-------
is highly significant. Clearly, the method by which hourly emissions are
estimated has an important impact on the mean estimated for a 1-month
or 3-month period. It is interesting that at the eight Chicago stations,
the mean Q estimates are all less than the variable Q estimates. At the
St. Louis stations, the three largest differences are the result of high
mean Q estimates. Since these differences are all attributable to the
emission algorithm, it is clear that the algorithm can act either to
increase or decrease the mean calculated value over a long-term period.
The non-linear correlation between changes in emission rates and changes
in the rate of diffusion through the atmosphere is indicated by these
results.
Since the means of the two sets of calculations are significantly
different, it is not surprising that the distributions about these means
are different. The standard deviation of the calculations for each
station is shown in Table 2. Except for one St. Louis station, the cor-
responding standard deviations for calculations using the two types of
emission rates are significantly different at the one percent level based
on the F-test statistic. The distributions were also compared by means of
the nonparametric Kolmogorov-Smirnov statistic. All the distributions
are significantly different at the five percent level using this statistic,
and all but two are significantly different at the one percent level. The
results of the Kolmogorov-Smirnov test are the more conclusive because it
is likely that the distributions are not normal as is assumed in the F-test.
-36-
-------
«* a
^6
rO
rt
£""
0
•43
CO
'43
g ^
1 S
D
§
1
CO
> -~ '43
O K^
o '"' to
^
'o
A
o-
0)
ro ^
^_ rt
ti 'C
_O rt
Q
1
XI
rt O'
^ S
(U
^
a
o
•43
a
to
•B
rt
^
OOOOOOOOOO OOOOOOOO
OOOOOOOOOO OOOOOOOO
muifqtxTj.oic\j 81 81 Si 8> 81 8> Si
. i3S3SS2S2
tocotototototot/Jtoto (JUUUUOUU
-37-
o
60
rt
o
U
O
*+-t
13
O
P.
a
o
E
rt
a
rt
M
'3
3
.
T3
(U
d
S
S
,3
rt
K,
o
X
a)
fci
S
+J
U.
bo
c«
1
(U
(U
(— t
4J
a
u
o
I
o
a
a
a
S
5
^
•M
rt
u
1
•K
(U
rt
a
o
•43
rt
TJ
t4
•S
a
5
W
^
ro
-------
I
* 3.1.2 Model -To -Measurement Comparji sons
• Having determined that tjhere is a significant difference between
calculations made using hourly variations in emission rates and calcula-
• tions made using mean values, it is of interest to determine whether one
_ gives significantly better agreement with corresponding measured values.
™ Table 3 shows some comparison statistics. For the St. Louis data, the
• use of variable emission rates yields a calculated mean closer to the
measured mean at 7 of 10 stations. For the Chicago data, the variable
• emission rate calculations are closer at only 3 of 8 stations. Three
sets of paired comparisons statistics are shown in Table 3, including
*• mean absolute error (MAE), root-mean-square error (RMSE), and correlation
• coefficient. The variable emission rate calculations show better agree-
ment with measured values based on MAE and RMSE than the mean emission
• rate calculations for St. Louis stations. For Chicago stations the
reverse is true. Using correlation coefficients, the variable emission
w rate calculations show the better agreement with measured values at both
m locations. Except for correlation coefficients, the mean emission rate
calculations are best for Chicago stations and the variable emission
• rate calculations are best for St. Louis stations.
The model -to-model comparisons shown in Tables 1 and 2 indicate
• that the model is sensitive to hourly variations in emission rates such
m as are represented by the algorithm which has been employed. However,
the validation results suggest that the algorithm has some, but limited,
ff skill in representing these hourly variations. Another possibility is
that other sources of error are so large that they mask the effects of
| changes in emission rates. Since the emission algorithm uses only
-38-
-------
I
I
I
I
1
I
I
I
I
I
I
t
I
I
I
1
I
I
1
"8
CM Jj
W tj
•o „
§«
a
T3 O
S 'g
11
_O W
"rt
tu -a
1 «
'rt 2
fi g1
a s
r>
1
a *J rt
1 g H
43 .S; H
rt O >
"""* *n
O o
U U O
4J
o
rt
a 8 1
rt B S
U W £
O j3
3 o* o
OS t/j ^
rt
S
O
1 1
-3 S
0 <- >
5 g
a w
rt
HI O
2 n
rt
(U
S
o
tu
4^
rt
•a
rt
g O
s §
2
1
M
O
o
•13
o
1
%.
i
o
oo
TH
TH
O
TH
CM
*
CM
O
CM
S
ro
T-i
(M
CO
\Q
o\
rH
ro
1
tx
VO
in
TH
ro
|
to
* •*
00 O
T* Ol
^ ro
O O
g R
O rt
o o
1
* *
•* ro
ro >O
CM O
O oo
CM rt
* *
tx -^
CO rt
TH
oo oo
# &
00 rt
rt 00
TH TH
00 •*
tx CM
ro rt
rt CM
in tx
TH TH
CO CO
to to
# *
rt ro
•* Cft
CM ro
O O
00 rt
rt O
O O
X
S5 S
•*
rt in
TH TH
*
CM ro
TH
.£,
CM ro
*
O CM
CTl tQ
TH
*
Tjl CO
O\ ro
00 tx
ro ro
CM ro
•3 "3
3 3
« +J
to to
\r «
ro in
CJi ro
CM CM
O 0
Rtx
CM
O rt
O O
*
rt 8
CM
^
in CM
O rt
TH ro
* *
m CM
rt O
rt CM
t*"* ^
ro CM
^L ^L
TH tX
^* o
TH CM
ro tx
co to
to to
N. ro
TH CO
^ O
TH
w in
'3 '3
& &
* *
O CM
rf ro
O O
in 3t!
ro O
o o
1
^j.
in ro
Oi m
*
Cl 00
•
*
O
O
8
CM
O
*
CM
CM
00
CO
TH
1
xj*
ro
^
QHj
CO
cj
i
\Q
CO
•3
3
•
tx
VO
0
in
in
o
o
o
j£
TH
10
o
y.
tx
CM
tx
ft
CM
CO
ro
TH
a
s
2
o
*
00
TH
0
TH
tx
O
o
o
ro
*
TH
TH
o
a\
#
ro
00
^.
tn
a
TH
•vH
03
a
s
2
u
* *
S R
ro ro
0 O
CM m
CM rt
O 0
80
ro
•* *
ro O
CM •*
CM rt
0 0
CM O
CM CM
•# •*
rt (O
tx CTl
•*
01 in
IX rt
ro ro
§ £
CM CO
TH CM
CO ^"*
ro Tf
a a
s s
S S
•$(•
tx O
IX 00
CM rt
0 O
m 01
in tx
ro rt
o o
o o
in o
•st •$
in tx
tx m
o o
oo in
* *
in tx
•* CM
oo oo
tM in
TH
oo ro
CM CO
IO CM
in
rt
tu
tJ
a
rt
tu
T3
• K
i a o
rt ^ 8
1 II
SSH
o
>, 5
rt -O
£ a
I* •*
i i
s 1
4) «
43 C
H <
CM ro
-------
temperature, time of day and day of the week as variables, it is possible
to examine the validity of the model with different combinations of these
variables and to determine whether there is any detectable bias. This
has been done in Table 4 for classes of hour of the day, day of the week
and temperature. With regard to hour of the day, the St, Louis measure-
ments show two peaks, one in late morning and one in late evening. The
calculations using variable emission rates also show two peaks, but they
precede the measured peaks by one class. The amplitude of the calculated
cycles, although small, agrees with the amplitude of the measured cycles.
Calculations using a mean emission rate show a single unlikely peak
occurring during the early morning hours. In the Chicago data, the
measurements also show two peaks, although the second peak in the early
evening is very weak. The calculations using variable emission rates
are in phase with the Chicago measurements but the amplitude of the
cycle is greatly magnified. The single cycle of the calculations using
mean emission rates agrees more closely with the Chicago measurements.
There does not appear to be any systematic variation in the
measured or calculated class means by day of the week. However, both
measured and calculated class means show a trend from high to low values
with increasing temperature. The trend is more pronounced in the Chicago
measurements than in the St. Louis measurements. The temperature depen-
dence is exaggerated in the calculations based on variable emission rates.
It appears that many sources were incorrectly assumed to have a high
degree of temperature dependency.
-40-
-------
Table 4. Variations in Mean Concentrations for Selected Classifications
Classification
Hours
00-04
04-08
08-12
12-16
16-20
20-24
Day of the Week
Monday
Tuesday
Wednesday
Thursday
Friday
Saturday
Sunday
Temperature
<15F
15F to 24. 9F
25F to 34. 9F
35F to44.9F
45F to 54. 9F
55F to 64. 9F
>65F
No. of
Cases
1578
1595
1613
1480
1540
1585
1294
1330
1376
1367
1338
1377
1309
806
1795
2748
2099
1359
495
89
Mean 3- Month
Concentration (jig/m )
Over 10 St. Louis
Stations
Measured
145
161
165
128
153
173
154
166
164
141
153
148
155
161
150
154
168
144
139
124
Calculated
Variable Q
148
167
140
123
175
154
141
157
159
148
151
130
173
217
202
161
129
95
73
84
MeanQ
209
180
138
132
185
188
164
171
180
153
176
155
208
178
191
181
163
156
138
132
-KT ,,(
No. ol
Cases
908
896
893
904
897
685
836
850
703
707
716
742
901
738
1000
2365
848
309
180
15
Mean 1-Month
Concentration (/ig/rrr)
Over 8 Chicago
Stations
Measured
79
113
106
92
98
95
87
88
124
102
110
91
82
128
95
89
102
85
73
44
Calculated
Variable Q
152
213
123
103
154
133
123
140
229
140
121
109
160
280
155
122
125
95
46
23
Mean Q
86
101
69
64
84
86
72
76
96
90
72
66
96
93
81
82
72
95
52
36
-41-
-------
In view of the sensitivity shown by the model-to-model compar-
isons, the variable emission rate is a desirable approach where pollution
emissions have significant time-dependent variations. It appears that
the emission algorithms used for St. Louis and Chicago are introducing
significant errors, and that data must be sought which can be used to
make more reasonable estimates of the variations in emission rates.
3.2 COMPARISONS AMONG TEN MODEL VARIATIONS USING NEW YORK CITY DATA
A validation study of the three Gaussian plume models and the
simplified Gifford-Hanna model, described in Section 2.0, was carried
out using data applicable to the vicinity of New York City for 1969.
There are various ways of preparing input for use in these models. In
particular, the emission rates, stability conditions or wind speed may
be assumed to be constant throughout the data period or varied from
hour to hour. Altogether 10 combinations of models and data condition-
ing were analyzed. The material which follows describes the combinations,
describes the New York City data, presents the method of analysis, presents
and discusses the results, and summarizes the findings.
3.2.1 Description of the Ten Model and Data Conditioning Combinations
For the SCIM, area source emissions and the meteorological
conditions of atmospheric stability and height of the mixing layer
(grouped together) are treated either as varying from hour to hour or
-42-
-------
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
• Area source emission rates, atmospheric stability and
height of the mixing layer constant.
For the simplified Gifford-Hanna Model (GHM), area source
emissions and wind speed are treated as both varying from hour to hour
or as both being constant throughout the data period. In addition, the
calculated concentration at a receptor due to point sources (as estimated
by SCIM with variable atmospheric stability and mixing layer height) are
either added or not added to the GHM calculations. This results in four
variations of this model, including:
• Constant area source emission rates and wind speed,
without point sources
• Variable area source emission rates and wind speed,
without point sources
• Constant area source emission rates and wind speed,
with point sources
t 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
-43-
-------
I
for these calculations are included in this report for comparison with
V the other models. Calculations for AQDM (no variations) also were
furnished by Mr. Turner and are included for comparison.
3.2.2 Description of the Data
£ Three types of data for New York City which were used to test
_. the validity of the models are described here, including air quality
* measurements, pollutant emission information and meteorological measure-
A ments. The air quality and meteorological measurements were made during
1969, and the emission information is applicable to 1969.
3.2.2.1 Air Quality Data
• The air quality measurements consist of annual mean concentra-
tions of SO-, and particulate aerosols and hourly concentrations of S00.
12 2
The annual mean concentrations were measured at 127 locations in the
£ New York - New Jersey - Connecticut Air Quality Control Region, S02 was
measured at 75 locations and particulate aerosols were measured at
If 114 locations. Hourly mean concentrations of SOg were measured at
39 locations in New York City. The 10 locations shown in Figure 21,
£ representing different geographic characteristics, were selected for use
_ in this analysis. All 127 of the locations measuring annual concentra-
* tions were used. The locations in the immediate vicinity of the New York
• City limits are shown in Figure 21.
m 3.2.2.2 Emission Data
The emission information consists of: (1) emission inventory
• data prepared for the Implementation Planning Program (IPP) for the
• .44,
1
-------
I
I
t
I
I
I
I
I
I
I
I
I
t
I
I
I
I
I
I
Annual Concentrations Only
4-Hour SO and Annual
Concentrations
4530
4520
4510
4500
4490
4480
570 580 590 600
Figure 21. Locations of Air Quality Measurements in the Immediate Vicinity of New York City
-45-
-------
New Jersey - New York - Connecticut AQCR, and (2) time and temperature
dependent weighting factors which when multiplied by the annual emission
rate of area sources provide an estimate applicable to a particular hour.
The IPP data includes 675 point sources and 871 area sources.
For each point the available information includes Transverse Mercator
Projection Coordinates (TMPC), Standard Industrial Code (SIC), annual
S02 and particulate emission rates, stack height, stack diameter, stack
gas exit velocity and stack gas exit temperature. Each area source is
2
defined for a specified square area varying between 1 and 100 km . The
available information includes the TMPC of the southwest corner, the
size of the area, the annual S02 and particulate emission rates, and
the effective height of the emissions.
The weighting factor used to determine area source emissions
for a particular hour is a function of temperature and time of day; its
development is described in Appendix A. The area source emission rate
algorithm is
«1 = QA Ki
where
Q. = emission rate for the ith hour of the year (yg/irfysec)
Kn- = [(l-Fh) J.L1 + Fh (T.-T.) I.KL2] for T. < T.
(22)
Kn- = [(l-Fh) J.L^ for T. >. T.
p
0. = mean annual area source emission rate (yg/m /sec)
F, = 0.29 = fraction of emissions attributed to space heating
requirements
-46-
-------
I
Ji = base level emission coefficient for hour i (day/hr)
• TJ = threshold temperature for hour i (°F)
T. = observed temperature for hour i (°F)
I Ii = temperature sensitivity factor for hour i (Index/°F)
• K = constant H 0.0002337 yr/Index hr
LI = unit conversion = 24 hr/day
I L2 = unit conversion = 8760 hr/yr.
• The numerical values of the terms J., T. and I. are given in Appendix A.
A value of K. was computed for each 3-hour temperature observation
I obtained for 1969.
| 3.2.2.3 Meteorological Data
_ The meteorological measurements consist of 3-hourly synoptic
™ weather observations reported for LaGuardia Airport and twice daily
A radiosonde observations reported for Kennedy Airport, The observations
from the 3-hourly weather report for LaGuardia Airport which were used
| in this study were:
I • Wind direction (nearest 10° azimuth)
0 Wind speed (observing height of 6.1 m)
| • Temperature
». t Cloud cover (tenths)
• Height of cloud ceiling.
•
m
I
I
The first observation period of each day was 0100 EST. Twelve hours for
which the reported wind speed was calm or less than 1 m/sec were excluded
-------
from the data set. The validation was limited to the use of 3-hourly
observations since this is what is routinely available from NOAA's
National Weather Records Center in Ashville. As a result, the validation
findings indicate what can be expected from routine as opposed to research
use of the models tested.
The radiosonde observations were used to estimate the height
of the surface-based mixing layer. The details of the parcel method
used to define the height are described in Appendix B. The height
derived for the morning sounding was assumed to be applicable to the
hours of 0100, 0400 and 0700. The height derived for the afternoon
sounding was assumed to be applicable to the hours of 1300, 1600, 1900,
and 2200. An average of the morning and afternoon values was used for
1000. If a sounding was missing, the height assigned to the previous
24-hour period was used. An annual mean mixing height was determined
by averaging all the 3-hourly values. Heights in excess of 5000 m were
treated as being 5000 m. The resulting annual mean was 979 m which is
about midway between the annual morning and afternoon mixing layer
heights recently reported by Holzworth (1972) for New York City.
Two other annual mean meteorological values which were used
in this study were 5.1852 m/sec for wind speed and neutral (Pasquill-
Turner class 4) for atmospheric stability.
3.2.3 Comparison Statistics
Model-to-measurement comparison statistics were obtained for
each of the 10 model and data conditioning combinations. The statistics
are listed and compared in the following sections. All the statistics
-48-
-------
I
are standard and require no special explanation. Error as used in these
• statistics refers to calculated minus measured concentration.
• 3.2.4 Results of Model -toJIeasurement Comparisons
The ability of all four models to estimate long-term concentra-
| tions accurately is evaluated. Only the SCIM and GHM are appropriate
_ for testing for short-term concentration estimates. The short-term
' concentration calculations, which are compared with measured hourly mean
•
m
concentrations of S0p» are presented first.
3.2.4.1 One-Hour SOz Comparisons for SCIM and GHM
Using SCIM and GHM, the SOp concentration was calculated at
each of 10 receptor locations for every third hour of the year, for which
a measured concentration was available, beginning 0100 on January 1, 1969.
For each hour selected, SCIM calculations were made in three ways:
• With variable area source emissions (using Equation (4)
for Ki) and variable stability and mixing layer height
(variable Q, S, and H)
t With mean area source emissions (K-j=l) and variable
stability and mixing layer height [mean Q, variable S and H)
t With mean area source emissions, stability (Pasqui11-Turner
class 4) and mixing layer height (979 m) (mean Q, S, and H).
The GHM calculations were made in two ways:
• With variable area source emissions and wind speed, and
without the inclusion of any contributions from point
sources
• With variable area source emissions and wind speed, and
with point source contributions as calculated by SCIM
added to the basic GHM calculations.
-49-
-------
The frequency distribution of 1-hour concentrations for each of
the three types of SCIM calculations and for the measured values are shown
on log-normal graphs for each of the 10 stations in Figures 22 through 31.
The corresponding frequency distributions for the two types of GHM calcula-
tions and for measured values are similarly plotted in Figures 32-41. The
curves are close to straight lines; therefore, it is reasonable to expect
that they may be approximated by log-normal distribution functions. This
hypothesis will be tested during Phase II. The GHM graphs clearly show that
the model calculations with point source concentrations added give better
agreement with the frequency distribution of measured concentrations at 8 of
the 10 New York City stations. Of the other two (Stations #0 and #1), only
Station #0 shows better agreement with calculations without point source con-
centrations. The SCIM frequency distributions based on variable S and H
differ very little and agree best with the measured frequency distributions
at Stations #14, 17, 27, 28, and 31. Only at Station #0 does the SCIM fre-
quency distribution based on mean Q, S and H agree best with measured values.
A summary of model-to-measurement comparison statistics is given
in Tables 5 and 6. Table 5 includes a comparison of the means and standard
deviations of measured and calculated concentrations for each station. The
root-mean-square errors (RMSE), the mean absolute errors (MAE), and the error
(calculated minus observed) concentration corresponding to the measured
maximum concentration are also included for each set of model calculations.
Table 6 shows correlation coefficients and linear regression characteristics
for model-to-measurement comparisons. The tabulated regression characteristics
of measured on calculated concentrations include the variance attributed to
the regression relationship and the two slope and intercept regression
coefficients.
-50-
-------
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
X
'
X
r/^
/
//
//
/
T
/
m
ff
T
/
0
/
/
j
/
r
i
/
//
— ff
^
4
I ft
7
r
/
Key
- 8 Observed
• Variable Q, S,
x Mean Q, :>, H
t
lean Q, Va
riabl
H
e H
S
RS
80
70
60
•in
^n
in
?n
is
10
5
10
100
1,000
10,000
Concentration
Figure 22. Frequency Distribution of SCIM Calculated and Measured One-Hour Concentrations for New York City Station #0
i—|-~r-iX2>rr
Key
8 Observed
t Variable Q, S, H
x Mean Q, S, H
+ Mean 0. Variab e H. S
99
98
95
90
85
80
70
60
50
40
30
20
15
10
10 100 • 1,000 10,000
Concentration (yg/m3)
Figure 23. Frequency Distribution of SCIM Calculated and Measured One-Hour Concentrations lor New York City Station #1
-SI-
-------
t77
99
98
95
90
85
80
70
60
50
40
30
20
15
10
Key
6 Observed
• Variable Q, S, H
x Mean Q, S, H
+ Mean Q, Variable H, S
10
100 • 1,000
Concentration (ug/mj)
10,000
Figure 24. Frequency Distribution of SC1M Calculated and Measured One-Hour Concentrations for New York City Station #3
99
98
95
90
85
80
70
60 re
50 «
o
40 ^
n
30 S
3
20 ?>
fD
15 ^
**
10 -
//
??
/
Key
6 Observed
• Variable Q, S, H
x Mean Q, S, H
Mean Q, Variable H, S
10
100 • 1,000
Concentration (pg/m3)
10,000
Figure 25. Frequency Distribution of SCIM Calculated and Measured One-Hour Concentrations for New York City Station #10
-52-
-------
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
Pr
Key
« Observed
• Variable Q. S. H
x Mean q, S, H
Mean q, Variable H, S
99
98
95
90 g
85 I
OJ
so £
o5
70 ^
60 S
50
40
30
o
-t>
o
20 =
n
15 „
w
10 "
10 100 • 1,000 10,000
Concentration (tjg/m3)
Figure 26. Frequency Distribution of SCIM Calculated and Measured One-Hour Concentrations for New York City Station #14
ty
7
f- --
fr
r
Key
8 Observed
• Variable O.S, H
x Mean q, s; H
Mean q, Variable H, S
99
98
95
50
30
20
15
10
j
I
4
o
10 100 ' 1,000 10,000
Concentration (\ig/m3)
Figure 27. Frequency Distribution of SCIM Calculated und Measured One-Hour Concentratlonl for New York City Station 117
-53-
-------
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
Key
8 Observed
i Variable q, S. H
x Mean Q, S, H
Mean q, Variable H, S
99
98
95
90
85
80
o
70 31
50
40
30
20
15
10
o
-h
O
i
10 100 • 1,000 10,000
Concentration (pg/m?)
Figure 28. Frequency Distrlbutl«n of SCIM Calculated and Measured One-Hour Concentration! tor New York City Station 027
-*~T
3
Key
0 Observed
t Variable q, S, H
x Mean q. S. II
Mean 0. Variable H. S
99
98
95
90 o
85 ^
80 2
<
n
70 3
(D
60 I
50 Q
O
40 •"
o
n
30 9
10 100 ' 1,000 10,000
Concentration (pg/m3)
Figure 29. Frequency Distribution of SCIM Calculated and Measured One-Hour Concentration! fcr New York City Station »28
-54-
-------
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
T
/,
y
^
/
\'
— T
/
/
/
7x
L/7
/ \. fi.
V /
t
9*
I
/ f
J
Y
I/
y1
ni
/
^/^
r
//
~fy
//
'/
Key
' Observed
• Variable C
x Mean Q, S
<• Mean Q. Vc
!.HS,
riabl
H
e H
S
95
90 *"*
|
rt-
on -•*
5
_. 70 J
i*
- - 50 **
o
n
in c
- - JU ^
3
-.20 o
- 15 ~»
_. 10 -
-. 5
10 100 • 1,000 10,000
Concentration wg/m3
FiRiire 30. Frequency Distribution of SCIM Calculated and Meaiured One-Hour Concentration! tor New York City Station *31
^
^
J
.
I
/^
>*
X
^
/
/
'/
/
1
' J[
I
/
/j
•t
J;
y
f
/
//
/
^
F^
e o
• V
x N
t M
//
t
Key
bserved
ariable Q, S
ean Q, S, II
ean Q, Vartat
. H
>1e H
S
95
90
• 7W o
85 1
3
- 70 -n
fin c
- - 50 £
0
_ - 40 -*
o
o
20 1
15 S
»«
-.10 ~"
- . 5
.. 2
10
10,000
100 • 1,000
Concentration (yg/m3)
Figure 31. Frequency Diminution of SCIM Calculated and Meaiured One-Hour Concentration! for New York City Station #36
-55-
-------
^
I
- 85 —
- 80 ~
£
- 70 ?
3
_Q
--50 1
o
- 40 -
n
_.30 S
T
rt>
-.20 S
S
_. 15 -~
«
10
-- 5
. •>
10 100 1,000 10,000
Concentration (ug/m3)
Figure 32. Frequency Distribution of GHM Calculated and Measured One-Hour Concentrations for New York City Station #0
•
y
^
^^
X/
J.
If
7
in
7
i
//
i
if
L
f/\
/
j
^
'/
i
+
•
V
Ob
Mi
Mi
Key
served
th Point S
thout Poin
ource
t Sou
s
rces
95
90
85 §
Ol
3
70 -n
3
A
--60 =
3
so C
o
AO "**
o
o
-. 30 S
3
20 3
S
15
«
10 *"*
5
_. 2
10
100
1,000
10,000
Concentration
Figure 33. Frequency Dlitrlbutlo^ of CUM Calculated and Measured One-Hour Concentration* for New York City Station 41
-56-
-------
Key
6 Observed
+ With Point Sources
• Without Point Sources
99
98
95
90
85
80
70
60
50
40
30
20
15
10
10
100
1,000
10,000
Concentration (ug/m3)
Figure 34. Frequency Distribution of CUM Calculated and Measured One-Hour Concentrations for New York City Station »3
y.
Key
6 Observed
+ With Point Sources
• Without Point Sources
99
98
95
90
85
80
70
60
50
40
30
20
15
10
£>
C
ro
I
o
10
100
1,000
10,000
Concentration (pg/m3)
Figure 35. Frequency Distribution of CUM Calculated and Measured One-Hour Concentration! for New Yolk City Station #10
-57-
-------
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
b*
Key
8 Observed
+ With Point Sources
• Without Point Sources
99
98
95
90
85
80
70
60
50
40
30
20
15
10
10
100 ' 1.000
Concentration (ug/m3)
10,000
Figure 36. Frequency Distribution of CHM Calculated and Measured One-Hour ConcentradoDj for New York City Station #14
100
' 1,000
Concentration (pg/m3
10,000
Figure 37. Frequency Distribution of CHM Calculated and Measured One-Hour Concentrations for New York City Station #17
-58-
-------
1
"
1
1
1
1
1
1
]
1
1
1
1
1
1
1
1
1
1
1
X
/
//
A
n
/
//
///
///
iff
' if
/
L
V
n
p
I
//
V
/
/
r
f
9
*
Key
Observed
With Point
Without Po
Sour
int S
ces
ourc
es
99
- 95
90
85
- on
- 70
- 50
30
20
15
- 10
-- 5
•>
10 100 ' 1,000 10,000
Concentration (ug/m3)
Figure 38. Frequency Distribution of CHM Calculated and Measured One-Hour Concentrations for New York City Station #27
- ^ A OU
/jr<
1
/
/
/
/
t
1
4
1
/
1 i
t
f
y
t
/
J
' 1 1
' j j
^ / /
y (J
f
/
7
1 1
/
r
/
'
Key
8 Observed
+ With Point Sources
• Without Point Sources
- • 95
- 90
- - 85
- . 70
- . fin
- 50
40
- 30
_. 20
-. 15
_. 10
5
. . J
o 100 • 1,000 10,000
Concentration (yg/m3)
Figure 39. Frequency Distribution of CHM Calculated and Measured One-Hour Concentrations for New York City Station »28
-59-
o
o
-*
o
-------
1
I-L
1
1
1
1
1
1
1
I_
1
I
•
1
(
J
f
1 /
<{'
/
/
/
/ /
//
' //
/ //
//
/ f
f f
f
./ .
/
/
/
*
/
//
r
r
Wt
?
i
/
7
9
+
•
Key
Observed
With Point
Without Po
Soure
nt Sc
es
>urc
es
«. 99
95
-.90 r>
- 85 £•
a>
--80 2
S
--70 ?
n
J3
- - fin S
--50 «
0
-.40 2
o
n
-.30 c
2
-.20 o
- . 15 "•
«
-.10 ""
5
1
LO 100 ' 1,000 10,000
Concentration (yg/m3)
Figure 40. Frequency Distribution of CHM Calculated and Measured One-Hour Concentration! for New York City Station *31
•
i/v on
/
J
//
1 \
/
1
1,
J
/«•
/
f
/
/
X1
/
A
/
/
^ .
/
/
/
^ _
xT
/
1
1
f
}
/
//
/" f
/ 1
' ®
/
ri/
//
/
c/
™
I
//
r
7
« Ob
+ W1
• Wi
Key
served
th Point Sot
thout Point
irces
Sourc
.es
- 95
- 90 <•>
85 -•
- • 03 Q,
80 2
•80 s
- 70 ?
J
- 50 5
o
-ti
- 40 o
S
- 30 5
- 20 2
- 15 C1
. 10
5
10 100 • 1,000 10,000
• Concentration pg/m3)
Figure 41 . Frequency Dlrtilbudon of Cl IM Calculated and McniuKcl One-Hour Conccntntloiil br Niw Yolk City Station 136
• -60-
1
-------
a «
o c
'•°
O
U
O
=F §
^ w
M-l
o t5
a S
g w
u
s PS
-
§!
si
rA T3
™ -5 i?
1 c? o- « 1
1 g S o
C w Co r* t-j
^ 2 QJ fl] T3 ^3
^ *|5 5- "5J *"> ''j!
| D u u a 3
2 oo oo oo O O
(cui/2r/) ui3aj^
in TH a\ (Ti co CM
O CM CO TH CM O
00 IX >H ID O TH
o a* oo & vo CM
oo o rji tn TH TJI
oo oo oo tn T^i m
CTl in CM tx TH TH
CM in in oo CTI
CTI in o^ tx ^* co
TH fo CM 00 00 O
tx CO CO CO 00 VD
CM CM Tf tx VO tx
CM 00 tx CO CO T*
CO CM CM TH CM CM
CM 00 00 CO CO TH
O^ oo O VO rx oo
tx in CM oo m o
in O TH Tj< CM Tjl
O TH TH O O C\
CM CM CM tx O O
CM Tf Tl< TH CM CM
*^L
t>> ,• ffi jT
(> > Vi 'g "Q
Q) ^ •» A* ri
"B > S S 'S 'S
! o o o ss
A 00 00 00 O O
(gui/STf)
TH ' T$I in TJ* TJ<
M m vo TJ in
oo tf> oo &i tf
00 C\ CM O O
TH oo m T* ix
m vo vo TJI m
.N rv) tn Tj< CM CM CM
~ £"
O > oo -g g"
» • • B« B
2 2 2 2 S
D)-H HH LlJ JZ!
U U S K
oo oo oo (J O
( /»ttt /Sn)
JOJkfrT 3Jl?lA^V
tX VO O TH T*
CO T^l iH TH O
o^ vo o oo 01
in vo TH o o\
CM oo vo at m
TH TH C> tx tx
VO ffl 00 CO tx
^* in TH o fJi
CM vo in co oo
in in CM o o\
VO O tx tx OO
Tjl VO Tjl fO TH
VO IX IX tx CO
in tx in o TH
CM CM CM CM CM
00 T* 00 tx VO
CM in vo vo vo
CM N TH TH TH
CM O\ CM TH TJI
TH CM TH TH TH
co a\ oo T* o
CO 00 00 I/) VO
CM CM TH TH TH
^ > "if
m ~ • o. a
3 o o g 'g
lillf
"5! "5 ^| M »^i
0 D D i i
oo oo oo O O
(pUi/3T/) 10113
s
o\
CM
tx
O
VO
CO
oo
oo
tx
TH
«
TH
in
CO
TH
TH
tx
VO
CO
CM
tx
O
in
CO
TH
CM
CM
mm Measured (ug/m3)
£
I
TH VO VO O Tj.
T* o\ o in TH
i i i i i
CO CM >H O\ TH
1 1 1 TH |
tx tx CM tx tx
CM CM CO 00 00
1 1 1 1 1
Ix TJI (M VO CM
vo co m vo o
1
in CM co o m
Tj" in in co CM
i i i i i
SOO OO T* CM
tX TH T* CM
1 | TH TH TH
CM CM TH ft 00
STH TH CM in
in in TJI jo
VO VO VO VO VO
1 1 1 1 1
CO O^ O) O co
vo vo tx 5> tx
^J* ^J* I/) ^ T^(
1 1 1 1 1
TH 00 O ** CM
in o "* co 05
co in cji TH oo
1 1 1 TH |
1
•* oo vo co co
co in ^ tx CM
o» CM in co TH
I TH TH TH
1 1 1
C? t> IA 'g g~
2 9 S Q
1 3 | | g
D D D a a
on oo oo O O
pamstfaj/v umot
-prej/\[ JE 10113
-61-
-------
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
.
u
.
B
2 fi
C rt
rt *r^
a >
o
U j-T
S £
II
3 J2
4, g
Z
*o
TH
00
tx
tx
•*
O
CO
i— 1
O
K
CO
GO
TH
T-H
00
(f)
T— 1
(M
o
TH
CO
^O
T-H
^
T-1
vo
T-l
R
T— 1
Ol
00
TH
CO
i— I
rH
CM
a
(M
iber of Comparisons
tt
3
vo Oi oo co co
CM TH TH CO TH
O O O O O
TH 00 Cl TH T^
CM TH TH CM CM
O O 0 0 O
tx CM in a\ rj<
TH TH O TH TH
0 O 0 0 0
01 -* in O co >O TH
CM T-I t-i TH CM
O O 0 0 O
o o o •* vo
ro CM CM ro co
O O O O O
CO O co VO «3
T}< CM i-l Tf ^J<
0 0 O 0 0
S"
K. ">~~
^ J X •g'
- ^ - s _
0 > w -S K
a) « ~ & o
3 o o- 3 -5
oi n a § &
•fl rt rt J3 J!
I ^ ^ S s
b E s $ $
2 2 S S S
1 — 1 »— 1 1 — 1 [^ Lj
U u o K X
w oo oo U O
juapi^aoo
UOPB-[3JJCO
8 S 8 3 S
o o o o o
s s s s §
o o o o o
CO CM O •* CM
O 0 O O 0
0 O 0 0 0
00 CM CM CTi CTl
O O O O O
O 0 0 0 0
SCM CM
O > w -g y
o ^ . a, a
3 o o- 3 -g
rt fl fl g &
•fi rt rt .s rj
J3 £ 2 3 3
§ § § s s
D u u K x
oo t/j oo O U
cioissaj3a>j jB3ur[
Xq aowsuEA
jo uoponpa-^j
CM on -H CO CO
o o o o o
CO K O CO TH
CM T-I uo in vo
0 O O 0 O
o Tji 10 oo in
CM O O CO CO
O O 0 0 O
1-1 oo in oo CM
TH O TH TH CM
o o o o o
CM 00 T-I CM TH
TH O CM Tj< Tj<
O O O O 0
CM O tx O CTv
CM TH TH in ^f
o o o o o
^K
E>~
oo" 4 ffi J-
M rt •» fl
o > w -5 e"
5, . . 6, s
3 o o- g •£
|| I 5 il
> ^ E ^ ^
SUss
D u u pc ffi
00 00 00 O O
aui-j uoissai8a-a
jo ado^s
in T-t c^ fj, co
CM CO CM 00 CO
TH CM TH T-i -^
K K 10 in m
O\ T-I co O fx
TH (M CM O O
tx CM CO 00 CTi
in K 10 CM TH
O U5 Tj< 00 TH
tX K CX Tjl I^H
TH 00 CM CO O
tri (Ti oo tx in
O K tx Oi TH
VO O O tx tx
rf tx G^ tx tx
m 10 m -^t1 CM
TH TH CJ\ O O
oo 01 rx co CM
o oo in oo co
TM m in •* co
£
^"'~~
w" ,; X 5-
•> «* « «
o > w 'S F
u . ~ 6, s
SO-O-^-S
•e S S I
| •S 3 2 ^
> S S £ u
S 2 S 2 S
o u u x x
00 00 00 O O
auji uotsssiSa^j
jo jdaojajuj
-62-
-------
Several general results may be noted by comparing the statistics
for the three SCIM calculations. The results obtained using variable Q,
S, and H are very similar to the results obtained using mean Q, variable S,
and H inputs; however, higher correlations and slightly smaller errors are
observed using variable Q. When the variable Q, S, and H results are
compared with the mean Q, S, and H results, the differences are more strik-
ing. The calculated mean is closer to the measured mean at 7 out of 10
receptors using the variable inputs, while the root-mean-square error is
smaller at all receptors using the mean inputs. The correlation coefficients
are higher at all receptors using the variable inputs. The frequency dis-
tributions of errors for each set of calculations was also examined. It
was found that the variable input calculations produce some large overpre-
diction errors not found with mean input calculations. In view of the
similar results obtained using mean and variable Q with variable S and H,
these large errors must be due to inaccurate estimates of the atmospheric
stability or the mixing layer height.
The GHM results are slightly, but consistently, better for the
calculations with point source concentrations than for the calculations
without point sources. The correlation coefficients with point sources
are higher at 7 of the 10 stations; the RMSE is smaller at 6 stations;
the error at the minimum measured concentration is smaller at 9 stations.
The results in Tables 5 and 6 suggest that SCIM with Variable Q,
S and H and GHM with point sources gave the best overall results. Of these
two, SCIM produced better results for estimating the mean and the maximum
measured concentration. GHM produced smaller RMSE's and better correla-
tion as indicated by the correlation coefficient and the regression charac-
teristics.
-63-
-------
I
In order to get more insight into the validation results, the
I results were stratified by classes of five different parameters. For
each parameter, the mean of each class was determined for each set of
| model calculations and for the measured values. The class means for
• hour of the day and day of the week are shown in Table 7. The class
means for temperature, wind speed, and stability are shown in Table 8.
H With regard to day of the week, observations for Sunday were
available for only 1 station (#0). The means for Sunday were made
| comparable to other days of the week by multiplying the Sunday mean by
_ the ratio of the annual mean for all stations to the annual mean for
* Station #0. None of the six columns in Table 7 show any important
• variation by day of the week.
The results by hour of the day are listed in Table 7 and plotted
g in Figure 42. With small exceptions all five sets of calculated results
follow the general trend of the measured values. Calculations using
• SCIM with mean Q, S, and H underestimate the magnitude of the cycle of
• measured values while the other four sets of calculations overestimate
the cycle. This result suggests there is a need to include model param-
• eters with diurnal variations, and that the diurnal variation is easily
overestimated. Clearly, the diurnal cycle of stability, mixing layer
I depth and emissions as represented in SCIM exaggerate the diurnal cycle
m of measured values. The 6HM calculations which do not include a stability
parameter show a better correspondence to the diurnal cycle of measured
I values. The addition of point sources in GHM calculations, while not
changing the amplitude of the diurnal cycle, brings it into better agree-
I ment with measured values.
I -64-
I
-------
I
I
I
I
I
I
I
I
I
I
I
I
1
I
I
I
I
I
1
Table 7. Comparisons of 1-Hour SC>2 Concentrations by
Hour of the Day and Day of the Week
Classification
Hour
01
04
07
10
13
16
19
22
Day
Monday
Tuesday
Wednesday
Thursday
Friday
Saturday
Sunday*
Number of
Cases
2143
2116
2111
2077
2134
2146
2146
2141
1796
3191
3350
3362
3193
1710
412
Measured
(Mg/m3)
209
237
303
258
223
214
231
226
232
242
232
233
238
240
(215)
SCIM (Mg/m3)
Variable
Q, S, H
269
311
444
223
166
163
226
247
239
245
229
264
241
306
(270)
MeanQ,
Variable S and H
341
372
336
223
191
179
219
268
253
257
243
272
252
308
(269)
MeanQ,
S andH
169
183
190
168
142
136
137
154
154
155
149
158
152
182
(152)
GHM (Mg/m3)
Without
Points
156
180
302
188
139
139
176
171
171
175
165
184
176
206
(178)
With
Points
194
221
352
236
180
177
211
209
210
216
205
224
217
254
(218)
* Sunday values were observed only for Station #0. The tabulated values were normalized by multiplying
the Sunday mean by the ratio of the mean of all station measurements to the mean of all Station #0
measurements.
-65-
-------
I
I
I
I
I
I
I
I
I
I
I
I
1
1
I
I
I
I
1
Table 8. Comparisons of 1-Hour SC>2 Concentrations by
Temperature, Wind Speed and Stability Classes
Temperature (°F)
<15
15 to 24.9
25 to 34. 9
35 to 44. 9
45 to 54.9
55 to 64.9
>65
Wind Speed f m/sec)
<1.5
>1.5 to 2.0
>2.0 to 2.5
>2.5 to 3.0
>3.0 to 4.0
>4.0 to 5.0
>5.0 to 6.0
>6.0 to 8.0
>8.0 to 10.0
>10.0
Stability
(Turner-Pasquill)
1
2
3
4
5
Number
of Cases
36
511
2595
2573
2401
3316
5582
55
724
728
989
2570
2901
2952
4699
1271
271
17
559
1784
11667
2987
Measured
(jUg/m3)
363
322
299
306
279
210
167
226
258
254
278
252
228
230
228
231
230
143
213
235
236
248
SCIM (jug/m3)
Variable
Q, S, H
248
316
268
290
258
261
225
937
799
506
446
325
232
210
146
114
106
162
374
284
188
482
Mean Q,
Variable S, H
146
180
196
238
258
315
295
1054
893
544
467
338
246
211
146
103
83
255
434
300
185
533
Mean
Q, S, H
132
123
131
147
154
170
179
625
402
285
248
194
159
135
107
82
67
332
273
201
133
218
GHM (/ig/m3)
Without
Points
288
281
230
221
185
162
141
680
456
342
288
221
166
148
123
111
114
316
239
193
160
247
With
Points
320
306
256
255
223
202
194
763
515
396
335
266
210
190
158
136
132
340
302
248
198
284
-66-
-------
I
I
I
I
I
1
500
400
300
bo
a
o
o
e
o
O
200
100
Measured
SCIM (Variable Q, S, H)
& SCIM (Mean Q, Variable S, H)
SCIM (Mean Q, S. H)
GHM (.without points
SHM (with points)
10 13 16
Hour of the Day
22
Figure 42. Variations in Measured and Calculated Mean One-Hour SO2 Concentration by Hour of the Day
-67-
-------
I
Temperature-dependent trends are tabulated in Table 8 and shown
• graphically in Figure 43. The importance of the temperature-dependent
emission algorithm is suggested by the observation that the two sets of
• calculations based on mean Q run counter to the trend of the three sets
m of calculations using a variable Q and the measured values.
All five sets of calculated results show an important wind
• speed dependence in Table 8. This dependence is not evident in the
measured concentrations. It is difficult to explain the measured results,
• since they are contrary to results observed at other locations such as
m Chicago, and St. Louis (Koch and Thayer 1971). Normally, if the wind speed
is increased one would expect the pollutant concentration to decrease
ff because the pollutant is diluted by a larger volume of air. However,
for this particular data some compensating effect occurred, such as a
jj decrease in diffusion or an increase in emissions. Clearly, the compen-
g sating effect is not accounted for in the models.
* Stratifying the calculated and measured concentrations by
• stability class (Table 8) also shows some systematic errors in the models.
The results for class 1 are inconclusive due to the small number of cases,
f less than two per station. For the remaining four stability classes, all
_ five model results show the same pattern in varying magnitudes. The
™ means decrease from class 2 to class 4 and then increase for class 5.
• The magnitude of the variations are especially large for the two SCIM
results with variable stability inputs. The pattern contradicts the
• pattern of measured concentrations which show a small consistent increase
from class 2 to class 5. Since wind speed tends to be light with very
• stable and unstable conditions, it is likely that the difference between
I
1
-------
1
I
I
I
1
1 40°
1
™ 300
ro
f ^
• 1
£
O
I.P-*
•M
a
ts
s
o
f<5
200
1
f
100
I
1 .
1
Figu
•
1
4
E
<§
\
^
t--—-.
/
/
S
s
3 —
\
^
£
1
15
/\
~~~~-^
•>
%
\
\
<^
il
)— *
^
^s
V
>-=
Ik
\T
V
^
^^
il
r"
— c
^
i
k
^
^V-
V
H-
®-
A
•_
i.
"N
"x
\
i
JK
\
) —
—
k.
\
^
k
*\
V
--$.
\
r'
I/
— ^
\
^
^
><^
^
/
•x
[r
r^
^
Key
\
S>
S.
\
s
— 4
V<$
^
J
>
N
)
£L Measured
XX «
y^- •.
-ra «
^S> S
A r
^r t
• r
t
iCIM (Variable Q, S, H)
CIM (Mean Q, Variable S, H)
CIM (Mean Q, S, H)
.HM (without points)
.HM (with points)
10 20 30 40 50 60 70
Temperature, °F
re 43. Variations in Measures and Calculated Mean One-Hour SC^ Concentrations by Temperature
-69-
-------
I
^* measured values and the three sets of calculated means without a stability
• dependent parameter (SCIM with mean Q, S and H, and the two GHM's) are
due to the effects of wind speed on the model calculations. Although
J the use of a mean wind speed is shown in Section 3.2.4.4 to improve
_ annual mean concentrations calculated using the GHM with point sources
^ added, this seems undesirable because it amounts to suppressing a measured
• variation in atmospheric conditions. It is more desirable to find modi-
fications to the models or to other model inputs which represent the
I compensating effects and leave the model compatible with results for other
locations.
• In view of the interrelationships between the parameters considered
* above, it is not possible to isolate the sources of error. However, it
appears that some hypotheses can be drawn which are worthy of future study.
• The rate of change of mean calculated concentration with temperatures for
the two best performing models (SCIM with variable Q, S and H, and GHM
m with points) is less than the rate of change of mean measured concentra-
• tion with temperature. These models also show a larger mean diurnal varia-
tion than the measured values. These results suggest that the emission
• algorithm overestimates the diurnal factor and underestimates the fraction
of emissions which are , temperature-dependent. Another hypothesis is that
• SCIM overestimates diffusion effects associated with changes in atmospheric
M stability. This might be due to the inadequacy of the meteorological mea-
surements based on the Turner-Pasquill index to represent stability varia-
M tions. Of course, there are other possible sources of error which have
-------
not been investigated. These include the effective height of the area
| source emissions, the allocation of emissions to geographical areas in
_ the emission inventory, and the diurnal variations in the mixing layer
^ height, to mention a few.
9 3.2.4.2 Twenty-four Hour SO? Comparisons for SCIM and GHM
• The calculated and measured concentrations were averaged for
each 24-hour period and compared. The frequency distributions for each
• of the three sets of SCIM calculations and the set of measured values
are shown in Figures 44 through 53 by receptor location. These figures
I agree with the results of the preceding figures for hour concentrations.
M They suggest that both sets of calculations based on variable S and H
inputs more closely approximate the distribution of measured concentra-
I
tions than the calculations based on mean S and H inputs. The frequency
distribution of 24-hour SCL concentrations for the two sets of GHM
B calculations and the measured values at each of 10 New York City sampling
_ stations are shown in Figures 54 through 63. Of the two model calcula-
* tions the frequency distribution of calculations using point sources more
• closely coincides with the frequency distribution of measured values at
7 of 10 stations. For two stations (#0 and #1), the calculations without
£ points show better agreement with measured values. The measurements at
— the other station (#3) are about equally matched by both sets of calcula-
' tions.
• Statistical comparisons of the five sets of model calculations
with measured values are given in Tables 9 and 10. As with the hourly
• concentrations, the 24 hourly concentrations show that, of the three sets
I
f
-------
99
98
95
85
80
70
60
50
40
30
20
15
10
Key
Observed
Variable Q, S, H
x Mean Q, S, H
Mean Q, Variable H, S
10
100
10,000
1,000
Concentration (ug/m3)
Figure 44. Frequency Distribution of SCIM Calculated and Measured 24-Hour Concentrations for New York City Station 10
10
100
j
If
u
1
/,
^s
t
/
t
/
/
/
/ <6
/m
' /I
7
r
t
//
/
«
/
\
\
i
i
7
'
^
^
'O
/
/
1
/
y
8
X
?
1
Ob:
Vat
Me;
Mec
//
>/
Key
erved
•iable Q,
n Q, S, H
n Q, Vari
5, H
able t
<. s
QQ
95
90 _
85 1
D*
on r+
70 -n
f.n c
50 Q
o
40 ^*
o
o
30 g
^ 3
0
15 "
X
10 ~
5
1,000
10,000
Concentration (yg/m3)
Figure 45. Frequency Distribution of SCIM Calculated and Measured 24-Hour Concentrations for Hew York City Station #1
-72-
-------
I
I
I
I
I
I
f
I
1
I
1
I
1
I
1
I
I
t
I
t
i\
Key
9 Observed
• Variable Q, S, H
x Mean Q, S, H
+ Mean Q, Variable H, S
n
99
98
95
90
85
80
70
60
50
40
30
20
15
10
10
100
10,000
1,000
Concentration (yg/m3)
Figure 46. Frequency Distribution of SCIM Calculated and Meatured 24-Hour Concentration! lot New York City Station 13
Key
8 Observed
• Variable Q, S, H
x Mean Q, S, H
+ Mean Q, Variable H, S
99
98
95
90
85
80
70
60
50
40
30
20
15
10
10 100 1,000 10,000
Concentration (pg/m3)
Figure 47. Frequency Distribution of SCIM Calculated and Measured 24-Hour Concentration* for New York Ctty Station #10
-73-
-------
I
I
t
I
I
I
f
I
1
I
I
I
I
I
1
I
I
I
I
100
10,000
1,000
Concentration (ng/m3)
Figure 48. Frequency Distribution of SCIM Calculated and Measured 24-Hour Concentrations for New York City Station #14
A/
i
Observed
• Variable Q, S, H
x Mean Q, S, H
+ Mean Q, Variable H, S
99
98
95
90 o
85 H,
o»
80 2
n>
70 ^
60 §
50 %
o
40 ~*
o
o
30 £
20 =
o>
15 _
10 ""
10 100 1,000 10,000
Concentration (yg/m3)
Figure 49. Frequency Distribution of SCIM Calculated and Measured 24-Hour Concentradoni for New York Ctty Station #17
-74-
-------
Key
8 Observed
Variable Q, S, H
* Mean Q, S, H
Mean Q, Variable H. S
99
98
95
90 o
85 |
80 ?
70
60
50
40
30
20
15
10
3
JD
C
ro
10
100
10,000
1,000
Concentration (tjg/rn ^
: SO. Frequency Distribution of SCIM Calculated and Measured 24-Hour Concentrations forNew York City Station #27
^
4-
m
Key
e Observed
• Variable Q, S, H
x Mean Q, S, H
+ Mean Q, Variable H, S
99
98
90 r>
» I1
80 ?
70
50
30
o
-*
o
20 8
15 S
10 "^
10 100 . 1,000 10,000
Concentration (pg/m3)
Figure 51, Frequency Distribution of SCIM Calculated and Measured 24-Hour Concentration! forNew York City Station #28
-75-
-------
I
I
I
I
I
I
I
I
I
I
I
I
t
I
I
I
I
I
I
9 Observed
» Variable Q, S, H
x Mean Q, S, H
+ Mean Q, Variable H, S
10,000
Concentration
Figure 52. Frequency Distribution of SCIM Calculated and Measured 24-Hour Concentrations for New York City Station #31
/
/
/
/
/
/
/
7-
-H*
A
/
/
/
//
/
/
I
1
I /
f Jy
//
/ i
$
> - -
r
/
7
//
///
/
i
7
/
/f
Y
'/
9
t
X
t
,7
//
Ubse
Varf
Mean
Mean
Key
rved
able Q, S, H
Q. S, H
Q, Variable H,
S
0 100 1,000
10,
QS
90
DC
70
40
30
20
1 C
10
5
000
•<
o
Concentration (pg/m3)
Figure 53. Frequency Distribution of SCIM Calculated and Measured 24-Hour Concentrations for New York City Station #36
-76-
-------
1
I
1
1
1
1
I
1
1
I
I
1
1
1
1
1
1
1
1
t
1
0
Figure 54.
/
r
J i
/ 1
100
Frequency Distribution of CHM
/
r/
/
/ /
//
1 j
1
I
/
i
/i
7
[
/I
I
?
I
!f\
1
1
Concentration (yg/
Calculated and Measured 24-Hour
- -- - -fd- <•>
4
/
//
//
1
I
II"
/
//
il
/
i
!
(I
1
'/
/
/
Key
% Observe
+ With PC
t Without
d
int Sour
Point S
ces
ource
s
- 99
- 95
90
85
- 70
-60
50
-- DU v
40
- 30
20
- 15
10 '
5
000 10,000
»3)
SO Concentradoni for N«w York City Station 10
. QQ
Key
•
Observed
With Poin
Without P
t Sou
Dint
rces
Sour
ces
95
- 85 1.
DJ
on rf
IV
- 70 -n
00 fj
- 50 •<
o
o
o
- 30 S
T
15 "
- 10 ^
. 5
. 2
0 100 1,000 10,000
Concentration (pg/m3)
ire 55. Frequency Distribution of CUM Calculated and Measured 24-Hour SO,, Concentration! for New York City Station 11
-77-
-------
1
1
1
I
1
1
1
1
1
1
1
1
1
1
1
I
t
1
1
0
_
~
of
1
ft
// \
///
-.- -.
'/
1
VI
1
/
r
—
/
/ j
7
f
A
It
t
I
1
PI
Key
9 Obse
* With
• With
rved
Poin
3.Ut P
t So
j1nt
ur(
Sc
.es
>ur
ces
99
- - 95
-. 90
o
-.85 I
- - 80 rt
1
--70 -„
,n *
n
- - 50 D
-. AO °
o
30 S
20 1
15 8
..10 =
. . 5
100 1,000 10,000
Concentration (ug/m3)
Figure 56. The Frequency Distribution of 24-Hour SO2 Concentrations for the Two Sets
of CHM Calculation and Measured Values for New York City Station 13
. ^ A oa
//
'
^
/
/ ;
//
/
/
/
I
J
*
it
ji
f
k
1
/
i /
f
%
^
1
Key
Observed
With Pol
Without
nt So
Point
urces
Sources
95
90 o
85 —
03 o.
sn 2
80 1
70 7
en ' S
50 ^
O
40 •*
0
o
30 0
-i
3
20 g
' 8
15 ~-
M
10 ~"
5
0 100 1,000 10,000
Concentration (yg/ra3)
Figure 57. The Frequency Distribution of 24-Hour SO2 Concentrations for the Two Sets
of CHM Calculations and Measured Values for New York City Sudon 110
-78-
-------
1
1
1
1
1
1
1
1
1
I
I
t
t
1
1
1
1
0
1
I
/
1 1
J
:
Figure 38. Tl
of CM
. . _
/
/
i
J
/.
II
1 I
II
1 4
1 I
T y
? f
I
I
LOO
(
ie Frequency I
A Calculatlonl
/
/
/
/ /
/ //
III
ur SOj Conceatradona, foi the Two Sect
el for New York City Station 114
on
9
!
Key
Observed
With Point
Without Pol
Sourc
nt Sc
es
urc
es
. 95
90
• o
. 85 i
80 £
70 -n
,0 !
60 |
50 4J
. 40 °
,0 i?
20 a
o
. 15 "
. 10 ^
s
10 100 . 1,000 '• 10,000
t Concentration (pg/m3)
Figure 59. The Frequency Distribution of 24-Hour SOj Concentration forth* Two Sett
of CHM Calculation! and Measured Values for New York City Station 117
I
1
-------
I
I
I
I
I
I
I
I
1
I
t
I
I
I
I
I
I
I
I
10
TL
Key
t Observed
+ With Point Sources
• Without Point Sources
99
98
95
90
85
80
70
60
50
40
30
20
15
10
2
§
100 1,000
Concentration (ug/m3)
10,000
Figure 60. The Frequency Dlftrtbutloii of 24-IJour SOj Concentnttou.fcn tin Two Sea
of CHM Calculationi and Meuured Valuei for New Yolk City Station *27
if
J
1
1
7^
-4
/
7
—r ~ 4-
1 t
4 -1
7
/(_
I
II
1!
I
1 1 j
/
I
/
/•
1
t
8
+
•
Key
Observed
With Point Sources
Without Point Sourc
0 wo i.ooo
es
- - 80
- - 60
- • 50
10,000
3
J3
3
Concentration (ii9/m3)
Figure 61. The Frequency Dlltrltuition of 24-Hour SOj ConccntnUonl fcr tha TWo S«n
of CUM Calculatloni anJ Meaurcd Valuel tor New York City Station 128
-80-
-------
r
99
98
95
90
85
80
70
60
50
40
30
20
15
10
10
10
Key
8 Observed
+ With Point Sources
• Without Point Sources
100
1,000
Concentration
Figure 62. The Frequency Distribution of 24-Hour SO2 Concentration! for.the Two Sell
of CHM Calculation! uil Meaiured Value) tor New York City Station »31
100 1,000
Concentration (ug/m3)
Figure 63. The Frequency Distribution of 24-Hour SOj Concentrationl forth* Two Sett
of CUM Calculation! and Measured Value! forNcw York City Station 136
10,000
1
1
1
I
1
1
1
1
/
™
/
/
4
f
I
P
,
'
/
/
1
/
/
'
;
f
J
/
j
/
f /
/
/
/
/
/
It
1
I
r
1
8
+
•
Key
Observed
H1th Point
Without Poi
Sourc
nt So
es
urces
- - 95
- 90
85
80
- . 70
50
. . 40
. . 30
. . 20
. . 15
10
- - 5
.. 2
10,000
c
t
3
-81-
-------
Table 9. Model-to-Measurement Comparisons of 24-Hour NYC SC>2 Concentrations
by Mean, Standard Deviation, RMSE, MAE and Error at Maximum Measurement
Number of Comparisons
^
4
DO
a
S
a
\n
a
rt *-^
(tj QJcf)
I 3 6
•* 3 -5.
£ CT< 60
U \J\ w
o 3»
ps
SnP
f— 1 s
< A
a «-
rt O
dj c
.1 "«
X 3 ^~.
rt «ro^
rt S ~J0
t-> P ^>
fc 3
w S
Measured
SCIM ( Variable Q, H, S)
SCIM (MeanQ, Var. H, S)
SCIM (MeanQ, H, S)
GHM (without points)
GHM (with points)
Measured
SCIM ( Variable Q, H, S)
SCIM (MeanQ, Var. H, S)
SCIM (MeanQ, H, S)
GHM (without points)
GHM (with points)
SCIM ( Variable Q, H, S)
SCIM (MeanQ, Var. H, S)
SCIM (MeanQ, H, S)
GHM (without points)
GHM (with points)
SCIM (Variable Q, H, S)
SCIM (MeanQ, Var. H, S)
SCIM (MeanQ, H, S)
GHM (without points)
GHM (with points)
Maximum Measured (/ig/m )
SCIM ( Variable Q, H, S)
SCIM (MeanQ, Var. H, S)
SCIM (MeanQ, H, S)
GHM (without points)
GHM (with points)
NYC Station Number
0
364
303
421
440
283
307
350
163
241
258
106
119
121
247
314
192
143
150
183
242
161
113
121
917
118
41
-528
-555
-443
1
213
212
281
289
159
196
227
139
276
276
106
83
92
261
284
164
124
121
164
199
129
84
87
933
-203
-267
-619
-715
-628
3
257
292
367
384
246
268
310
163
246
274
112
111
115
274
309
194
190
184
175
206
140
135
134
1313
-1112
-1129
-1208
-1125
-1117
10
268
420
314
324
226
375
429
220
188
186
99
159
166
271
300
313
226
227
196
224
232
157
165
1613
-1025
-1166
-1300
-804
-749
14
266
229
170
180
93
106
131
98
137
150
47
43
45
133
148
161
151
126
106
118
138
126
103
607
-161
-142
-499
-485
-444
17
285
194
193
197
101
130
169
89
150
148
58
52
64
150
161
137
111
95
112
115
109
86
72
674
-451
-433
-605
-520
-480
27
281
185
211
211
118
128
156
98
160
159
60
52
55
158
179
126
108
95
107
125
100
83
72
547
-235
-286
-380
-249
-223
28
233
126
99
102
49
66
90
74
115
110
30
25
31
120
123
108
91
81
84
90
83
68
58
446
-284
-318
-384
-352
-333
31
283
186
225
229
122
99
158
76
213
202
63
40
72
212
204
105
115
87
113
122
82
96
64
474
-359
-375
-375
-376
-316
36
245
144
169
178
102
37
96
80
141
156
69
14
57
141
162
100
130
100
94
106
75
107
76
413
-301
-286
-308
-326
-314
-82-