immediate
vicinity of a freeway.
An area in Forest Park, shown in Figure 5, was
chosen for these studies.
With the prevailing summertime wind flow from
the south, the air moves progressively from a well-mixed region with no
15
-------
concentrated CO sources over the six-lane Daniel Boone Expressway and
into the adjacent Forest Park area.
Measurements were made during the
early morning hours during the peak freeway traffic period, and traffic
within the park did not influence the local concentration.
The measure-
ments were of four types:
(1) time series at a fixed point, (2) vertical
profiles to 11-m height; (3) horizontal profiles to 200 m downwind of the
freeway, and (4) horizontal traverses throughout the park.
b.
Fixed Installation, Street Canyon
Location of Equipment.
Photographs of the street canyon
experimental site have already been presented in Figures land 2.
A
diagram of the location of the various street canyon sensors and air
sampling inlets is shown in Figure 6.
These locations have also been
indicated in Figure 3.
Air sampling inlets were placed at fifteen
locations in each of the two street canyons and at five heights on the
FIGURE 5
FREEWAY MEASUREMENT SITE, HIGHWAY 40
AT FOREST PARK, ST. LOUIS
16
-------
tower seen in Figure 6.
The inlets on the tower were at about 85, 93,
100, 116, and 131 m above street level.
The base of the television
tower was about 70 m above street level.
Three-component wind sensors were located at the top and
bottom of the vertical arrays in the street canyons, as shown in Figure 6.
Propeller-vane wind sensors were placed at two levels on the tower, 90 m
and 131 m above the street.
Temperature profile measurements (five-level)
were made on the south side of the Locust Street canyon at the same
heights as, and adjacent to,the CO profile, and also on the tower at 86,
95, 101, 116, and 130 m above the street.
Two traffic counters were
used to count the vehicles that passed on the two streets where the
equipment was located.
The CO analyzer used with the air sampling inlets on the
TV tower was located in a room just below the base of the tower.
Too
electronics required to process the tower wind and temperature signals
were also located in this room.
All the rest of the electronic equipment
and CO analyzers were located in a room on the tenth floor of the Boatmen's
Building (Figure 7).
As can be seen in Figure 2, the tubing and cables
from the air inlets and the wind, temperature, and traffic sensors in the
two street canyons were brought to this central location for signal con-
ditioning, processing, recording, and preliminary analysis.
Carbon Monoxide. Nondispersive infrared analyzers were
*
used to measure the CO concentrations.
They have a 40-inch-long
absorption
cell and use an optical filter to remove the effects of
of water vapor interference.
They have a sensitivity of about 0.5 ppm.
It was necessary to calibrate the instruments twice daily.
The room at
the foot of the television tower was not air-conditioned, and more frequent
*
Beckman Instruments, Model 3l5-AL.
17
-------
WIND DIRECTIONS GIVEN RELATIVE
~ 'I TO BROADwAY, TRUE
\ ARE 17° GREATER,
\18"
:;
~ FEDERAL
LOJ i I RE:.RVE BANK
FI~ ~ BO:TMEN'S
NATlCii..-A'L reo BUILDING
BANK
'k
5
°
14
I-"
(X)
FEDERAL
RESERVE
BANK
4°
3°
2°
If
°
11
°
12
LOCUST
DIRECTIONS
~
+ 10
°
IS
+°9
+°8
°
13
+°7
+'16
NORTH
FIGURE 6
o co INLET
iii 3 - COMPONENT WIND SENSOR
...L 2- COMPONENT WIND SENSOR
+ TEMPERATURE SENSOR
o 2 4 6 B 10 IIETERS
I . " , ' , '
o 10 20 30 FEET
-
BOA TMEN' S
BUILDING
'I
'~:~~~"'I'
NATIONAL
BANK
SOUTH
WEST
TOWER ON BOATMEN'S BUILDING
(NOT TO SCALE)
'H
10
°
14
°
15
...L+o
'k
5
°4
°3
°2
f
9°
8°
7°
6111,
11
°
12
.
13
BROADWAY
LOCATION OF STREET CANYON SENSORS AND AIR INLETS
...L+o
+0
+0
+0
Ri
'~~~~;';~'~'.~
BUILDING
EAST
T8-8563-1 HI
-------
FIGURE 7
EQUIPMENT IN BOATMEN'S BUILDING
19
-------
calibrations were required for that instrument.
Zero adjustments were
made using high purity helium.
The span was adjusted using a mixture
of 21-ppm CO in nitrogen prepared and analyzed by the Matheson Company.
All instruments were calibrated with gases from the same sources.
Air
was brought to the analyzers from the various sampling points through
1/4-inch-ID polyethylene tubing.
Tests conducted earlier (Johnson et al.,
1971) indicated that polyethylene tubing as used on this program does not
interfere with the CO measurements.
Filters at the tubing inlets
removed particulates that might affect CO analyzer operation.
The
tubing lengths ranged from 40 m to 100 m; tests showed that air enter-
ing the longest tube reached the CO analyzer in about 20 seconds.
The
air sampling inlets were connected in groups of five to manifold systems.
Each manifold system used computer-controlled solenoid-actuated valves
to switch the CO analyzer from one inlet to another.
The four levels
that were not being sampled at any given time were continually purged
with an auxiliary pump.
Thus, the CO analyzer pump did not need to
exhaust dead air from the tubing when air from a new level was switched
to the instrument.
Wind.
Three-component winds were measured at two levels
on the four vertical arrays:
at 4.5 m height and just above the roof
level; the sensors protruded 3 m from the buildings.
Each instrument*
used three orthogonally oriented, low-inertia propellers to measure
three wind components (u,v,w).
Winds on the tower were measured with propeller and vane
type
**
sensors.
The propellers on these instruments, like those on the
three-component sensors, have a low starting speed, about 0.2 m s-l.
*R. M. Young Co. Model 27002/27302
**R. M. Young Co. Model 35002-35402/35602
20
-------
Vertical Temperature Gradient.
Platinum wire resistance
*
elements
were used to measure vertical temperature gradients.
These
elements were mounted in 0.125-inch' stainless steel tubes.
These tubes
were housed in radiation shields of silvered, double-walled glass cylinders
-1 **
similar to Dewar flasks and were ventilated at a rate of about 5 m s
The time constant of the aspirated, steel-housed sensor was about 40
seconds.
Temperature differences of as little as'O.OloC between sensors
could be detected.
Traffic Counters.
The capability to count traffic and
record the counts with the other data was added to the system used for our
earlier studies in San Jose (Johnson et al., 1971).
Vehicle passage is
sensed by contact closure in a pneumatically operated switch (like
those used in gasoline stations to alert the operator of the presence
of a customer). These closures are detected by a buffer circuit with
80-m~11isecond response time and then counted by an 8-stage binary
counter. This digital number is converted to an analog voltage by a
***
commercially available digital-to-analog converter.
The traffic
count is accumulated for approximately 90 seconds and then entered into
the system.
Within 0.3 second, the counter is reset and the count for
the next time interval is started.
A maximum of 127 vehicles can be
counted in 90 seconds.
-1
At a speed of 10 m s (22 mph) , separate
vehicles can be detected if they are spaced at least 80 cm apart.
Control and Data Acquisition System.
****
purpose computer
A small general-
with a magnetic tape recorder eontrolled the data
*Rosemount Engineering Model 104 MK-57-BB-CC
**
R. M. Young Co. Model 43404
***
Zeltex Co.
****
Data General Corp., NOVA.
21
-------
collection and CO analyzer manifold switching operations.
Nine separate
remote units accepted commands from the computer via a 2Q-mA dc circuit,
digitized the voltage inputs from the sensors, and transmitted the data
back to the computer via another 20-mA dc circuit.
The computer was
connected to the two dc circuits through a line coupler.
The organiza-
tion of the various components of the data acquisition system is shown
schematically in Figure 8.
The system has been described in detail in
an appendix of an earlier report (Johnson et al., 1971).
The computer sends a command message, consisting of two
characters, to all the remote units.
The first character is an address
code that is recognized by only one remote unit.
On recognition of the
address code, the activated remote unit sends its data message to the
computer.
The second character of the command message (inlet code)
indicates which inlet is to be opened by those units controlling the
air sampling inlet manifolds.
The data message returned to the computer
from the remote unit consists of the inlet code stored at the time of
the preceding command, followed sequentially by the digitized data from
the sensors connected to that unit.
The data message ends with the
identifying character of that control unit.
Each measurement was trans-
mitted as a 7-bit binary number plus even parity.
After the computer received the complete message, it
checked the identification code to verify that it corresponded to the
address that was transmitted.
Then another unit was commanded to report.
The timing of command messages was derived from the computer's rea1-
time clock in a programmed sequence.
The sequence for the St. Louis
experiment (shown schematically in Figure 9) was as follows:
The three-
component winds were sampled for about 8 seconds, or about 2 times each;
these data were recorded on magnetic tape.
This was followed by the
sampling of about 2 seconds of temperature and traffic data.
This, in
turn, was followed by 44 seconds of CO concentrations and tower winds--
22
-------
From
Locust
.AiLSample. .~
Inlets
-.
-.
-.
-.
~
-.
~
~
.......
.~
~
-.
-.
--.
From
Broadway
Air Sample
Inlets
'-,
....
~
--.
~
-.
~
-+-
-+-
-+-
-+-
~
-.
~
INLET
MANIFOLD
INLET
MANIFOLD
INLET
MANIFOLD
INLET
MANIFOLD
INLET
MANIFOLD
INLET
MANIFOLD
-
1
1
I"" -
-
-
l
1-
....
Inputs From
North Locust
3-Component
Wind Sensors
{----.
.
Inputs From {
South Locust
3-Component
Wind Sensors
Inputs From {
East Broadway
3-Component
Wind Sensors
Inputs From {
West Broadway
3-Component
Wind Sensors
Inputs From
Locust Temperature {
Sensors and Locust
and Broadway
Traffic Cou nters
CO
ANALYZER
~
CO
ANALYZER
-
-
- -_..
CO
ANALYZER
-
Inlet Manifold Control Circuit
CO
ANAL YZER
-
CO
ANAL YZER
CO
ANALYZER
-
Inlet Manifold Control Circuit
NOTE:
Remote units 6 and 9 were located in room below TV tower;
other equipment were located in a room on the tenth floor 01
the Boatmen's Building.
FIGURE 8
Command
Line
.
REMOTE REMOTE
--.. ~ UNIT No.6
--
Inlet Manifold
Control Circuit
'.
} From Tower
Air Sample
Inlets
TB-8563-111 R
ORGANIZATION OF DATA ACQUISITION SYSTEM AS USED IN ST. LOUIS
.
-
REMOTE ...........
UNI:!" No. 7...,--,-
'--
V
REMOTE <...........
UNIT No.8..""-
"""--
j> REMOTE
-~ UNIT No.9
I"-~
CO
ANAL YZER
INLET
MANIFOLD
j11tt
-------
~
tJ1
w
~
u
>-
u
u.
o
I-
r:r:
~ 188
(J)
w
U
Z
(J)
o
w
(J)
~ 278
~
w
~
z
o
u
w
(J)
DATA RECORDED
DATA RECORDED
8
I uvwread\
~___L-4- START
. .
\ traffic and temperatures read
, -
~ uvw read ---.......------
------
------------- SWITCH TO
. .---. INLET 3 '
\ CO, tower winds read "-
+' traffic and temperatures read u v w read - - - ~ - - - - - " 3
---------
------------ ~reHro
. ~- INLET 4
'" CO, tower winds read
traffic and temperatures read ~
+, uvw rea~_____---"1111(-------
------
------------- ~reHro
. -.wi INLET 5
CO" tower winds read
"traffic and temperatures read ~
~' uvwread_____---4----- 5
------------- SWITCH TO
. . ------ INLET 1
CO, tower winds read
, \ traffic and temperatures read ~
6 ...-..........................
. ..............................................
..........................
. IlVo'''''''''''''''''''' ..---5 printout on TTY
........................................ 480 SECONDS, START NEW CYCLE
CO, tower winds read
MANIFOLD
SWITCHED TO
INLET2~
2
98
368
458
FIGURE 9
SCHEMATIC DIAGRAM OF DATA COLLECTION SEQUENCE
53
w
~
U
>-
143 u
u.
o
I-
r:r:
-------
about 16 values of each.
At the last sampling of the CO data, the inlet
manifold valves were switched to another level, and the accumulated data
were recorded on magnetic tape.
The three-component winds were read for
44 seconds (about ten times eac~, then traffic and CO again.
After the
CO values had been recorded for the fifth time, average values of all
parameters (except traffic) for the preceding 7 minutes were displayed
on a teletype, so that equipment performance could be monitored.
Too
complete sampling cycle began anew after the teletype printout.
The digitizing of the data for transmission, processing,
and recording imposed certain resolution limits on the observations.
We have tried to keep these consistent with the accuracies of the
sensors.
Table 3 summarizes the resolution imposed by the analog-to-
digital conversion for the various parameters.
Operations.
The equipment for the street experiment was
shipped to St. Louis and installed during the early weeks of August 1971.
The system was tested and the necessary calibrations were completed;
some limited data collection began August 26.
This was before all
the installation was complete on Locust Street.
(Data from the tower
sensors were recorded on the teletypewriter commencing August 23 in
order to support the mobile measurement program.)
We continued to
collect data while making the final installations, which were completed
by September 2, 1971.
The next few days were spent checking to see
that the equipment was all operating properly, in anticipation of
starting round-the-clock data collection.
On the afternoon of September 5, a lightning strike on
the tower atop the Boatmen's Building did extensive damage to the
system.
Data collection was not resumed until September 12, and then
only on a limited basis.
Repair work continued during the following
week and by September 19 the system was returned to near normal
operation.
There were malfunctions of various pieces of equipment
26
-------
Table 3
RESOLUTION LIMITATIONS IMPOSED BY THE ANALOG-TQ-DIGITAL CONVERTER
Parameter Range Resolution
Wind components -20.3 to +20.3 m s-l 0.16 m s-l
Wind speed 0 to 20.3 m s-l 0.16 m s-l
Wind direction 0 0
0 to 360 3
Temperature difference -1 to 0 0
+1 C 0.008 C
Carbon monoxide 0 to 50 ppm 0.4 ppm
concentration
Traffic count 0 to 127 vehicles 1 vehicle
during the ensuing weeks and many of these appeared to be related to
the damage sustained September 5.
We had originally planned to operate the traffic counters
24 hours per day, but we encountered some difficulties.
The hoses that
stretch across the street were damaged by street sweepers during the
night on several occasions, so it was necessary to bring them in
during the late night and early morning hours.
Calls to the Street
Department and wrapping the hoses with high visibility tape minimized the
problem, so that traffic data were successfully collected during the
early morning hours on many days during the last three weeks of operation.
Table 4 summarizes the hours of data collection for the
street experiment.
Data collection began August 26 and ended October 15.
The equipment was taken down, packed,and returned to California during
the last weeks of October and early November.
27
-------
~
00
Table 4
APPROXIMATE NUJIBER OF HOURS OF USABLE
DATA OBTAINED FRaI THE STREET CANYON INSTRU!oIENTATION
East West I Tower Mid- East West Mid-
Date N. Locust S. Lac'us t Broadway Broadway Tower Locust N. Locust 5. Locust Locus t Broadway Broadway Broadway Tower Locus t Broadway
UVW UVW UVW UVW Winds Tempe i Temp CO CO CO CO CO CO CO Traffic Traffic
Aug 26 4 11 11 11 11 11 r 11 3 3 3 11 11 11 11 4 4
27 13 13 13 13 13 13 13 13 13 13 13 13 13 0 9 9
28 13 13 13 13 13 9 I 13 9 9 9 13 13 12 8 13 13
,
29 12 12 12 12 12 12 I 12 12 12 12 12 12 12 11 10 10
I
30 13 13 13 13 13 13 13 13 13 13 13 13 13 13 12 12
31 14 14 14 14 14 14 14 14 14 14 14 14 14 14 12 12
Sep 1 17 17 17 17 16 17 17 17 17 17 17 17 17 17 15 15
2 16 16 16 9 16 16 16 16 16 16 14 16 16 14 15 15
3 16 16 16 16 15 16 15 16 16 16 16 16 16 16 13 13
4 14 14 14 14 14 14 14 14 14 14 14 14 14 14 13 13
5 6 6 6 6 3 6 6 3 6 6 0 6 5 6 6 6
12 5 0 5 5 0 0 0 0 5 0 5 5 5 5 0 0
13 10 0 18 18 0 9 9 11 18 9 11 17 17 10 7 7
14 0 0 18 18 0 18 0 14 16 15 13 16 13 12 16 16
15 12 12 24 24 0 24 0 23 24 24 24 24 24 8 15 15
16 24 24 24 24 0 24 0 20 24 20 24 14 24 24 17 17
17 24 24 24 24 7 24 7 22 24 0 24 0 24 12 16 16
18" 16 16 16 16 16 16 16 16 16 6 16 6 16 16 0 0
19 24 24 24 24 24 24 24 24 24 24 24 24 24 24 14 14
20 20 20 20 20 20 20 20 20 20 20 20 20 20 19 20 20
21 24 24 24 24 24 18 24 14 21 22 16 19 19 16 16 16
22 24 24 24 24 24 24 24 24 24 24 14 24 24 24 24 24
23 24 24 24 24 24 24 24 19 23 18 24 24 24 11 24 24
24 24 24 24 24 24 24 24 17 24 22 24 17 24 9 12 12
25 11, 11 11 11 11 11 11 11 11 4 11 11 11 0 13 13
30 12 12 12 12 12 12 12 12 12 12 11 12 12 1 12 12
Oct 1 21 21 21 21 21 21 21 21 20 21 15 21 21 21 9 9
2 23 23 23 23 23 23 23 17 23 16 23 23 23 15 16 16
3 15 15 15 15 15 15 15 9 15 9 15 15 15 15 24 24
4 23 23 23 23 23 23 23 18 23 18 23 23 23 20 21 21
5 24 24 24 24 24 24 24 24 24 24 24 24 24 22 16 16
6 22 22 22 22 22 22 22 22 22 15 22 22 22 21 13 13
7 16 16 16 16 16 16 16 15 16 16 16 16 16 16 8 8
8 24 24" 24 24 24 24 24 24 24 24 24 24 24 24 18 18
9 13 13 13 13 13 13 13 13 13 13 13 13 13 13 13 13
10 24 24 24 24 24 24 24 24 24 24 24 24 24 24 24 24
11 24 24 24 24 24 24 24 24 5 24 24 24 24 24 24 24
12 24 24 24 24 24 24 24 24 0 24 24 24 24 24 24 24
13 24 24 24 24 24 24 24 24 12 24 24 24 24 24 22 22
14 24 24 24 24 10 24 20 24 24 24 24 24 24 9 24 24
15 11 11" 11 11 0 11 9 11 11 11 11 11 11 0 10 10
otals 704 696 749 742 i 613 725 645 671 675 640 704 690 736 587 594 594
T
-------
3.
Data from Other Sources
a.
Meteorological
Conventional meteorological surface observations are taken
at Lambert Field, about 22 km northwest of the downtown area.
Rawinsonde
measurements are made twice daily, nominally at 0500 and 1700 CDT, 110 km
east of St. Louis at Salem, Illinois.
Special urban soundings are taken
by the National Weather Service at the Arch, about 700 m southeast of
Locust Street and Broadway.
These soundings are made Monday through
Friday twice daily at about 0700 and 1230 CDT.
In addition to the conventional National Weather Service
data mentioned in the preceding paragraph, there were a number of special
meteorological measurements during the summer in connection with the
METROMEX program (Chagnon, 1971).
These measurements included detailed
pilot balloon studies of winds and 1idar soundings of the depth of the
polluted layer.
The 1idar measurements have been quite useful in the
evaluation of the mixing depth submode1.
b.
Air Pollution
Air pollution measurements are made at several locations
in the St. Louis area by the St. Louis Pollution Control Commission.
There is also a station of the Continuous Air Monitoring Program (CAMP)
in St. Louis.
Data from these sources have been used only minimally in
this study, largely because of the difficulties in integrating them with
the project's magnetic tape data records.
C.
Preparation and Analysis of Data
1.
Mobile Systems
The van and helicopter instrumentation systems employed iden-
tical sensors for the measurement of air temperature and carbon monoxide
concentration; details are given in an earlier report by Johnson et ale
29
-------
(1971).
The voltage outputs of the sensors were amplified and recorded
on strip chart recorders in both systems, while the van system also
*
used an instrumentation magnetic tape recorder (H-P Model 3960A ) in
parallel as a backup device.
Preceding and following each helicopter flight, calibration
checks were made of the CO analyzer electronic reference signals as well
as the outputs from zero (high purity nitrogen) and span (10-ppm CO) gases.
In-flight checks of the reference signal and zero gas were also made.
This
provided a reasonably comprehensive history of instrument performance
during the course of each flight.
Atmospheric mixing serves to filter
small-scale variations of both temperature and CO, and so it was possible
to obtain representative profiles by relatively low-frequency digitization
of the analog chart record:
For the vertical profiles, l5-m increments
were used to 150 m height and 30-m increments at higher levels; l40-m
intervals were used along the horizontal traverses.
The digital chart
values and the pre-, post-, and in-flight calibration and check data were
**
entered into a Wang 720/702 Programmable Calculator/Plotter
for proc-
essing and display.
All input and identification data were displayed in
addition to the calculated value and its location.
For the traverse data,
segment averages were automatically computed and listed.
The van data were processed somewhat differently because of the
larger and more frequent fluctuations that occur near to the earth's
surface and
also near the CO sources.
When the van operated in a mobile
configuration, spatial averages were required over specific route segments,
while time averages were required when the van was stationary.
One-hour
*
Hewlett-Packard, Palo Alto, California
**
Wang Laboratories, Inc., Tewksbury. Massachusetts
30
-------
averages (compatible with fixed-station values) were used while studying
along-street variations in the Locust and Broadway street canyons.
The
chart records were marked to identify the periods,and a simple, graphical
average was computed.
Calibration data were applied as with the helicopter
analysis.
Observations at the freeway site were both limited and of a
special, nonroutine nature and so an alternative procedure was used.
The
analog chart record was digitized so that both the mean and standard
deviation of the observed CO concentrations could be computed.
The data records on magnetic tape served as a backup for those
periods when the chart recorder failed.
The data were then retrieved by
one of two methods, depending on the purpose of the observations.
For
most cases, the analog output from the tape recorder was used to generate
a new chart record which was analyzed by the methods described above.
For statistical studies of fine-scale fluctuations, the analog data were
input to a probability density analyzer which digitizes the signal at
a prescribed rate and stores the data in the form of a probability density
histogram.
The "density" is then transferred automatically through a
hardware interface to the programmable calculator which can analyze the data
to obtain the mean, variance, skewness, and kurtosis of the distribution.
This procedure was not used routinely, but provided a check on the mean
values determined by more conventional arithmetic or graphical averaging.
2.
Street Canyon
The street canyon data are quite extensive, and it was necessary
to perform some editing and condensation.
The data handling began by
translating the binary data as originally recorded to the binary-coded
decimal used for subsequent processing.
During this preliminary processing,
hourly average values were calculated for the various measured parameters
along with standard deviations.
Although these averages did not include
some of the corrections employed in later processing, they did allow us
31
-------
to assess the quality of the data that had been recorded on magnetic
tape and make necessary corrections. Some of the preliminary processing
was done in St. Louis using the facilities of the McDonnell Douglas
Corporation.
This allowed the staff in st. Louis to check equipment
operation quite rapidly.
During the translation of the tapes the recorded voltages were
converted to engineering units.
Next, corrections were made for drift
in the output of the CO analyzers.
The following procedure was adopted.
The true CO readings were assumed to be represented by
(CO) = c + d(CO)
true read
( 1)
Immediately following calibration, the values of c and d were taken to
be 0 and 1, respectively, because the calibration results in the reading
of the true concentration.
During calibration the reading of the instru-
ment before adjustment was recorded while it was sampling CO-free gas.
This reading was used as -c for the period just prior to the calibration
time.
Before adjusting the span setting the change in reading was
determined when a gas of known CO concentration (21 ppm) was introduced
to replace the CO-free gas.
For the period just prior to calibration,
the factor
d equaled the ratio of the change in reading that should
have been observed for the calibration gas to the change that actually
was observed.
Between calibrations, the values of c and d were assumed
to change linearly with time.
When the recorded CO concentrations were corrected, some
data were eliminated.
These included cases where adjustments may have
been made while instruments were still on line.
Actually, most equipment
remained on line all the time; the recorded data included a "mode code"
(see Johnson et al., 1971, Appendix A) whose value specified whether data
were good or were to be ignored.
Switches on the equipment changed the
mode code values appropriately, but sometimes the switches were set
improperly.
Generally, these cases were discovered and noted in the
32
-------
station log.
It was at this point that the appropriate corrections were
made in the recorded figures.
Next, the data were consolidated into a master tape similar to
the summary tape described in Appendix D of the Comprehensive Report
(Johnson et al., 1971).
Although this tape is similar to that described
in the earlier report, there are differences.
On the St. Louis summary
tape, carbon monoxide observations have been summarized in much the same
way as the wind components were summarized on the earlier tape;
i.e.,
for each period, the sum of the observed values, the sum of the squares
of the observed values, and the number of measured values for each inlet
are recorded.
On this new tape we have also included sums of the cubes
of observed CO readings so that skewness of the frequency distribution
of observed values can be determined if desired.
Wind component
summaries are also included, as before.
There are other differences between the current summary tape
and the earlier one.
Each record of the earlier version summarized the
data collected during a 5-minute period; the St. Louis data tape records
cover 7-minute periods, because of the longer sampling cycle.
We
have reduced the considerable redundancy of the earlier summary; e.g.
the date appeared numerous times in a single record.
Also some of the
identification numbers for CO analyzer inlets that were previously
recorded explicitly are now implicit in the formatting of the records.
The summary records for St. Louis contain traffic data that
are quite useful in interpreting the data.
We have also included standard
deviations of the wind directions measured on the tower.
These have been
related to the stability categories, as suggested elsewhere (Pasquill,
1961, or Slade, 1968).
The standard deviations were determined by first
calculating the vector-average wind direction and then the root-mean-
square deviation between the individual wind directions and this average.
33
-------
Some of the traffic data contained in the records were not
measured.
The traffic flow data for those periods when measurements
were not available were based on the available measured values for the
same hour of the day and day of the week.
The observed traffic data
indicate that there are variations in hourly averaged traffic of about
10 to 20 percent about the average.
Copies of the summary tape are available to interested persons
from the National Climatic Center in Asheville, North Carolina.
The summary tape has been used for the calculation of all the
averages presented in subsequent sections of this report.
Hourly averages
were calculated for the entire period and these have been used for com-
p~~~sons with values determined from the model and its components.
Too
data for each 7-minute period have also been stratified according to
the values of different parameters.
For instance, averages were cal-
culated for all those cases where certain combinations of wind speed and
wind direction (as measured by the TV tower instrumentation) were observed.
The averages of these stratified data have helped us to analyze the
importance of different variables.
3.
Weather Service Data
The hourly meteorological data obtained from the National
Weather Service have been used primarily for the application of the
diffusion model and its submodels.
The information from records of
hourly observations for the period August 20 to October 15, 1971 at
Lambert Field was
copied at the Weather Service office.
The data
copied included those necessary for use with the model, i.e., wind speed
and direction, temperature, and opaque cloud cover.
Eventually the data
needed by the model were transferred to punched cards.
The significant levels of the Salem, Illinois rawinsonde
station were obtained in punch card form from the National Climatic
34
-------
Center in Asheville, North Carolina.
These cards could be used directly
by the mode 1.
The low level sounding data were copied at the National Weather
Service Office in St. Louis.
These sounding data are not the type that
the model was designed to use. However, they have provided good, inde-
pendent estimates of the atmospheric mixing height over the downtown
area for comparison with mixing heights determined by the model using
Salem, Illinois data.
35
-------
III
EVALUATION OF THE MODEL
The diffusion model is composed of a number of submodels.
This
design allows us to check the performance of these submodels individually
and to make improvements where necessary without disrupting the overall
structure.
In the following sections the outputs of the most important
submodels are compared with data specially collected during this program
or with other independent sources of data.
Weaknesses are diagnosed and
improvements made where possible.
.
A.
Mixing Height Submodel
The mixing height submodel has three distinct parts.
First, there
is the part that calculates the mixing height for the early morning hours.
Second is that part that calculates the maximum mixing height during the
day. and finally is the routine that interpolates to obtain values for
the other hours of the day.
Both morning and afternoon mixing heights are based on the nearest
radiosonde temperature profile that is collected at about dawn.
Bot hare
determined from the intersection of this sounding and the dry adiabat
that passes through the temperature at the surface.
For the afternoon
mixing height, the maximum temperature measured at the airport is used.
For the morning value, the downtown urban temperature is determined using
the sounding and an empirical relation developed by Ludwig et ale (1970).
The rationale underlying the model comes from the fact that the lapse
rate within a well mixed layer will be adiabatic in the absence of con-
densation.
The afternoon mixing height calculations also assume no
change in the temperatures above the mixing layer during the period
following the morning sounding.
37
-------
The interpolation during daylight hours is based directly on the
surface temperature variations, and the rationale is the same as discussed
above. During the hours from midnight until dawn, the mixing height is
taken to remain constant at its predawn value.
This is based on obser-
vations that the urban heat island is generally stable, with relatively
constant urban-rural temperature differences (Ludwig and Kealoha, 1968).
From sundown through midnight, time-based interpolation is used.
In this section, the different parts of the model will be evaluated
using different sources of independent data.
*
collected on another SRI project
Several days of lidar data
were available and probably provide
the best check on all aspects of the model.
The lidar mixing height was
determined as the level where there was a sudden decrease in backscattered
energy.
This sudden decrease corresponds to a drop in aerosol concen-
tration.
Since most of the aerosol particles are generated at the surface,
the lidar is essentially measuring the height to which they and other
pollutants are mixed.
This is a direct measurement of the parameter.
Figure 10(a) shows values of mixing height determined from lidar
observations plotted against the values calculated from the model using
the morning sounding taken at Salem, Illinois and hourly temperatures
from Lambert Field.
The figure shows that the model tends to over-
estimate midday values and underestimate dawn values.
Of the 19
cases available, 16 (84 percent) calculated values are within a factor
of 2 of the lidar observations and 12 (63 percent) are within a factor
of 1. 5 .
Temperature soundings made in the urban area are another source
of independent mixing depth measurements. Using the fact that the mixing
layer should be characterized by lapse rates near the adiabatic
*
National Science Foundation Grant GA 30435, "Lidar Observation in
St. Louis."
38
-------
5000
(a)
on
....
III
..
III
E
I ,X
a: xMx X)O(
~
c 1000 X
:J
X X
::i:
0
a:
u..
I-
J:
~
W .
J: .
~ .
Z
X . .
::i:
100
5000
on
..
III
t
E
I
~
z
o
z
5 1000
en
J:
U
a:
~
::i:
o
a:
u..
I-
J:
~
W
J:
~
Z
X
::i:
100
FIGURE 10
(b)
...
X
X
.
. .
. /
X .
.:xx ~ X
"x-, Jf<
X~ X
t
X X
XX
X
x..
X
X
X
X
X
X
100 1000
MIXING HEIGHT CALCULATED WITH MODEL - meters
T A-8563-118
5000
COMPARISONS OF CALCULATED AND OBSERVED
MIXING HEIGHTS
39
-------
(0.010 em-I), the top of the mixing layer should be marked by a sig-
nificant deviation from this value. The objective method that was used
to define the mixing depth selected the bottom of that layer where the
lapse rate falls below 75 percent of the adiabatic.
This criterion was
applied to 43 EMSU soundings taken near midday at the St. Louis Gateway
Arch during the months of August, September, and October 1971.
lowest 100 meters were ignored because it was felt that that layer was
The
too likely to be affected by local heating and hence not representative.
The soundings taken at dawn were often affected by the local sur-
face cooling in the park area surrounding the launch site.
Thus most of
these soundings were typified by an inversion beginning at the surface
and extending to heights of 200 to 750 m.
It was assumed that the
mixing layer over the urbanized area was within that inversion layer.
There were 33 such cases and the mixing height determined from the model
was less than the height of the top of the inversion in all but three
cases.
In the worst of these exceptions, the model specified an 800-m
mixing height, 280 m above the inversion top.
There were ten soundings that could be used to determine mixing
depth at dawn, i.e., they had either no surface inversion or only a very
shallow one.
For these soundings the mixing height was taken to be the
bottom of the lowest elevated fit able layer.
The mixing heights obtained
using this criterion and the midday mixing heights obtained by the
objective technique
described earlier are plotted in Figure 10(b) versus
the corresponding values specified by the model.
The agreement is
generally good.
In 83 percent of the cases the model values are within
a factor of
2
of the values obtained from the sounding; in 68 percent
the agreement is within a factor of 1.5.
This is comparable to the
results that Wuerch (1970) obtained when predicting midday mixing depths
using the dawn sounding at the Arch (as opposed to the Salem, Illinois
sounding) and the midday surface temperature at the Arch (as opposed to
40
-------
the Lambert Field surface temperature).
Thus it appears that the model
does about as well with routine data as is likely to be possible.
Continuous lidar measurements of the mixing height during four
August days were used to evaluate the interpolation scheme used by the
model.
The mixing heights from the two sources are compared in
Figure 11.
The first three days had relatively few clouds that attenuated
the insolation, while the fourth day shown was overcast for most of the
morning and early afternoon hours. As a result surface temperatures rose
rather rapidly after sunrise on the first three days. Therefore, the
calculated value of mixing height increased rapidly because the inter-
polation scheme is based on surface temperature.
The lidar observations
show a much slower increase in mixing height during these hours.
It is
apparent, and reasonable, that the postdawn morning mixing height. which
is a product of solar heating, responds to that heating with considerable
lag.
The model should probably include a provision of this sort, but at
present the available theory and experimental information are both
inadequate to formulate such a feature with confidence that it would be
generally applicable.
On August 23 and 24 the lidar observations continue into the
evening hours and another feature of the mixing layer behavior is
evident.
As cooling at the lowest layers begins in the late afternoon
and early evening, a stable layer develops near the surface.
This stable
layer caps the mixing, so the transition from the afternoon to the
evening mixing heights is nearly discontinuous; there is an abrupt drop.
The performance of the model could undoubtedly be improved if this
behavior were included in the interpolation scheme.
Unfortunately: the
available theory and experimental data again do not permit this to be
incorporated at this time.
The interpolation scheme used by the model does a good job of
showing the diurnal mixing height trend on the overcast day, August 25.
41
-------
2000
1500
1000
500
o
2000
~ 1500
'"
...
'"
E
I 1000
I-
J: 500
C)
w
J:
~ C) 0
l" Z
X 2000
~
1500
1000
500
o
1000
500
o
o
- - - LIDAR
13 August 1971
'"
/
/
/
CALCULATED
23 August 1971
"'" ..........
,- ./ ....
- --
"
\
",.
24 August 1971
,.
I
-_/
..-
\
"-
-,
,
/
..... -
'"
.."
- -
- --
25 August 1971
--
-'.;' "
---
-----
- -
'"
- - --"""
--
2
3
24
5
6
4
7
8
9
10
11 12 13
TIME - CST
14
15
16
17
18
19
20
21
23
22
T A-8563-117
FIGURE 11
VARIATIONS OF CALCULATED AND LlDAR-OBSERVED MIXING HEIGHTS WITH TIME
-------
Thus any future modifications of the mixing height submodel designed to
reflect the morning lag in the response of mixing to solar heating
should probably be made dependent on the rate of change of temperature
or on the insolation strength.
Since the latter is already determined
in the stability submodel, it might be the most convenient.
The abrupt
ev~ning mixing height decline is probably dependent on rate of cooling
and hence cloud cover, although we do not have the data to confirm this.
The sensitivity of the model to mixing height variations has been
discussed in an earlier report (Ludwig et al., 1970).
Since that dis-
cussion, the functions used to describe the variability of a
z
with
stability and travel distance have been changed (Johnson et al., 1971) ,
and this somewhat changes the model's sensitivity to mixing height.
The
results are qualitatively the same as before, but differ slightly in
detail.
Figure 12 shows normalized concentration as a function of mixing
height and stability class, using the new a
z
functions.
Figure 12(a)
assumes source strength to be uniform and equal to 1 from the receptor
to 32 km upwind, where it becomes zero.
The average concentration is
the same in Figure 12(b) , but decreases linearly from 2 at the receptor
to zero at 32 km.
The latter case is more nearly like the source field
surrounding a receptor at the center of the city.
For small mixing heights the concentration will be nearly pro-
portional to the reciprocal of the mixing height.
As the mixing height
increases, the dependence of concentration on it is reduced.
Too
dependence is less pronounced for the more stable cases and for the more
realistic source field.
This is because the mixing height dependence is
the result of those parts of the calculation based on the box model.
The
Gaussian model is used for calculations of concentrations due to emissions
from segments close to the receptor.
The box model, if used at all,
applies to the farther segments.
In the linearly decreasing source
43
-------
1000
800
600
400
200
o
-
U 100
::J
80
~ 60
~
40
SLIGHTLY STABLE
20
SLIGHTLY UNSTABLE
MODERATELY UNSTABLE -----
EXTREMELY UNSTABLE
la) UNIFORM SOURCE STRENGTH
NEUTRAL
50
500
1000
50
MIXING HEIGHT - m
1000
10
100
SLIGHTLY STABLE
NEUTRAL
Ib) SOURCE STRENGTH DECREASING
LlNEARL V FROM RECEPTOR
100
500
TA-8563-38R
FIGURE 12
NORMALIZED CONCENTRATION AS A FUNCTION OF STABILITY AND MIXING HEIGHT
-------
configuration the emission rates in the more distant segments are less
than those in segments close to the receptor.
Thus, the contributions
that the box model makes to the results are smaller than those for the
uniform area case.
This decreases the effect of mixing depths upon the
normalized concentrations.
The results displayed in Figure 12 suggest
that,for some values of mixing depth, the results obtained from the
model are nearly independent of stability and, for certain stability
types, changes in the mixing depth produce no changes in the results, if
the mixing depth is sufficiently large.
Thus, the model's dependence on
mixing height will always be less than proportional to the reciprocal of
the mixing height.
The effects of mixing height are further reduced
when the street canyon model is included.
Since the concentrations con-
tributed by local sources and street canyon effects are independent of
mixing height and are often as large or larger than the concentrations
arising from more distant sources, the percentage effects of mixing
height will be further reduced.
The above discussion is not intended to minimize the importance
of improving the mixing depth submodel.
It should be realized though
that mixing depth errors do not translate directly into concentration
errors.
The uncertainty factor of 1.5 in mixing height probably
results in uncertainties in street canyon concentrations of only about
20 percent or less.
B.
Stability
The purpose of the stability submode1 is to specify the variations
of the vertical dispersion parameter cr with travel distance. The
z
model's selection of stability category depends on the radiative balance
at the surface and the low-level wind speed, as suggested by Pasquil1
(1961).
Radiative balance is deduced from hourly cloud cover obser-
vations and the solar elevation angle.
As we have already noted,
fluctuations of wind direction are also related to the Pasqui11 stability
categories (Slade, 1968).
45
-------
Standard deviations of the wind directions observed on the tower
were determined for 7-minute observation periods.
These standard devi-
ations were calculated as the root-mean-square difference between about
75 individual observations and the vector average direction for the
period.
These 7-minute standard deviations may be slightly smaller than
values obtained for the recommended 10 minute to 1 hour sampling periods.
The 7-minute standard deviations were averaged over an hour period and
compared with the stability categories determined from Turner's (1964)
algorithm, using hourly meteorological observations from the National
Weather Service at Lambert Field.
Table 5 is a contingency table based
on more than 690 cases.
It shows percentage of occurrence of various
combinations of Turner stability class and the corresponding values
based on wind direction variability.
In nearly 80 percent of the cases
the two agree within one class.
The table shows that Turner's algorithm
has a bias such that it tends to define the stability as somewhat greater
than actually observed in this urban location.
This is not surprising
because the city's roughness is likely to cause greater standard devi-
at ions of wind direction than would be expected in rural areas from which
the data were obtained to define stability classes in terms of wind
direction variations.
In earlier phases of this program, an attempt was made to develop
a simpler method for determining stability.
The approach taken was to
specify the level of daytime insolation in terms of the insolation
parameter ~, described by the following equation:
~ = (1 - 0.5 N) sin a
(2)
where N is the fraction of the sky covered by opaque clouds and a is the
solar elevation angle.
Originally we classified the daytime insolation
so that the categories covered equal ranges of the value of ~, e.g., less
than 0.33 was slight insolation, greater than 0.67, strong.
During this
46
-------
,j:>.
....:j
Table 5
PERCENT OCCURRENCE OF VARIOUS COMBINATIONS OF WIND
DIRECTION VARIABILITY AND TURNER STABILITY CLASS
Standard Deviation of Wind Direction (Degrees)
and Corresponding Stability Class
Stabili ty > 22.55 17.55-22.55 12.55-17.55 7.55-12.55 3.75-7.55 < 3.75
Class
Extreme ly Moderately Slight ly Neutral Slight ly Moderately
Unstable Unstable Unstable Stable Stable
Extremely
Unstable 0.1 0.1 0.1 0.0 0.0 0.0
1
Moderately
Unstable 2.0 0.4 1.2 0.1 0.0 0.0
2
Slight ly
Unstable 4.2 1.6 4.9 1.2 0.3 0.3
3
Neutral 3.5 5.1 15.9
4 20.8 7.2 0.6
Slight ly
Stable 0.7 1.0 2.5 11.8 11.0 3.5
5
Moderately
Stable Not used for ur :>an cases
6
-------
program, 141 hours of pyrheliograph data were collected to check the
suitability of these categories.
The categories,
strong, moderate,
and slight insolation, were originally defined by Pasquill (1961).
Strong insolation was defined as
"...sunny midday conditions in mid-
summer in England."
According to Chandler (1965), the corresponding
value of insolation measured at Kew Observatory is 1.0 cal cm-2 min-1.
Using this as a guide the following insolation classes would fit Pasquill's
definition of strong insolation:
slight insolation:
< 0.4 cal cm-2 min-1
moderate insolation:
0.4 - 0.8
strong insolation:
> 0.8 .
However, Pasquill has also specified slight insolation to be typical
sunny, midwinter, midday conditions, which would be about 0.6 cal cm-2
min-1 according to Chandler.
This seems inconsistent, in that its use
would result in a very large class for slight radiation and a very small
moderate insolation class.
The class intervals given above seem more
reasonable and lead to reasonably good agreement with Turner's insolation
categories based on solar elevation, cloud height, and sky cover; as shown
in Table 6.
This contingency table shows the relative frequencies of
Table 6
JOINT OCCURRENCE OF TURNER'S NET RADIATION
INDEX AND OBSERVED INSOLATION
Turner Net Insolation, -2 min-1
Cal cm
Radiation
Index < 0.4 0.4 - 0.8 ~ 0.8
o or 1 29.8% 19.9% 5.0%
2 2.1% 24.1% 9.9%
3 0.7% 2.8% 5.7%
48
-------
various combinations of measured insolation, classed as above, and
Turner's net radiation index.
As can be seen from the table, there
are a substantial number of instances where moderate or strong insolation
is observed
but given a low radiation index by Turner's algorithm.
This
generally occurs during overcast conditions.
The method tends to over-
estimate the effect of low overcast clouds on the insolation.
Table 7 shows the results of cross classifying measured insolation
with equal intervals of ~.
This table shows that these intervals of ~
generally lead to greater underestimates of insolation than does Turner's
algorithm.
This suggests that the ~ categories should be revised. It
would seem reasonable to use a classification system for ~ that was
closely related to the classes of observed insolation.
Maximum tropical
Table 7
JOINT OCCURRENCE OF INSOLATION CLASSES
(after Ludwig et al., 1970) AND OBSERVED INSOLATION
Insolation, Cal cm-2 min-1
~
< 0.4 0.4 - 0.8 ~ 0.8
< 0.33 27.7% 6.4% 0.0%
0.33 - 0.66 5.0% 37.6% 18.4%
> 0.66 0% 2.8% 2.1%
sea level insolation is about 1.5 cal cm-2 min-1 (derived from Smithsonian
Meteorological Tables, 1951).
We have already concluded that the lower
bound of the strong insolation class is at a value slightly more than half
this maximum possible.
Therefore, a value of ~, slightly more than half
its maximum possible value, should be appropriate for a class boundary.
Table 8 shows the results obtained when the class intervals of ~ are
revised accordingly.
The revised categories result in very good agreement
49
-------
Table 8
JOINT OCCURRENCE OF REVISED INSOLATION CLASSES
AND OBSERVED INSOLATION
Insolation, Cal cm-2 min -1
~
< 0.4 0.4 - 0.8 ~ 0.8
< 0.3 27.0% 2.8% 0.0%
0.3 - 0.55 5.7% 35.5% 7.1%
> 0.55 0.0% 8.5% 13.5%
with the measured insolation.
Accordingly, the stability submodel has
been changed to incorporate these new insolation categories:
slight insolation:
~
< 0.3
moderate insolation:
0.3 ~ ~ ~ 0.55
strong insolation:
~
> 0.55
Another revision, based on Pasquill's (1961) article, has also been
incorporated.
Now, overcast (opaque cloud cover ~ 9/10) conditions,
night or day, are taken to have neutral stability.
Neutral stability
o
also prevails during the day when the sun is within 15 of the horizon.
o
The 15 solar elevation criterion has replaced the earlier one (Johnson
et al. , 1971) based on number of hours before sunset or after sunrise.
The two are virtually identical in their effect. Table 9 below summarizes
the revised criteria used in the stability submodel.
The revised model
was applied to the hourly meteorological data from Lambert Field so that
it could be compared with the wind direction standard deviations as was
done with Turner's stability categories.
Table 10 shows the results of
this comparison.
Reference to Tables 5 and 10 shows that the stabilities
obtained with the revised submodel are in somewhat better agreement with
the wind variabilities than are the Turner values.
More than 87 percent
50
-------
Table 9
REVISED STABILITY CATEGORIES
Opaque Cloud Cover Nighttime
Surface Daytime 0 ;::: 9/10, Day or Night
Solar Elevation Ang 1e > 15
Winds or
(knots) Strong Moderate Slight Solar Elevation Angle ;::: 5/10 ~ 4/10
Insolation Insolation Insolation ~ 150 Clouds Clouds
~ 3 1 2 2 4 5 5
3-6 1 2 3 4 4 5
6-10 2 3 3 4 4 4
10-12 3 3 4 4 4 4
;::: 13 3 4 4 4 4 4
of the stabilities determined by the revised submode1 were within one
category of those based on wind observations; with Turner's method 80 percent
of the cases were within one category.
There is a slight bias in this method,
in the direction that would be expected for urban wind direction fluctuations.
It was expected that the wind direction fluctuations prevailing
during a given stability condition, as determined by the model, would vary
according to the roughness of the underlying surface.
To test this, the
data were stratified according to wind direction.
The experimental location
is near the eastern edge of the central business district, with generally
open areas farther to the east, as can be seen in Figures 1 and 4.
Thus
winds with a component across Broadway from the east would have a consider-
ably smoother upwind fetch than those with a component across Broadway from
the west.
The wind direction measured at the upper level of the tower was used
to stratify the data according to whether there was an upwind urban fetch
(326 hours of data) or a nonurban fetch (309 hours).
Tables 11 and 12
summarize the results.
.
The difference between the two is rather striking.
51
-------
Table 10
PERCENT OCCURRENCE OF VARIOUS COMBINATIONS OF WIND
DIRECTION VARIABILITY AND REVISED MODEL STABILITY CLASS
CJ1
t.:>
Standard Deviation of Wind Direction (Degrees)
and Corresponding Stability Class
Stability > 22.55 17.55-22.55 12.55-17.55 7.55-12.55 3.75-7.55 < 3.75
Class
Extremely Moderately Slight ly Neutral Slight ly Moderately
Unstable Unstab Ie Unstable Stable Stable
Extremely
Unstable 2.9 0.9 1.1 0.3 0.0 0.0
1
Moderately
Unst ab Ie 3.6 1.4 5.0 1.0 0.3 0.1
2
Slight ly
Unstable 2.0 2.6 7.5 5.2 0.3 0.0
3
Neutral 1.0 2.9 9.8 21.4
4 10.8 1.1
Slight ly
Stable 0.9 0.7 1.4 6.2 6.6 3.0
5
Moderately
Stable Not used for urJ an cases
6
-------
Table 11
PERCENT OCCURRENCE OF VARIOUS COMBINATIONS OF WIND
DIRECTION VARIABILITY AND REVISED STABILITY CLASS
WITH URBAN FETCH
tJ1
CIJ
Standard Deviation of Wind Direction (Degrees)
and Corresponding Stability Class
Stability > 22.55 17.55-22.55 12.55-17.55
Class 7.55-12.55 3.75-7.55 < 3.75
Extremely Moderately Slight ly Neutral Slight ly Moderately
Unstable Unstable Unstable Stable Stable
Extremely
Unstable 1.8 0.3 0.3 0.0 0.0 0.0
1
Moderately
Unstable 4.0 0.9 1.8 0.6 0.0 0.0
2
Slight ly
Unstable 3.4 3.1 10.7 4.3 0.0 0.0
3
Neutral 1.2 3.1 13.5 25.5 4.9 0.0
4
Slight ly
Stable 1.2 0.6 1.2 7.1 7.7 2.8
5
Moderately
Stable Not used for ur an cases
6
-------
Table 12
PERCENT OCCURRENCE OF VARIOUS COMBINATIONS OF WIND
DIRECTION VARIABILITY AND REVISED STABILITY CLASS
WITH NONURBAN FETCH
t1I
,po.
Standard Deviation of Wind Direction (Degrees)
and Corresponding Stability Class
Stability > 22.55 17.55-22.55 12.55-17.55 7.55-12.55 3.75-7.55 < 3.75
Class
Extremely Moderately Slightly Neutral Slight 1y Moderately
Unstable Unstab Ie Unstable Stable Stable
Extreme ly
Unstable 2.9 1.3 1.6 0.3 0.0 0.0
1
Moderately
Unstable 3.2 2.3 7.8 1.3 0.3 0.3
2
Slight ly
Unstable 1.0 1.3 4.9 6.1 0.6 0.0
3
Neutral 1.0 1.6 4.2 18.1 16.5 2.6
4
Slight ly
Stable 0.6 1.0 1.3 6.5 7.4 3.9
5
Moderately
Stable Not used for u ban cases
6
-------
With an urban fetch more than 88 percent of the cases have wind
variabilities within one category of that predicted from the hourly
meteorological observations by the revised stability model; however,
the bias is quite pronounced.
In only 15 percent of the cases is the
wind variability less than predicted; in more than 38 percent it is
greater.
For the
nonurban
fetch cases the situation is nearly reversed.
More than 86 percent of the cases are still within one category of that
predicted, but almost 43 percent have wind direction variability less
than the appropriate value.
Less than 22 percent have greater vari-
ability.
There is little question that the rougher urban surface
causes significantly greater wind direction fluctuations and hence more
rapid diffusion than is found over the smoother areas.
These results,
at least qualitatively, support the adoption of different cr
z
functions
for urban and rural modeling.
This was done in the previous report for
this project (Johnson et al., 1971), and the functions derived in that
report to fit urban conditions will continue to be used in the model.
The results of these studies suggest that future investigations of the
effects on diffusion of different types of urban and rural surfaces
could be quite profitable.
However, at present, the combination of the
revised stability model and the urban cr functions seems to be a very
z
good compromise for use in urbanized areas.
c.
Emissions Submodel
There are at least three potential sources of error in calculating
the emissions from the traffic data.
These are:
(1) inaccurate estimates
of hourly traffic volume, (2) inaccurate estimates of average vehicular
speeds on the links, and (3) an inappropriate equation relating emission
rates to the preceding two factors.
These factors were all considered
in the preceding report (Johnson et al., 1971), but data collected in
St. Louis makes their reconsideration worthwhile.
55
-------
1.
Traffic Volumes
The detailed traffic data from San Jose indicated that the
traffic on any given street link, for a given hour of the day, does
not vary much from one weekday to another.
The standard deviations of
day-to-day values of half-hour traffic volumes in downtown San Jose
at a given time of day were found to range from about 5 to 15 percent
of the average volume.
The St. Louis data collected at the street
canyon experimental site showed somewhat greater variability, but this
may only reflect the fact that a new freeway was opened at the southern
edge of the downtown section at about the time the experiment started,
so traffic patterns may have been in a state of flux.
In any event,
the conclusion reached in San Jose that the variability of 24-hour
traffic volumes is less than 15 percent still seems reasonable.
This
could not be checked accurately because of previously mentioned dif-
ficulties in collecting overnight traffic data.
However, enough data were collected to construct mean diurnal
traffic cycles that we believe are more reliable than our earlier sources.
The results are shown in Figure 13 where they are compared with two
weekday cycles obtained from other sources.
The curve marked "original"
is a composite diurnal emission cycle that has been used with the model.
The curve marked "statistically determined" was derived from observed CO
data at the St. Louis Continuous Air Monitoring Program (CAMP) Station,
using a least-squares approach to give the best agreement between model
calculations and observed values.
It is not a traffic cycle, but an
emission cycle.
The corresponding traffic cycle would show lower peaks,
because emissions per vehicle mile are increased by the slower speeds
that accompany rush hour traffic.
Similarly. off-peak traffic volumes
are greater than those shown for emissions by the curves.
56
-------
en
..J
«
I-
o
I-
>-
«
c
~
w
w
:5:
w
C)
«
II:
w
>
«
u.
o
z
o
i=
<..>
«
II:
u.
en
..J
« 0.08
I-
0
I-
>-
«
c
~
w
w
:5: 0.06
w
C)
«
II:
w
>
«
u.
0
Z 0.04
o
i=
<..>
«
II:
u.
0.02
0.06
WEEKEND MEASURED, BROADWAY AND LOCUST
0.04
SATURDAY....... ......"',
......., , '"
, '--".....' \ ....,
-' \' ,
,.. \, ,
, " I
I ~-J
,
I
I
I
,
,
,
,
,
!I. ,
~~....~...~
0.02
o
WEEKDAY
0.10
~,
STATISTICALL Y :..,'\
. \' ,
DETERMINED......... I , ' \
-.....,' ,
I ,
I ,
I ,
I ,
I , ~-
:-,.. , \ "
, 1 \
MEASURED, I' ..,'
BROADWAY;: \ .'\ './-""'."""
AN D LOCUST......... .' \ / . A,'
-....., : .-' V',
. I
,
,
I
,
I
,
I
I
I
I
I
I
I
I
I
I
. I
\ :
. I
\ :
. I
\:
~
~
I.
:\
\\ ''''''''''.
\
\
,
,
,
,
,
,
\
" ,...,
',' "
ORIGINAL ""-
o
o
5
10
15
20
HOUR OF DAY
T A-8563-70R
FIGURE 13
DIURNAL EMISSION PATTERNS FOR ST. LOUIS
57
-------
The traffic cycle derived from the combined data collected
on Locust and Broadway during the project is also shown in the figure.
The measured cycle shows the high midday traffic that was implied by
the statistically determined curve, but the morning and afternoon peaks
are not so large.
The predawn traffic is also greater than expected.
Saturday and Sunday traffic cycles are also shown in Figure 13.
These curves are expressed in terms of fractions of total weekday traffic.
Therefore, the sums of the fractfons for each hour do not total one.
The
total Saturday traffic is about 67 percent of the weekday totals.
The
total Sunday traffic is about 40 percent of the weekday value.
The model has been revised to apply the weekday traffic cycle
measured during this project to all the downtown streets and to the
street canyon submode1 (see Section III-E).
The data collected on two
days in San Jose (Johnson et a1. ,1971) showed very much the same general
daytime trends as those displayed in this St. Louis data.
It therefore
seems reasonable to adopt this diurnal cycle for downtown traffic in
genera1,in lieu of any better information.
However; it is recommended
that the model use local data if they are available.
2.
Traffic Speeds
During the course of the mobile CO measurement program, an
accurate log was kept of the movement of the van in the downtown area.
Travel times along the route segments were noted to the closest 0.1 minute;
the average segment length was 0.71 km.
Observations were made on five
days and during five different periods on each day:
(1) early morning
rush (0730-0830 CDT) , (2)
midmorning (0930-1030), (3) noon (1145-1245),
(4) afternoon (1430-1530), and (5) evening rush (1630-1730).
An average
of five circuits were made each period around the downtown loop i1lus-
trated in Figure 4.
The route has its fair share of lights, stop
signs, crosswalks, and heavy traffic and is quite typical of traffic
58
-------
Tab Ie 13
A VERAGE VEHICLE SPEEDS (mi h -1) IN DOWNTOWN ST. LOUIS
DURING THE PERIOD 24 AUGUST TO 2 SEPTEMBER 1970
tJ1
to
Broadway Market 12th Delmar
Average
Period (CDT) Ave. No. of Ave. No. of Ave. No. of Ave. No. of Period
Speed Cases Speed Cases Speed Cases Speed Cases Speed
0730 - 0830 9.2 25 8.9 24 9.6 24 8.6 22 9.1
0930 - 1030 10.1 20 7.8 20 9.6 20 7.5 20 8.8
1145 - 1245 9.7 18 7.8 18 8.7 18 8.0 16 8.5
1430 - 1530 10.0 23 9.0 23 10.5 22 7.8 21 9.3
1630 - 1730 8.8 22 9.0 22 8.6 22 7.5 20 8.5
Average
Segment 9.1 8.5 9.4 7.8 8.8
Speed
-------
flow in the central business district.
Average speeds for the various
periods and route segments are given in Table 13.
When the uncertainty
in the segment time is considered together with the overall average van
speed of 8.8 mph, the uncertainty in the individual computed speeds is
on the order of 3 percent or 0.3 mph.
The downtown average speed of 8.8 mi h-1 is less than the 14.1
mi h-1 average determined earlier (Johnson et al., 1971) for San Jose and
considerably less than the assumed st. Louis suburban arterial speed
of 20 mi h-1.
In practice, these localized speed values are quite easy
to obtain and it is recommended in using the model that they be measured
in the area of interest in view of the variations in traffic conditions
among cities and the sensitivity of the model to them.
The traffic speeds
appropriate for application of the model in St. Louis are summarized in
Table 14.
Although we have obtained refined values of vehicle speeds, we
have not collected data under this program that can be used for eval-
uation of the speed-dependence of the emissions submodel.
The mode 1 is
relatively sensitive to errors in speed, particularly at low values.
The overall performance of the model indicates that the emissions cal-
culations are reasonable, but additional work is desirable to thoroughly
evaluate the current submodel as well as possible alternatives.
3.
Emissions Formula
The validation of the emissions submodel has centered on the
evaluation of the predicted emissions [using Eq. (A4) in Appendix AJ
with independent measurements of the parameter.
Traffic speeds were
obtained in the manner described in the preceding section.
Daily
traffic volume in the downtown area was determined from Missouri State
Highway Department (1970) data.
Estimates of hourly averages were
derived from Figure 13.
Vehicular emissions could not be measured
directly, but were computed on the basis of a CO mass budget analysis
using surface (van) and aerial (helicopter) measurements around a 0.7-km
square encompassing much of the St. Louis downtown area.
60
-------
Table 14
VEHICLE SPEEDS FOR ST. LOUIS
*
Traffic Link Type Average Off-Peak
Vehicle Speed (mi h-1)
Freeway:
Downtown 43
Suburban 53
Arterial:
Downtown 9
Suburban 20
Local 12
* Peak-hour speeds are assumed to equal
85 percent of the off-peak values.
A major difficulty in applying the budget technique is the
definition of the wind field within and over the downtown district.
One
is faced with the dilemma that the largest concentrations and gradients
of CO concentration are found near the lower boundary which, however,
is beneath the zero-plane displacement height of the vertical wind
profile.
The mean areal wind structure up to a level of approximately
75 m is difficult, if not impossible,to define.
Furthermore, the air
in moving from the relatively smooth suburban and rural environs to the
aerodynamically rough urban core undergoes marked horizontal and vertical
accelerations.
Therefore, the three-dimensional structure of the urban
wind field is poorly defined and variable in both time and space.
Wind measurements available for the analysis are of two types:
(1) the tWice-daily profiles obtained at the EMSU station located 0.8 kID
to the southeast of the center of the study area, and (2) continuous
61
-------
observations at the 90-m and 131-m levels on the specially instrumented
tower on the east side of the study area.
The EMSU data were used to
describe the vertical wind structure for those analyses made near the
times of the soundings.
For lack of more detailed data and knowledge
on low-level urban wind structure, the effective horizontal transport
of CO in the lowermost stratum was taken as the average of the transport
at the bottom and top of the layer.
The same procedure was used for
the upper strata.
The mean vertical CO flux at the 335-m level was
assumed to be zero; the net horizontal flux of the uppermost layer is .
then equal to the vertical flux through the bottom level of that layer.
By working down to the surface, the vertical CO-flux profile is obtained
as well as the areal average of surface emissions.
These can then be
compared with submodel predictions from the traffic data.
Of necessity, a streamlined procedure was necessary when EMSU
wind data were not available.
For these cases the tower winds were
used and it was assumed that the transport in the lowermost stratum
could be estimated by using an effective layer wind equal to two-thirds
of the 90-m observation.
Winds in the upper layers were assumed to be
constant and equal to the value of the 130-m tower observation.
These
assumptions are necessarily coarse, but probably result in a minimum
probability for biasing the computations.
An example of a sequence of concentration profiles obtained by
the helicopter on 31 August is
illustrated in Figure 14; corresponding
vertical suburban temperature profiles are shown in Figure 15.
Because
of the assumptions on the wind structure discussed above, it is practical
to analyze only those cases with relatively shallow mixing depths and
high concentrations aloft in order to minimize the influence of the
surface (van) observations on the computations.
The upper level measure-
ments are more representative of a true line average, because of the
integrating effect of atmospheric motions, while the van observations
62
-------
8
7
6
5
4
E
a.
a. 3
Z 2
o
i= 4
«
OC
f-
Z 3
LU
U
Z
0 2
u
LU
0 1
x
0') O
w z 3
o
:2:
z 2
o
IX!
OC 1
«
u
0
3
2
...
.,
1
1
1
1
I
........ -__--__1
-.c.. ..::.'"""--
.-.-./ '-.
... "., -.
..
~.-'7--
---
....' '-',
. ._._._._._._._.--~ - -'-215m .,
./ ~~~ =-=---.:
---~.:-=~- 500 m
115 m
1145 CDT
~
o
6
(SE)
7
9
10
12
14
17
18
19
20
21
(NE)
15
16
(NW)
13
1
(NE)
2
4
5
8
11
(SW)
TRAVERSE POSITION NUM8ER
3
T A-8563-122
FIGURE 14
SPACE AND TIME VARIATION OF CO CONCENTRATION OVER DOWNTOWN ST. LOUIS,
31 AUGUST 1971
-------
100
650
600
550
500
450
400
E
I 350
I-
J:
CJ
W 300
J:
0810 CDT 0945
250
200
150
50
o
-9
-6
-8
-7
-5
-3 -2 -1 0 1 2
RELATIVE TEMPERATURE - oK
3
-4
4
6
7
8
5
TA-8563-127
FIGURE 15
ST. LOUIS SUBURBAN TEMPERATURE PROFILES
64
-------
can easily be biased by the influence of sources close to the sampling
intake. This effect can deteriorate the quality of the cross-town
gradient computed at the surface.
Furthermore, it is necessary to
I
limit the analyses to cases where the emissions and meteorological
fields do not vary rapidly with time.
Observations made at 0800 CDT, 31 August satisfy these require-
ments sufficiently to permit a mass budget analysis for the estimation
of the surface CO flux.
The concentration and meteorological data are
tabulated in Table 15 and the resulting fluxes are illustrated in
Figure 16.
The average surface emission rate of 571 gm (CO)
-1
sec '
computed from the analysis is 2.5 times the value calculated from Eq. (4)
using traffic data.
Another analysis was made using data collected during the
evening rush period on 23 August (see Table 15).
The surface flux
computed from the budget analysis is about 75 percent of the value pre-
dicted by the submodel from the traffic data for the period.
An inter-
esting feature of the results of the analysis (Figure 17) is the net
horizontal influx of CO in the bottom layer.
The winds in that layer
are a factor of 2-to-3 times larger than those in the previous case.
Under these conditions, it is possible that there is a pronounced
lifting of the air near the surface as it strikes the relatively dense
face of the tall buildings in the urban core region.
This could cause
the higher upwind concentrations observed at the surface and the higher
downwind values (as one would expect) aloft.
A more striking example
of this phenomenon is presented in the following analysis.
On the morning of 24 August, an interesting and unusual CO
distribution pattern was observed:
Street level concentrations observed
by the mobile van were consistently high (~ 35 ppm) and decreased down-
wind across the study area; concentrations measured by the helicopter
65
-------
Table 15
CARBON MONOXIDE AND WIND DATA
FOR BUDGET ANALYSES
CO Concentration (ppm) Street-Oriented
Height Wind Components
(m) Broadway Market 12th Delmar -1
u(m s ) v(m s-l)
0 23.7 22.1 22.6 22.7 -0.7 -0.8
115 3.3 4.5 7.1 5.7 -1.4 +1.6 Case I:
150 3.3 3.9 6.5 4.8 -0.5 +2.2 0800 CDT
215 2.7 2.6 2.5 2.7 +0.9 +3.0 31 August
275 2.4 2.4 2.3 2.3 +1.6 +3.8
335 2.4 2.4 2.4 2.4 +2.3 +4.5
0 18.3 16.3 15.4 17.2 -1.0 +1.8
Case II:
115 2.0 2.3 1.8 2.5
-0.8 +3.2 1730 CDT
150 1.6 1.9 1.3 1.8 23 August
-0.8 +3.2
215 0.9 1.1 1.4 1.4
-0.8 +3.2
275 1.0 1.1 1.4 1.8
-0.8 +3.2
335 0.3 0.3 0.7 0.5
0 37.0 34.6 34.8 34.4 -0.7 +1.9
115 0.4 0.4 0.6 0.5 +1.4 +3.8 Case III:
137 0.4 0.6 0.8 0.6 +2.1 +3.7
150 4.3 2.4 4.2 4.4 +2.6 +2.7 0730 CDT
215 1.8 1.1 2.1 0.9 +3.5 +3.5 24 August
275 0.5 0.5 0.5 0.5 +3.9 +3.9
335 0.1 0.2 0.2 0.1 +4.2 +4.2
66
-------
.
u - Transport out
u - Transport in
v - Transport in
gm (CO) see-I
.
~..........
I......."".~
v - Transport out
335
o
274
0
5
........... 3
213 2
] 44
I- .........~ 61
:I:
"
W
:I: 152
107 112
..........~ 63
115
282
227
.........~ 62
TA-8563-121
FIGURE 16
CARBON MONOXIDE MASS BUDGET ANALYSIS, ST. LOUIS
0800 COT, 31 AUGUST 1971
67
-------
.
u - Transport out
u - Transport in
gm (CO) sec.1
t
............ v Transport in
..........~ v - Transport out
335 0
10
..........~ 68
274
78
21
..........~ 83
213
182
] 6
I- ..........~ 15
:I:
C)
w
:I: 152
203 II( 5
..M.......~ 8
115 206
II(
160
..........~ I I 0
"
TA-8563-128
FIGURE 17
CARBON MONOXIDE MASS BUDGET ANALYSIS, ST. LOUIS
1730 COT, 23 AUGUST 1971
68
-------
were low and uniform at the l15-m and l35-m levels, but then increased
dramatically at 150 m and remained high at 215 m, dropping again to
consistently low values at 275 m and 315 m.
These patterns are illus-
trated in Figure 18 together with the corresponding vertical profile of
temperature; the wind and average CO values are summarized in Table 15.
In spite of the stable conditions, the winds were moderately strong from
the south through southwest.
Again, the explanation for the pattern may
be the f~rographic" lifting of the air as it encoupters the wall of
buildings.
This would have the effect of decreasing the surface concen-
trations downwind, while the tilt of the flow field could have the effect
of raising the CO content of the air at the intermediate levels.
It is,
however, puzzling that only the two intermediate levels show the high
concentrations.
This pattern might occur if the tilt of the updraft
were just right and the vertical diffusion sufficiently damped.
These
arguments are only offered in the way of conjecture, but the situation
does illustrate some of the complexities of understanding and simulating
urban dispersion.
Figure 19 illustrates the results of a budget analysis retaining
the restriction of zero flux through the uppermost level, while Figure 20 is
ap alternative analysis where the zero-flux restriction has been modified
to ensure no negative vertical fluxes.
Traffic data were not yet avail-
able from the street experiment on this date and it has been assumed that
conditions were similar to those experienced during the 0800 CDT case,
one week later.
Under the zero-flux assumption, the budget calculation
is 25 percent of the submodel prediction; under the alternative analysis
assumption,the submodel value is only 80 percent greater.
The budget analyses are interesting and serve to illustrate the
features of urban-scale dispersion, but they have not provided data of
sufficient quality to objectively analyze the performance of the emissions.
69
-------
E 5
a.
a.
Z
0
i= 4
-------
.
u - Transport out
u - Transport in
v - Transport in
...........-
v - Transport out
I gm(CO) S.C-I
c
~..........
335 0
12
.......... II
274
23
27
.......... 16
213 66
20
]
I- .........~ 173
J:
C)
w
J: 152 87 4
..........~ 47
137 130 13
..........~ 5
115 122
........... 78
TA-8563-129
FIGURE 19
CARBON MONOXIDE MASS BUDGET ANALYSIS, ST. LOUIS
0730 CDT, 24 AUGUST 1971
71
-------
.
u - Transport out
u - Tra nsport in
gm{CO )see-I
(
...................
v - Transport in
v - Transport out
..........I~
335 66
"III 12
.......... II
274 43
27
........... 16
213 0
20
]
I- .........~ 173
J:
1.7
w 152
J: 153 4
..........~ 47
137 196 13
.........~ 5
115 188
TA-8563-130
FIGURE 20
CARBON MONOXIDE MASS BUDGET ANALYSIS, ST. LOUIS
0730 COT, 24 AUGUST 1971
72
-------
submode 1.
To do this would require an experimental program where all
the variables could be monitored with the precision necessary to fully
describe the traffic and meteorological conditions and the resulting
pollution concentrations.
D.
Winds
-
Winds are an important input to many aspects of the model.
The
wind direction is used to determine those sources that will contribute
to concentrations at the receptor.
The direction is also used in the
application of the street canyon effects submodel.
Wind speed enters
as a dilution factor in both the Gaussian and box model formulations;
it is also required for the stability and street submodels.
Recognition of the importance of more detailed wind (and temperature)
information for the prediction of atmospheric dispersion has prompted the
National Weather Service to develop the EMSU program; as these data
become available on a routine basis, additional model improvements can
be anticipated.
However, the current program is a limited one--
measurements are made at only a few cities and even then on a relatively
infrequent basis (twice 'daily on weekdays).
Therefore, in formulating
the model, we have assumed that winds measured at the airport will be
the only measured values generally available for use with the model.
These winds have been assumed to be uniform over the urban area.
In
this section, the validity of this assumption will be discussed.
Much
of this discussion is based on a recent study in St. Louis by Wuerch
(1971) .
Wuerch has compared airport winds with the "transport" winds computed
from the EMSU radiosonde data obtained at the Gateway Arch.
The radio-
sonde site is about 21 km southeast of Lambert Field.
The transport winds
are a vector-average of the winds in the mixing layer.
The airport wind
speed that is used for comparison is a scalar average over I-hour time
periods, and therefore differs somewhat from the data used as inputs to
73
-------
the model.
The model input data are the routine, subjective one-minute
averages as determined by a weather observer.
Wuerch has studied the relationships between airport and transport
winds at both dawn and midday.
The dawn results are based on 134 sets
of data collected during the period of November 1969 through June 1970.
The midday results are based on 150 data pairs collected from September
1969 through June 1970.
For the midday cases, Wuerch found that the airport wind speed u
a
was h~ly correlated (correlation coefficient, 0.82) with the transport
wind speed u. The standard error of estimate was 1.8 m s-l about the
t
regression line,
u
t
= 0.576 + 1.174 u
a
(3, midday)
The dawn cases were slightly less correlated (0.76) and had a some-
what greater standard error of estimate (2.1 m s-l) about the regression
line,
u
t
= 2.421 + 1.179 u
a
(4, dawn)
The equations show that the airport wind speeds tend to be less than the
mixing layer transport wind, as would be expected.
Wuerch infers from
the generally high correlations that the transport wind speed does not
vary much between downtown and the airport.
However, for low wind speeds
in the morning, the differences may be more important.
The transport wind speed would be the appropriate speed for use with
the box model, which assumes the pollutants to be spread through the
mixing layer.
The appropriate wind speed for the Gaussian model would
be less, because for those regions to which that model is applied the
pollutants have not yet mixed to the higher levels with their correspond-
ingly larger wind speeds.
Furthermore, in the lower layers over the cit~
the frictional effects of the rough urban surface will cause the wind
speeds to be less than those observed at comparable heights over the
smoother airport (rural) surface.
74
-------
There are some very important computational advantages to the use
of the same value of wind speed in both the box and Gaussian models.
The
wind speed can then be factored out of many of the expressions and the
calculations that are required are greatly reduced, particularly when a
long series of concentrations are to be determined (see, e.g., Ludwig
et al., 1970, pp. 58-62 and 163-166).
For this reason, it seems best to
use only a single value of wind speed in the general model.
In the ab-
sence of other information, the wind speeds measured at the airport seem
to be the best compromise choice.
They are also the most suitable for
use to determine the stability category.
For application of the street
model, a wind speed of half the airport value has been found to give
good results in the complete model.
A minimum value of 0.75 m s-l was
used in this application.
The question of wind direction variability within the mixing layer
was also addressed by Wuerch (1971~.
He compared the vector-averaged
wind speeds in the layer with scalar averages; the greater the differ-
ence between these two averages, the greater the directional variability
in the layer.
On the basis of these comparisons, Wuerch concluded that
changes of wind direction through the mixing layer are generally quite
small at midday, but of greater significance during the morning.
The variation in wind direction between the airport and downtown
was checked using the data collected on this program.
The reported air-
port wind directions (excluding calm winds) were compared with those
observed at the upper tower level.
The root-mean-square difference
between the directions measured at these two locations was 310.
In
assessing these results it should be remembered that the airport winds are
reported to only the nearest 50; this fact accounts for at least a small
part of the apparent differences.
The aerodynamic influence of the tower
(Dabberdt, 1968a,b) may also be a contributing factor.
The variability
of winds in the area should be better defined when the detailed studies
75
-------
from the Metropolitan Meteorological Experiment (METROMEX) (Chagnon, 1971)
become available.
At the time of this writing, the analyses of the multi-
balloon wind measurements in the St. Louis area are not yet complete.
E.
Street Canyon
The data collected in the street canyon experiment were stratified
according to wind direction and wind speed, because in our earlier
studies
in San Jose (Johnson et al., 1971) these two variables had been
found to be very important.
The stratifications were based on the winds
measured at the upper tower level.
Twelve 300 wind classes were used
and three wind speed classes.
The wind speed categories were:
1 to 3
m s-1, 3-6 m s-1, and greater than 6 m s-1.
Sufficient data were avail-
able for analysis in all but one case (600, > 6 m s-1).
The resulting
average CO concentrations were computer analyzed and the results are shown
in Figures 21 and 22.
In the coordinate system used for the winds, Broadway is oriented
o 0 0 0
along a 0 -180 line, and Locust along a 90 -270 line.
The street canyon
model predicts that cross-street CO gradients should be zero for winds
parallel to the street, and concentration should decrease with height.
For roof-level winds perpendicular to the street, street-level concen-
trations are expected to decrease in the downwind direction; the vertical
gradients should be zero in front of the buildings that face the wind,
while on the opposite side of the street the concentrations should decrease
with height.
In all cases, the gradients should be more pronounced on
streets with greater traffic emissions than on lightly traveled streets.
From this latter fact it can be deduced that the street canyon effects
would be more pronounced on heavily traveled Broadway than on Locust,
which generally has only about 40 percent the traffic that Broadway has.
76
-------
The figures show that as expected, the gradients on Broadway are gen-
erally greater than those on Locust.
In Figure 21, the Broadway results
can be seen to show the features predicted by the model.
The Broadway
o 0
analyses for the 90 and 270 wind cases show the street-level concen-
trations increasing across the street in the proper sense.
There is very
little change of concentration with height in front of the buildings that
face the wind.
On the other side of the street, there are pronounced
decreases of concentration with height.
These same features are equally
o
pronounced for winds from 30 either side of the direction perpendicular
to the street--with the exception of the 3000 case.
At the lower wind
speeds the 3000 case does not exhibit the expected features.
Instead the
concentrations are more like those that we would expect with winds parallel
to the street.
o 0
For 180 and 360 winds, parallel to Broadway, the cross-street
concentration gradients are much less than the vertical gradients.
The
gradients are generally seen to decrease with increasing wind speed as
expected.
o
For winds 30
either side of the parallel directions, the CO
distributions are of the parallel-wind type, with the exception of the
o
150
o
and 210
low-wind cases, which show considerable evidence of the
effects of the cross-street wind component.
As already noted, the lighter traffic on Locust results in generally
smaller concentration gradients.
In many cases these smaller gradients
make the expected features of the distribution less well defined.
The
o
parallel-wind cases, 90
o
and 270 , are much as expected with generally
horizontal isopleths of concentration.
o
The cases with winds within 30
of the street alignment show more influence of the cross-street component
than was found for Broadway. In fact, these analyses show the influence
o 0 0
of the cross-street component for all cases except 90 , 270 , and 180 .
The last named direction is perpendicular to Locust Street,and at first
it appears quite surprising that it does not show the typical pattern for
77
-------
WIND DIRECTION - relative to Broadway
300 000 900 1200 1500 1800
)' 5, ~~ 4... .4
4
~ 5
~
1-3 ms-1 ~ ~
..L-- ;~
6,1;;- ~.
L
> )' J ! ) \
It
0 J
C1
w ~
~ ~ ~
U -1 <4
0 3-6 ms 3~
...:J w 2~ V-
W
oo Q.
U)
0 1-- ~
z >--- ~
3: 9~
Boatman's
Bank 3 (. ~
First National 3~
Bank
>6 ms-1
24 m ~ ~ y--- ~
~ ~ ~ ~
--- WEST EAST .
TA-8563-112a
FIGURE 21 DISTRIBUTION OF CO CONCENTRATION IN BROADWAY STREET CANYON
-------
WIND DIRECTION - relative to Broadway
2100 2400 2700 3000 3300 3600
j \ /4 /3
4 ;7
~ ~ ~ ~
1-3 ms-1
~ ~
5 4 ~ 6 6 ~
~n:: ~ 6 ~ r--
./
> ) '\ T' )3
a:
0
C)
w y-----
I- ~ :\
6 ms-1
~
4 3 r--
s----- ~ ~
~
-- WEST EAST-
TA-8563-112b
FIGURE 21
DISTRIBUTION OF CO CONCENTRATION IN BROADWAY STREET CANYON (Concluded)
-------
WIND DIRECTION - relative to Broadway
300 600 900 1200 1500 1800
V ~ / /
4
1-3 ms-1 ~ ~ ~
~ ~
5 ~
".---3 ~ ~
>- 3 \ 1
a:
0
CJ
w
!( 5-----
u
a 3-6 ms-1
00 w
0 w
...
(I)
a V--
z ~ y
~ 4 7;---:---
--..
Boatman's ~
Bank
>6 ms-1 Federal <3 ~
Reserve Bank /
18 m--- 3/\ ~
4 5
--... r-- --
- NORTH SOUTH-
T A-8563-113a
FIGURE 22 DISTRIBUTION OF CO CONCENTRATION IN LOCUST STREET CANYON
-------
WIND DIRECTION - relative to Broadway
2100 2400 2700 3000 3300 3600
4V 4" U ')
1-3 ms-1 ~
~ 5 6
~~ 5
y- ~
10 8
>- ~ / ~ V 3
II:
0
C:J
w
I-
« <3
u -1 / 3
C 3-6 ms ~ >3
00 w
...... w y ~
IL 3
UI ( ~
C
Z
~ 7/ 5"'--
13 2
/ ~ <3
>6 ms -1
>3
y;: ~ ~ ~
'--...3
- NORTH SOUTH-
TA-8563-113b
FIGURE 22 DISTRIBUTION OF CO CONCENTRATION IN LOCUST STREET CANYON (Concluded)
-------
such cases.
However, the observed patterns can apparently be explained
by the distribution of buildings in the immediate area.
The south wall of the Locust Street canyon is formed qy the Boatmen's
Bank Building which is 11 stories high on Locust Street.
The south half of
this same building rises to 23 stories and thus forms a wal~ one-half block
to the south that towers above the street canyon.
The distribution of CO
in the Locust Street canyon during south winds suggests that this wall
deflects the air flow and causes eddies in its lee such that the expected
south-to-north airflow across the canyon top does not occur.
Without such
flow there is nothing to initiate the helical circulation that produces the
cross-street gradients.
As a result, ordinary turbulent diffusion processes
prevail and produce the vertical concentration gradients shown in the
o
analyses. If the wind shifts as little as 30 , the concentration patterns
expected of cross-street winds are observed, as can be seen in the analyses
o
for 150
o
and 210
winds.
o
The average CO distribution in the Locust Street canyon for 300
winds
and low wind speed is anomalous and deserves some comment.
The distribution
shown represents the average of about one hour total data and is therefore
very subject to unmeaningful bias.
In this particular case, the very high
average concentrations occur near a bus stop.
On occasion, when two or
more buses stopped simultaneously, the exhaust from one of the buses would
be below the lowest CO sampling inlet.
In this small sample one or two such
events would be enough to cause the observed bias.
However, such occurrences
were infrequent enough that they should not have seriously affected the
other averages.
It has been suggested (Chang et al., 1971) that light winds with light
turbulence can produce a double helical circulation with the two vortices
rotating in opposite directions in a street canyon that is deeper than it
is wide.
This would lead to cross-street CO gradients at low levels that
would be opposite to those expected with a single vortex.
Furthermore,
82
-------
since the transfer of surface-generated CO would be slow between the
upper and lower vortices, a "concentration front" might be expected at
the interface between the two circulations.
There would be relatively
uniform high concentrations at lower levels, then an abrupt drop between
vortices to relatively uniform lower concentrations.
None of the average
CO distributions shown in Figures 21 and 22 substantiate the existence of
a double helix.
Nearly all are consistent with a single-helical cir-
culation.
The air motion measurements in the street canyon do not always agree
with the hypothesis of a single-helical circulation, but the agreement
was found to be much better in st. Louis than it was in San Jose,and the
observations are in better agreement with the single-helix hypothesis than
with the double-. It appears, as it did in San Jose, that the lower sensors
were located in a region of eddies near the "corner" of the street canyon.
The expected low-level cross-street directions of flow were observed in a
greater percentage of cases in St. Louis than jn San Jose; this may reflect
the fact that the sensors were located about one meter higher in St. Louis
and were therefore somewhat less likely to be in the corner eddy.
The
upper sensors were located, of necessity, almost on a level with the
building cornices and were therefore subject to all the eddies associated
with airflow over these edges.
The evidence suggests that the street canyon submodel developed from
the San Jose data (Johnson et al., 1971) is fundamentally correct and
requires only relatively minor revision.
Some revision is required because
the results have shown that the air does not retain the same CO concen-
tration from rooftops to street level as it flows down in front of the
buildings facing the wind.
Apparently there is substantial entrainment
of recirculating polluted air and this leads to greater concentrations at
lower 1eve 1s .
This feature was observed in San Jose also; the total
body of data indicates that it is not anomalous and should be considered
in the formulation of the model.
83
-------
It is reasonable that the air in the down draft should be typified
by a 'box model" concentration as was done in the formulation of the
street model.
We have chosen to retain this feature.
To incorporate
the observed increase in concentration from roof level to street level,
a term has been introduced that is linear with height.
We have assumed
that the CO augmentation of concentration from roof level to street level
is the same as was given by the equation developed from the San Jose data
(Johnson et al., 1971):
t.C
W
=
-0.75
0.1 KNS
W (u + 0.5)
ppm,
(5)
where t.C
W
effects in front of the buildings facing the roof-level wind; K is a
is the CO added to roof-level values to account for street
constant; Sand N are the average traffic speed (miles per hour) and
volume (vehicles per hour); W is street width (m); and u is wind speed
(m s-1) above roof level. The development of this parameterization has
been discussed by Johnson et al. (1971) and will not be repeated here.
If the value of t.C is taken to be zero at roof level, and if its street-
W
level value is given by the above equation (which is independent of
height, z) a linear variation with height gives
t.C
W
=
O 1 -0.75
. KNS
W (u + 0.5)
. (H - z)
H
(6)
where H is the depth of the street canyon.
There is no evidence to justify changing the equation used to describe
the augmentation of CO concentration (t.C ) by street effects on the opposite
L
side of the street,
i.e.,
in front of the buildings facing downwind.
This
equation, as developed by Johns,on et al.,is
t.C
L
=
0.1 KNS -0 . 75
2 2 1/2
(u + 0.5) [(x + z ) + 2J
(7)
84
-------
where x is the horizontal distance to the nearest traffic lane.
This
formulation retains the same cross-street gradients given by the earlier
pair of equations for small values of z, but the cross-street gradients
at higher levels are greater than with the earlier version of the model.
For parallel winds the average of 6C and 6C is used for both sides
L W
of the street. The equation for winds parallel to the street is
6C = 0.5 (bC + 6C )
I L W
=
-0.75
0.05 KNS
(u + 0.5)
[ 1 H-Z]
( 2 2)1/2 + ~
x +z +2
(8)
The value of the exponent of S is subject to change with changes in
exhaust emission devices and the mixture of model years on the road.
The
value -0.75 is appropriate to a 50/50 mix of cars whose model years are
before and after the 1968 introduction of exhaust emission control devices
(1965 in California).
The constant 0.1 in Eqs. (5) and (6) is also a
factor that relates to automotive emission controls and the model year mix.
The distributions of CO concentrations shown in Figures 21 and 22
suggest that the helical circulation develops with rather small cross-
street roof-level components and therefore the application of the parallel-
wind equation for 6C should be restricted to a smaller angular segment
I
than was previously suggested.
Ninety-degree segments were to be used for
o
general applications; for wind directions within 45 of the street direction,
the "parallel" formulation would be used, and for other wind directions the
"cross-wind" formulation.
However, when the San Jose data were analyzed,
o
the cross-wind cases prevailed over 120
wind direction intervals and the
parallel-wind over 600 intervals for the two midblock street canyon stations.
The sectors were not always symmetric about the parallel and perpendicular
directions.
This was ascribed to the effects of nearby buildings, which, as
we have seen in the St. Louis data, can be important.
However, the St. Louis
85
-------
data also indicate that cross-wind patterns often develop with winds
blowing at a relatively small angle to the street direction, and so it
o
appear~ that 60 wind segments are appropriate for application of the
parallel-wind model.
In lieu of other information, which would not be
available for most general applications, the segments should be arranged
sYmmetrically about the street direction.
The street submodel predicts the difference between roof-level and
street-level CO concentrations.
The submodel was applied to the data
shown in Figures 21 and 22 to estimate its accuracy and to compare it
with the version that had been developed from San Jose data. The
o
anomalous case (300 wind direction, low wind speed) was excluded from
the data set.
The upper-tower average winds, and the average traffic
volumes were used as inputs for application of the submodel.
The average
traffic volumes vary from 90 to 384 vehicles per hour on Locust and 211
to 904 vehicles per hour on Broadway, so a wide range of traffic conditions
is represented.
An average vehicle speed of 8.5 mph was used for the Broadway cal-
culations.
This value comes from the measurements discussed in an earlier
section.
For Locust Street a value of 5 mph was used.
This low speed was
chosen because there is a stop sign at the end of the short block where
the measurements were made and vehicles enter the block after stopping
for another stop sign, so the average speeds must be less than on Broadway.
The original and the revised street model were applied to the averaged
data, and the calculated values of ~C were compared with the corresponding
observed differences between street- and roof-level CO concentrations.
The
correlation coefficients and root-me an-square differences (RMSD) between
observed and calculated values were determined for several values of K
between 3 and 10.
Both the original and the revised street submodels
showed minimum RMSD's between observed and calculated ~C's for K approx-
86
-------
imately equal to 7, the same as used in. San Jose.
Using this value,
the RMSD for the original submodel was 1.4 ppm.
For the revised model,
the RMSD was 1.1 ppm.
The respective correlation coefficients were 0.55
and 0.60; the revisions have improved the performance of the model
somewhat.
The RMSD's and correlations were also determined for the observed and
calculated differences between concentrations measured on opposite sides
of the street at roof level and at street level.
Since the original
and revised models give virtually the same cross-street differences at
street level, whatever differences there are between the two versions of
the model must arise from their performances in predicting cross-street
differences at roof level.
The revision in the submodel reduces the
predicted cross-street CO concentration differences.
The original model
predicts the cross-street differences with an RMSD of 1.12 ppm and a
correlation coefficient of 0.57.
The revised model's RMSD is 1.07 and
its correlation coefficient is 0.63.
As with the calculations of 6C there
has been an improvement in the performance of the submodel.
In order to avoid the complications of wind circulations that may
exist in the immediate .vicinity of intersections, the fixed wind and CO
installations were located at midblock.
However, we also made measure-
ments of the along-street variability of CO concentrations in order to
assess the general applicability of the midblock analyses.
The mobile
van system was used for this purpose; the van was parked during 13 daytime
periods of up to 7 hours each at a variety of locations on Locust Street
and Broadway as shown in Figure 23.
Hourly van concentrations were
computed for comparison with the 4-m CO observations made at the fixed
installation on the same side of the block. Figure 24 shows the results of
the comparisons for those periods during which both van and fixed-station
CO data were available; the horizontal wind components measured at the
4-m level are also shown.
We could not use a larger variety of sites due
to local parking restrictions, but there are still enough sites to
warrant confidence in the results.
87
-------
FIGURE 23
u> 4
E
o I
Z >- 0
~ !::
u
o
..J
LU
> -4
20
E
0.
0.
I 10
o
u
FIGURE 24
'\
N:J
Lu
Reserve Ban k
@
00@
STREET
First
National
Bank
>-
«
~
o
«
o CD
a:
en
Boatmen's
Building
@
CD
@
. Fixed Installations
I
T A-8563-143
LOCATIONS OF ALONG-STREET CO MEASUREMENT SITES
"''''''0'''''''''''''''''''''''''' 0.. 0... "0"""'" 0"00" 0.. v
~u
,
,
I
22 SEPTEMBER 1971
(a)
15
5
......... V
~....... -- an -
---- ~ ~.- --- ---
..... ~.--;:.....--- Fixed Station
-
o
0900
,
I
1000
1100
1200 1 300 1400
TIME - CDT
1500
1600
1700
TA-8563-131
COMPARISON OF ALONG-STREET CO CONCENTRATIONS
WITH MIDBLOCK, 4-m OBSERVATIONS OF CO AND WINDS.
Locations of along-street measurement sites are given in Figure 23.
88
-------
'" 4
E
o I
Z >- 0
~!::
u
o
...J
UJ
> -4
20
E
c.
c.
I 10
o
u
o
'" 4
E
o I
Z >- 0
~!::
u
o
...J
UJ
> -4
20
E
c.
c.
I 10
o
U
FIGURE 24
,
....................................~.........................
u
.............
v
,
,
,
,
,
23 SEPTEMBER 1971
(b)
15
5
.' Van
..
..
. ~
~ ~
- - ~ .~.../:.. .
...~ ..
~........ ... " Fixed Station
........ "...... '"
--------
,
.................... ....... .... .......... v
...................~ -.............
u
-
,
24 SEPTEMBER 1971
,
(e)
15
5
....~:a...,...1 Van
............" ----...........
-----~ Fixed Station
o
0900
1500
1600
1100
1200 1300 1400
TIME - CDT
1700
1000
TA-8563-132
COMPARISON OF ALONG-STREET CO CONCENTRATIONS
WITH MID BLOCK, 4-m OBSERVATIONS OF CO AND WINDS.
Locations of along-street measurement sites are given in
Figure 23. (Continued)
89
-------
~ 4
I
o I
Z >- 0
~ t::
u
o
...J
W
> -4
20
E
0.
0.
I 10
o
u
o
'" 4
E
o I
Z >- 0
~t::
u
o
...J
W
> -4
20
E
0.
0.
I 10
o
u
FIGURE 24
,
................................................. ...... v-
--
-..............
-u
,
15
,
30 SEPTEMBER 1971
..'..
.. .
... .. ..-......
.. .. "..
. .. ""..
... ..
".. ..'
""" ..--
~
Idl
Van
5
"
----, /
"""
Fixed Station
,
.
.... ......... ................... ..... ...... . .. .. . . . .. .... ... .... ..... v
-u
,
.
.
1 OCTOBER 1971
Ie)
15
..
..
..
..
.
.
.
.
.
.
.
.
.
.
.
''':
"..
"..
. Van
Missing
Missing
5
...------
/ '''.
Fixed Station
o
0900
,
1000
1100
1200 1300 1400
TIME - COT
1500
1600
1700
TA-8563-133
COMPARISON OF ALONG-STREET CO CONCENTRATIONS
WITH MIDBLOCK, 4-m OBS~RVATIONS OF CO AND WINDS.
Locations of along-street measurement sites are given in
Figure 23. (Continued)
90
-------
~ 4
E
c I
~ > 0
i!: t
u
o
...J
~ -4
20
E
Co
Co
I 10
o
u
o
~ 4
E
c I
~ ~ 0
i!: -
u
o
...J
W
> -4
20
E
Co
Co
I 10
o
U
FIGURE 24
...1
.....-
................. .................... ................... v
~ u
.
,
I
,
,
3 OCTOBER 1971
(f)
15
5
/ Fixed Station
~4t4t // .' Van
4t4t ..
'... ~....-- ..
---.4t A.
~# 4t...-
'# ...........
#
.
.
.
.
u
.............................. v
,
.
4 OCTOBER 1971
(g)
15
5
...~fft'#"--
~
-..,.....-,-
.-- -- ~
Fixed Station Van
o
0900
,
.
1200 1300 1400
TIME - CDT
1500
1600
1700
1000
1100
TA-8563-134
COMPARISON OF ALONG-STREET CO CONCENTRATIONS
WITH MIDBLOCK, 4-m OBSERVATIONS' OF CO AND WINDS.
Locations of along-street measurement sites are given in
Figure 23. (Continued)
91
-------
"' 4
E
o I
z >-
3:!: 0
u
o
..J
W
> -4
20
15
E
a.
a.
10
o
u
5
o
"' 4
E
01
z>- 0
3:!:
u
o
..J
W
> -4
20
15
E
a.
a.
10
o
u
5
FIGURE 24
............ ..... m.... ....................... ......... ....... .... ..,... .,... .... v
u
,
,
,
5 OCTOBER 1971
(hi
~~
. ..~-----
.." --. V
Missing an
- --- ... ............. ... ------. ... .
Fixed Station
,
.
,
u
.................................................. ..... ................... ......... v
,
I
,
,
6 OCTOBER 1971
iii
Fixed Station
---
- -- ..es...r. - --
........~~ .-......... Van
o
0900
.
I
,
1000
1200 1300 1400
TIME - COT
1100
1600
1500
1700
T A-8563-1 35
COMPARISON OF ALONG-STREET CO CONCENTRATIONS
WITH MIDBLOCK, 4-m OBSERVATIONS OF CO AND WINDS.
Locations of along-street measurement sites are given in
Figure 23. (Concluded)
92
-------
Agreement between the van and fixed-station measurements is generally
quite good; the two usually agree within 1-2 ppm and show no systematic
differences.
This agreement may. in part, be attributable to the along-
street uniformity of the source.
This, together with the generally
larger wind components parallel to the street, will tend to minimize the
along-street gradients, as opposed to the cross-street gradients which
can be appreciable, as the discussions on the preceding pages have shown.
Two cases of relatively large along-street variations were found.
These are shown in Figure 24(d) and (e) where systematically larger values
(on the order of 6 ppm) were measured with the van.
Correlation with the
fixed-station measurements, however, is quite good.
We can offer no
explanation for this phenomenon, but several observations may be in order.
The differences do not appear to be a peculiarity of the site.
The obser-
vations illustrated in Figure 24(b) and (c) were made in the same area
(see Figure 23), yet show good agreement.
In all four cases, the magni-
tude of the cross-street wind component was less than 0.5 m s-l. while
the along-street component ranges from 0.5 to 3.0 m s-l.
Furthermore,
the case of good and the case of poor agreement were observed on the same
day of the week, thereby minimizing effects of traffic differences.
TM
only systematic difference is that during the periods of large differences
in concentrations measured at van and fixed sites, the cross-street wind
was consistently from the north, while good agreement was achieved for
zero and southerly cross-street winds.
Finally, in view of the good
correlation observed during the periods of poor absolute agreement, the
question of instrument malfunction cannot be ruled out entirely. although
calibrations of both the fixed and van systems were made routinely.
Figure 24(g) is an example of the correspondence that can be expected
-between the two different measuring systems.
On this day. the van intake
was approximately 1 m from that of the fixed installation.
The differ-
ences (~ 0.5 ppm) are within the uncertainties of the two measurement
systems.
93
-------
In summary, we may conclude that the midblock observations are quite
representative of conditions throughout the block.
These data show that
the street effects submodel is capable of predicting CO concentrations in
the street canyon with an accuracy that is comparable to the observed
variations of CO along the street.
F.
Freeway
In order to assess the potential application of the model to areas
in the immediate vicinity of major roadways, the mobile van monitoring
system was used to make selected observations of CO concentrations down-
wind of the six-lane Daniel Boone Expressway at Forest Park (see Figure 5).
Specifically, these measurements were of four types:
(1) temporal obser-
vations at a fixed location, (2) horizontal profiles immediately downwind
of the freeway, (3) vertical profiles adjacent to the roadway, and (4)
mobile measurements in the park area to a distance of about 1 km downwind
of the freeway.
There are a number of questions that need to be answered with regard
to the source characteristics of a modern, multilane freeway.
In the
first approximation, the freeway may be treated as an infinite line-
source with an effectively constant emission rate for periods up to one
hour.
More realistically, the modern freeway is better represented by
two roadways--one in each direction of travel--with significantly differ-
ent emission rates and traffic patterns.
The mechanical turbulence in-
duced by the usually fast moving vehicles disperses the emissions quite
rapidly at the source, thereby effectively creating an extended volume
source.
Other roadway characteristics must also be considered, such as
the height displacement of the roadway with regard to the surrounding
terrain.
The mobile van facility was used for a preliminary investigation
of the manifestations of these features on the ~sulting distribution of
CO near a major, ground-level freeway.
94
-------
Observations made during two early morning, peak traffic periods are
illustrated in Figures 25 through 27; meteorological conditions derived
from the near-simultaneous EMSU data are also presented.
Figure 25
illustrates the temporal variation of CO observed during a half hour
period at a height of 3 m and at a distance of about 3-4 m downwind of
the nearest (westbound) traffic lane; traffic during this period is a
factor of about three times heavier in the more distant, eastbound lanes.
Atmospheric conditions were slightly stable during the period and the
winds were steady and nearly perpendicular to the roadway, although the
speed was estimated to be about half the value of 6 m s-l reported at
the EMSU station.
The observed CO concentrations ranged from about 12
to 22 ppm.
Five-minute average concentrations differed by about 3 ppm
over the period; during the last 20 minutes there was only 1.5 ppm
variation.
When a single monitoring system is used to obtain horizontal CO
profiles immediately downwind of the freeway; care must be exercised to
separate the horizontal gradients from temporal variations.
To minimize
this difficulty, horizontal profiles were made by moving the van in
discrete steps toward the source during a period when the source strength
could be expected to decrease and the turbulent diffusion to increase
with time.
The horizontal profile obtained in this manner to a distance
of 200 m is shown in Figure 26.
Five-to-seven minute observation periods
were made at each site, except for the near-freeway site, where two con-
secutive observations were made.
Both the absolute value and the standard
deviation of the concentrations decrease in an exponential fashion with
distance from the source.
Qualitatively, the data are as expected, but
the lack of information on ambient CO levels and temporal emission trends,
as well as the lack of micrometeorological data, precludes quantitative
analysis.
95
-------
24
[ 22
0.
z
o
- 20
~
a:
I-
z
w
u
~ 18
u
w
o
x
~ 16
o
~
z
o
III
~ 14
u
12
0735
FIGURE 25
29 SEPTEMBER 1971
Meteorological Data:
0700 CDT, 29 September 1971
Surface Wind: 205°/6.0 ms-1
Transport Wind: 241°/7.3 ms-1
Mixing Depth: 270 m
0740
0745
0750
0755
0800
TIME, CDT
T A-8563-126
TEMPORAL VARIATION OF CO CONCENTRATION ADJACENT TO THE
DOWNWIND EDGE OF THE SIX-LANE DANIEL BOONE EXPRESSWAY AT
FOREST PARK. ST. LOUIS
centration at the near-freeway site; the data were collected sequentially
Figure 27 is an illustration of the vertical structure of CO con-
at the various levels over a 65-minute period under meteorological con-
ditions similar to those given above.
Concentrations were measured at five
The observations began and ended with measure-
levels from 3 m to 11 m.
The 3-m measurements (which differed by 3 ppm)
ments at the 3-m level.
were used to establish the temporal trend by linear interpolation.
The
values.
observations at other heights were referenced to the interpolated 3-m
The significant feature of these observations is the apparent
pronounced elevated CO maximum at 5 m.
Two arguments may be offered
96
-------
IS)
-.!
E
a.
a.
z 11
o
i=
«
a:
~
z
~ 9
z
o
u
o
u
15
13
(0848)
(0841)
7
Daniel Boone
Expressway - 6 Lanes
(I/> : 275°/095°)
.
Meteorological Data:
0700 CDT, 29 September 1971
Surface Wind: 205°/6.0 ms-1
Transport Wind: 241°/7.3 ms-1
Mixing Depth: 270 m
(0830)
5
o
20
40
60 80 100 120 140 160
x - DISTANCE DOWNWIND OF FREEWAY - m
T A-8563-125
180
200
220
FIGURE 26
DISTRIBUTION OF CO CONCENTRATION AT 3-m HEIGHT DOWNWIND OF THE DANIEL BOONE
EXPRESSWAY AT FOREST PARK, ST. LOUIS
-------
12
4
Meteorological Data:
0700 CDT, 27 September 1971
Surface Wind: 180°/7.0 ms-1
Mixing Depth: 200 m
10
E 8
I
~
J:
e"
w
J: 6
(0732)
14.9 ppm
(0714, 0819 CDT)
2
-4
-3
-2 -1 0 1
RELATIVE CO CONCENTRATION - ppm
2
3
TA-8563-123
FIGURE 27
HEIGHT VARIATION OF CO CONCENTRATION ADJACENT TO THE DOWNWIND
EDGE OF THE SIX-LANE DANIEL BOONE EXPRESSWAY AT FOREST PARK,
ST. LOUIS
that this feature is an artifact caused by:
(1) temporal variations of
the emissions and/or meteorological fields which were not accounted for
by the interpolation, or (2) aerodynamic effects of the van structure.
The observations presented in Figures 25 and 26, as well as the personal
evaluation of the observers, indicate that conditions did not change
significantly with time.
The latter, however, may be significant in
that the body of the van could have caused a tilting of the local wind
traj ectories .
However, in view of the relatively light winds, it is
somewhat doubtful that this is the total cause of the phenomenon. Alter-
natively, one may suggest some characteristic of the freeway itself.
In this regard, air flowing from one side of the roadway to the other
may encounter a shelterbelt effect of the "wall" of vehicles in the
98
-------
downwind lanes.
Also, the strong mechanical turbulence induced by the
traffic movement causes the emissions to be rapidly dispersed into some
effective volume near the source, thereby increasing the effective
height of the source.
These limited freeway observations serve to illustrate the com-
plexity and large variability of the distribution of vehicle-generated
pollutants in the freeway corridor.
Additional research is indicated
to better define concentration patterns near major roadways and to
evaluate the applicability of the APRAC-IA model in these areas.
The
situation is somewhat similar to the importance of street effects in the
downtown area, and an analogous freeway submodel may be required.
99
-------
IV
PERFORMANCE OF THE MODEL
In the preceding section, the various components of the composite
model were examined.
Their initial outputs were compared with values of
the corresponding parameters obtained in St. Louis by different and inde-
pendent methods, and in some cases minor revisions were made.
The
algorithm for obtaining the stability index was changed slightly on the
basis of the St. Louis data, as was the street-effects submodel.
Also,
the emissions model now uses a new diurnal traffic cycle for downtown
streets, both arterial and freeway.
This is not a fundamental change,
but simply reflects the availability of better data.
The changes resulting from the st. Louis studies have not been major;
the major changes in the model were made at the conclusion of the San Jose
study.
In that experiment, the traffic data were not used to apply the
model in its complete form.
Concentrations resulting from emissions at
distances greater than 2 km from the receptor were not calculated.
Instead,
measured concentrations were used to simulate this part of the model.
The
San Jose evaluations used winds specially measured in the downtown area for
the application of the model and its submodels; also special measurements
had been used to determine the wind direction sectors where the helical
street canyon circulation developed.
All these factors should have given
the model somewhat better performance than would be expected if it used a
complete traffic network, airport meteorological measurements, and the
standard street model wind direction sectors.
For this report, the model has been applied in the manner that would
be employed by its potential users.
The extensive traffic network for
St. Louis that was used in our earlier studies (Ludwig et al., 1970) has
been updated using the most recent data available to us (Missouri State
101
-------
Highway Department, 1970).
Hourly airport meteorological observations
and the routine radiosonde measurements from Salem, Illinois have been
used. When the model is applied in this way and the results are compared
with the special CO measurements made in downtown St. Louis, its cap-
abilities can be accurately defined.
Figure 28 shows the observed hourly average CO concentrations for
the operating periods at the four low-level street canyon sites; these
values were measured at heights ot about 3 and 4 m from the building faces.
This same figure shows the concentrations calculated with the model using
hourly meteorological data.
The calculated values are shown for all hours,
but the observations were not always made continuously (see Table 4).
It
is obvious that the model generally reproduces the observed trends very
well.
Figure 29 shows the scatter diagrams where the observed CO con-
cent rations are plotted against those calculated with the model.
T~
root-mean-square differences between the observed and the calculated CO
concentrations range from 2.9 ppm at the east side of Broadway to 3.9 ppm
at the west side of Broadway.
At roof level the observed (the average of
the four measurements at the top of each street canyon) and the calculated
concentrations have root-mean-square differences of 2.5 ppm.
These values
compare quite favorably with the 3-ppm accuracy found in San Jose using
special data.
They certainly exceed the performance of the model before
it was revised.
In earlier studies (Ludwig et al., 1970) using st. Louis
CAMP station data, the original model was found to have root-mean-square
differences with the observations of 5 to 9 ppm.
The model revisions
made on the basis of the San Jose and St. Louis field studies have
resulted in considerable improvements in the performance of the model.
As demonstrated earlier (Ludwig et al., 1970), the results can be
improved considerably if special statistical regressions are undertaken.
102
-------
E
0.
0. 10
z 5
o
~ 0
~ 30
z
w
~ 25
o
u
o 20
u
30
25
N. LOCUST
Observed
Calculated
~
. .
". ",:
,II ~ :
.,.
20
15
10
5
o
30
25
S. LOCUST
~.:...:
. " I
" .'
: !.
. "
",
r~'f!
,II',
: r P,
60
o
20
60
80
80
100
120
140
160
40
100
120
140 160hr
20
15
E. BROADWAY
15
10
5
o
30
25
W. BROADWAY
!-MOI\I-+-TUES+WEO+THUR+FRI+ SAT+ SUN ~ r-M..Hoj ..!+-rUES+WEO+THUR+FRI+ SAT+ SUN-1
20
15
10
5
o
o
20
40
23-29 AUGUST 1971
FIGURE 28
30 AUGUST-5 SEPTEMBER 1971
T A-8563-136
OBSERVED AND CALCULATED CO CONCENTRATIONS AT 4 m
103
-------
30
25 N. LOCUST - - - -- Observed
Calculated
20
15
10
5
0
30
25 S. LOCUST
20
15
E
Co
Co 10
Z 5
0
i=
« 0
a: 30
I-
Z
w
CJ 25 E. BROADWAY
Z I
0
CJ ,
20 '
0 '.
"
CJ '.
"
. 'j
15 .
.
.
"
.,~
10
5
0
30
25 W. BROADWAY
20
15
10
5
o
o 20 40 60 80 100 120 140 160 hr 0 20 40 60 80 100 120 140 160 hr
I--MON+TUES+WEO+THUR+FRI+ SAT+ SUN -~ I--MON-+-TUES+WEO+THUR+FRI + SAT+ SUN-.J
13-19 SEPTEMBER 1971
20-26 SEPTEMBER 1971
TA-8563-137
FIGURE 28
OBSERVED AND CALCULATED CO CONCENTRATIONS AT 4 m
(Continued)
104
-------
30
25 N. LOCUST Observed
Calculated
20
15
10
5
0
30
25 S. LOCUST
20
15
E I:
~ 10 f
z 5
0
f= 0
c(
a: 30
I-
Z
w 25 E. BROADWAY
Co)
z
0
Co) 20
0
Co)
15
10
5
0
30
25 W. BROADWAY
20
15 .t i
.' .
.' .
10 : .
5
o
o
20
40
60
80
100
120
140
160 hr 0
20
40
50
80
100
120
140
160 hr
!-MON+TUEs+WED+THUR+FR'+ SAT+ sUN-.j !-MON-+-TUEs+WED+THUR+FR'+ sAT+ sUN-.j
27 SEPTEMBER-3 OCTOBER, 1971
4-10 OCTOBER 1971
TA-8563-138
FIGURE 28
OBSERVED AND CALCULATED CO CONCENTRATIONS AT 4 m
(Continued)
105
-------
FIGURE 28
30
25 N. LOCUST Observed
Calculated
20
15
10
5
0
30
25' S. LOCUST'
20
j
15 Ii
E
it 10
z 5
0
t= 0
«
::30
z
w E. BROADWAY
~ 25
0
0 20
0
0
15
10
5
0
30
25 W. BROADWAY
20
15
10
5
o
o
60
20
40
80
100
120
140
160 hr
r-MON+TUES+WEO+THUR+FR'+ SAT+ SUN-1
11-15 OCTOBER 1971
TA-8563-139
OBSERVED AND CALCULATED CO CONCENTRATIONS AT 4 m
106
(Concluded)
-------
o
u
o 15
UJ
>
II:
~ 10
to
o
FIGURE 29
30
BROADWAY, ROOF LEVEL
25
Observed
2.6 + 0.67 (Calculated)
20
15
10
5
o
30
EAST BROADWAY, 4 m
25
E
Co
Co
I 20
Observed = 2.6 + 0.64 (Calculated)
:. .' . .
'.:"
o
30
WEST BROADWAY, 4 m
25
Observed
3.5 + 0.50 (Calculated)
20
15
10
5
o
o
LOCUST, ROOF LEVEL
Observed = 2.7 + 0.60 (Calculated)
~
NORTH LOCUST, 4 m
. .
SOUTH LOCUST,4 m
Observed = 2.6 + 0.68 (Calculated)
"
'.. .
10
15
25
30
20
15
20
25 30 0 5
CALCULATED CO - ppm
TA-8563-140
5
10
SCATTER DIAGRAMS SHOWING OBSERVED CO CONCENTRATIONS VERSUS
THOSE CALCULATED WITH THE MODEL
107
-------
However, special regressions cannot be depended upon to apply to all
cities or sets of conditions, only those for which they were derived.
Since the model was developed from physical principles, the error figures
in the preceding paragraph should be a good indication of the model's
general applicability.
The revisions and additions that have been made
to the model have brought its accuracy to a point where it is comparable
to the ~ 2-to-4-ppm values achieved earlier only through special fitting
processes (Ludwig et al., 1970).
In some special situations, reliable measurements are available,
and these could be used as a basis for statistical regressions approp-
riate to certain locations.
Such relationships developed from the data
collected in St. Louis are shown in Figure 29.
Use of these relation-
ships would reduce root-mean-square errors somewhat.
At street levels,
the relationships shown in Figure 29 reduce the root-mean-square errors
on the east side of Broadway to 2.5 ppm (from 2.9) and on the west side
to 3.3 ppm (from 3.9); these are the extremes.
At roof level, the
regression lines reduce the errors to 1.6 ppm (from 2.5).
The correlation coefficients between calculated and observed values
of CO concentration vary from 0.40 for the North Locust site to 0.67 for
the East Broadway site.
Although these values are not as high as we would
like, they do represent an improvement over the values found in our
earlier St. Louis comparisons.
These earlier values varied from 0.16 to
0.45 for the months studied.
For comparison, the earlier correlations
for the months of August through October 1964 ranged from 0.16 to 0.27.
Figure 29 shows certain systematic differences between the observed
CO concentrations and those calculated with the model.
The model generally
underestimates concentrations below about 7 or 8 ppm and overestimates
those at the higher values.
The overestimation of high concentrations may
reflect a systematic inadequacy in the treatment of conditions conducive
108
-------
to high concentrations.
If the source of error were the use of a min-
imum effective wind speed during periods of observed calm, then the
data suggest that a slightly larger minimum wind speed would be more
appropriate.
Another likely source of bias is the emissions model.
High concentrations are frequently found during periods of peak traffic.
The model presumes that the average speeds during peak traffic periods
are 85 percent of the average speeds at other hours.
If this were an
underestimate or if the exponent used in the emission equation [see
Appendix A, Eq. (A-4)J were incorrect, then emissions might be over-
estimated during periods of high concentration; overestimates of the
emissions would lead directly to overestimates of the concentration.
The fact that the model underestimates CO concentration when the
concentration is low may be attributable to the effects of sources beyond
32 km from the receptor.
Virtually all the important sources in the
St. Louis metropolitan area are within this distance and should be
accounted for, so it seems unlikely that differences of about 2.5 ppm
would arise from CO entering the area from the rural regions.
If the
emission model overestimates the emissions for high emission situations,
it may also underestimate for the low emission situations that account
for many of the periods of observed low concentration.
The tests that
have been conducted to check the performance of the emissions model
have not been sufficiently sensitive to detect errors of the magnitude
that might be associated with the observed differences between calculated
and observed concentrations.
Figures 30 and 31 illustrate the model's ability to predict the
frequency distributions of hourly average CO concentrations.
The model
does well; median values have been defined within about 2 ppm of the
observed values.
Ninety- and 98-percentile values are within about 3 and
2-1/2 ppm, respectively.
The tendency of the model to overestimate high
concentrations and underestimate low is reflected in these figures.
Since
109
-------
20
E
Q,
Q,
I 10
z
o
~ 6
«
a:
~ 4
!oJ
U
z 3
o
u
BROADWAY, ROOF LEVEL
64B hours
10
B
6
:::;.-~
~/ "",
"
"
"
"
Regression """
"
.-.-
"
"
.......~ Model
"
",
4
3
2
20
EAST BROADWAY. 4m
701 hours
8
4
~ "
"
~ .......
~~ "
,....-: ",
...-
"
~ ~ .....-
...-
-- "
Regression '-- ~ """"
~ "
~ "
",..,.. ~ ....-
"
"
". "
......
"
"
"
"
"
"
"
"""" "Model
"
"
2
20
WEST BROADWAY, 4m
687 hours
10
8
6
4
3
2
2
5
10 20 30 40 50 60 70 80 90
PERCENT OF HOURLY VALUES ~ ORDINATE VALUE
95
98
TA-8563-141
FIGURE 30
CUMULATIVE FREQUENCY DISTRIBUTIONS
OF HOURLY CO CONCENTRATIONS
ON BROADWAY
110
-------
E
a.
a.
I 10
z
o
~ 6
-------
there are many more low concentration cases than there are high, the
curve based on the model calculations falls consistently below the
observed values.
However, because of the model's tendency to over-
estimate at the higher concentrations, the model curves have steeper
slopes than the observed and approach them at the higher cumulative per-
cent ages.
Figures 30 and 31 show the importance of the emissions within the
street canyon, particularly at higher concentrations.
Comparisons of
the 90-percentile concentrations at roof and street level indicate that
the street-level concentrations are nearly twice those at roof level;
the differences in the medians are not so pronounced, but still show
that the local street emissions are of considerable importance in deter-
mining the observed CO concentrations at street level.
The curves marked "regression" in Figures 30 and 31 indicate the
improvements that can be achieved through the use of the linear regres-
sion equations shown in Figure 29.
Use of these relationships halves
the errors in specifying the median and 90-percentile values; observed
concentrations and those calculated with the regression equations have
root-mean-square differences of only about 1.3 ppm.
112
-------
v
CONCLUDING REMARKS
A.
General
The results of the St. Louis experimental and analysis program
for the validation of the APRAC-1A urban diffusion model are extremely
encouraging.
The model had been refined and evaluated earlier during
the San Jose program, with good success in predicting urban levels of
CO concentration as a function of the meteorological conditions, traffic
characteristics, and geographical location.
On the basis of these
results, the st. Louis study was formulated to evaluate the model in a
significantly different environment.
Specifically, the following
objectives of the study have been accomplished:
(1) analysis of the
street canyon-flow and pollution-concentration regimes in conjunction
with the evaluation of the street-effects sUbmodel, (2) analysis of the
representativeness of airport wind data, (3) further evaluation of the
procedure for the estimation of urban stability and mixing depth, and
(4) validation of the emissions submodel.
This study has shown that
the street-effects, stability, and mixing-depth submodels perform very
well, but we are less certain of the performance of the emissions submodel.
On the basis of the results of the evaluation programs, it is con-
eluded that the APRAC-1A urban diffusion model is suitable for use as an
operational tool by air pollution control personnel, city planners,
traffic engineers, and the like. The model has demonstrated its capacity
to predict the distribution of concentrations of vehicle-generated, inert
pollutants (e.g., CO) on both the urban mesoscale and the street-canyon
microscale.
It may therefore be used to evaluate objectively the impact
of variations of meteorological conditions, vehicle emission levels, and
traffic distributions, either singly or in combination.
113
-------
Even with these highly significant capabilities, the APRAC-1A model
provides the answers only to some, but not all, questions of vehicle-
generated pollution.
Three areas exist where additional development is
required:
(1) the distribution of pollutants on the near-freeway micro-
scale, i.e., within the freeway corridor; (2) evaluation of vehicle
emissions for traffic patterns on the local scale, i.e., of the appli-
cability of the standard composite test route to local conditions; and
(3) consideration of chemically reactive pollutants.
As the basic APRAC-1A model is currently structured, the best
spatial resolution available is on the order of 65 m.
This is sufficient
to account for variations on the urban scale, but it may be inadequate to
simulate large microscale variations in the immediate vicinity of strong
emission sources.
It was for this reason that a submodel was developed
for street canyon effects.
Similar work appears necessary for the near-
freeway situation.
This need is acknowledged by the federal requirements
for environmental impact statements to assess in detail the effect of
freeway development on air quality within the freeway corridor (a lateral
distance of 350 m), as well as the mesoscale effect.
Although the model
can handle the latter, additional work is required to simulate the micro-
scale structure.
Another area where additional refinement may be required is the
emissions formulation.
The current submodel for emissions is based on
a relationship that uses average speed over a composite urban route as
the independent variable.
This composite route speed includes idling
time at intersections as part of the average speed and, therefore, appears
to give high readings for emissions along a relatively uncongested freeway.
A potential for improvement of the emission rate model has been suggested
by Curry and Andersen (1971).
In their emission model, emission rates for
CO and hydrocarbons are presented as functions of average traffic speed
114
-------
over a short uniform highway segment, and corrections to the rates are
presented that account for signal stops and for speed changes caused by
congestion.
As currently structured, the APRAC-IA model predicts only the
distribution of CO, although the treatment of other vehicle-generated
pollutants is both desirable and feasible, especially on a limited scale.
The very small size of the lead particulates found in the urban atmos-
phere and in the vicinity of freeways (Ludwig and Robinson, 1965;
Robinson and Ludwig, 1967) means that they will be dispersed much like
the relatively inert gas CO.
For the short travel distances in the free-
way corridor, the hydrocarbons should also disperse similarly to CO, in
that many of them react quite slowly (Ott, 1971).
On the other hand, NO,
NO , and 0 react very rapidly, and additional research about their dis-
2 3
tribution in the immediate vicinity of the freeway would be desirable.
Qualitative analysis of these chemical reactions suggest that the problem
is reasonably simple on the freeway microscale, and a solution for this
special case appears tractable.
Additional research in these several areas would expand the scope
of the problems to which the model can be applied but would not invalidate
the usefulness of the model in the areas for which it has been developed
and tested.
115
-------
Appendix A
GENERAL DESCRIPTION OF THE MODEL
A-l
-------
Appendix A
GENERAL DESCRIPTION OF THE MODEL
1.
Introduction
The basic nature of the APRAC-IA diffusion model has not changed
appreciably since it was first formulated (Ludwig et al., 1970).
How-
ever, there has been considerable refinement in a number of areas.
The
model has been developed to be applied in virtually any city.
As a
result a number of submodels have been formulated to convert conventional,
readily available meteorological data into the specialized inputs required
by the mode 1.
The results of this program are intended primarily to be
used as tools in planning activities for predicting the pollution
patterns in any urban region that will result from planned traffic
changes from predicted growth, or to be used in an operational mode for
short-term
predictions.
Because of the emphasis on traffic-generated pollutants, it has also
been necessary to develop a submodel to describe the behavior of these
effluents in the street canyons where they are often generated.
Since a
substantial portion of the pUblic is exposed to pollution in these street
canyons, it is quite important that the effects of the localized air cir-
culations found in them be understood.
One of the major achievements of
the second phase of the research (Johnson et al.,1971) was the development
of a method of describing these effects.
The current model is appropriate to quasi-inert traffic-generated
pollutants.
The submodel that is used to convert traffic data to
emissions is for the specific pollutant carbon monoxide (CO).
In prin-
ciple, another submodel could be substituted that would be appropriate
to another pollutant such as lead aerosol.
A-3
-------
The modular nature of the model is one of its important features.
It has been designed for relative ease in incorporating new findings
without requiring complete reorganization of the model.
Thus when
better methods or better data are available for determining atmospheric
stability. mixing depth, or other parameters, they can be used to design
submodels to replace the submodels now used.
This Appendix describes the model in general terms.
The detailed
development of the equations is given in the earlier reports (Ludwig
et al., 1970; Johnson et al., 1971) and in the body of this report.
The
listing of the computer programs that actually constitute the APRAC-IA
model is given in the user's manual (Mancuso and Ludwig, 1972).
2.
General Organization of the Model
a.
The Basic Model
The model uses a combination of the "Gaussian plume" and "box
model" diffusion formulations.
Basically, the Gaussian plume model
assumes that the vertical concentration profile from a cross-wind line
source is Gaussian in shape as shown in Figure A-I.
The spread of this
vertical concentration distribution is described by its standard devia-
tion, a .
z
On the basis of experimental data, the vertical diffusion
parameter cr
z
is taken to have the form
cr
z
b
= ax
(A-I)
where x is the downwind distance and the parameters a and b depend upon
atmospheric stability.
The functions used in the model are based on the
analysis of urban tracer tests (see Johnson et al., 1971).
Figure A-2
shows how cr varies with x and atmospheric stability.
z
The curves in the
figure represent conditions from extremely unstable (A) to moderately
stable (E).
When vertical mixing is inhibited, the box model is applied to
emissions from sources relatively distant from the receptor.
The receptor
A-4
-------
LINE
SOURCE
FIGURE A-1
103
~
j!!
Q)
E
ON
102
FIGURE A-2
UZ DEPENDS UPON
. TRAVEL DISTANCE
. ATMOSPHERE STABILITY
t
:/::
GAUSSIAN
VERTICAL
CONCENTRATION
PROFILE
. .
X1
X2
DISTANCE
X3
.
T A-8563-49
VERTICAL DIFFUSION ACCORDING TO GAUSSIAN FORMULATION
104
10
10
102
103
DOWNWIND DISTANCE - meters
104
105
T A-8563-98
VERTICAL DIFFUSION AS A FUNCTION OF TRAVEL DISTANCE
AND STABILITY CATEGORY, AS REVISED FOR URBAN CONDITIONS
A-5
-------
is the point for which the concentrations are being calculated.
Emissions
tend to become uniformly distributed in the vertical up to the limiting
mixing height after sufficient travel has taken place.
The models are applied to ten area sources, each of which is
assumed to have source emissions spread uniformly throughout.
These area
sources are oriented in the upwind direction as shown in Figure A-3.
The
logarithmic spacing of the area boundaries allows the nearby sources to be
considered in greater detail than the farther sources, whose individual
contributions tend to be merged during their longer travel.
The contri-
but ions of each of the ten area sources to the CO concentration at the
receptor are computed individually with one of the simple formulations
given below.
For the closer segments the Gaussian formulation is used
to obtain the concentration C. resulting from emissions in the ith segment:
1
u a
j
(1 - b.)-1 (l-bj
J x. 1
1+
I-bj)
- x
i
(A"';2)
C
i
0.8 Q
Ai
=
where Q is the average area emission rate (g m-2 s-1), u is the
Ai
transport wind speed (m s-1), and a
j
riate to the segment and atmospheric
and b are the constants approp-
j
stability class j. The x's are
the distances to the closest boundary of the segment designated by the
subscript, i.
The model changes from the Gaussian formulation to the box
model at a distance where the two (in their respective line source
formulations) give equal surface concentration values.
The box model
concentration is given by the following equation:
C
i
= QAi
x - x
i+1 i
uth
(A-3)
A-6
-------
To
32 km
~
I
'I
22.50
16 km
SURFACE
WIND
EXPANDED VIEW OF
ANNULAR SEGMENTS
WITHIN 1 km OF
RECEPTOR
H 1000 m
o
45
FIGURE A-3
DIAGRAM OF SEGMENTS USED FOR SPATIAL PARTITIONING OF EMISSIONS
RECEPTOR
POINT
RECEPTOR
POINT
TA-7874-1R
-------
where h is the mixing height.
The contributions of the emissions in each
segment, as determined from Eqs. (A-2) and (A-3), are summed to obtain
the concentration at the receptor.
are:
The input variables required for application of the basic model
(1) traffic emissions, (2) mixing height, (3) atmospheric stability
type, and (4) transport wind speed and direction.
The model is designed
to be applicable to any city, where traffic data and conventional (airport)
weather observations are available.
None of the required input variables
are directly observed.
The airport surface wind speed and direction have
been applied as a first approximation to the transport wind, but suitable
adjustments need to be made to account for urban effects.
Separate sub-
models had to be developed to estimate mixing depths and stability
categories from the available airport observations, and emissions from the
daily traffic volumes on the major city streets.
The basic model treats the dispersion process only on the urban
mesoscale, whereas concentrations at specific locations may be strongly
influenced by microscale effects.
One of the principal accomplishments of
this program has been the development and validation of a street-scale
submodel that describes microscale effects and thereby substantially
improves the performance of the overall model.
b.
Emissions Submodel
Instead of an emissions inventory, such as is used for other
pollutants, the CO model uses a traffic inventory.
An example of a
network of traffic "links," in this case for Chicago, is shown in
Figure A-4.
Each straight road segment, or link, is assigned an average
daily traffic volume, based on historical or forecast data.
Each link
is identified in the computer memory by its length and the geographical
locations of its endpoints and is designated as a particular road type.
To calculate emissions for a given time, daily traffic volumes are
A-8
-------
I N. \ ~\\ \
\L ' .. !-\~ .~~
,.'~ /~I \1 \ '~
~ r... I i'-..JI \ ~ ~
:5r- u,~ ~. i! I i. ~l l ~
I 'I Ir~ ~
r:::: >~ ~-H \ I
I" ...:: "-j ---+-.. " i-
I ...
_L'~ ~ ~'I\
~~ -1..(-, -- ~ '-
'" N-..-L.
"-K.l rr I"...
i----.....'-.'''toJ I
I " )- - 1\
'1- J ...... '.
\. ~~ y
( / I I
",....
L --Jj'1 lPI -L ~..... ~
L--t1 r-:: ~.- l-~ IiaJ
7 '/' I -
'-l /~ ~~ ~ (r- //
~~ ~7/
- :Y '"
48 MI LES
v 4-
I I
L[J
V
-.......
I
- ~,
~
l
If
-\
\ ~-
t-~~
~
~
r.o;
~
p
I
!
1\
\
f -.- \...
1/ J
1 t- .,.....
..L I ['-.
- -
TA-7874-78
FIGURE A-4
COMPUTER DISPLAY OF TRAFFIC LINKS FOR CHICAGO
A-9
-------
multiplied by a temporal adjustment factor to obtain traffic volumes
for a particular hour.
Then the emission rate, E (g-CO vehic1e-mi-1),
is estimated from the mean vehicle speed, S (mi h-1) , by an empirical
equation of the form
E = a s-\3
(A -4)
Here a and \3 are constants that depend on the characteristics of the
emission control devices installed and the mixture of old and new
model cars on the road.
For the calculations given in this report a was
taken to be 700 and \3 to be 0.75.
These values are appropriate to a
mixture of about half pre-1968 (1966 for California) and half newer cars
(Johnson et al., 1971).
For cars produced since 1968, the value of \3
has been 0.48.
Existing and potential legislation requires a to decrease
with time, as shown in Table A-I.
For future years, the effective values
of a and \3 for use in Eq. (A-4) have to be determined on the basis of the
fraction of the total cars represented by each model year.
Table A-1
VALUES OF a FUR CARS PRODUCED AFTER 1970
Model Years a
1972 - 1974 160
1975 - 1979 16
After 1980 8
The values of S in Eq. (A-4) are based on the road type and
the time of day (whether during peak or off-peak traffic conditions).
The total hourly CO emission for a given traffic link is the product of
three factors:
the emission rate, the length of the link, and the
hourly traffic volume.
The emissions from all the links and parts of
links that fall within one of the segments in Figure A-3 are summed to
determine the average emission rate for that segment.
A-10
-------
Mixing Height Submodel
c.
The determination of mixing height is based on the physical
characteristic that a mixed layer of the atmosphere has an adiabatic
lapse rate.
Thus, by using the observed morning lapse rate obtained
from the nearest National Weather Service radiosonde station and the
surface temperature at a given hour during the day, it is possible to
determine at which height a parcel of air lifted adiabatically would
reach a temperature observed by the radiosonde.
This is the mixing
height, and the method of determination is the one commonly used (e.g.,
Holzworth, 1967).
During the daylight hours the observed surface temperatures
at the airport observing site are used as a basis for the determination
of mixing height.
During the predawn hours the airport temperatures are
augmented by an amount determined from an empirical relationship that
accounts for urban heat island effects.
During the post sunset hours
the model interpolates between the afternoon mixing depth and that for
the predawn hours of the following morning.
d.
Stability Submodel
In order to select the proper function to use for cr , it is
z
necessary to categorize the atmospheric stability.
Pasquill (1961) in
originally proposing the Gaussian plume approach to describing atmos-
pheric diffusion, suggested that wind fluctuation data be used.
Wind
fluctuations are not routinely measured and cannot be used as inputs for
a model designed for widespread practical application.
Therefore, the
stability submodel has employed an alternative approach suggested by
Pasquill; this alternative method makes use of wind speed, cloud cover,
and
intensity of insolation.
The last factor is not measured, but
the St. Louis experiments indicate that it can be estimated with suf-
ficient accuracy from solar elevation angle and opaque cloud cover.
A-ll
-------
In essence, the submodel consists of a table wherein daytime
stability class is related to five classes of surface wind speed and
three classes of insolation strength.
Nighttime stability is tabulated
according to wind speed and two classes of cloud cover.
In general,
daytime stability increases with wind speed and decreases with increased
solar heating.
At night, the stability increases with the increased
surface cooling that accompanies clear skies.
Higher nighttime winds
will diminish the effectiveness of the cooling and reduce the stability.
The table used by the model reflects these effects.
It is based on
Pasquill's table which summarizes the interpretation of numerous field
tracer studies.
e.
Winds
-
The model's treatment of the winds that advect the pollutants
is quite simple, of necessity.
First, the wind is taken to be uniform
in speed and direction over the entire urban area.
This assumption is
dictated by the fact that many cities have only one wind measurement site.
For areas of relatively simple geography, it does not appear to be a
serious shortcoming.
It is most serious for areas that are subject
to land-sea breeze circulations.
Two wind speeds are used with the model; both can be derived
from the airport value.
The first wind speed is the transport wind
speed, or the average wind speed through the mixing layer. This has
been found in St. Louis to be reasonably well correlated with airport
measurements (Wuerch, 1971) and for reasons given in the body of the
report
we have chosen to use the value measured at the airport.
The
other wind speed that is important in the model is the "roof-level"
wind.
It tends to be less than the airport wind, but it is relatable
to it (Johnson et al.,197l; Schnelle et al.,1969).
The roof-level wind
affects the dilution and small-scale transport of pollutants released
within the street canyons and serves as one of the inputs to the street
canyon effects submodel discussed below.
A-12
-------
f.
Street Canyon Effects Submodel
The street canyon effects submodel was developed as a result of
studies in San Jose, California.
These studies indicated that CO con-
cent rat ions could change by a factor of 2, or several parts per million
from one side of the street to the other.
With such large gradients, it
is apparent that the effects must be included if reliable comparisons are
to be made with observed concentrations in such areas.
Observations have shown that when the roof-level wind blows
o
within about ~60 of the cross-street direction, a helical circulation
develops in the street.
At street level the cross-street component is
opposite the roof-level wind direction, causing a down flow of relatively
clean air in front of the 'downwind" buildings that face the roof-level
wind, and an upflow across the street.
Resulting low-level CO concen-
trations in front of the downwind buildings are appreciably less than
those observed across the street.
When winds are approximately parallel
to the street, cross-street gradients of CO are quite small.
A simple model based on physical principles is used to describe
the observed distribution of CO in the street canyon. First, the model
assumes that the emissions from the local street traffic are added to
the CO already present in the air that enters from roof level.
These
additive concentrations are proportional to the local street emissions,
QL(g m-1 s-l), and inversely proportional to the roof-level wind
speed u (m s-l), augmented by a small amount (0.5 m s-l has been found
to work well) to account for the mechanically induced air movement
caused by traffic.
In front of the downwind buildings the air begins its downward
flow at roof-level concentration.
Pollutants are gradually entrained as
the air sinks toward street level, so CO concentrations on this side of
the street increase slightly in the downward direction.
The CO that is
added should be inversely proportional to the width of the street Wand
A-13
-------
the wind speed u, because these two factors govern the volume of air
available for dilution of the emissions from the vehicles in the street.
The added pollutants should be directly proportional to the rate Q at
which these emissions are released.
Entrainment has been assumed to
vary linearly with height z through the depth (H) of the street canyon.
Thus the amount of CO that is added to the roof-level concentrations is
given by the following equation for the side of the street on which the
buildings face the wind:
Q.L
b.C=K
W(u + 0.5)
H-z
H
(A-5)
On the upwind side of the street the model uses box model
reasoning.
The volume into which the emissions are mixed is limited by
the air circulation toward the buildings and upward.
As the air moves
from the street-level source, the volume into which the pollutants are
mixed increases, so the concentration is taken to be inversely propor-
tional to the slant distance, r, between the receptor and the nearest
traffic lane.
For concentrations in front of the upwind buildings, the
equation is:
QL
t::,C = K
r(u + 0.5)
(A-G)
The constant, K, is the same for both equations. When the wind
blows nearly parallel to the street, the additive concentration, t::,C, is
described by the average of the values from the above two equations and
is the same on both sides of the street.
The development of these
equations is discussed in greater detail in Johnson et al.(1971) and
in the body of this report.
A-14
-------
3.
Uses of the Model
a.
General
The model has several available outputs, each designed to
satisfy particular needs.
Hourly concentrations for a few locations
can be calculated as a function of time.
Certain simplifications can
be incorporated so that hourly concentrations can be calculated for
periods of time up to several years.
This provides a basis for deter-
mining frequency distributions and other statistical features of the
calculated concentrations for the location.
Finally, the concentrations
can be calculated for various locations in a geographical grid, using
current or forecast meteorological data.
From such calculations detailed
horizontal concentration patterns can be developed for operational or
planning purposes.
The model can be applied to current or forecast traffic
patterns.
Thus, there is the very important capability to determine
in advance what the effects of planned highways will be.
These effects
can be determined as an area-wide pattern for specific meteorological
conditions, or on a climatological basis for a limited number of
arbitarily chosen special points.
In the following sections some
examples are given of these various applications.
b.
Hourly Concentration Sequences and Climatological Outputs
Figures A-5 and A-6 are examples of calculated concentrations
for extended periods; Figure A-5 is for conditions in San Jose using
an earlier version composite model, while results from the first
version of the basic model are shown in Figure A-6 for Cincinnati.
Figure A-6 also shows observed winds and cloud amounts and the mixing
heights and stabilities determined by the model.
This type of cal-
culation is most useful for evaluating the performance of the model
A-15
-------
25
20
E
0.
0.
~ 15
f=
oCt
a:
I-
Z
~ 10
z
o
u
o
u
FIGURE A-5
3-m HEIGHT
--- Calculated
- Observed
9 DEC 1970
10 DEC 1970
\
\
\ 11 DEC 1970
\
\
,
\
,
,
\
,
5
\
\
\
\
,
,
" ,...-,
,I "
,I "
o
08
13
18 08
18
18 08
13
13
TIME - PST
TA-8663-90R
CALCULATED AND OBSERVED CO CONCENTRATIONS
FOR A MIDBLOCK LOCATION IN SAN JOSE
and comparing it with the observed patterns.
If such calculations are
applied to several years of weather data, then the results can be
interpreted as shown in Figure A-7.
This figure shows the calculated
frequency distribution of CO concentration for weekdays, Saturday, and
Sunday for a downtown St. Louis location with 1965 traffic conditions.
If the model is reapplied in the same manner, but with
emissions appropriately projected for traffic conditions at some future
time, then frequency distributions can be forecast.
Figure A-8 shows
the results for the same St. Louis location, using a forecast of 1990
traffic and assuming that the auto population has the emission controls
that can be anticipated from current federal regulations.
This figure
also shows the effects of changing the averaging time.
As the averaging
time is increased, the spread of the frequency distribution decreases.
A-16
-------
o
J: 'lOCO
l-
n.
10.1
E 0
- ~ 2000
X
~
o
~ '100
I-
- U
0>10.1
.. cr
~c200
o
~
~
o
10.1
10.1
n.
- If)
'"
.... 0
E z
- ~
~ ~
- '"
o 0
.!! ~
X If)
10.1 J:
o I-
Z Z
- 10.1
>- I-
':: 0
...J ::>
- 0
III ...J
ct U
I-
If)
1'1
12
eu_ue OBSERVED
- CALCULATED
E
0.
0.
.
.
"
10
z
o
i
I-
Z
IIJ
o
Z
o
o
o
o
8
"
"
"
"
"
"
"
"
"
:'
. II '"
~:: ~~
"I .--".,
1,1 I I
." ,
" ,
I I
I ,
~_'
,~ ~
,,1 ~ :'
t - -. , ~
: ~ ,~ I I
~ ~:- - ~
I '" I
I. "
, ,
I I
'--'
6
..
"
"
.-_1 "
I ,
I ,
"
.,,'
2 I,(
--'
,
I
,
I
,
'I
.'... -
'..-.-.-.' .. . .
, .,
..,...''''''''''''':''
...... ..-:
'. ...-.-........-'"
'.,------
- '
,_-".M,-
"
, -'
",-. -,..' .
- .
o
20
10
o
10
5
o
o 20 40 60 80 100 120 140 160 HOURS
I-- MON +- TUES + WED + THURS-!-- FRI -t- SAT -+-- SUN -1
CINCINNATI. OHIO DATA (DECEMBER 14-20. 1964)
TA-7874-50
FIGURE A-6
OBSERVED AND CALCULATED CO CONCENTRATIONS AT
THE CINCINNATI CAMP STATION, 14-20 DECEMBER 1964
A-17
-------
60
..J 50
cr
>
cr
!oJ
I- 40
~
en
en 30
cr
..J
U
...... 20
I-
Z
!oJ
U
cr 10
!oJ
Q.
0
60
..J 50
cr
>
cr
!oJ
I- 40
~
en 30
en
cr
..J
U
...... 20
I-
Z
!oJ
U 10
cr
!oJ
Q.
0
60
50
..J
cr
>
cr
!oJ 40
I-
~
en 30
en
cr
..J
U 20
......
I-
Z
!oJ 10
u
cr
!oJ
Q.
0
0.1 0.2
CO
(a) WEEKDAY-HRS
(b) SAT.- HRS
(c) SUN.-HRS
0.5
2
5
10
20
50
CONCENTRATION - ppm
TA-7874-113R
FIGURE A-7
CALCULATED ST. LOUIS CAMP STATION CO CONCENTRATION
FREQUENCY DISTRIBUTION FOR 1965 TRAFFIC CONDITIONS-
WEEKDAY, SATURDAY, AND SUNDAY HOURS
A-IS
-------
...J 100
«
>
a::
~ 80
z
~ 60
«
...J
u
....... 40
I-
Z
I.IJ
U
:5 20
Il.
100
...J
«
>
a::
I.IJ 80
I-
Z
UI 60
UI
«
...J
u
....... 40
I-
Z
I.IJ
~ 20
I.IJ
Il.
100
...J
«
>
a:: 80
I.IJ
I-
Z
UI 60
UI
«
...J
~ 40
I-
Z
I.IJ
~ 20
I.IJ
Il.
FIGURE A-a
(0) ALL HOURS
o
(b) 8- HR MEAN
o
(c) 24- HR MEAN
o
0.1
10
20
2
5
50
0.2
05
CO CONCENTRATION - ppm
TA-7874-67
CALCULATED ST. LOUIS CAMP STATION CO CONCENTRATION
FREQUENCY DISTRIBUTION FOR 1990 TRAFFIC CONDITIONS-
1-HOUR, a-HOUR, AND 24-HOUR AVERAGES
A-19
-------
Maps of Concentration Distribution
c.
For operational applications, the ability of the model to
produce a map of expected CO concentration in an urban area is quite
useful.
This capability is illustrated in Figure A-9.
This figure also
illustrates the ability of the model to calculate results for different
grid spacing.
In the upper part of the figure, concentrations at 625
grid points spaced at 1-mile intervals were calculated,and from these
the computer analysis that is shown was prepared.
The calculations were
repeated for a smaller area with smaller grid-point spacing resulting in
the more detailed picture of CO distribution shown in the lower part of
the figure.
poses.
The grid-point calculations can also be used for planning pur-
Figure A-10 shows the results of applying the model for three
different emission fields while maintaining the same meteorological
inputs.
In this case, the three fields were calculated for morning
rush hour traffic, relatively stable atmospheric conditions, a mixing
height of 120 m, and a 2 m s-l west wind.
Figure A-10(a) shows the
field for 1964 traffic; Figure A-10(b) shows the field for 1990 traffic
with no emission controls on the vehicles; and Figure A-10(c) illus-
trates the effects of introducing the anticipated emission control
devices.
A-20
-------
12
10
..
~ 8
E
6
z
0
f: 4
«
I-
(I) 2
c..
~
«
u 0
LL
0
J: -2
I-
a:
0 -4
z
UJ
u -6
z
«
I-
!!2 -8
CI
-10
(>11.5
-12
-12 -10 -8
..
~ 08
E .
I 0.6
z
o
f: 0.4
«
I-
(I) 0.2
c..
~
;) 0
LL
o -02
J: .
I-
a:
o -0.4
z
~ -0.6
z
«
I-
!!2 -0.8
CI
FIGURE A-9
~5
dJ.5
05
1500-1600 CDT
15 OCTOBER 1964
WIND 310"/1.5 m ,-I
{;0. MIXING DEPTH 1670 m
~ UNSTABLE
-6
-2
10 12
8
-4
o
2
4
6
DISTANCE EAST OF CAMP STATION -- miles
TA-7874-26
(a) 1-MILE (1.6 kml GRID SPACING
1.2
1.0
Q
~
-1.0
u
~o9.0
-1.2
-1.2 -1.0 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1.0 1.2
DISTANCE EAST OF CAMP STATION - miles
TA-7874-24
(bl 0.1-MILE (0.16 km) GRID SPACING
CALCULATED ST. LOUIS CONCENTRATION PATTERNS FOR
TWO GRID SIZES
A-21
-------
en
w
..J
:2
....
N
/
Q
Q./
,"-"--~ -,
/
/
'.
,r-"'-
- -?t:~f:26t.
//
(a)
1964, BASED ON HISTORICAL TRAFFIC DATA
.--.-4 ._-~
(b)
1990 FORECAST TRAFFIC WITHOUT EMISSION
CONTROLS
FIGURE A-10
'-/ ..:,/
.oJI"
'f
.GV
/~l
i
--/1
",~'
,f,) ,
" ..
-', "/~
, /'
~- ,,--~< "'1
""~~';;:~t
, 1
"
(e)
1990 FORECAST TRAFFIC WITH EMISSION
CONTROLS
T A-653583-16
CALCULATED CO CONCENTRATIONS IN ST. LOUIS FOR HISTORICAL AND
FORECAST TRAFFIC CONDITIONS
A-22
-------
Appendix B
LIST OF SYMBOLS
B-1
-------
Symbol
Unit
Appendix B
LIS T OF SYMBOLS
Definition
Equation
Number(s)
(A-I), (A-2)
a, a.
J
b, b.
J
C, Ci
c
d
E
H
h
m
Dimensionless
-3
gm m
ppm
Dimensionless
-1
gm(vehicle-mile)
m
m
Coefficient used in
representing the
vertical dispersion
coefficient (for
stability class j)
as an exponential
function of travel
distance from source
Exponent used in
representing the
vertical dispersion
coefficient (for
stability class j)
as an exponential
function of travel
distance
CO concentration
arising from
th
emissions in the i
segment (see Fig. A-3)
Term used to correct
for zero offset in CO
instrumentation
Coefficient used to
correct for span
changes in CO
instrumentation
CO emission rate
Depth of street
canyon
Mixing height
B-3
(A-I), (A-2)
(A-2),(A-3)
(1)
(1)
(A-4)
(6), (8), (A-5)
(A-3)
-------
Equation
Symbol Unit Defini tion Number(s)
K Dimensionless Proportionality (5),(6),(7),(8),
constant used in (A-5) , (A-6)
street canyon
submodel. K
found to equal 7
in San Jose and
St. Louis
-1 (5), (6), (7), (8),
N Vehicles hr Traffic volume
(A-5) , (A-6)
N Dimensionless Fraction of sky (2)
covered by opaque
clouds
-2 -1 (A-2),(A-3)
QAi gm m s Area source
emission rate
for ith segment
-1 -1 (A-5),(A-6)
QL gm m s Line source
emission rate
r m Straight-line (A-6)
distance between
receptor and
nearest traffic
lane
-1
S mi hr Average vehicle (5),(6),(7),(8),
speed on traffic (A-5),(A-6)
link
-1
u m s Wind speed at (5), (6), (7), (8) ,
roof level (A-4)
-1
u m s Wind speed measured (3), (4)
a
at airport
-1
Ut m s Transport wind (3), (4), (A-2),
speed in mixing (A-3)
layer
B-4
-------
Symbol
u,v,w,
W
x
x
x
i
z
C'i
ex:
s
Unit
-1
m s
m
m
m
m
m
-1
gm(vehicle mi)
Degrees
Dimensionless
Defini tion
Wind components
measured in street
canyon experiment; u,
parallel to Locust
Street, positive
toward east; v
parallel to Broadway
positive to north; w
positive upward (see
Fig. 3)
Street width
Travel distance
from source to
receptor
Cross-street
distance from
receptor to nearest
traffic lane
Distance from
receptor to nearest
boundary of the ith
segment (see Fig. A-3)
Height of receptor
Coefficient in
exponential
equation describing
automotive emissions
as a function average
vehicle speed
Solar elevation angle
Exponent in equation
describing automotive
emissions as a function
of average vehicle speed
B-5
Equation
Number(s)
(5),(6),(8),
(A-5)
(A-I)
(7),(8)
(A-2),(A-3)
(6), (7), (8),
(A-5)
(A-4)
(2)
(A-4)
-------
Equation
Symbol Unit Definition Number(s)
b.C ppm Concentration at (A-5),(A-6)
receptor minus roof
level concentration
b.CI ppm Some as b.C, but (8)
specifically for
0
winds within 30
of the street
direction
b.C ppm Same as b.C, but (7) , (8)
L
for receptors in
front of buildings
facing in the
direction the
roof-level wind
is blowing
b.C ppm Same as b.C, but for (5), (6) , (8)
W
receptors in front
of buildings
facing toward the
roof-level wind
O"z m Vertical dispersion (A-I)
coefficient (see
Fig. A-I)
ij) Dimensionless Insolation parameter (2)
B-6
-------
REFERENCES
Chagnon, S. A., Jr., 1971 Operational Report for
National Laboratory, University of Chicago,
Water Survey, and University of Wyoming.
METROMEX, Argonne
Illinois State
Chandler, T. J., 1965: The Climate of London, Hutchison and Co.,
Ltd., London, 292 pp.
Chang, ~. C., P. N.
a city street,
December 7-10,
New Mexico.
Wang, and A. Lin, 1971: Turbulent diffusion in
Air Pollution, Turbulence and Diffusion Symposium,
1971, New Mexico State University, Las Cruces,
Curry, D. A., ~nd D. G. Andersen, 1971: Procedures for estimating
highway user costs and air and noise pollution effects, Final
Report, Highway Research Board Project 7-8, Stanford Research
Institute, Menlo Park, California.
Dabberdt, W. F., 1968a: Tower-induced errors in wind profile
measurements, J. Appl. Met., Z, 359-366.
Dabberdt, W. F., 1968b: Wind disturbance by a vertical cylinder
in the atmospheric surface layer, J. Appl. Met., Z, 367-371.
Holzworth, G. C., 1967: Mixing depths, wind speeds, and
tion potential for selected locations in the United
J. Appl. Met., ~, 1039-1044.
air pollu-
States,
Johnson, W. B., W. F. Dabberdt, F. L. Ludwig and R. J. Allen, 1971:
Field study for initial evaluation of an urban diffusion model
for carbon monoxide, Comprehensive Report for Coordinating
Research Council and Environmental Protection Agency, Contract
CAPA-3-68 (1-69), Stanford Research Institute, Menlo Park,
California, 240 pp., National Technical Information Service
No. PB 203469.
Junge, C. E., 1963: Air Chemistry and Radioactivity, Academic
Press, New York, 392 pp.
R-l
-------
Ludwig, F. L., 1970: Urban air temperatures and their relation to
extraurban meteorological measurements, paper presented at
Semiannual Meeting of the American Society of Heating,
Refrigerating and Air-Conditioning Engineers, January 19-22,
1970, San Francisco, California.
Ludwig, F. L., W. B. Johnson, A. E. Moon, and R. L. Mancuso, 1970:
A practical multipurpose urban diffusion model for carbon monoxide,
Final Report, Coordinating Research Council Contract CAPA-3-68,
National Air Pollution Control Administration Contract CPA
22-69-64, Stanford Research Institute, Menlo Park, California,
184 pp., National Technical Information Service No. PB 196003.
Ludwig, F. L., and J. H. S. Kealoha, 1968: Urban Climatological Studies,
Final Report, Contract CD-DAHC-20-67-C-0136 (Work Unit l235A),
Stanford Research Institute, Menlo Park, California, 195 pp.,
National Technical Information Service No. AD 672 071.
Ludwig, F. L., and E. Robinson, 1965: Size distribution studies with
the aerosol spectrometer., J. Air Poll. Cont. Assoc., 15, 102.
Mancuso, R. L., and F. L. Ludwig, 1972 (in preparation): User's
Manual for the APRAC-lA Diffusion Model Computer Program, Coor-
dinating Research Council and Environmental Protection Agency,
Contract No. CAPA-3-68{1-69), Stanford Research Institute, Menlo
Park, California.
McElroy, J. L., and F. Pooler, Jr., 1968: St. Louis dispersion
Vol. II--Analysis, 51 pp., National Air Pollution Control
Administration Publication No. AP-53, 51 pp.
study,
Missouri State Highway Department, 1970: Traffic map of St. Louis,
St. Louis County and St. Charles City.
Ott, Wayne, 1971:
variation of
thesis, Dept.
153 pp.
An urban survey technique for measuring the spatial
carbon monoxide concentrations in cities. PhD
Civil. Eng., Stanford University, Stanford California
Pasquill, F., 1961: The estimation of the dispersion of windborne
material, Meteorological Mag. (London), 90, pp. 33-49.
R-2
-------
Pooler, F., 1966: A tracer study of dispersion over a city, J. Air
Poll. Contr. Assoc., 11, pp. 677-681.
Robinson, E., and F. L. Ludwig, 1967: Particle size distribution
of urban lead aerosols, J. Air Poll. Cont. Assoc., 17, 664.
Schnelle, K. B., F. G. Ziegler, and P. A. Krenkel, 1969: A study of
the vertical distribution of carbon monoxide and temperature
above an urban intersection, APCA Paper No. 69-152, Air Pollution
Control Association.
Slade, D. H. (ed.), 1968: Meteorology and Atomic Energy, U.S. Atomic
Energy Commission, Oak Ridge, Tennessee.
Smith, M. (ed), 1968, Recommended Guide for the
Dispersion of Airborne Effluents, American
Engineers, New York.
Prediction of the
Society Mechanical
Smithsonian Institution, 1951:
527 pp.
Smithsonian Meteorological Table,
Stanford University Aerosol Laboratory and The Ralph M. Parsons Co.,
1953a: Behavior of aerosol clouds within cities, Joint Quarterly
Report No.3, January-March 1952, Contracts DA-18-064-CML-1856
and DA-18-064-CML-2282, DDC No. AD31711.
Stanford University Aerosol Laboratory and The Ralph M. Parsons Co.,
1953b: ibid; Joint Quarterly Report No.4, April-June 1953,
DDC No. AD31508.
Stanford University Aerosol Laboratory and The Ralph M. Parsons Co.,
1953c: ibid; Joint Quarterly Report No.5, July-Sept., 1953,
238 pp., DDC No. AD31507.
Stanford University Aerosol Laboratory and The Ralph M. Parsons Co.,
1953d: ibid; Joint Quarterly Report No.6, Vol. I, October-
December, 1953, DDC No. AD31510.
Stanford University Aerosol Laboratory and The Ralph M. Parsons Co.,
1953e: ibid; Joint Quarterly Report No.6, Vol. II, October-
December 1953, DDC No. AD31711.
Turner, D. B., 1964:
Met., ~, 83-91.
A diffusion model for an urban area, J. Appl.
R-3
-------
Wuerch, D. E., 1970: A comparison of observed and
mixing depths, Environmental Sciences Services
Technical Memorandum WBTM CR-36.
calculated urban
Administration
Wuerch, D. E., 1971: An investigation of the resultant
wind within the urban complex, National Oceanic and
Administration Technical Memorandum NWS CR-44.
R-4
transport
Atmospheric
------- |