EPA-650/4-75-018-b
EVALUATION
OF THE MULTIPLE SOURCE
GAUSSIAN PLUME DIFFUSION MODEL -
PHASE II
by
Robert C. Koch and Scott D. Thayer
Geomet, Inc.
15 Firstfield Road
Gaithersburg, Maryland 20760
Contract No. 68-02-0281
ROAP No. 21ADO
Program Element No. 1AA009
EPA Project Officer: D. Bruce Turner
Meteorology Laboratory
National Environmental Research Center
Research Triangle Park, North Carolina 27711
Prepared for
U.S. ENVIRONMENTAL PROTECTION AGENCY
OFFICE OF RESEARCH AND DEVELOPMENT
WASHINGTON, D. C. 20460
April 1975
-------
EPA REVIEW NOTICE
This report has been reviewed by the National Environmental Research
Center - Research Triangle Park, Office of Research and Development,
EPA, and approved for publication. Approval does not signify th^t the
contents necessarily reflect the views and policies of the Environmental
Protection Agency, nor does mention of trade names or commercial
products constitute endorsement or recommendation for use.
RESEARCH REPORTING SERIES
Research reports of the Office of Research and Development, U.S. Environ-
mental Protection Agency, have been grouped into series. These broad
categories were established to facilitate further development and applica-
tion of environmental technology. Elimination of traditional grouping was
consciously planned to foster technology transfer and maximum interface
in related fields. These series are:
1. ENVIRONMENTAL HEALTH EFFECTS RESEARCH
2. ENVIRONMENTAL PROTECTION TECHNOLOGY
3. ECOLOGICAL RESEARCH
4. ENVIRONMENTAL MONITORING
5. SOCIOECONOMIC ENVIRONMENTAL STUDIES
6. SCIENTIFIC AND TECHNICAL ASSESSMENT REPORTS
9. MISCELLANEOUS
This report has been assigned to the ENVIRONMENTAL MONITORING
series. This series describes research conducted to develop new or
improved methods and instrumentation for the identification and quanti-
fication of environmental pollutants at the lowest conceivably significant
concentrations. It also includes studies to determine the ambient concen-
trations of pollutants in the environment and/or the variance of pollutants
as a function of time or meteorological factors.
This document is available to the public for sale through the National
Technical Information Service, Springfield, Virginia 22161.
Publication No. EPA-650/4-75-018-b
11
-------
TABLE OF CONTENTS
1.0 INTRODUCTION AND SCOPE
2.0 SUMMARY OF THE PHASE I REPORT 6
2.1 Models Compared - 6
2.2 Model Comparisons 7
2.3 Conclusions from the Phase I Report 9
3.0 PHASE II WORK REPORTED AND ACCOMPLISHED ELSEWHERE 12
3.1 SCIM Computer Program and User's Manual 12
3.2 Training of EPA Personnel 14
4.0 CONSIDERATIONS OF EMISSION DATA, EMSU DATA AND
SHORT-TERM MAXIMUM CONCENTRATIONS 15
4.1 Diurnal and Seasonal Variations in Emissions 15
4.2 Use of EMSU Data as Meteorological Input 21
4.3 Procedure for Calculating Annual Short-Term
Maximum Concentrations 35
5.0 REFERENCES 46
Appendix A-l METHOD OF ESTIMATING THE HEIGHT OF THE
MIXING LAYER
Appendix A-2 MIXING HEIGHT INTERPOLATION
-------
Section 1.0
INTRODUCTION AND SCOPE
-------
Section 1.0
INTRODUCTION AND SCOPE
This report represents a summation, for record purposes of a
variety of types and phases of work conducted under EPA Contract Number
68-02-0281, "Evaluation of the Multiple-Source Gaussian Plume Diffusion
Model." Because the contract work covered an extended period of time,
and because its products were documented in a number of reports, it is
considered desirable to have this summation document as a matter of com-
plete record of the work; of course, where the work has been formally
reported elsewhere, this report will contain only a summary.
The scope of this report will be to briefly cover Phase I by
reproducing the introduction, model description, and conclusions of the
report of that work (Number EF-186). Phase II will be more definitively
covered by presenting a summary of the report (Number EF-261) of the com-
puter program and user's manual and describing the training provided EPA
staff, and by accumulating the work reported elsewhere related to special
aspects of the model (handling variations in emission data input, using
EMSU data as meteorological input, and calculating short-term maximum
concentrations). Phase I is covered in Section 2.0 of this report, the
computer program, user's manual, and staff training in Section 3.0, and
the special aspects in Section 4.0.
For convenience of reference, and to complete the introductory
remarks, the contract scope of work is quoted in the following paragraphs.
-1-
-------
For cross-referencing purposes, the following relationships are given
relating the scope of work's paragraphs 1 and 2 (Phase I), and 3 through
7 (Phase II) with this report's sections:
Contract Scope of Work Report Section
Phase I 1, 2, 3 (Phase II) 2.1
1, 2 2.2
Phase II 3 3.1
4 3.2
5 4.1
6 4.2
7 4.3
SCOPE OF WORK
Background:
This program shall be a continuation of work previously
supported under Contract No. CPA 70-94, "Validity and Sensi-
tivity of the Gaussian Plume Urban Diffusion Model." (Avail-
able from NTIS as PB 206-951).
This previous work developed a short-term, steady-state
Gaussian plume model for urban diffusion and evaluated this
model using three-months' data for St. Louis and one-month's
data for Chicago. By proper selection of input data, this
model can also be used for long-term average concentrations.
Since this model is expected to replace a currently used
annual model, it is necessary to make direct comparisons with
two other models with the same data base used by these two
models.
It is required to completely document the model so that
the dispersion modeler can completely understand its steps of
operation in detail including underlying assumptions and the
-2-
-------
Abstract:
Phase I:
technical user can properly assemble required input data and
interpret correctly the concentrations resulting from the model.
The Contractor shall conduct a research program for further
evaluation and documentation of a Gaussian Plume Urban Diffusion
Model.
For both St. Louis and Chicago data (as considered in the
previous contract) calculate 1-hour (Chicago) and 2-hour
(St. Louis) concentration frequency distributions using
GEOMET (Mean Q) and using sound statistical techniques
compare with the results of GEOMET (Variable Q), calcu-
lated by the previous contract and with measured concen-
trations.
Calculate concentrations for locations in the New York
area for 1969 and make comparisons with measurements for
the averaging times and the models indicated by the sponsor.
Similar calculations will be made for particulate matter
for annual averages only.
Mean annual emission rates for all point and area sources
and stack characteristics for the point sources for the New York
region will be furnished by EPA in the format used for IPP.
Meteorological information for the year 1969 consisting
of an observation each three hours will also be furnished by
EPA and will be used as the meteorological data base.
For each calculation of concentration at a receptor, a
value for the concentration due to point sources will be retained
as well as the calculated total concentration. All estimates of
concentrations made will be stored on magnetic tape (hourly and
24-hour concentrations as time series) and delivered to EPA for
possible subsequent analysis at the conclusion of the contract
period.
Data on measured S02 concentrations will be furnished the
contractor by EPA. Frequency distributions for measured con-
centrations and for calculated concentrations will be determined
-3-
-------
Phase II:
and compared by log-probability plots and by appropriate sta-
tistical techniques for each station. (There are approximately
40 locations with sufficient S02 measurements to obtain fre-
quency distributions.) In consultation with the project
officer, the contractor will select 10 for study of frequency
distributions.
Appropriate subsets of calculations will be used to vali-
date the usefulness of proportionate stratified sampling in
obtaining frequency distributions. Resulting frequency distri-
butions from these subsets need only be compared with the cal-
culated frequency distribution using all data.
For each station a linear correlation coefficient, the
variance (the square of the correlation coefficient), the
slope and intercept of a least-squares regression line will be
determined considering the calculated concentration as the inde-
pendent variable and the measured concentration as the dependent
variable. Considering error as the calculated value minus the
measured value, the mean absolute error, the root-mean-square
error, and the distribution of errors will be determined. For
all but the distribution of errors, there will be a value of
the above numerated statistics for each station for both 1-hour
and 24-hour averaging times for each applicable model. For
annual averages, the pairs of calculated and measured concentra-
tions for all stations will be included to calculate one value
of each of the above statistics for each pollutant for each
model.
3. Considering the results of the sensitivity analysis (pre-
vious contract) and the results of the evaluation in Tasks 1
and 2, restructure the computer programs used in GEOMET
(Variable Q) to minimize the digital computer execution
times. One version of these programs must be compatible
with the IPP. (Information on the IPP will be furnished
by EPA.) If simplification will reduce computer time
without significant loss of accuracy, a separate model shall
be suggested and documented for annual average concentra-
tions. Prepare a user's manual for the use of these opti-
mized computer programs for use with or without the IPP.
4. Train 2 to 3 members of the Model Development Branch,
Division of Meteorology, on the operation of the optimized
models resulting from Task 3, on a computer used by NERC,
North Carolina. This shall include a demonstration of the
-4-
-------
compatibility with the IPP. (EPA personnel shall be
responsible for operation of all phases of the IPP not
directly connected with the dispersion model.)
5. Write procedures to be used in preparing emission data so
that diurnal and seasonal variations in emissions can be
used.
6. Evaluate the use of Environmental Meteorological Support
Unit (EMSU) data for determination of meteorological param-
eter values for input to the optimized model. Enumerate
procedures for the use of such data.
7. Examine the statistical portion (Larsen transform) of the
AQDM and suggest alternative procedures to be utilized
with the optimized dispersion model to estimate short-term
(1-hour, 3-hours, and 24-hour) maxima that occur with a
frequency of once per year.
-5-
-------
Section 2.0
SUMMARY OF THE PHASE I REPORT
-------
Section 2.0
SUMMARY OF THE PHASE I REPORT
The report of Phase I (Number EF-186) contains complete docu-
mentation of the work on the basic model, and evalution of its valida-
tion performance against monitored air quality data in comparison to
other models potentially usable for similar purposes. The work is sum-
marized here by reproducing brief descriptions of the models compared,
analyses performed, and conclusions of that report.
The work accomplished during Phase I concentrated on validation
of a steady-state Gaussian plume urban diffusion model which uses sampled
chronological input data. The model was developed for EPA by GEOMET in
previous work (Contract Number CPA 70-94). The model has been compared
with three other models (using the same data base).
2.1 MODELS COMPARED
The model studies include three variations of the multiple-
source, Gaussian plume, meteorological diffusion model. Two of the models,
the Air Quality Display Model (AQDM) and the Climatological Dispersion
Model (COM)5 are primarily designed to calculate long-term mean concentra-
tions. The third model, developed by GEOMET under previous EPA sponsorship,
is designed to calculate both the long-term mean concentration and the
frequency distribution of short-term concentrations using selected chrono-
logical data (SCIM). The frequency distribution is determined by concen-
trations calculated for a statistically selected set of short-term periods.
Representative meteorological characteristics and simulated time-dependent
-6-
-------
emission characteristics are determined for each selected period. The
other two models use mean emission characteristics and a specified set
of combinations of meteorological characteristics (wind direction, wind
speed and stability). To determine the long-term mean, the calculations
for each combination of meteorological conditions are weighted by the
relative frequency of occurrence of the combination.
In addition to the three Gaussian plume models, the simplified
version of the Gifford-Hanna model recently described by Hanna (1971) was
included. The model is
- r Q
x - C u
where
x = concentration at a receptor location (yg/m^)
Q = area source strength surrounding the receptor (ug/m^/sec)
u = wind speed (m/sec)
C = dimensionless constant.
2.2 MODEL COMPARISONS
Calculations using the four models were compared against each
other and against measured values. Each model was run in its normal mode.
In addition, certain simplifications were made by averaging the inputs
used for the model. The model comparisons include consideration of 10 dif-
ferent variations of model and inputs.
Two model comparison tasks were carried out. The first task was
to compare calculations for SCIM which were obtained in a preceding program
(Contract No. CPA 70-94, "Validity and Sensitivity of the Gaussian Plume
-7-
-------
Urban Diffusion Model") with calculations made using the same model and
data set, except that mean rather than time-dependent emission rates were
used for all sources. These calculations involved a 3-month data period
for St. Louis and a 1-month data period for Chicago, for sulfur dioxide
emissions.
The second task was to compare 10 different combinations of
variations in input and the four models described above with each other
and with measured values using a 1-year data set for New York City. The
comparisons include 3-hourly, 24-hourly, and annual concentrations of
sulfur dioxide emissions, and annual concentrations of particulate matter
emissions.
For the SCIM, area source emissions and the meteorological con-
ditions of atmospheric stability and height of the mixing layer (grouped
together) were treated either as varying from hour to hour or as being
constant throughout the data period. Three combinations of input data
conditioning were analyzed, including:
• Area source emission rates, atmospheric stability and
height of the mixing layer variable
• Area source emission rates constant, but atmospheric
stability and height of the mixing layer variable
e Area source emission rates, atmospheric stability and
height of the mixing layer constant.
For the simplified Gifford-Hanna Model (GHM), area source emis-
sions and wind speed were treated as both varying from hour to hour or as
being constant throughout the data period. In addition, the calculated
concentration at a receptor due to point sources (as estimated by SCIM
-8-
-------
with variable atmospheric stability and mixing layer height) were either
added or not added to the GHM calculations. This results in four varia-
tions of this model, including:
• Constant area source emission rates and wind speed,
without point sources
o Variable area source emission rates and wind speed,
without point sources
e Constant area source emission rates and wind speed,
with point sources
« Variable area source emission rates and wind speed,
with point sources.
Calculations for COM, which treat atmospheric stability and
height of the mixing layer as either both variable or both constant, were
furnished by Mr. D. B. Turner of the Division of Meteorology, EPA/NERC/RTP.
Statistical results of model-to-measurement comparisons for these calcula-
tions were included for comparison with the other models. Calculations for
AQDM (no variations) also were furnished by Mr. Turner and were included
for comparison.
2.3 CONCLUSIONS FROM THE PHASE I REPORT
Conclusions (1-5) regarding the use of the Sampled Chronological
Input Model (SCIM), a multiple-source Gaussian plume model, to estimate
short-term SO,, concentrations (e.g., 1-hour and 24-hour concentrations) are
based on model-to-measurement comparisons for 1 month of Chicago, 3 months
of St. Louis and 1 year of New York City (NYC) data. The model was analyzed
using NYC data for three types of inputs, including: (1) variable emission
-9-
-------
rates, stability classifications and mixing heights (variable Q, S, H),
(2) mean emission rates and variable stability classifications and mixing
heights (mean Q, variable S, H), and (3) mean emission rates, stability
classifications and mixing heights (mean Q, S, H). The model was analyzed
using St. Louis and Chicago data for the first two types of input. For
comparison purposes, an analysis was also made of the use of the simplified
Gifford-Hanna Model (6HM).
1. Comparing the results for the three types of input to SCIM,
it is concluded that:
0 Use of a mean, rather than a variable, emission rate may
either increase or decrease the root-mean-square error
(RMSE) at a receptor but will decrease the correlation
with measurements (observed at 10 of 10 St. Louis receptors
for 2-hour concentrations, 5 of 8 Chicago receptors for
1-hour concentrations, and 10 of 10 NYC receptors for
1-hour and 24-hour concentrations).
e Based on comparisons using NYC data and mean emission rates,
the use of a neutral stability classification and a mean mixing
height will decrease the correlation with measurements but will
also decrease the RMSE at a receptor (observed at 9 of 10 recep-
tors for 1-hour concentrations and 8 of 10 receptors for 24-hour
concentrations).
o Based on NYC comparisons, the combined use of a mean emis-
sion rate and mixing height and a neutral stability classi-
fication will decrease the correlation with measurements
but will decrease the RMSE (observed at 10 of 10 receptors
for 1-hour concentrations and 7 of 10 receptors for 24-hour
concentrations).
2. In evaluating GHM, it was concluded that adding point source
contributions (i.e., calculated using SCIM) to GHM calculations improved
the results for this model. The RMSE was smaller at 6 of 10 NYC receptors,
the correlation coefficient was higher at 6 of 10 receptors, and the standard
deviation of calculated concentrations was closer to the standard deviation
of measured concentrations at all 10 receptors.
-10-
-------
3. Comparing SCIM and GHM using hourly calculations and S02
measurements, NYC data, SCIM produced the least annual mean error at 6
of 10 receptors, the closer agreement, between standard deviations of cal-
culated and measured concentrations at 5 of 10 receptors, the least error
in estimating the maximum measured concentrations at 6 of 10 receptors,
and the highest correlation coefficient at 3 of 10 receptors; GHM produced
the least RMSE at all 10 receptors. Results for comparison of 24-hour S02
concentrations are similar but slightly more favorable to SCIM.
4. There is a need to improve the input data used with the
multiple-source Gaussian plume type model, particularly atmospheric sta-
bility information, since the model is very sensitive to the rather gross
changes in stability which are routinely introduced. SCIM calculations on
the average, greatly overestimated concentrations associated with Turner-
Pasquill stability classes 2 and 5.
5. Calculations based on a NYC emission algorithm developed in
this report, particularly when applied with GHM, generally agree with diur-
nal and temperature dependent trends in measured S02 concentrations. Fur-
ther improvements in this algorithm are desirable but require more detailed
information.
Conclusions (6-8) regarding the use of several versions of the
multiple-source Gaussian plume model and GHM to estimate long-term mean
concentrations of S02 and particulates are based on model-to-measurement
comparisons for the same data periods and locations.
6. The use of variable emission rates for SCIM and GHM are not
able to demonstrate any conclusive improvement in model validity over the
use of mean emission rates. It is inferred that this result is due to the
failure to properly treat other causes of variance, such as those associated
with atmospheric stability.
7. Based on results for NYC, the Climatological Dispersion Model
(CMD) and SCIM versions of the multiple-source Gaussian plume model produce
a smaller station-to-station RMSE than the Air Quality Display Model (AQDM)
version (i.e., RMSE's of 52 and 59, respectively, compared to 92, with an
overall mean of 135 yg/m3 of S02; RMSE's 22 and 22 compared to 36 with an
overall mean of 82 yg/m3 of particulates).
8. Although the NYC validation statistics for GHM, COM, and
SCIM are similar for S02, GHM results for particulates have a much higher
station-to-station RMSE than do COM and SCIM (i.e., RMSE of 60 compared to
22, with an overall mean of 82 yg/m3).
-11-
-------
Section 3.0
PHASE II WORK REPORTED AND ACCOMPLISHED ELSEWHERE
-------
Section 3.0
PHASE II WORK REPORTED AND ACCOMPLISHED ELSEWHERE
3.1 SCIM COMPUTER PROGRAM AND USER'S MANUAL
The treatments called for in the contract scope (Section 1.0)
were performed on the SCIM computer program. The program itself and its
use were documented in GEOMET Report EF-261, as briefly indicated in the
excerpts from that report which follow.
The Sampled Chronological Input Model (SCIM) is an urban air
pollution simulation. It is designed to provide the user with a method
of estimating the air quality characteristics of a particular pollutant
over a specified control area. Both the mean long-term concentration and
the frequency distribution of short-term concentrations are estimated
using conventional emission inventory and meteorological data.
The objective of User's Manual is to:
e Briefly describe the SCIM computer program and its
intended applications
e Provide guidance and sample programs to process
conventional data into the input forms required by
the SCIM program
e Describe how to set up and operate the SCIM program.
The SCIM computer program provides the user with a tool for esti-
mating short-term maxima of pollutant concentrations in addition to long-
term means. This is done by calculating concentrations for a sample of
short-term periods selected from a specified long-term period. The sample
is then used to estimate the long-term mean concentration, the geometric
-12-
-------
standard deviation and the statistical frequency distribution of short-
term concentrations for all specified locations. The expected annual
maximum concent! ;tion may be determined from the frequency distributions
or by means of the geometric standard deviation (e.g., see Larsen 1971).
The calculations are made for specified receptor locations.
The calculations are made using a multiple-source Gaussian
plume model. Emissions from large stationary sources are represented by
elevated point sources. All other emissions are represented by an area
source. Contributions from the area source to concentrations at a recep-
tor are calculated using a numerical technique to evaluate the integral
equation which must be solved. The narrow plume concept which implies
that crosswind variations in emission rates may be neglected is an impor-
tant assumption in the numerical technique. This assumption is valid as
long as the distance between variations in the area source emission rate
is large compared to the crosswind diffusion parameter (a ). A critical
characteristic of the numerical technique is the spacing of grid points
for which emission rates per unit area are determined. Model sensitivity
findings show that a spacing of one-quarter mile is important in areas of
high spatial variations of emissions. More generally, spacings of 1 km
or more are satisfactory.
A significant feature of this program is that varying patterns
of emissions are linked to a chronology of weather observations so that
related variations in emission rates and in the dispersive capability of
the atmosphere can be taken into account. In the emission algorithm pre-
sented here emission rates are related to ambient air temperature and to
-13-
-------
hour of the day. This algorithm is especially applicable to emissions
which are related to space heating requirements. The program has been
tested and found applicable to estimating sulfur dioxide and particulate
air quality characteristics.
The program inputs are prepared from conveniently available
data, including.Implementation Planning Program (IPP) or Air Quality
Display Model (AQDM) emission data and standard weather data which is
available on punched cards or magnetic tape from the National Weather
Records Center. The program analyzes the air quality of a region of inter-
est by calculating a sample of hourly concentrations at specified locations.
The user controls the sample size by specifying the sampling interval
between successive hours for which calculations are made. The standard
program outputs consist of a data file containing the concentrations cal-
culated for each specified location for each selected hour and a printed
statistical summary of the air quality characteristic of each location
and of all locations combined. In addition, the user may choose to use a
version of the program which will generate a Source Contribution File in
the correct format to interface with IPP.
3.2 TRAINING OF EPA PERSONNEL
The final training of EPA personnel called for in the contract
scope was provided at EPA by GEOMET staff in July of 1973. Ten to fifteen
staff members of the Office of Air Quality Planning and Standards and of
the Meteorology Laboratory were given instruction in the use of the pro-
gram. This instruction was augmented by subsequent extensive interaction
by phone and in person between GEOMET and EPA staff.
-14-
-------
Section 4,0
CONSIDERATIONS OF EMISSION DATA, EMSU DATA AND
SHORT-TERM MAXIMUM CONCENTRATIONS
-------
Section 4.0
CONSIDERATIONS OF EMISSION DATA, EMSU DATA AND
SHORT-TERM MAXIMUM CONCENTRATIONS
4.1 DIURNAL AND SEASONAL VARIATIONS IN EMISSIONS
The SCIM program is primarily designed to analyze air quality
levels associated with emissions of stable pollutants such as sulfur
dioxide, particulates and carbon monoxide. The emissions from any given
source will vary with hour of the day, day of the week and season of the
year. Standard emission factors have been developed for most pollutants
which allow estimates of emission rates to be established as a function
of fuel consumption rates or of processing rates for various industrial
activities. When these fuel consumption rates and processing rates are
described as functions of times, the emission rate of each pollutant is
well defined.
Unfortunately, information on variations of emissions with time
are not usually available. However, when emissions result from the con-
sumption of fuel for space heating, the emissions will vary with temper-
ature. Variations in these emissions with time can be estimated from
local temperature records which are available for almost all locations.
The consumption of fuel for space heating accounts for a certain percent-
age of the emissions of a pollutant from a particular source. A general
algorithm used in SCIM which describes emissions as a function of
-15-
-------
I
parameters which can be related to temperature and other activity indexes
• is the following:
| Q(t) = QA K(t)
K(t) = (1-F) A(t) + F [H(t) - T(t)] S(t), T(t) < H(t)
T(t) >.H(t) K(t) = (f-F) A(t).
where
Q(t) = emission rate at time t
Q. = average annual emission rate
K(t) = time dependent emission factor
F = fraction of emissions which result from space heating
requirements
A(t) = activity factor which defines activity level for time t
relative to annual average activity level for activities
which control emissions not related to space heating
requirements
H(t) = temperature threshold for space heating requirements for
time t
T(t) = temperature at time t
S(t) = sensitivity factor which defines rate of emission per
degree below temperature threshold at time t relative
to annual average rate of emission per degree below
temperature threshold.
In the above algorithm the parameters A(t), H(t) and S(t) may
vary with time of day, day of the week and week of the year. The informa-
tion required to determine these parameters as functions of time for every
point source and every square mile of an area source is far too detailed
-------
for what is normally economically feasible to collect and analyze. However,
it may be useful to derive city-wide parameters which can be applied to
area sources.
The ideal data for estimating the above parameters would be fuel
consumption records and process operating records for a large number of
sources. Lacking this, other less desirable data might be used. In the
Phase I report of this project, a large set of S0? concentration measure-
ments was used (12 years of almost continuous hourly observations) for
New York City. Blade and Ferrand (1969) summarized these measurements by
hour of the day, day of the week, and week and month of the year. The mean
hourly SO^ concentrations for each month of the year were correlated with
mean hourly temperature for each month of the year for the same data period.
Following the methods described in the Phase I report (Appendix A), the
parameter values presented in Table 4.4-1 were developed. On the basis
of this same data, it was estimated that 29 percent of the emissions are
dependent on temperature variations (i.e., F = 0.29).
SCIM is programmed to use the above algorithm and the parameter
values in Table 4.1-1 to estimate diurnal variations in area source emis-
sions. There are some drawbacks to using this data. The parameter values
for the emission algorithm are specifically applicable to S02 emissions in
New York City. It is not known how applicable these are to other cities.
Furthermore, since the emission rates were derived from S02 measurements
the parameter values may contain diurnal variations which are associated
with diurnal variations in meteorological conditions. The diurnal varia-
tion in activity factors shown in Table 4.1-1 does not make much sense
-17-
-------
Table 4. 1-1. Emission Parameters Developed from New York City SO2 Data
Hour of
the Day
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
A ctivity
Factor, A(A)
1.0272
1.0008
0. 9576
0. 9576
1.00344
1.1784
1.3032
1.3200
1.2624
1.1616
1.0104
0.9336
0. 8760
0. 8232
0.8160
0.8160
0.8232
0. 8424
0.8832
0. 9264
0.9744
1.0104
1.0272
1.0344
Temperature
Threshold, H(t), °F
56
55
55
55
56
58
59
61
63
64
65
65
65
65
65
65
65
65
65
65
65
64
62
60
Sensitivity
Factor, S(t), "^F"1
0.0330
0.0280
0.0280
0. 0293
0.0371
0.0717
0. 1208
0. 1437
0.1214
0. 0977
0. 0893
0.0841
0.0841
0.0852
0. 0856
0. 0886
0. 0983
0. 1060
0.1120
0.1142
0.1114
0. 1024
0. 0696
0.0416
Mean = 0.0826
-18-
-------
when considered in terms of normal variations in business activities. The
validation analysis reported in the Phase I report showed that, when annual
averages of measured and SCIM calculated concentrations were compared for
different hours of the day, the SCIM calculations overestimate the measured
value by the greatest amount at 7 A.M. and underestimate by the greatest
amount at 1 P.M. and 4 P.M. These correspond to maximum and minimum values
of A(t), respectively. These results suggest that a uniform value of A(t) = 1
may be more appropriate than the values in Table 4.1-1. It is therefore recom-
mended that SCIM be run with A(t) = 1, rather than the values in Table 4.1-1.
The sensitivity factors shown in Table 4.1-1 have a logical basis
when considered in terms of people's diurnal activities. The sensitivity
factors are highest in the early morning with a peak value at 7 A.M. This
is when a major portion of the population arises. Residential fuel con-
sumption and probably certain commercial and industrial fuel consumption
is increased greatly relative to other hours of the day. Therefore, the
sensitivity of fuel consumption relative to the temperature deficit from
the threshold space heating temperature is likely to be very great. There
is a secondary maximum in the sensitivity factor in the early evening hours.
This corresponds with the time that residential fuel consumption is likely to
be adjusted to temperature changes (i.e., the end of the working day when
people return to apartments and homes). The low sensitivity factors in the
early morning hours after midnight are times of minimum residential fuel
consumption and low heat replacement in commercial and institutional sources
due to opening doors. Thus, there is a qualitative basis for using the
sensitivity factors shown in Table 4.1-1. Although the sensitivity factors
-19-
-------
and temperature thresholds in Table 4.1-1 were developed for New York City,
they are probably qualitatively applicable to other cities. If other
information is not available, they are a reasonable approximation to what
can be expected in other large cities.
The fractions of emissions which are temperature sensitive is
likely to be variable from one city to another, depending primarily on
how cold the climate is. One gross assumption which could be made is
that the fraction of SCL emissions which are temperature sensitive is
directly proportional to the climatological mean degree days which occur
at a given location. However, it is recommended that estimates of the
fuel use and climatological mean degree days be obtained for several dif-
ferent climates before attempting to define such a relationship. One other
source of data is available from a study by Argonne National Laboratory
.(Roberts, et. al 1970). From this report it is estimated that 72 percent
of S0? emissions in Chicago are temperature sensitive. By way of compari-
son, it is noted that the annual mean degree days are approximately 5000°F
days and 6200°F days for New York City and Chicago, respectively. The
two available estimates of 29 percent for 5000°F days and 72 percent for
6200°F days are not very consistent. ' Of course, other factors, such as the
relative mix of industrial, commercial and residential fuel users in the
area sources, affect the relationship. More data on the amount of fuel use
which is related to temperature considerations is needed. For the time
being, one might reasonably assume that, for large cities with normal total
heating degree days (with a base of 65°F) of 5000 to 6000 degree days, S02
emissions from area sources are 50 percent temperature sensitive and use
the temperature thresholds and sensitivity factors in Table 4.1-1.
-20-
-------
4.2 USE OF EMSU DATA AS METEOROLOGICAL INPUT ;
During the late 1960's and early 1970's, Environmental Meteor-
ological Support Units (EMSU) were established by the National Oceanic
and Atmospheric Administration in roughly 20 U.S. cities. The purpose
of these units was to take observations, prepare forecasts, and provide
advice on the present and future meteorological conditions which 'affect
air pollution levels. An analysis is made in this study of how data
reported by EMSU's could be used to determine meteorological parameters
for SCIM and how the selected values compared with values determined from
conventional airport weather observations. Three meteorological param-
eters analyzed were mixing height, atmospheric stability and wind speed
and direction.
The EMSU data consist of radiosonde observations of temperature,
relative humidity and wind direction and speed from a slow rise balloon
(i.e., about 65 meters per minute). The soundings are taken from urban
areas and generally provide useful estimates of the temperature, moisture
and wind profiles over large cities. The soundings are generally taken
at times of expected minimum (near sunrise) and expected maximum (early
afternoon) dispersion conditions. Additional soundings may also be
available for intermediate hours.
4.2.1 Mixing Height
Mixing heights were calculated for EMSU (slow rise) radiosonde
data and for standard radiosonde data using the method described in
Appendix A-l. The data used included all days in August and December
-21-
-------
of 1969 for which both EMSU and standard RAOB data were available for
New York City and St. Louis.
New York City represents a site at which both standard and
EMSU data are available for the same city. The EMSU data are obtained
from releases at Laguardia Airport which is located well within the
New York Metropolitan area. The standard data are obtained from releases
at Kennedy Airport which is located on the edge of the metropolitan area.
Mixing heights corresponding to EMSU observation times are determined from
the standard data using an interpolation scheme described in Appendix A-2.
The 12Z mixing height is taken to be representative of 0600 local time
and the OOZ mixing height is taken to be representative of 1400 local time.
Linear interpolation with time is used between 0600 and 1400. The com-
puted mixing heights, interpolated values and differences between mixing
heights calculated using EMSU and standard RAOB data are presented in
Table 4.2-1.
St. Louis represents a site at which standard RAOB data are not
available, but for which EMSU data are available. The mixing height for
St. Louis was estimated using an average of heights calculated for Peoria,
Illinois, and Columbia, Missouri. The computed mixing heights, calculated
averages, interpolated values and EMSU less standard RAOB differences are
presented in Table 4.2-2.
The question of whether the EMSU data provides information about
mixing heights which is significantly different from that available from
standard radiosondes may be examined using the data in Tables 4.2-1 and
4.2-2. For each location and each month, the difference between mixing
-22-
-------
Table 4.2-1. New York City Mlxjng Height Estimates from Standard RAOB and EMSU Data
Date (1969)
August 1
August 4
August 5
August 6
August 7
August 8
August 11
August 12
August 13
August 14
August 15
August 18
August 19
August 20
Hour
0600
0700
1100
1900
0700
1100
1900
0600
0700
1100
1900
0600
0700
1100
1900
0600
0700
1100
1900
0500
0700
1100
1900
0600
0700
1100
1900
0600
0700
JOOO
1900
0600
0700
1100
1900
0600
0700
1100
1900
0600
0700
1100
1900
0600
0700
1100
1900
0600
0700
1100
1900
0600
0700
1100
1900
Kennedy Airport RAOB
(100)*
107
(2446)
4200
107
(2446)
4200
(4200)
4200
(4200)
4200
(4200)
4200
(1861)
107
(118)
156
(307)
420
(628)
567
(325)
143
(238')
261
(352)
420
(219)
219
(216)
213
(119)
138
(214)
270
(100)
100
(255)
372
(4200)
4200
(4200)
4200
(100)
107
(203) '
275
(1303)
1172
(647)
254
•(100)
423
(2581)
4200
Laguardia Airport EMSU
100
763
100
125
100
187
244
2248
100
791
100
138
100
100
100
881
100
2770
100
172
126
106
100
156
127
930
100
1027
EMSU Minus RAOB
0
-1683
-7
-2321
-4100
-4013
-3956
387
-18
484
-528
-187
-138
-252
-119
665
-19
2556
0
-83
-4074
-4094
0
-47
-1176
282
0
-1554
* Values in parentheses are interpolated (see text).
(Continued)
-23-
-------
Table 4.2-1. New York City Mixing Height Estimates from Standard RAOB and EMSU Data (Continued)
Date (1969)
August 21
August 22
August 25
August 26
August 27
August 28
August 29
December 1
December 2
December 3
December 4
December 5
December 8
December 9
December 10
Hour
0600
0700
1200
1900
0600
0700
1100
1900
0600
0700
1100
1900
0600 '
0700
1100
1900
0600
0700
1100
1900
0600
0700
1100
1900
0600
0700
1100
1900
0700
1200
1900
0700
1200
1900
0700
1200
1900
0700
1200
1900
0700
1200
1900
0700
0800
1200
1900
0700
1200
1900
0700
1200
1900
Kennedy Airport RAOB
(292)*
452
(1253)
1573
(100)
168
(93S)
1510
(105)
167
(417)
605
(4200)
4200
(4200)
4200
(642)
643
(649)
653
(162)
180
(255)
310
(155)
149
(125)
107
382
(1666)
2179
797
(1027>
1119
277
(2522)
3420
1057
(1476)
1644
599
743
800
481
(449)
(320)
255
519
(575)
597
285
(153)
100
Laguardia Airport EMSU
100
1281
100
1486
100
671
263
735
190
1194
115
846
192
782
589
561
646
403
1257
1060
1381
913
1274
428
453
544
673
351
352
EMSU Minus RAOB
-192
28
0
551
-5
254
-3937
-3465
«
-451
545
-47
592
37
657
-1077
-236
-381
126
-1265
3
-95
314
530
•
-21
134
26
99
66
199
* Values in parentheses are interpolated (sec text).
(Continued)
-24-
-------
Table 4.2-1. New York City Mixing Height Estimates from Standard RAOR and EMSU Data (Concluded)
Date (1969)
December 11
December 12
December 15
December 16
December 17
December 48
December 19
December 23
December 24
December 29
December 30
Hour
0700
0700
1200
1900
0700
1200
1900
0700
1200
1900
0700
1200
1900
0700
1200
1900
0700
1200
1900
0700
1200
1900
0700
1200
1900
0700
1300
1900
0700
1400
1900
Kennedy Airport RAOB
400
117
(557)*
733
1172
(546)
295
1154
(1135)
1128
878
(1019)
1076
332
(1078)
1377
3342
(26S3J
2377
448
(547)
587
440
(332)
289
460
(5061
514
171
451
451
Laguardia Airport EMS'!
436
221
3638
1103
1240
100
1327
618
1104
219
370
2382
2644
422
743
560
377
531
147
224
327
EMSU Minus RAOB
36
105
3081
-68
693
-1054
192
-260
85
-113 .
-709
-960
-9
-26
196
120
45
71
-359
53
-124
* Values in parentheses are interpolated (see text).
-25-
-------
Table 4.2-2. St. Louis Mixing Height Estimates from Standard RAOB and ESMU Data
Date (1969)
August 11
August 12
August 13
August 14
August 18
August 19
August 20
August 21
August 22
August 25
August 26
August 27
August 28
August 29
Hour
0600
1300
1800
0600
1000
1400
1800
0600
1300
1800
0600
1300
1800
0600
1100
1300
1800
0600
1300
1800
0600
1300
1800
0600
1300
1800
0600
1200
1800
0600
1000
1300
1400
1800
0600
1000
1300
1800
0600
1000
1400
1800
0600
1000
1400
1800
0600
1300
1800
RAOB
Peoria
Sounding
227
3269
236
3473
231
725
260
3464
363
3338
255
3163
401
3945
443
1333
250
1336
213
2223
217
2214
228
426
225
3412
233
812
Columbia
Sounding
259
3314
310
2033
294
800
298
3071
316
3093
3230
309
25?5
304
964
777
4056
284
1317
276
3813
280
3574
310
4014
291
3704
284
1565
Average
243
(2910)*
3292
273
(1513)
(2753)
2753
263
(700)
763
279
(2894)
3268
340
3093
(3270)
3284
282
(2523)
2844
352
(2191)
2454
610
(2434)
2694
267
(1062)
1326
245
(16il)
(2671)
(3018)
3018
249
(1571)
(2563)
2894
269
(.1244)
(2220)
2220
258
(1908)
(3558)
3558
258
(1072)
1189
EMSU
Sounding
170
2780
168
1301
162
211
567
192
2626
322
3021
336
2467
345
2345
2820
1774
165
154
155
1736
1865
251
157
1926
2005
178
2018
2002
170
3050
2013
173
3080
Minus
Average RAOB
-72
-130
-105
-212
-2591
-51
-133
-87
-268
-18
-249
54
-57
-7
153
2210
-660
-101
-908
-90
105
-806
-2767
-92
355
-559
-91
773
-218
-87
1142
-1545
-89
2008
* Values in parentheses are interpolated (see text).
(Continued)
-26-
-------
Table 4.2-2. St. Louis Mixing Height Estimates from Standard RAOB and EMSU Data (Concluded)
Date (1969)
December 4
' December 5
December 8
December 9
December 11
December 12
December 15
December 16
December 17
December 18
December 19
December 22
December 23
December 24
How
0600
1300
, 1800
0600
1300
1800
0600
1300
1800
0600
1300
1800
0600
1200
1800
0600
1300
1800
0600
1300
1800
OfiOO
1200
1800
0600
1300
1800
0600
1300
1800
0600
1300
1800
0600
1200
1800
0600
1200
1800
0600
1200
1800
RAOB
Peoria
Sounding
299
284
302
557
492
246
301
1414
733
1214
559
805
344
566
231
396
841
712
580
920
483
1238
529
639
902
1552
294
612
Columbia
Sounding
285
1014
544
809
307
341
375
355
821
1196
902
412
988
1070
379
663
799
472
826
444
293
1062
998
1044
1398
1297
912
603
- Average
292
(604)*
649
423
(651)
683
400
(307)
293
338
(816)
885
777
(1098)
1205
'730
(624)
608
666
(799).
818
305
(474)
530
820
(621)
592
703
(648)
682
388
(105S)
1150
764
(822)
841
1150
(1356)
1425
603
(607)
608
St. Louis
EMSU
Sounding
216
928
455
897
942
1313
257
1074
1737
717
1098
1003
921
1109
426
802
318
786
1085
914
345
1794
1175
965
1207
1851
305
753
St. Louis
Minus
Average RAOB
-75
324
32
246
543
1006
-81
258
960
-381
368
380
255
310
121
327
-502
164
381
266
-43
740
412
143
57
495
-297
146
* Values in parentheses are interpolated (see text).
-27-
-------
heights estimated from EMSU and standard RAOB data is summarized in
Table 4.2-3 for sunrise and for mid-day observation times. Overall
these comparisons show that the mean differences (-205m) is about
20 percent of mean mixing height (1044m). However, there is a large
amount of variability for individual comparisons as demonstrated by
the large root mean square difference of 1146m. These results sug-
gest that for the overall climatological average, the EMSU information
may not be important. However, for day to day variations, there is
important information available in the EMSU data. Mixing height is
most important in determining dispersion conditions during the day.
It is less important near sunrise when stable or neutral stability
conditions prevail. An examination of the data in Tables 4.2-1 and
4.2-2 shows that in 46 percent (18 comparisons) for New York City
and 47 percent (16 comparisons) for St. Louis, the EMSU mixing height
estimate differs from that derived with standard RAOB data by over
50 percent of the standard RAOB estimate. As a result, it is con-
cluded that, when EMSU data is available, it should be used in place
of or to supplement the standard data.
The following procedures are suggested for using the EMSU
data. Use the interpolation scheme described in Appendix A-l as a model
of diurnal variation in mixing height. The following steps may be
followed.
1. If a sunrise EMSU sounding is available, use the mixing
height from it for the period from Midnight to 6 a.m.
If not, use an estiamte from standard RAOB data.
-28-
-------
w
I,
•F-(
o
X
M
a
s
M
o
2
U
rt
"§
rt
•M
oo
T3
00
S
«
a
a
0)
00
U
o
c
0 H ^
S 5:
a) u
-g
U O
M X
^ u
O O
JX {M
«: ^
.
-g
H
0
H
-29-
-------
2. If one or more mid-day EMSU soundings are available,
obtain mixing heights from each. Use linear inter-
polation with time to estimate mixing heights for
hours between EMSU soundings.
3. If more than one EMSU sounding is available these may
be linearly extrapolated with time to estimate mixing
heights over the period from 6 a.m. to 2 p.m.
4. If only one EMSU sounding is available, compute the
mixing height for other hours in the period from
6 a.m. to 2 p.m. by substituting the hourly surface
temperature for the surface temperature in the sounding
and computing the mixing for the revised sounding.
4.2.2 Stability Class
Since EMSU sounding data is obtained from a slow rise bal-
loon, it should be useful in characterizing the temperature and wind
profiles of the lowest layers of the atmosphere, which determine dis-
persion conditions. Several ways of characterizing the stability of
the atmosphere using EMSU data are compared with the Pasquill stability
classes determined by a method suggested by Turner using surface wea-
ther observations of cloud cover and wind speed. The extent to which
the EMSU data suggest stability classifications different from the
Turner-Pasquill categories is discussed. In conclusion a method of
integrating the two types of data is'proposed.
Bulk Richardson number, calculated over three different
heights, was used to characterize stability. The three heights were
from the lowest height in the EMSU sounding with both wind and temper-
ature data to (1) the top of the mixing layer, (2) 140 meters, as used
by McElroy (1969) to classify measurements of a and a , and (3) the
next lowest height with wind and temperature data. The'wind speed data
-30-
-------
within the mixing layer were fitted to a power law profile by the method
of least squares. This too was used to characterize the atmospheric
stability. These estimates were obtained from St. Louis EMSU soundings
for August and December 1969. The three bulk Richardson numbers, the
wind speed profile power law and the mean mixing layer wind speed and
direction, for each EMSU sounding with reasonably complete data, are
listed in Table 4.2-4. For comparison the Pasquill stability class and
surface wind determined from the closest (in time) three-hour surface
weather observation are also listed.
In order to compare the sounding stability characteristics with
the Pasquill stability classes, the correspondence shown in Table 4.2-5
was assumed. The correspondence in Table 4.2-5 is hypothetical and was
selected to be reasonably consistent both with the data shown in
Table 4.2.4 and with information reported by other investigators. Using
these correspondences, the best agreement between EMSU stability data and
the Pasquill stability classes (listed in Table 4.2-4) is obtained using
the bulk Richardson number over 140 meters. This gives 19 hours out of
38 in agreement. The next best was the bulk Richardson number over the
lowest 2 heights, which gives 18 hours of agreement out of 44 comparisons.
Since 50 percent of the compared hours differ between EMSU and
surface data stability classifications, there is probably a significant
amount of additional information available in the EMSU data. However,
it is difficult to see how to use the EMSU data, except to modify the
single hour for which EMSU stability classifications are obtained. The
stability changes so rapidly from hour to hour during ti.e periods over
-31-
-------
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
Table 4.2-4. Stability and Wind Characteristics for St. Louis
Date
(1969)
Aug. 25
Aug. 25
Aug. 25
Aug. 26
Aug. 26
Aug. 26
Aug. 27
Aug. 27
Aug. 27
Aug. 28
Aug. 28
Aug. 28
Aug. 29
Aug. 29
Dec. 4
Dec. 4
Dec. '-5
Dec. 5
Dec. 8
Dec. 8
Dec. 9
Dec. 9
Dec. 10
Dec. 11
Dec. 12
Dec. 12
Dec. IS
Dec. 15
Dec. 16
Dec. 16
Dec. 17
Dec. 17
Dec. 18
Dec. IS
Dec. 19
Dec. 19
Dec. 22
Dec. 22
Dec. 24
Dec. 24
Dec. 30
Dec. 30
Dec. 31
Dec. 31
Hour
1000
1200
2000
0500
0900
1200
0500
1000
1300
0500
1000
1400
0500
1300
0500
1200
0500
1200
0500
1200
0500
1200
1200
1200
0500
1200
0500
1200
0500
1200
0500
1200
0500
120O
0500
1200
0500
1200
0500
1200
0500
1200
OSOO
1200
Bulk Richardson Number
Over
Mixing
Layer
1.690
-0. 550
0.036
0.146
0.136
-0,632
0.851
4.620
1.125
0.119
7.968
0.200
0.090
0.389
0.027
0.016
0.031
0.051
0.271
0.051
0.026
0.294
0.238
-0.044
0.138
-0.012
0.168
0.039
0.322
-0. 007
0.030
-0. 795
0.020
-0. 035
-0.022
0.094
-0.53S
0.065
0.226
0.069
0.308
-0.544
-1.904
-5.829
Over 140 m
Layer
0.072
-
0.044
0.101
0.131
-0.038
1.028
-0.134
-0.019
0.109
-0.007
0.003
0.097
0.029
0.033
-0. 070
0.009
0.005
0.038
0.010
0.030
0.016
-0.004
-0.007
0.010
-0. 060
0.004
-
0.177
-0. 180
0.034
-0.057
-0. 264
-0. 220
0.091
-
0.330
0.670
-
-
0.233
0.112
0.226
-
Lowest Two
Sounding
Heights
0.009
0.015
0.003
0.149
-O.069
-0.079
0.307
-0. 156
-0. 103
0.073
-0.016
-0.040
O.OS8
-0.043
0.017
-0.060
-0.002
-0.004
0.021
-0.007
0.026
0.007
-0.083
-0.001
0.007
-0. 077
-0.002
-0.038
0.094
-0.138
0.030
-0. 165
0.002
-0.008
-0.012
-0.063
0.170
0.330
0.167
-0.066
0.184
0.035
-0.005
-0.315
Pasquill
Stability
Class
B
C
E
E
C
A
E
B
A
D
D
B
E
D
E
C
D
D
D
C
D
D
D
'D
D
D
E
D
E
C
E
D
D
D
E
D
D
D
D
D
D
D
D
D
Wind Speed
Profile Power
Law
-0.06
0.06
0.28
0.22
0.15
-0.01
0.08
-0.06
0.14
0.34
-0.04
0.06
0.28 '
0.03
0.05
0.04
0.15
0.08
0.17
0.14
0.33
0.17
0.32
0.05
0.08
0.16
O.06
0.07
0.19
0.02
0.38
0.01
0.33
0.08
0.17
0.28
0.10
0.15
0.82
0.26
0.04
0.02
0.08
0.18
Win 1 Lvirection/
Speed (m/sec)
Mixing
Layer
Mean
47/4
65/5
59/6
48/5
70/6
62/5
133/2
169/3
162/7
203/6
228/8
179/9
223/8
210/6
32/12
42/7
157/14
155/11
269/13
254/13
190/11
200/16
69/10
298/14
272/5
301/7
13/19
355/10
85/4
180/8
355/9
175/6
253/15
324/13
318/11
313/21
356/10
158/8
90/3
184/11
7/14
352/13
315/10
271/4
Surface
10/3
10/4
30/3
calm
350/3
70/2
calm
160/5
190/2
calm
160/5
180/2
150/2
200/4
30/3
40/4
130/6
130/6
240/4
230/5
160/4
170/7
30/3
290/7
260/4
310/4
10/3
350/6
60/1
160/3
260/3
130/3
250/5
280/6
230/3
300/4
300/S
100/4
100/4
140/5
320/5
320/S
260/5
280/4
-32-
-------
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
1
I
I
I
which EMSU data is available that the surface data is much better that
an extrapolation of EMSU data. The EMSU data has limited use. The data
would be more useful if soundings were available every three hours.
Table 4.2-5. Proposed Correspondence Between Three Types of Stability Classifications
Time of Day
Day
Day
Day
Day
Night
Night
Pasquill Class
A
B
C
D
D
E
Bulk Richardson Number
< -0. 05
-0.05 to -0.031
-0.03 to -0.011
>-0.01
< 0.01
> 0.01
Wind Speed Profile Power Law
<0.05
0.05 toO. 12
0. 13 to 0. 17
>0. 17
<0.22
>0. 22
Another uncertainty with stability data related to vertical
temperature and wind profiles is that the relation of this data to the
commonly used Pasquill dispersion parameters is not well established.
In the light of the above considerations, it is recommended that EMSU
data not be used to determine stability characteristics.
4.2.3 Hind Direction and Speed
The data given in Table 4.2-4 shows comparisons between the
surface wind speed and direction and the vector mean wind speed and
direction for the mixing layer. It is clear from these comparisons that
the EMSU data provides significant additional information on the wind
profile. Of particular interest is the frequent occurrence of a notice-
able turning of the wind with height. This can have a significant
effect on model calculations. The need for a detailed study of how to
-33-
-------
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
B
I
I
I
use wind profile data in estimating the wind direction and speed used
in model calculations is clearly indicated by this data. No attempt
has been made to develop general techniques from the limited data pre-
sented in Table 4.2-4. However, the data does suggest some possible
ways of using EMSU data to improve the wind direction and speed estimates
used in modeling. During some periods the vertical wind shear remains
nearly fixed from one EMSU sounding to the next. This suggests that an
average shear could be derived and applied to all surface wind observations
between the EMSU observation times. Another possibility is to develop
diurnal patterns of wind shear from other data sources (e.g., St. Louis
micromet tower data of 1964) and use the EMSU data to identify and scale
the patterns. The turning of the wind with height should be taken into
account in dispersion models.
For the present it is suggested that the following tentative
procedure be used to acco""it for turning of the wind.
1. Determine the mean wind direction for the mixing layer
for each EMSU sounding. If the sources being modeled
are mostly elevated sources, use this as the wind dir-
ection; if the sources are mostly ground sources, use
the surface wind direction.
2. If the succeeding EMSU sounding is less than 24-hours
away and the mixing layer mean wind direction has changed
by less than 90°, estimate the wind direction for inter-
vening hours by linear interpolation. Use these wind
directions in place of the surface wind direction if
mostly elevated sources are being modeled. If the wind
direction has shifted by 90° or more or ground sources
are being modeled use the reported surface wind directions
for intervening hours. If the time period between sound-
ings exceeds 24 hours, use the reported surface wind
directions for intervening hours.
-34-
-------
4.3 PROCEDURE FOR CALCULATING ANNUAL SHORT-TERM MAXIMUM CONCENTRATIONS
4.3.1 Introduction
The operational implementation of the Gaussian plume model of
this study for purposes of evaluating proposed air quality strategies
could impose a severe computational burden. The model might be used to
calculate all hourly concentrations within a long-term period, e.g., one
year, for which both emission and meteorological data are available. This
is repeated for each of a number of stations in each of many control
regions for each of several proposed air quality control strategies. A
statistical sampling procedure was devised to reduce the amount of computa-
tions and was tested on 10 stations in New York City, the procedure con-
sists of reducing the number of hourly concentrations calculated; the
number of stations, air quality regions, and control strategies do not
change. The reduction is achieved by selecting a sample of the variable
hours in a systematic manner and using the sample to calculate the param-
eters required for evaluating air quality strategies (e.g., mean annual
concentration, daily value exceeded only once a year, etc.).
The test on the 10 stations consisted of choosing samples of
various sizes and determining the loss in accuracy is given by the dif-
ference between a parameter value calculated from all available hours and
the value of this same parameter calculated from a sample. Six air quality
control parameters were chosen for analyses, and differences were obtained
for each of 29 sample sizes. Tables and graphs of a function of these
differences, are presented in a fashion to provide information on the
tradeoff between reduction of computations and loss in accuracy. These
-35-
-------
I
1
furnish guidance for choosing a sample size, and thereby reducing compu-
| tations, in any operational implementation of the model for evaluating
. air quality strategies.
The test procedure, and the results of the test are described
• below including: method of calculating the six air quality control param-
eters, the sampling scheme, and development of the function of the dif-
| ferences which serves as an overall measure of accuracy.
I
4.3.2 Test Procedure
4.3.2.1 Air Quality Parameters
In current EPA practice, it is generally assumed that air pollu-
tion concentration values follow the log-normal distribution. This assump-
tion was adopted in this study. Although other distributions have been
advanced, and may in fact eventually replace the log-normal assumption, it
is our opinion that results obtained here would not be changed substantially.
For any set of hourly concentrations (e.g., all 2920 third hours in a year
or a sample thereof), the log-normal distribution was fitted by calculating
the mean, 7, and the standard deviation, Sy, of the logarithms of the
concentrations. Three air quality values were then derived as follows:
Mean = exp (7 + 0.5 sy2)
Value exceeded once in 1000 hours = exp (Y + 3.091 SY)
Value exceeded once in 2920 hours = exp (Y + 3.396 sv)
-36-
-------
Three additional values were calculated in a similar manner
from daily mean concentrations, i.e., the arithmetic mean of the eight
hourly concentrations for the day. The log-normal distribution was fitted
to these means; the parameters are:
Mean = exp (I + 0.5 sz2)
Daily value exceeded once in a 100 day = exp (I + 2.33 sz)
Daily value exceeded once in a year = exp (I + 2.776 s^)
Where Z and s7 are, respectively, the mean and standard deviation of the
logarithms of the daily values.
4.3.2.2 Sampling Procedure
The sampling procedure for the daily values (average of 8 three-
hourly values) is presented first because it is simpler. Two terms require
definition: a sampling interval is the number of values from one selected
value to the next (e.g., a sampling interval of two means that every other
day is included in the sample); the initial time indicates the starting
day of the sample. Twenty-nine sampling intervals were chosen, each value
from 2 through 30. For each interval, from 2 to 8 different samples were
selected by varying the initial time. Thus, for a sampling interval of
2, two samples were drawn: the" first consisted of days 1, 3, 5, ...,
365 and the second of days 2, 4, 6, ..., 364. The number of samples for
a sampling interval is given by the maximum of [sampling interval or eight].
Thus, for sampling intervals up to eight the number of samples is equal to
the sampling interval; for sampling intervals beyond eight exactly eight
-37-
-------
I
I
samples were drawn from the 365 days by varying the initial time from
• 1 through 8. The value of eight has no significance; it simply reduces
• the amount of computation.
When the sampling procedure is applied to the 3-hourly data
• some unequal sampling intervals may result, e.g., for interval two every
- other value within one day is selected, but from the last value of one
™ day to the first value of the next, the interval is either 1 or 3. How-
fl ever, it was deemed more important to ensure equal representation of each
of the eight hours of the day than to maintain a consistent sampling
• interval. Again, 29 sampling intervals were used but this time they do
^ not proceed by steps of one but range from 2 to 249. They are listed in
" Table 4.3-1. As before, for each interval from two through eight the
ft number of samples equals the sampling interval and beyond eight exactly
eight samples were selected.
I
4.3.2.3 Measure of Accuracy
• At each of the 10 stations, three air quality parameters were
calculated using all 2920 hourly concentrations and an additional three
using the 365 daily mean concentrations. For each sample selected from
f* the 2920 hourly values, three air quality parameters were calculated and
differences were taken between them and the corresponding parameter values
using all 2920 hours. The same procedure was followed for samples drawn
from the 365 daily concentrations. The differences were combined to obtain
a measure of accuracy for each sampling interval for each air quality
parameter.
-38-
-------
Table 4.3-1. Measures of Accuracy - Hourly Concentrations
Sampling
Interval
2
• 3
4
5
6
7
9
11
13
IS
17
19
21
23
25
27
29
31
39
47
55
63
71
79
95
119
159
199
249
Average No.
of Cases
in Sample
1460
973
730
584
487
417
325
266
225
195
172
154
139
127
117
108
101
95
75
63
54
47
42
37
31
25
19
15
12
Proportion
of Cases
in Sam pie
0.500
0.333
0.250
0.200
0.167
0.143
0.111
0.091
0.077
0.067
0.059
0.053
0.048
0.043
0.040
0.037
0.034
0.032
0.026
0.021
0.018
0.016
0.014
0.013
0.011
0.008
0.006
0.005
0.004
Measures of Accuracy for
Mean
0.011
0.015
0.032
0.045
0.033
0.056
0.050
0.057
0.056
0.125
0.101
0.077
0.113
0.109
0.080
0.115
0.122
0.106
0.127
0.189
0.232
0.157
0.173
0.260
0.201
0.336
0.234
0.336
0.403
1/1000
0.020
0.024
0.082
0.102
0.077
0.166
0.097
0.144
0.149
0.323
0.287
0.218
0.257
0.380
0.216
0.261
0.258
0.257
0.312
0.406
0.812
0.408
0.535
0.491
0.476
0.810
0.615
0.976
0.842
1/2920
0.023
0.026
0.097
0.113
0.085
0.195
0.109
0.165
0.175
0.376
0.336
0.254
0.290
0.457
0.246
0.291
0.290
0.295
0.362
0.457
0.997
0.481
0.615
0.530
0.524
0.910
0.693
1.105
0.982
-39-
-------
I
I
I
i
i
i
i
i
i
t
I
i
i
i
i
i
I
i
Let A., denote the measure of accuracy for air quality parameter
j (j = 1, 2, ..., 6) for sampling interval k (k = 2, 3, ..., 29). Then,
10
_ _
10 • N, -> -4
i i
m=l n=l
x
jlmn
10
Where
X.,
Xilmn
N, =
the value of air quality parameter j
for sampling interval k
at station m
for sample n
same as above with sampling interval of one (i.e., using
all available data)
number of samples for sampling interval k.
The numerator in Equation 4-1 is the root-mean-square of the
differences between air quality parameters calculated from a sample and
the corresponding air quality parameters calculated using all available
data. The denominator is the mean over the 10 stations of the parameter
calculated by using all available data. The measure of accuracy, A., ,
JK
is similar to the coefficient of variation (standard deviation/mean)
except that the numerator is a root-mean-square rather than a standard
deviation.
-40-
-------
4.3.3 RESULTS
Table 4.3-1 gives values of A., for the hourly concentrations
j K
and Table 4.3-2 contains results for the daily concentrations. Both
tables indicate considerable savings in computational effort with reason-
ably small losses in accuracy. In Table 4.3-1, the loss in accuracy, as
defined by A.,,, is below 20 percent for all three air quality parameters
JK
for sampling intervals up to 13 (i.e., sample size only 0.056 as large as
all available 2920 hours). The loss in accuracy is less for the mean than
it is for the two exceedance values. This is consistent with statistical
theory which indicates greater accuracy in estimating the mean of a dis-
tribution than the tails. In Table 4.3-2, the loss in accuracy for the
daily concentrations is greater than for the hourly concentrations for
the same proportion of cases in the sample. But even here, the loss is
below 31 percent for sampling intervals up to 10 days. Again the loss is
less for the mean than for the two exceedance levels.
To facilitate use of the results, the measures of accuracy were
plotted against proportion of total cases in the sample, and lines were
fitted by least squares. Figure 4.3-1 contains the measures from the hourly
concentrations and Figure 4.3-2 the measures calculated from the daily con-
centrations. Only the first several values are plotted because, as can be
seen in Tables 4.3-1 and 4.3-2, the measures show large fluctuations for
small proportions of cases. In an operational problem, these graphs can
be entered with an hypothesized proportion of cases to estimate what loss
in accuracy would occur. It must be cautioned, however, that the graphs
are based on S02 at 10 stations in New York City. It is our subjective
-41-
-------
Table 4.3-2. Measures of Accuracy - Daily Concentrations
Sampling
Interval
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
Average No.
of Cases
in Sample
183
122
91
73
61
52
46
41
37
33
31
28
26
25
23
22
21
20
19
18
17
16
16
15
14
14
13
13
13
Proportions
of Cases
in Sample
0.501
0.334
0.249
0.200
0.167
0.142
0.126
0.112
0.101
0.090
0.085
0.077
0.071
0.068
0.063
0.060
0.058
0.055
0.052
0.049
0.047
0.044
0.044
0.041
0.038
0.038
0.036
0.036
0.036
Measures of Accuracy for
Mean
0.047
0.081
0.094
0.100
0.122
0.090
0.124
0.096
0.149
0.127
0.235
0.245
0.192
' 0.275
0.162
0.125
0.125
0.233
0.208
0.271
0.228
0.313
0.257
0.227
0.229
0.221
0.225
0.283
0.427
1/100
0.088
0.190
0.191
0.181
0.256
0. 202
0.234
0.218
0.270
0.202
0.472
0.515
0.407
0.514
0.283
0.250
0.279
0.488
0.371
0.602
0.366
0.704
0.328
0.516
0.379
0.360
0.376
0.585
0.916
1/365
0.100
0.228
0.230
0.205
0.305
0.244
0.276
0.262
0.309
0.230
0.576
0.629
0.488
0.616
0.333
0.291
0.328
0.596
0.453
0.748
0.411
0.903
0.363
0.673
0.432
0.405
0.422
0.754
1.173
-42-
-------
CQ
OJ
-------
X X
CM CO
cn oo
CO LO
CM
CO «=1-
a
o
a
o
o
g
u
0)
.
a
rt
o
•43
O
(N
I
ro
jo
-44-
-------
judgement that they are applicable to other pollutants and other sites,
but this remains to be proven.
-45-
-------
Section 5.0
REFERENCES
-------
Section 5.0
REFERENCES
Blade, E. and E.F. Ferrand. 1969. Sulfur Dioxide Air Pollution on New
York City: Statistical Analysis of Twelve Years. Journal of the
Air Pollution Control Association. Volume 19, Number 11, pp. 873-878.
Davidson, B. 1967. "A Summary of the New York Urban Air Pollution
Dynamics Research Program." Journal of Air Pollution Control
Association, 17, pp. 154-158.
Haltiner, G.J. and F.L. Martin. 1957. Dynamical and Physical Meteorology.
McGraw-Hill, New York.
Hanna, S.R. 1971. A Simple Method of Calculating Dispersion from Urba
Area Sources. Journal of the Air Pollution Control Association,
Volume 21, Number 12, pp. 774-777.
Larsen, R.I. 1971. A Mathematical Model for Relating Air Quality Measure-
ments to Air Quality Standards. Environmental Protection Agency,
Office of Air Programs, Research Triangle Park, N.C. 1971.
Ludwig, F.L., W.B. Johnson, et al. 1970. A Practical Multipurpose
Urban Diffusion Model for Carbon Monoxide. Contracts CAPA-3-68 and
CPA 22-69-64. Menlo Park, California: Standford Research Institute.
McCaldin, R.O. and R.S. Sholtes. 1970. Mixing Height Determinations by
Means of an Instrumented Aircraft. Contract No. CPA 22-69-76. Gaines-
ville, Florida: University of Florida.
McElroy, J.L. 1969. A Comparative Study of Urban and Rural Dispersion.
Journal Applied Meteorology 8, pp. 19-31.
Roberts, J.J. et al. 1970. Chicago Air Pollution Systems Analysis
Program: A Multiple-Source Urban Atmospheric Dispersion Model,
ANL/ES-CC-007, Argonne National Laboratory, Argonne, Illinois.
Saucier, W.J. 1955. Principles of Meteorological Analysis. The University
of Chicago Press, Chicago, Illinois.
-46-
-------
APPENDICES
Appendix A-l
METHOD OF ESTIMATING THE HEIGHT OF THE MIXING LAYER
Appendix A-2
MIXING HEIGHT INTERPOLATION
-------
Appendix A-l
METHOD OF ESTIMATING THE HEIGHT OF THE MIXING LAYER
Mixing heights may be estimated using either standard or EMSU
radiosonde data. The data may be obtained from the NOAA National Weather
Records Center in Asheville, North Carolina on magnetic tapes. The
method outlined here consists of determining the mixing height by a par-
cel displacement method. The temperature and mositure content of a repre-
sentative parcel are defined for ground level. The reported surface
temperature may be used, or a more representative temperature from a
nearby urban location, or another time may be selected. The moisture
content is defined by the maximum mixing ratio in the vertical profile.
The temperature change which will occur if the parcel is displaced upward
is traced until the parcel temperature is 1°C less than the sounding
temperature. The temperature change is assumed to be adiabatic between
the ground surface and the mixing condensation level, and pseudo-adiabatic
above the mixing condensation level. The following seven steps are used
to determine the mixing layer height for each observation time.
1. Read and store the height, pressure, temperature, and
relative humidity of each data level.
2. Convert all relative humidities to mixing ratios using
the following equations (Saucier 1955):
M = 0.01 U S
(A-l)
r _ 0.62197 f E _n goo E ,. .
P - f E ~0.62Z pr (A-2)
A-l
-------
where
E = 6.11 (10)
7.5 T
T + 237.3
(A-3)
M = mixing ratio
U = reported relative humidity (percent)
•
S = saturation mixing ratio
f w 1 = correction factor for departure from ideal gas laws
E = saturation vapor pressure of water (mb)
P = reported pressure (mb)
T = reported temperature (°C)«
3. Find the maximum mixing ratio for the observation time (M ).
A
4. Find the mixing condensation level by the following equa-
tions (Saucier 1955):
7 _ 1000 ,, n ,
Zc - "O1 (To ' V
(A-4)
where
Z = mixing condensation level (m)
v»
T = ground level temperature (°C)
D = ground level dewpoint (°C).
In order to account for evaporation of dew during the early morning, it
is assumed that the rrr'xed atmosphere will contain moisture equal to that
indicated by Mx. D0 is determined from Mx by means of "Equations A-2 and
A-3 using S = Mx and T = D0:
D
0
237.3
7.5 -
log10
Iog10
" M P
X 0
6.11 (0.622)
f MxPo 1
6.11 (0.622)_
(A-5)
A-2
-------
where
P = ground level pressure.
•
5. Using the reported data levels to define layers, find the
layer (Z-j_] to Zj) containing the top of the mixing layer. The top of
the mixing layer is identified by the parcel method. When the reported
vertical temperature profile exceeds the temperature of a parcel lifted
from the surface by 1°C, this is assumed to be the top of the mixing
layer (Zm). The layer containing the mixing layer height is identified
by testing if
Ti i Ti + ! (A-6)
where
T. = temperature of ith level (°C)
Tr = temperature of parcel lifted to ith level (°C)
The parcel temperature is calculated as follows:
y- = -0.0098, Z. < Z,.
I — C
T , L.$J:I (A-?)
-0.0098
i o .
L o. -t
cp Rv ^i
where
' = temperature lapse rate (°C/m, Haltiner and Martin 1957)
L = 2500 = latent heat of vaporization (joules/g)
A-3
-------
$il] ~ saturation mixing ratio of parcel lifted to (i-l)th level,
estimated from Equations A-2 and A-3 using P. , and T.',
Rd = 0.287 = gas constant for dry air (joule/g/°C)
c = -1.003 = specific heat of dry air at constant pressure
P (joule/g/°C)
R = 0.461 = gas constant for water vapor (joule/g/°C).
6. Estimate the height of the mixing layer by linear interpo-
lation as follows:
Zn, = Zi-l * (V\r) -'(Vl " Vl'1 " HS
-------
Appendix A-2 j
MIXING HEIGHT INTERPOLATION
The vertical mixing ceiling is defined as that height above
ground level at which there is a marked reduction in vertical diffusion.
Such barriers are observed as a sharp drop in the concentration observed
i
in a vertical sounding (e.g., Davidson, 1967). It may be observed as a
delineation between the smoke-filled layer and cleaner air aloft over
many cities in the early morning. Much higher ceilings typical of
afternoon hours are clearly visible to air travelers climbing to or
descending from cruising altitudes. The ceiling may vary from 100 meters
at night to over 1500 meters during the day. Hourly estimates of the
ceiling are required for use in the model.
Unfortunately, this mixing ceiling is not always visibly dis-
cernible and no routine systems for taking vertical soundings of pollu-
tant materials are in operation. Therefore, the ceiling is generally
inferred from temperature soundings which are routinely observed twice
daily at certain airports by the National Oceanic and Atmospheric Admin-
istration (NOAA). These observing locations are separated by about
200 km on the average and are usually located outside the urban area.
The mixing layer is generally characterized by a near adiabatic lapse
rate extending from the ground to some latitude at which a deep, (several
kilometers) more stable lapse rate exists. However, the vertical temper-
ature structure of the atmosphere is frequently not this well defined.
As a result, considerable judgement may be required to define where, in
a vertical temperature profile, an effective mixing ceiling exists.
A-5
-------
Unfortunately very little data have been collected on the relationship
between vertical pollution and temperature profiles which could be used
to develop and substantiate rules for defining the mixing ceiling over
an urban region.
The procedure which is generally used to define the mixing
ceiling is the following: Determine the general rural vertical temper-
ature profile from the nearest appropriate (same air mass) radiosonde,
or by interpolation of two or more nearby radiosondes. Estimate mini-
mum morning and maximum afternoon air temperatures which are representa-
tive of the urban area. The afternoon temperature may be obtained directly
from airport observations or other available data. In most cases the
morning urban temperature will exceed the rural temperature. The follow-
ing equation (Ludwig, et al . , 1970) may be used to estimate morning
urban temperatures (T ) from rural temperatures (T ) using the urban
population $ and the radiosonde temperature lapse rate (dT/dp) as param-
eters:
Tu = Tr + 4>°'25(0.0633 - 0.298 ^1) (A-8)
Construct adiabatic temperature profiles from the urban temperatures
which intersect the rural temperature profile. The height of these inter-
sections are assumed to be the minimum and maximum mixing ceilings.
A-6
-------
A method of interpolating between these values to give hourly estimates
is to: :
\
1. Use the morning minimum from midnight to 6 a.m.
2. Linearly interpolate between the minimum and the maximum
between 6 a.m. and 2 p.m. \
i
3. U.se the afternoon maximum between 2 p.m. and midnight.
This pattern of diurnal variations is illustrated in Figure A.2-1.
f
•1, ooz*
*1-1
wJ Sounding
I
to Ceiling |
4
9999999®
-------
Limited simultaneous observations of temperature and S0? or
particle concentration profiles reported by Davidson (1967), Roberts,
et. al (1970) and McCaldin and Sholtes (1970) attest to the general
validity of this approach.
A-8
-------
TECHNICAL REPORT DATA
(Please read Instructions on tlie reverse before completing)
1. REPORT NO.
EPA-650/4-75-018-b
2.
3. RECIPIENT'S ACCESSION-NO.
4. TITLE AND SUBTITLE
Evaluation of the Multiple Source Gaussian Plume
Diffusion Model - Phase II
5. REPORT DATE
April 1975
6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
Robert C. Koch
Scott D. Thayer
8. PERFORMING ORGANIZATION REPORT NO.
GEOMET Report No. EF-467
9. PERFORMING ORGANIZATION NAME AND ADDRESS
GEOMET, Incorporated
15 Firstfield Road
Gaithersburg, Maryland
10. PROGRAM ELEMENT NO.
1AA009
11. CONTRACT/GRANT NO.
20760
68-02-0281
12. SPONSORING AGENCY NAME AND ADDRESS
Meteorology Laboratory
National Environmental Research Center
Research Triangle Park, North Carolina 27711
13. TYPE OF REPORT AND PERIOD COVERED
Final
14. SPONSORING AGENCY CODE
15. SUPPLEMENTARY NOTES
Phase I report was published April 1973 as GEOMET Report Number EF-186,
16. ABSTRACT
This report summarizes work done to compare a computer model for estimating air
pollution concentrations from multiple sources with measured S02 and particulate
concentrations and with other model calculations. The model is capable of estimating
short-term and long-term concentrations, and produces results which are equivalent in
validity to results produced with other models. Since the model represents hourly
variations in both emissions and meteorological conditions, this report considers
available sources of data and how these can best be used to estimate parameters for
the model. Use of temperature and industrial and commercial activity indexes to
estimate seasonal and diurnal variations in emissions is discussed. Use of slow-
rise balloon soundings taken in urban areas is discussed as a possible supplement
to conventional weather data. Finally, the applicability of using sampled calcula-
tions when estimating short-term maximum concentrations is evaluated.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.IDENTIFIERS/OPEN ENDED TERMS
c. COS AT I Field/Group
Air Pollution
Urban Areas
Atmospheric Diffusion
Diurnal Variation
Emission
Sequential Sampling
Air Quality Model
1302
0401
3. DISTRIBUTION STATEMENT
Release Unlimited
19. SECURITY CLASS (This Report)
UNCLASSIFIED
21. NO. OF PAGES
64
20. SECURITY CLASS (Thispage)
UNCLASSIFIED
22. PRICE
EPA Form 2220-1 (9-73)
------- |