J?
fl
Technical Basis for the EPA's Development of
the Significant Impact Thresholds for PM2.5
and Ozone
-------
-------
EPA-454/R-18-001
April 2018
Technical Basis for the EPA's Development of the Significant Impact Thresholds for
PM2.5 and Ozone
U.S. Environmental Protection Agency
Office of Air Quality Planning and Standards
Air Quality Assessment Division
Research Triangle Park, NC
-------
Contents
1 Bootstrap examples 6
2 Ozone results 12
2.1 2013-2015 ozone bootstrap results 12
2.2 2012-2014 ozone bootstrap results 15
2.3 2011-2013 ozone bootstrap results 18
2.4 2010-2012 ozone bootstrap results 21
2.5 2009-2011 ozone bootstrap results 24
2.6 2008-2010 ozone bootstrap results 27
2.7 2007-2009 ozone bootstrap results 30
2.8 2006-2008 ozone bootstrap results 33
2.9 2005-2007 ozone bootstrap results 36
2.10 2004-2006 ozone bootstrap results 39
2.11 2003-2005 ozone bootstrap results 42
2.12 2002-2004 ozone bootstrap results 45
2.13 2001-2003 ozone bootstrap results 48
2.14 2000-2002 ozone bootstrap results 51
3 Air quality variability results for years 2002-2013 for PM2.5 54
3.1 2013-2015 PM2.5 bootstrap results 54
3.2 2012-2014 PM2.5 bootstrap results 59
3.3 2011-2013 PM2.5 bootstrap results 64
3.4 2010-2012 PM2.5 bootstrap results 69
3.5 2009-2011 PM2.5 bootstrap results 74
3.6 2008-2010 PM2.5 bootstrap results 79
3.7 2007-2009 PM2.5 bootstrap results 84
3.8 2006-2008 PM2.5 bootstrap results 89
3.9 2005-2007 PM2.5 bootstrap results 94
3.10 2004-2006 PM2.5 bootstrap results 99
3.11 2003-2005 PM2.5 bootstrap results 104
3.12 2002-2004 PM2.5 bootstrap results 109
3.13 2001-2003 PM2.5 bootstrap results 114
3.14 2000-2002 PM2.5 bootstrap results 119
4 Comparison plots of nearby sites 124
5 Comparison of air quality variability for ozone sensitivity tests 208
5.1 All available data, no quarterly subsets 208
5.2 All available data, with quarterly subsets 210
6 Analysis of temporal lag on ozone data and results from a blocked bootstrap sensitivity
analysis 212
6.1 Analysis procedure and results 212
7 Results from cluster analyses and other spatial groupings 218
7.1 Cluster analyses 218
7.1.1 Cluster analysis with latitude, longitude, and variability values 218
7.1.2 Cluster analysis with time series of variability values (2014-2016) 223
7.1.3 Cluster analysis with time series of variability values (2012-2016) 227
7.2 Spatial analysis using NOAA Climate Regions 231
1
-------
List of Figures
1 Example from site 10732003 7
2 Example from site 21700008 8
3 Example from site 60195001 9
4 Example from site 481410053 10
5 Example from site 560210001 11
6 Summary of ozone bootstrap results for 2015 13
7 Distribution of ozone bootstrap results for 2015 14
8 Summary of ozone bootstrap results for 2014 16
9 Distribution of ozone bootstrap results for 2014 17
10 Summary of ozone bootstrap results for 2013 19
11 Distribution of ozone bootstrap results for 2013 20
12 Summary of ozone bootstrap results for 2012 22
13 Distribution of ozone bootstrap results for 2012 23
14 Summary of ozone bootstrap results for 2011 25
15 Distribution of ozone bootstrap results for 2011 26
16 Summary of ozone bootstrap results for 2010 28
17 Distribution of ozone bootstrap results for 2010 29
18 Summary of ozone bootstrap results for 2009 31
19 Distribution of ozone bootstrap results for 2009 32
20 Summary of ozone bootstrap results for 2008 34
21 Distribution of ozone bootstrap results for 2008 35
22 Summary of ozone bootstrap results for 2007 37
23 Distribution of ozone bootstrap results for 2007 38
24 Summary of ozone bootstrap results for 2006 40
25 Distribution of ozone bootstrap results for 2006 41
26 Summary of ozone bootstrap results for 2005 43
27 Distribution of ozone bootstrap results for 2005 44
28 Summary of ozone bootstrap results for 2004 46
29 Distribution of ozone bootstrap results for 2004 47
30 Summary of ozone bootstrap results for 2003 49
31 Distribution of ozone bootstrap results for 2003 50
32 Summary of ozone bootstrap results for 2002 52
33 Distribution of ozone bootstrap results for 2002 53
34 Summary of bootstrap results for 2015 55
35 Distribution of bootstrap results for 2015 56
36 Scatterplot of 50% CI bootstrap results for 2015 57
37 Map of the 50% CI bootstrap results for 2015 58
38 Summary of bootstrap results for 2014 60
39 Distribution of bootstrap results for 2014 61
40 Scatterplot of 50% CI bootstrap results for 2014 62
41 Map of the 50% CI bootstrap results for 2014 63
42 Summary of bootstrap results for 2013 65
43 Distribution of bootstrap results for 2013 66
44 Scatterplot of 50% CI bootstrap results for 2013 67
45 Map of the 50% CI bootstrap results for 2013 68
46 Summary of bootstrap results for 2012 70
47 Distribution of bootstrap results for 2012 71
48 Scatterplot of 50% CI bootstrap results for 2012 72
49 Map of the 50% CI bootstrap results for 2012 73
50 Summary of bootstrap results for 2011 75
51 Distribution of bootstrap results for 2011 76
52 Scatterplot of 50% CI bootstrap results for 2011 77
2
-------
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
78
80
81
82
83
85
86
87
88
90
91
92
93
95
96
97
98
100
101
102
103
105
106
107
108
110
111
112
113
115
116
117
118
120
121
122
123
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
Map of the 50% CI bootstrap results for 2011
Summary of bootstrap results for 2010
Distribution of bootstrap results for 2010
Scatterplot of 50% CI bootstrap results for 2010
Map of the 50% CI bootstrap results for 2010
Summary of bootstrap results for 2009
Distribution of bootstrap results for 2009
Scatterplot of 50% CI bootstrap results for 2009
Map of the 50% CI bootstrap results for 2009
Summary of bootstrap results for 2008
Distribution of bootstrap results for 2008
Scatterplot of 50% CI bootstrap results for 2008
Map of the 50% CI bootstrap results for 2008
Summary of bootstrap results for 2007
Distribution of bootstrap results for 2007
Scatterplot of 50% CI bootstrap results for 2007
Map of the 50% CI bootstrap results for 2007
Summary of bootstrap results for 2006
Distribution of bootstrap results for 2006
Scatterplot of 50% CI bootstrap results for 2006
Map of the 50% CI bootstrap results for 2006
Summary of bootstrap results for 2005
Distribution of bootstrap results for 2005
Scatterplot of 50% CI bootstrap results for 2005
Map of the 50% CI bootstrap results for 2005
Summary of bootstrap results for 2004
Distribution of bootstrap results for 2004
Scatterplot of 50% CI bootstrap results for 2004
Map of the 50% CI bootstrap results for 2004
Summary of bootstrap results for 2003
Distribution of bootstrap results for 2003
Scatterplot of 50% CI bootstrap results for 2003
Map of the 50% CI bootstrap results for 2003
Summary of bootstrap results for 2002
Distribution of bootstrap results for 2002
Scatterplot of 50% CI bootstrap results for 2002
Map of the 50% CI bootstrap results for 2002
150031001
180190006
180970078
190450019
220330009
271630447
320310016
350010023
420950025
421010047
421010055
440070022
490353006
100032004
110010043
130670003
150011006
Comparison
of
PM2.5
data
for
sites
150031001
and
Comparison
of
PM2.s
data
for
sites
180190006
and
Comparison
of
PM2.s
data
for
sites
180970078
and
Comparison
of
PM2.s
data
for
sites
190450019
and
Comparison
of
PM2.s
data
for
sites
220330009
and
Comparison
of
PM2.s
data
for
sites
271630447
and
Comparison
of
PM2.s
data
for
sites
320310016
and
Comparison
of
PM2.s
data
for
sites
350010023
and
Comparison
of
PM2.s
data
for
sites
420950025
and
Comparison
of
PM2.s
data
for
sites
421010047
and
Comparison
of
PM2.s
data
for
sites
421010055
and
Comparison
of
PM2.s
data
for
sites
440070022
and
Comparison
of
PM2.s
data
for
sites
490353006
and
Comparison
of
PM2.s
data
for
sites
100032004
and
Comparison
of
PM2.s
data
for
sites
110010043
and
Comparison
of
PM2.s
data
for
sites
130670003
and
Comparison
of
PM2.s
data
for
sites
150011006
and
3
-------
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
Comparison o
Comparison o
Comparison o
Comparison o
Comparison o
Comparison o
Comparison o
Comparison o
Comparison o
Comparison o
Comparison o
Comparison o
Comparison o
Comparison o
Comparison o
Comparison o
Comparison o
Comparison o
Comparison o
Comparison o
Comparison o
Comparison o
Comparison o
Comparison o
Comparison o
Comparison o
Comparison o
Comparison o
Comparison o
Comparison o
Comparison o
Comparison o
Comparison o
Comparison o
Comparison o
Comparison o
Comparison o
Comparison o
Comparison o
Comparison o
Comparison o
Comparison o
Comparison o
Comparison o
Comparison o
Comparison o
Comparison o
Comparison o
Comparison o
Comparison o
Comparison o
Comparison o
Comparison o
Comparison o
PM2.5 data for
PM2.5 data for
PM2.5 data for
PM2.5 data for
PM2.5 data for
PM2.5 data for
PM2.5 data for
PM2.5 data for
PM2.5 data for
PM2.5 data for
PM2.5 data for
PM2.5 data for
PM2.5 data for
PM2.5 data for
PM2.5 data for
PM2.5 data for
PM2.5 data for
PM2.5 data for
PM2.5 data for
PM2.5 data for
PM2.5 data for
PM2.5 data for
PM2.5 data for
PM2.5 data for
PM2.5 data for
PM2.5 data for
PM2.5 data for
PM2.5 data for
PM2.5 data for
PM2.5 data for
PM2.5 data for
PM2.5 data for
PM2.5 data for
PM2.5 data for
PM2.5 data for
PM2.5 data for
PM2.5 data for
PM2.5 data for
PM2.5 data for
PM2.5 data for
PM2.5 data for
PM2.5 data for
PM2.5 data for
PM2.5 data for
PM2.5 data for
PM2.5 data for
PM2.5 data for
PM2.5 data for
PM2.5 data for
PM2.5 data for
PM2.5 data for
PM2.5 data for
PM2.5 data for
PM2.5 data for
sites
150011012
sites
150012016
sites
150031001
sites
150032004
sites
180190006
sites
180970078
sites
190450019
sites
191032001
sites
191390015
sites
211110051
sites
220330009
sites
240150003
sites
240251001
sites
240290002
sites
240313001
sites
240330030
sites
261630001
sites
270031002
sites
270530963
sites
271630447
sites
290370003
sites
290470005
sites
290990019
sites
291893001
sites
295100007
sites
300490004
sites
300630024
sites
310550019
sites
320310016
sites
330050007
sites
330150018
sites
340171003
sites
340210008
sites
350010023
sites
360810124
sites
380570004
sites
420010001
sites
420030008
sites
420070014
sites
420110011
sites
420410101
sites
420450002
sites
420710007
sites
420910013
sites
420950025
sites
421010047
sites
421010055
sites
421250005
sites
421250200
sites
421255001
sites
421290008
sites
421330008
sites
440030002
sites
440070022
and
150011012
and
150012016
and
150031001
and
150032004
and
180190006
and
180970078
and
190450019
and
191032001
and
191390015
and
211110051
and
220330009
and
240150003
and
240251001
and
240290002
and
240313001
and
240330030
and
261630001
and
270031002
and
270530963
and
271630447
and
290370003
and
290470005
and
290990019
and
291893001
and
295100007
and
300490004
and
300630024
and
310550019
and
320310016
and
330050007
and
330150018
and
340171003
and
340210008
and
350010023
and
360810124
and
380570004
and
420010001
and
420030008
and
420070014
and
420110011
and
420410101
and
420450002
and
420710007
and
420910013
and
420950025
and
421010047
and
421010055
and
421250005
and
421250200
and
421255001
and
421290008
and
421330008
and
440030002
and
440070022
4
-------
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
196
197
198
199
200
201
202
203
204
205
206
207
209
211
214
215
216
217
221
222
225
226
229
230
232
Comparison
of
PM2.5
data
for
sites
450190048
and
450190048
Comparison
of
PM2.s
data
for
sites
450450015
and
450450015
Comparison
of
PM2.s
data
for
sites
450630008
and
450630008
Comparison
of
PM2.s
data
for
sites
482011035
and
482011035
Comparison
of
PM2.s
data
for
sites
490353006
and
490353006
Comparison
of
PM2.s
data
for
sites
490490002
and
490490002
Comparison
of
PM2.s
data
for
sites
490570002
and
490570002
Comparison
of
PM2.s
data
for
sites
530530029
and
530530029
Comparison
of
PM2.s
data
for
sites
530610005
and
530610005
Comparison
of
PM2.s
data
for
sites
530610020
and
530610020
Comparison
of
PM2.s
data
for
sites
530611007
and
530611007
Comparison
of
PM2.s
data
for
sites
550090005
and
550090005
Summary of ozone bootstrap results for 2013
Summary of ozone bootstrap results for 2013
Summary of ozone acf analysis for 2016
Summary of ozone pacf analysis for 2016
Summary of ozone blocked-bootstrap results for 2016
Distribution of ozone blocked-bootstrap results for 2016
Hierarchical and K-means clusters for the annual variability for 2016 . . .
Hierarchical and K-means clusters for the 24-hr variability for 2016 . . . .
Hierarchical and K-means clusters for the annual variability for 2014-2016
Hierarchical and K-means clusters for the 24-hr variability for 2014-2016 .
Hierarchical and K-means clusters for the annual variability for 2012-2016
Hierarchical and K-means clusters for the 24-hr variability for 2012-2016 .
Comparison of variability within NO A A climate regions for 2016
5
-------
1 Bootstrap examples
Bootstrap examples from selects PM2.5 sites for the 2008-2010 DV period. Top left, top right, and middle
left plots show the distribution of daily PM concentrations for 2008, 2009, and 2010, respectively. The
vertical red line shows the annual mean and the vertical blue line shows the annual 98th percentile. Middle
left plots show sample distributions of resampled data from 2008, along with the annual mean and the 98th
percentile from each resample. The bottom left plots show the distribution of the annual DVs from the
20,000 resampled DV periods (2008-2010). The bottom right plots show the distribution of the 24-hr DVs
from the 20,000 resampled DV periods (2008-2010)
6
-------
measurements, n = 360
ฆf 0.050-
c
0
~o
0.025-
ฆf0.050-
c
CD
T3
0.025 ฆ
0.000-
i
0 10 20 30 40
PM2.5 (ug/m3)
2010 PM measurements, n = 356
boot samples
2008
0.075-
PM measurements, n = 360
I I
o.ooo-
0.075-
>>
~ 0.050 -
c
0
"O
0.025-
0.000-
2009 PM
0.075-
10 20 30
PM2.5 (ug/m3)
0.06
w 0.04
c
0
ฆO
0.02
10 20 30
PM2.5 (ug/m3)
10 20 30
PM2.5 (ug/m3)
PM annual design value boot results
PM 24-hr design value boot results
900-
600-
300-
0-
400-
c
d
o
ฐ 200-
0-
-J
tk
12.0 12.5 13.0
PM2.5 (ug/m3)
13.5
25.0 27.5 30.0
PM2.5 (ug/m3)
32.5
Figure 1: Example from site 10732003.
7
-------
2008 PM measurements, n = 110
0.15-
ฆ0.10-
cn
c
0
"O
0.05-
0.00-
tnfbrrr
-m n n nn
i i i i
0 10 20 30
PM2.5 (ug/m3)
2010 PM measurements, n = 102
0.20-
_>>0.15 -
'55
ง0.10-
"D
0.05-
0.00-
ฆ
y
K n JML
-Qฆ1 J~1
l l l l l
0 10 20 30 40
PM2.5 (ug/m3)
0.10-
w
c
0
-D0.05-
0.00-
2009 PM measurements, n = 88
I
10 20 30
PM2.5 (ug/m3)
boot samples
0.15-
>>0.10 ฆ
" CO
c
0
0.05 ฆ
0.00-
i i i
10 20 30
PM2.5 (ug/m3)
500-
400-
300-
8 200-
100-
PM annual design value boot results
0-
Tin
5 6 7 8
PM2.5 (ug/m3)
PM 24-hr design value boot results
300-
200-
c
d
o
o
100-
o-
i i i i
20 25 30 35
PM2.5 (ug/m3)
Figure 2: Example from site 21700008.
8
-------
2008 PM measurements, n = 107
0.075-
w
0.025- /
0.000-
i i
20 40
PM2.5 (ug/m3)
0.06-
>>
ฆti 0.04 ฆ
w
c
CD
ฆO
0.02-
0.00-
2009 PM measurements, n = 295
I
I
I
If M -tUfliu
a d
i i i i
0 20 40 60
PM2.5 (ug/m3)
2010 PM measurements, n = 357
I
0.075-
<*> 0.050 -
c
0
"D
0.025-
0.000-
i i i i i
0 20 40 60 80
PM2.5 (ug/m3)
0.04-
>>0.03 ฆ
'(A
0 0.02 ฆ
ฆo
0.01 ฆ
o.oo-
boot samples
II
iii i
10 20 30 40
PM2.5 (ug/m3)
400-
300-
-4"
5 200-
O
O
100-
o-
PM annual design value boot results
'ฆฆill"
Tthw
16 17 18 19
PM2.5 (ug/m3)
PM 24-hr design value boot results
300-
| ?00-
o
o
100-
0-
""ftfff
i i i i
40 45 50 55
PM2.5 (ug/m3)
Figure 3: Example from site 60195001.
-------
0.09-
>>
"<75 0.06 -
c
0
"O
0.03-
0.00-
2008 PM measurements, n = 56
II
I
I
I
m
10
20
i
30
40
PM2.5 (ug/m3)
i
50
0.125-
0.100-
-^0.075-
w
c
.g 0.050-
0.025 ฆ
0.000-
2009 PM measurements, n = 57
I
I
I
m " n
i i i i i
10 20 30 40 50
PM2.5 (ug/m3)
0.06
^>0.04
w
c
CD
T3
0.02
0.00
2010 PM measurements, n = 49
0.15-
ฆ0.10-
cn
c
0
~o
0.05-
boot samples
i
10 20 30 40 50
PM2.5 (ug/m3)
o.oo-
i
i
20 40
PM2.5 (ug/m3)
300-
200-
4>
C
3
o
o
100-
o-
Figure 4: Example from site 481410053.
PM annual design value boot results
link.
i i i
14 16 18
PM2.5 (ug/m3)
PM 24-hr design value boot results
n
1
1 n
30 40 50
PM2.5 (ug/m3)
10
-------
c
0
"O
2008 PM measurements, n = 113
J
r
1
1
n v
/
/-
1
\
1
1
f
1
|
\
1
1
f"
1
hk
1
10
15
20
PM2.5 (ug/m3)
2010 PM measurements, n = 120
0.20-
>>0-15"
'55
ฉ0.10-
"D
0.05-
0.00-
\|
h-
0 4 8
PM2.5 (ug/m3)
12
1500-
PM annual design value boot results
1000-
c
o
o
500-
0-
4.0 4.5
PM2.5 (ug/m3)
2009 PM measurements, n = 117
0.20-
>>0-15 ฆ
"(/)
00.10-
ฆD
0.05-
0.00-
/
A
\
CL
^t~i
i
0
I I I
5 10 15
PM2.5 (ug/m3)
boot samples
i
20
0.20-
0.15-
>>
'55
c 0.10-
0
ฆo
0.05-
0.00-
ฆ i
ฆฆ
A /ง/11
11
II
/ Yf II
UN 11
^ 11
\l ""
II
if 11
/// II
\ 11
V 11
II
II
r 11
ฆ 1
r
i
10
15
20
PM2.5 (ug/m3)
PM 24-hr design value boot results
c
d
o
~
-
-|
rf
1
n-n
10
12
14
16
PM2.5 (ug/m3)
Figure 5: Example from site 560210001.
11
-------
2 Ozone results
Bootstrap results for ozone data from the years 2000-2013. Each section containts a single DV period, e.g.,
the results for 2015 include data from 2013-2015.
2.1 2013-2015 ozone bootstrap results
12
-------
CI limits
50% CI
68% CI
75% CI
95% CI
max/min
median
CI limits
50% CI
68% CI
75% CI
95% CI
max/min
median
Daily DV (ppb)
bootstrap metric
Figure 6: Bootstrap results for the ozone 2015 I >Vs. showing the 50%, 68%, 75%, and 95% CIs, along with
the mean and median bootstrap DVs. The top panel shows the DVs at the various CIs, the middle panel
shows the relative difference between the CI and the actual DV, and the bottom panel shows the distribution
of the relative differences between the CI and the actual DV.
13
-------
2015 bootstrap 50th percentile uncert
-i'
c
'ro
ฉ
o
a
CD
>
_ro
0
12.5%-
10.0%-
7.5% ฆ
5.0% ฆ
2.5% ฆ
0.0% ฆ
LONG
-80
-100
mean 1.391%
median 1.471%
-120
-140
40 60 80
4th 8-hr high NAAQS (ppb)
100
2015 bootstrap 50th percentile uncert
.Q
Q.
3 3
-i'
c
'co
ฆC 2
0
o
c
ฉ 1
Jt
o
w
CD
mean 0.922
median 1.000
40
60 80
4th 8-hr high NAAQS (ppb)
100
4th MDA8, rel uncert (%)
CD
50-
45-
40-
35-
30-
25-
4%
3%
2%
1%
0%
-120
-100
-80
long
Figure 7: Bootstrap results from the 50% CIs for the 2015 ozone DVs. The top panel shows the relative
difference between the CI and the actual DV, the middle panel shows the absolute difference between the
values for the DVs at each site and the CI, and the bottom panel shows the spatial distribution of the relative
difference between the 50% CIs for the 2015 ozone DV at each site.
14
-------
2.2 2012-2014 ozone bootstrap results
15
-------
40 60 80 100
Daily DV (ppb)
x; 100-
CL
CI limits
50% CI
68% CI
75% CI
95% CI
max/min
median
40 60 80 100
Daily DV (ppb)
CI limits
50% CI
68% CI
75% CI
95% CI
max/min
median
ii=|=
bootstrap metric
Figure 8: Bootstrap results for the ozone 2014 I >Vs. showing the 50%, 68%, 75%, and 95% CIs, along with
the mean and median bootstrap DVs. The top panel shows the DVs at the various CIs, the middle panel
shows the relative difference between the CI and the actual DV, and the bottom panel shows the distribution
of the relative differences between the CI and the actual DV.
16
-------
2014 bootstrap 50th percentile uncert
-i'
c
'ro
ฉ
o
a
CD
>
_ro
0
5% -
4% -
3% -
2% -
1%-
0%-
mean 1.491%
qgedian 1.471%
LONG
-80
-100
-120
-140
40
60 80
4th 8-hr high NAAQS (ppb)
100
2014 bootstrap 50th percentile uncert
.Q
Q.
3 3
-t'
c
'ro
ฆC 2
0
o
c
ฉ 1
o
w
CD
mean 1.020
median 1.000
40
60 80
4th 8-hr high NAAQS (ppb)
100
4th MDA8, rel uncert (%)
CD
50-
45-
40-
35-
30-
25-
4%
3%
2%
1%
0%
-120
-100
-80
long
Figure 9: Bootstrap results from the 50% CIs for the 2014 ozone DVs. The top panel shows the relative
difference between the CI and the actual DV, the middle panel shows the absolute difference between the
values for the DVs at each site and the CI, and the bottom panel shows the spatial distribution of the relative
difference between the 50% CIs for the 2014 ozone DV at each site.
17
-------
2011-2013 ozone bootstrap results
18
-------
CI limits
50% CI
68% CI
75% CI
95% CI
max/min
median
.Q
ง- 100-
>
Q
O. 75-
cc
s
o 50"
.Q
50 70 90 110
Daily DV (ppb)
70
Daily DV (ppb)
CI limits
50% CI
68% CI
75% CI
95% CI
max/min
median
0s
0)
O
c
E
5
fc
>
Q
25-
20-
15-
10-
5-
i i =4= cL
riii
i \ฐ
! \ฐ
r#5
. , v , ^ , \0'
^ ^ ^ jg
r#
1ฎ
I ^ I ^ I '
bootstrap metric
i i
-------
2013 bootstrap 50th percentile uncert
-i'
c
'ro
ฉ
o
a
CD
>
_ro
0
6% -
4% -
2% -
0%-
mean 1.534%
^median
1.449%
A
>
J "
_ *
y
m
LONG
-80
-100
-120
-140
50 70 90
4th 8-hr high NAAQS (ppb)
2013 bootstrap 50th percentile uncert
110
.Q
Q.
CL
-i'
C
'co
ฆc
0
o
c
ฉ
o
w
.Q
CD
9
mean
1.088
median 1.000
aaa
ป
f m !"
i Mil
l#9 99 9
m
ป
50
70
4th 8-hr high NAAQS (ppb)
90
110
4th MDA8, rel uncert (%)
CD
50-
45-
40-
35-
30-
25-
4%
3%
2%
1%
0%
-120
-100
-80
long
Figure 11: Bootstrap results from the 50% CIs for the 2013 ozone DVs. The top panel shows the relative
difference between the CI and the actual DV, the middle panel shows the absolute difference between the
values for the DVs at each site and the CI, and the bottom panel shows the spatial distribution ol the relative
difference between the 50% CIs for the 2013 ozone DV at each site.
20
-------
2.4 2010-2012 ozone bootstrap results
21
-------
125 -
X!
CL
Q. 100"
>
Q
Q.
CO
-4-J
W
o
o
.Q
75-
50-
50
70
90
CI limits
50% CI
68% CI
75% CI
95% CI
max/min
median
Daily DV (ppb)
0
o
c
2
a3
>
Q
20-
10-
o-
y
v,
4
70
Daily DV (ppb)
90
CI limits
50% CI
68% CI
75% CI
95% CI
max/min
median
0s
0)
O
c
E
5
fc
>
Q
25-
20-
15-
10-
5-
o-
,_\o ^ ^ ^
^ ^ ^
# . \<5*
^\o o\o
bootstrap metric
<$?+ ^
-------
2012 bootstrap 50th percentile uncert
-i'
c
'ro
ฉ
o
a
CD
>
_ro
0
6% -
4% -
2% -
0%-
mean 1
1.572%
^median 1.449%
m
LONG
-80
-100
-120
-140
50 70 90
4th 8-hr high NAAQS (ppb)
2012 bootstrap 50th percentile uncert
.Q
Q.
CL
-i'
C
'co
-e
0
o
c
ฉ
o
w
.Q
CD
mean
1.136
median 1.000
a aa m
4
4
MM
ป<
ป
>
"5
_
50
70 90
4th 8-hr high NAAQS (ppb)
4th MDA8, rel uncert (%)
CD
50-
45-
40-
35-
30-
25-
4%
3%
2%
1%
0%
-120
-100
-80
long
Figure 13: Bootstrap results from the 50% CIs lor the 2012 ozone DVs. The top panel shows the relative
difference between the CI and the actual DV, the middle panel shows the absolute difference between the
values lor the DVs at each site and the CI, and the bottom panel shows the spatial distribution ol the relative
difference between the 50% CIs lor the 2012 ozone DV at each site.
23
-------
2009-2011 ozone bootstrap results
24
-------
125 -
X!
CL
3 100-
>
Q
cl
CO
-4-ป
w
o
o
X!
75-
50-
70 90
Daily DV (ppb)
CI limits
50% CI
68% CI
75% CI
95% CI
max/min
median
110
20-
\\
m
70
Daily DV (ppb)
i \ฐ
^ . \o'
JS
1ฎN
1ฎ
rSV
I I \ฐ I
^ HP' qV
110
CI
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
25-
5^ on -
1
' ZD
0
c 15-
i
5-
Q
o-
5,1
1 _
' -
L J_ฑฑ Ad
r= i ^ n
-------
2011 bootstrap 50th percentile uncert
8% -
6% -
4% -
>
-i'
c
'ro
ฉ
o
a
=}
CD
>
ro 2% -
0
0%-
mean 1.557%
median 1.471%
50 70 90
4th 8-hr high NAAQS (ppb)
2011 bootstrap 50th percentile uncert
LONG
-80
-100
-120
-140
110
.a
Q.
Q.
= 4-
-ป'
c
'ro
ฆc
0
o
c
ฉ
-t
o
w
CD
CD
m
mean
1.097
median 1.000
a a
w
w
ป
50
70
4th 8-hr high NAAQS (ppb)
90
110
4th MDA8, rel uncert (%)
50-
45-
40-
35-
30-
25-
4%
3%
2%
1%
0%
-120
-100
-80
long
Figure 15: Bootstrap results from the 50% CIs lor the 2011 ozone DVs. The top panel shows the relative
difference between the CI and the actual DV, the middle panel shows the absolute difference between the
values lor the DVs at each site and the CI, and the bottom panel shows the spatial distribution ol the relative
difference between the 50% CIs lor the 2011 ozone DV at each site.
26
-------
2008-2010 ozone bootstrap results
27
-------
60 80
Daily DV (ppb)
25-
20-
0)
O
c
E
15-
5
10-
>
Q
5-
o-
icL4=
bootstrap metric
.Q
ง: 100-
Cl limits
50% CI
68% CI
75% CI
95% CI
max/min
median
T3
>
Q
10-
40 60 80 100
Daily DV (ppb)
CI limits
50% CI
68% CI
75% CI
95% CI
max/min
median
Figure 16: Bootstrap results for the ozone 2010 DVs, showing the 50%, 6851, 75%. and 95% CIs., along with
the mean and median bootstrap DVs. The top panel shows the DVs at the various CIs, the middle panel
shows the relative difference between the CI and the actual DV, and the bottom panel shows the distribution
of the relative differences between the CI and the actual DV.
28
-------
2010 bootstrap 50th percentile uncert
5% -
mean 1.558%
median 1.449%
4% -
>
-ป'
c
'ro
-e
g 3% -
c
CD 2% -
>
_ro
0
1%-
LONG
-80
-100
-120
-140
0%-
40
60 80
4th 8-hr high NAAQS (ppb)
100
2010 bootstrap 50th percentile uncert
.Q
Q.
3 3
-t'
c
'co
-e 2
0
o
c
ฉ 1
o
w
CD
mean 1.098
median 1.000
40
60 80
4th 8-hr high NAAQS (ppb)
100
4th MDA8, rel uncert (%)
CD
50-
45-
40-
35-
30-
25-
4%
3%
2%
1%
0%
-120
-100
-80
long
Figure 17: Bootstrap results from the 50% CIs lor the 2010 ozone DVs. The top panel shows the relative
difference between the CI and the actual DV, the middle panel shows the absolute difference between the
values lor the DVs at each site and the CI, and the bottom panel shows the spatial distribution ol the relative
difference between the 50% CIs lor the 2010 ozone DV at each site.
29
-------
2007-2009 ozone bootstrap results
30
-------
125 -
X!
CL
3 100-
>
Q
cl
CO
-4-ป
w
o
o
X!
75-
50-
75
Daily DV (ppb)
CI limits
50% CI
68% CI
75% CI
95% CI
max/min
median
>
Q
0s
0)
O
c
E
5
fc
>
Q
20-
15-
10-
5-
o-
25-
20-
15-
10-
5-
o-
oY^W
t
\X\
75
Daily DV (ppb)
i cL =T= i
w y MA
,\o- ,\o^
tp nP
100
ปs" y /
CI limits
50% CI
68% CI
75% CI
95% CI
max/min
median
<5*
4-
bootstrap metric
Figure 18: Bootstrap results lor the ozone 2009 DVs, showing the 50%,ฆ68%', 75%, and 95% CIs, along with
the mean and median bootstrap DVs. The top panel shows the DVs at the various CIs, the middle panel
shows the relative difference between the CI and the actual DV, and the bottom panel shows the distribution
of the relative differences between the. CI and the actual DV.
31
-------
2009 bootstrap 50th percentile uncert
-i'
'I 4%-
-e
ฉ
0
c
=}
1 2% -
"-4*
_ro
0
0%-
mean 1.ฃ70%
median 1.429%
LONG
-80
-100
-120
-140
50
75 100
4th 8-hr high NAAQS (ppb)
2009 bootstrap 50th percentile uncert
.Q
Q.
3 3
-t'
c
'ro
tr 2
0
o
c
ฉ 1
o
w
CD
mean 1.139
medtan 1.000
50
75
4th 8-hr high NAAQS (ppb)
100
4th MDA8, rel uncert (%)
ro
50-
45-
40-
35-
30-
25-
4%
3%
2%
1%
0%
-120
-100
-80
long
Figure 19: Bootstrap results from the 50% CIs lor the 2009 ozone DVs. The top panel shows the relative
difference between the CI and the actual DV, the middle panel shows the absolute difference between the
values lor the DVs at each site and the CI, and the bottom panel shows the spatial distribution of the relative
difference between the 50% CIs for the 2009 ozone DV at each site.
32
-------
2006-2008 ozone bootstrap results
33
-------
X!
120-
CL
Q.
*'
100-
>
Q
Q.
80-
CO
i
2
60-
o
o
.Q
40-
0
o
c
0)
0
t
>
Q
0s
0)
O
c
E
5
"D
>
Q
25-
20-
15-
10-
5-
o-
25-
20-
15-
10-
5-
Cl
75
Daily DV (ppb)
50
75
Daily DV (ppb)
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
CI
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
riii
o0x ^ -ฃป- *
, \ฐ , 0^ , \o , \
-------
2008 bootstrap 50th percentile uncert
mean 1.647%
median 1.471%
>>
"c
'ro
4% -
ฆe
ฉ
o
d
CD
>
2% -
CD
CD
LONG
-80
-100
-120
-140
0%-
50 75 100
4th 8-hr high NAAQS (ppb)
2008 bootstrap 50th percentile uncert
.Q
Q.
CL
-t'
C
'co
ฆc
0
o
c
ฉ
o
w
.Q
CD
mean 1.246 median 1.000
M
MNNW
ฆ
50 75
4th 8-hr high NAAQS (ppb)
100
4th MDA8, rel uncert (%)
ro
50-
45-
40-
35-
30-
25-
4%
3%
2%
1%
0%
-120
-100
-80
long
Figure 21: Bootstrap results from the 50% CIs lor the 2008 ozone DVs. The top panel shows the relative
difference between the CI and the actual DV, the middle panel shows the absolute difference between the
values lor the DVs at each site and the CI, and the bottom panel shows the spatial distribution ol the relative
difference between the 50% CIs lor the 2008 ozone DV at each site.
35
-------
2005-2007 ozone bootstrap results
36
-------
Daily DV (ppb)
25-
20-
0)
O
c
E
15-
5
10-
>
Q
5-
o-
bootstrap metric
X!
CL
Q.
>
Q
Q.
CO
100-
Cl limits
50% CI
68% CI
75% CI
95% CI
max/min
median
CI limits
50% CI
68% CI
75% CI
95% CI
max/min
median
75
Daily DV (ppb)
Figure 22: Bootstrap results for the ozone 2007 DVs, showing the 50%, 6851, 75%. and 95% CIs, along with
the mean and median bootstrap DVs. The top panel shows the DVs at the various CIs, the middle panel
shows the relative difference between the CI and the actual DV, and the bottom panel shows the distribution
of the relative differences between the CI and the actual DV.
37
-------
2007 bootstrap 50th percentile uncert
>>
"c
'ro
4% -
-e
ฉ
o
d
CD
>
2% -
-t'
CD
CD
mean 1
1.636%
median 1.449%
%
Ur*
iป _
o%-
25
50 75 100
4th 8-hr high NAAQS (ppb)
LONG
1-80
-100
-120
-140
125
2007 bootstrap 50th percentile uncert
.Q
g 4-
ฆjjj 3-
CD
tr
0
c 2-
ฉ
1 ฆ
o
w
-Q
ro o-
mean 1.267
median 1.000 <
25
50 75
4th 8-hr high NAAQS (ppb)
100
125
4th MDA8, rel uncert (%)
ro
50-
45-
40-
35-
30-
25-
-120
-100
-80
long
Figure 23: Bootstrap results from the 50% CIs for the 2007 ozone DVs. The top panel shows the relative
difference between the CI and the actual DV, the middle panel shows the absolute difference between the
values for the DVs at each site and the CI, and the bottom panel shows the spatial distribution of the relative
difference between the 50% CIs for the 2007 ozone DV at each site.
38
-------
2.10 2004-2006 ozone bootstrap results
39
-------
.Q
ฃ 120-
>
Q
Q.
CO
-4-J
W
o
o
.Q
90-
60-
m
40
60
80
Daily DV (ppb)
100
120
CI limits
50% CI
68% CI
75% CI
95% CI
max/min
median
n
80
Daily DV (ppb)
100
120
CI limits
50% CI
68% CI
75% CI
95% CI
max/min
median
25-
20-
0)
O
c
E
15-
5
10-
>
Q
5-
o-
riii
i \ฐ
! \ฐ
r#5
. , V , ^ , \0'
^ ^ ^ ^
r#
Iฉ
I ^ I ^ I '
^ ^0 o5b0,ฐ
bootstrap metric
i
ฅ
i i i ฆ
-------
-t'
c
'ro
ฉ
o
a
CD
>
_ro
0
2006 bootstrap 50th percentile uncert
5% - mean 1.655% npdian 1.449%
4% - * _
3% -
2% -
1%-
o%-
LONG
-80
-100
-120
-140
40
60 80 100
4th 8-hr high NAAQS (ppb)
120
2006 bootstrap 50th percentile uncert
.Q
Q.
CL
-t'
C
'ro
ฆc
0
o
c
ฉ
o
w
CD
1.262
mean
median*! ,0i0
40
60
80
4th 8-hr high NAAQS (ppb)
100
120
4th MDA8, rel uncert (%)
CD
50-
45-
40-
35-
30-
25-
4%
3%
2%
1%
0%
-120
-100
-80
long
Figure 25: Bootstrap results from the 50% CIs for the 2006 ozone DVs. The top panel shows the relative
difference between the CI and the actual DV, the middle panel shows the absolute difference between the
values for the DVs at each site and the CI, and the bottom panel shows the spatial distribution of the relative
difference between the 50% CIs for the 2006 ozone DV at each site.
41
-------
2.11 2003-2005 ozone bootstrap results
42
-------
.Q
9- 120 ฆ
cl
>
Q
Q.
CO
-4-J
W
o
o
X!
90-
60-
30-
75
Daily DV (ppb)
CI limits
50% CI
68% CI
75% CI
95% CI
max/min
median
0
o
c
0)
a3
'~o
>
Q
0s
0)
O
c
E
5
fc
>
Q
20-
15-
10-
5-
o-
25-
20-
15-
10-
5-
01
i \ฐ
50
i\i
75 100
Daily DV (ppb)
125
CI limits
50% CI
68% CI
75% CI
95% CI
max/min
median
. . . 1
4
i
2 1 T
b
d
L 4-4= 4-4
^VT
Li 1
?
i
-P\ฐ
\o
r#
Iฉ
<ฃ\ฐ
. . \o^
^ ^0 o5b0,ฐ
J? &
<$*>
bootstrap metric
Figure 26: Bootstrap results for the ozone 2005 DVs, showing the 50%, 68%, 75%. and 95% CIs., along with
the mean and median bootstrap DVs. The top panel shows the DVs at the various CIs, the middle panel
shows the relative difference between the CI and the actual DV, and the bottom panel shows the distribution
of the relative differences between the. CI and the actual DV.
43
-------
5% -
2005 bootstrap 50th percentile uncert
mean 1^50% median 1ป807%
>>
"c
4% -
'ro
-e
ฉ
o
3% -
d
CD
>
2% -
ro
0
1%-
LONG
-80
-100
-120
-140
50
75 100
4th 8-hr high NAAQS (ppb)
125
2005 bootstrap 50th percentile uncert
.Q
Q.
CL
-i'
C
'co
ฆc
CD
O
c
ฉ
o
w
.Q
CD
mean 1.435
median 4.5ฎงซ*
50
75 100
4th 8-hr high NAAQS (ppb)
125
4th MDA8, rel uncert (%)
ro
50-
45-
40-
35-
30-
25-
4%
3%
2%
1%
0%
-120
-100
-80
long
Figure 27: Bootstrap results from the 50% CIs for the 2005 ozone DVs. The top panel shows the relative
difference between the CI and the actual DV, the middle panel shows the absolute difference between the
values for the DVs at each site and the CI, and the bottom panel shows the spatial distribution of the relative
difference between the 50% CIs for the 2005 ozone DV at each site.
44
-------
2.12 2002-2004 ozone bootstrap results
45
-------
X!
CL
a
>
Q
Q.
CO
120 -
20-
0
o
c
2
a3
CI limits
50% CI
68% CI
75% CI
95% CI
max/min
median
CI limits
50% CI
68% CI
75% CI
95% CI
max/min
median
75
Daily DV (ppb)
50 75 100 125
Daily DV (ppb)
bootstrap metric
Figure 28: Bootstrap results for the ozone 2004 DVs, showing the 50%, 6851, 75%. and 95% CIs, along with
the mean and median bootstrap DVs. The top panel shows the DVs at the various CIs, the middle panel
shows the relative difference between the CI and the actual DV, and the bottom panel shows the distribution
of the relative differences between the CI and the actual DV.
46
-------
5% -
2004 bootstrap 50th percentile uncert
mean*1.869% median 1.829%
>>
"c
4% -
'ro
-e
ฉ
o
d
3% -
CD
>
2% -
ro
CD
1%-
LONG
-80
-100
-120
-140
50
75 100
4th 8-hr high NAAQS (ppb)
125
2004 bootstrap 50th percentile uncert
.Q
Q.
CL
-t'
C
'co
ฆc
CD
O
c
ฉ
o
w
.Q
CD
mean 1.496
median 1.5ฎ0ซซซ*1
50
75 100
4th 8-hr high NAAQS (ppb)
125
4th MDA8, rel uncert (%)
ro
50-
45-
40-
35-
30-
25-
4%
3%
2%
1%
0%
-120
-100
-80
long
Figure 29: Bootstrap results from the 50% CIs for the 2004 ozone DVs. The top panel shows the relative
difference between the CI and the actual DV, the middle panel shows the absolute difference between the
values for the DVs at each site and the CI, and the bottom panel shows the spatial distribution of the relative
difference between the 50% CIs for the 2004 ozone DV at each site.
47
-------
2.13 2001-2003 ozone bootstrap results
48
-------
40 60 80 100 120
Daily DV (ppb)
X!
CL
a
>
a
Q.
CO
120 -
CI limits
50% CI
68% CI
75% CI
95% CI
max/min
median
40 60 80 100 120
Daily DV (ppb)
CI limits
50% CI
68% CI
75% CI
95% CI
max/min
median
bootstrap metric
Figure 30: Bootstrap results for the ozone 2003 DVs, showing the 50%, 6851, 75%. and 95% CIs, along with
the mean and median bootstrap DVs. The top panel shows the DVs at the various CIs, the middle panel
shows the relative difference between the CI and the actual DV, and the bottom panel shows the distribution
of the relative differences between the CI and the actual DV.
49
-------
>
-t'
c
'ro
ฉ
o
a
CD
>
_ro
0
2003 bootstrap 50th percentile uncert
5% - mean 1.869%
4% - *
%
3% - %
2% - * ซ
1%- # t
0%-
median 1.818%
LONG
-80
-100
-120
-140
40
60 80 100
4th 8-hr high NAAQS (ppb)
120
2003 bootstrap 50th percentile uncert
.Q
Q.
CL
-t'
C
'ro
ฆc
0
o
c
ฉ
o
w
CD
mean 1.549
medtan 1 .0#
40
60 80 100
4th 8-hr high NAAQS (ppb)
120
4th MDA8, rel uncert (%)
CD
50-
45-
40-
35-
30-
25-
4%
3%
2%
1%
0%
-120
-100
-80
long
Figure 31: Bootstrap results from the 50% CIs for the 2003 ozone DVs. The top panel shows the relative
difference between the CI and the actual DV, the middle panel shows the absolute difference between the
values for the DVs at each site and the CI, and the bottom panel shows the spatial distribution of the relative
difference between the 50% CIs for the 2003 ozone DV at each site.
50
-------
2.14 2000-2002 ozone bootstrap results
51
-------
X!
CL
a
>
Q
Q.
CO
120 -
25-
Cp"
2-- 20-
0
S;ฑฑ4ปi-+A^A
i i i i i i i i
,i& 0e^ c0~ c&
, \ฐ^ , . \o^ . , \o^ ^ ^ , x#
rjb^ ^ ^ cf? of?
bootstrap metric
CI limits
50% CI
68% CI
75% CI
95% CI
max/min
median
CI limits
50% CI
68% CI
75% CI
95% CI
max/min
median
Daily DV (ppb)
50 75 100
Daily DV (ppb)
Figure 32: Bootstrap results for the ozone 2002 DVs, showing the 50%, 6851, 75%, and 95% CIs, along with
the mean and median bootstrap DVs. The top panel shows the DVs at the various CIs, the middle panel
shows the relative difference between the CI and the actual DV, and the bottom panel shows the distribution
of the relative differences between the CI and the actual DV.
52
-------
2002 bootstrap 50th percentile uncert
5% -
mean 1.815%
median 1.765%
E 4% -
ro
-e
ฉ
o 3% -
=}
CD
> 2% -
_ro
0
^ 1%-
LONG
-80
-100
-120
-140
50
75 100
4th 8-hr high NAAQS (ppb)
125
2002 bootstrap 50th percentile uncert
.Q
Q.
CL
-t'
C
'co
ฆc
0
o
c
ฉ
o
w
CD
mean 1.499
median 1.500ซซ*
50
75 100
4th 8-hr high NAAQS (ppb)
125
4th MDA8, rel uncert (%)
CD
50-
45-
40-
35-
30-
25-
4%
3%
2%
1%
0%
-120
-100
-80
long
Figure 33: Bootstrap results from the 50% CIs for the 2002 ozone DVs. The top panel shows the relative
difference between the CI and the actual DV, the middle panel shows the absolute difference between the
values for the DVs at each site and the CI, and the bottom panel shows the spatial distribution of the relative
difference between the 50% CIs for the 2002 ozone DV at each site.
53
-------
3 Air quality variability results for years 2002-2013 for PM2.5
Bootstrap results for PM2.5 data from the years 2000-2015. Each section containts a single DV period, e
the results for 2015 include data from 2013-2015.
3.1 2013-2015 PM2.5 bootstrap results
-------
Annual bootstrap results
Cl
CO
E
O)
>
Q
Q.
CD
L_
-4'
o
o
X!
Annual DV (ug/m3)
24-hr bootstrap results
ci
100-
75-
50-
25-
24-hr DV (ug/m3)
Difference between DVs and boot results, Annual NAAQ2?i
a)
o
c
2
0
'u
>
Q
40-
30-
20-
10-
0-
M
Annual DV (ug/m3)
Difference between DV and boot results, Daily NAAQS ci
VP
0s
0)
O
c
0
0
4
-a
>
a
60-
40-
20-
o-
Daily DV (ug/m3)
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
Pigure 34: Bootstrap results for the 2015 PMjDVs, showing the 50%, 68%, 75%, and 95งS GIs, along with
the mean and median bootstrap DVs. The top two panels show the values; for the DVs at the various GIs,
while the bottom two panels show the relative difference between the Gl and the actual DV.
55
-------
Annual NAAQS bootstrap summary
25-
20-
0
O 15-
c
0
0
^ 10-
T3
>
Q
5-
i
!
!
.
i ฆ
? _
1 .
1
1.
o-
I I I
00a
ง ฐ-
5 cl
3
LD
CN
LD
CN
O
ID
0
CL
Q_
O
ID
CO
CD
0
CL
CL
CO
CD
0 0
CL >
ง- o
ID
r^
LD
LD
C7)
0
CL
CL
ID
C7)
X
03
E
CD CD
0 --jj
E 0
E
c >
E S
ฆo'
4'
C/)
bootstrap metric
24-hr NAAQS bootstrap summary
0
O
C
0
0
u_
"O
>
Q
25-
20-
0 0
CL >
ง"
LD
CN
LD
CN
O
ID
0
CL
Q_
O
ID
0
CL
Q.
0 0
CL >
ง"
CO
CD
CO
CD
ID
r^
LD
r--
LD
CD
0
CL
CL
LD
cr>
X
CD
E
ฃ=
CD
0
E
ฃ=
CD
TD
0
E
"O
bootstrap metric
Figure 35: Bootstrap results for the 2015 PM2.5 DVs, showing distribution of the relative differences between
the bootstrap DVs and the actual DV at the 50%, 68%, 75%, and 95% CIs, along with the mean, median,
maximum, minimum, standard deviations of the relative differences.
56
-------
boot 50th percentile uncert, all sites
mean 1.705 %
boot 50th percentile uncert, all sites
5% -
median 1.648 %
= 3% -
5 10 15
annual NAAQS
boot 50th percentile uncert, all sites
mean 0.145
0.5-
*ie#ian 0.150
0.4-
>.
C
"TO
o 0.3-
C
3
C/)
-Q
03
0.2-
80% -
60% -
-)'
c
'ro
(D
O
ง 40%-
d)
>
-------
Annual NAAQS, rel uncert (%), all sites
50-
45-
40-
35-
30-
25-
5%
4%
3%
2%
1%
0%
-120
-100
long
24-hr NAAQS, rel uncert (%), all sites
-80
50-
45-
40-
35-
30-
25-
10.0%
7.5%
5.0%
2.5%
0.0%
-120
-100
-80
long
Figure 37: Spatial distribution of the relative difference between the C'J and the actual DY from the 50%
CIs for the 2015 PMa.s DVs.
58
-------
2012-2014 PM2.5 bootstrap results
59
-------
Annual bootstrap results
Cl
CO
E
100-
a>
75-
>
a
50-
Q.
CD
i_
25-
to
o
o
.a
a)
o
c
2
0
'u
>
Q
Annual DV (ug/m3)
24-hr bootstrap results
ci
24-hr DV (ug/m3)
Difference between DVs and boot results, Annual NAAQ2?i
40-
30-
20-
10-
0-
Annual DV (ug/m3)
Difference between DV and boot results, Daily NAAQS ci
VP
0s
0)
O
c
0
0
4
-a
>
a
60-
40"
20-
o-
Daily DV (ug/m3)
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
Figure 38: Bootstrap results for the 2014 I'DVs, showing the 50%, 68%, 75%, and 95SS GIs, along with
the mean and median bootstrap DVs. The top two panels show the values; for the DVs at the various GIs,
while the bottom two panels show the relative difference between the Gl and the actual DV.
60
-------
Annual NAAQS bootstrap summary
25-
20-
0
o 15-
c
0
i
0
it 10-
"O
>
Q
5-
o-
I
!
I
1
H
%
. i
. .
ป
1 _
L
L1
ฆ
r
J
L
4*
0
งฆ o
LO
CN
LO
CN
O
ID
1
1
1
l
1
0
0
0
0
0
CL
CL
Q_
Cl
_o
Q_
O
CL
ZS
=3
3
nP
O
0^
CO
O-
CO
0
LO
0s
LO
LO
CD
CD
h-
1^-
a>
0)
CL
CL
in
C7)
X
03
E
CD CD
0 --jj
E 0
E
c >
E S
ฆo'
4'
C/)
bootstrap metric
24-hr NAAQS bootstrap summary
Q
25-
20-
0 0
CL >
ง"
LO
CN
LO
CN
O
LO
CD
CL
Q_
O
LO
CD
CL
Q.
0 0
CL >
ง"
CO
CD
CO
CD
LO
r^
LO
r--
LO
CD
0
CL
Q.
LO
o>
X
CD
E
ฃ=
CD
0
E
ฃ=
CD
in
0
E
"O
c/)
bootstrap metric
Figure 39: Bootstrap results for the 2014 PM2.5 DVs, showing distribution of the relative differences between
the bootstrap DVs and the actual DV at the 50%, 68%, 75%, and 95% CIs, along with the mean, median,
maximum, minimum, standard deviations of the relative differences.
61
-------
boot 50th percentile uncert, all sites
boot 50th percentile uncert, all sites
5% -
>. 4% -
-I'
c
"iS
i
CD
o
c 3% -
=3
(D
>
jc
CD
2% -
1.684 %
mean
0
CO
CD
%
mediar
ป
\
t
,
\ V \
%
\
%
\
9
1%-
0.5-
i i i
5 10 15
annual NAAQS
boot 50th percentile uncert, all sites
mean 0.148
rfiedian 0.150
0.4-
O 0.3 -
mm m m
10 15
annual NAAQS
30% -
.E 20%-
i_
(D
O
C
=3
CD
>
p 10%-
v-* A
k549
r
riean h
D
median
4.286 %
V
It
0% -
20
40
24-hr NAAQS
60
boot 50th percentile uncert, all sites
6-
>.
c
'to
-c.
^ /I
o 4-
ฃ=
13
CO
-Q
CD
2-
0-
V
I.021
II.000
mean
median
a
A
i
m
w
W
4
mm* m
mm m
mm
---
A MA A JMA
V W W HV
20
40
24-hr NAAQS
60
Figure 40: Bootstrap results from the 50% CIs for PM2.5 DVs. The top two panels show the relative difference
between the CI and the actual DV and the bottom two panels show the absolute difference between the values
for the DVs at each site and the CI.
62
-------
Annual NAAQS, rel uncert (%), all sites
50-
45-
40-
35-
30-
25-
5%
4%
3%
2%
1%
0%
-120
-100
-80
long
24-hr NAAQS, rel uncert (%), all sites
50"
45-
40-
35-
30-
25-
10.0%
7.5%
5.0%
2.5%
0.0%
-120
-100
-80
long
Figure 41: Spatial distribution of the relative difference between the C'J and the actual DY from the 50%
CIs for the 2014 PMa.s DVs.
63
-------
2011-2013 PM2.5 bootstrap results
64
-------
Annual bootstrap results
Cl
CO
E
100-
a>
13
s'
75-
>
a
50-
Q.
CD
L_
25-
ฃ
o
o
.a
a)
o
c
2
0
'u
>
Q
1 1
5 10
15
Annual DV (ug/m3)
24-hr bootstrap results
CI
24-hr DV (ug/m3)
Difference between DVs and boot results, Annual NAAQ2?i
40-
30-
20-
10-
0-
Annual DV (ug/m3)
Difference between DV and boot results, Daily NAAQS ci
VP
0s
0)
O
c
0
0
4
-a
>
a
60-
40"
20-
o-
Daily DV (ug/m3)
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
Figure 42: Bootstrap results for the 2013 I'DVs, showing the 50%, 68%, 75%, and 95SS GIs, along with
the mean and median bootstrap DVs. The top two panels show the values; for the DVs at the various GIs,
while the bottom two panels show the relative difference between the Gl and the actual DV.
H
-------
Annual NAAQS bootstrap summary
25-
20-
0
O 15-
c
Q
5-
o-
I I I
00a
ง ฐ-
5 cl
3
ID
CN
ID
CN
O
ID
0
CL
Q_
O
ID
0
CL
CL
0 0
CL >
งฆ o
CO
CD
CO
CD
ID
r^
ID
ID
0>
0
CL
CL
ID
C7)
X
03
E
03
0
E
03
TD
0
E
>
0
-a
ฆa'
U)
bootstrap metric
24-hr NAAQS bootstrap summary
25-
U 15-
bootstrap metric
Figure 43: Bootstrap results for the 2013 PM2.5 DVs, showing distribution of the relative differences between
the bootstrap DVs and the actual DV at the 50%, 68%, 75%, and 95% CIs, along with the mean, median,
maximum, minimum, standard deviations of the relative differences.
66
-------
boot 50th percentile uncert, all sites
boot 50th percentile uncert, all sites
5% -
>. 4% -
-I'
c
"iS
i
0
o
C 3% -
=3
0
>
jc
CD
2% -
1%-
1.677
%
mean
miliar
i 1.667
%
V
%
\
Vj
\
W
,\.s
V\
i
ซ
0.5-
5 10 15
annual NAAQS
boot 50th percentile uncert, all sites
mean 0.152
median 0.150
0.4-
c
'03
0 n o
o 0.3 -
c
3
w
-Q
03
0.2-
0.1 -
mm mm
10
annual NAAQS
15
80% -
60% -
-I'
c
"iS
0
ฃ 40%-
=3
0
>
0
L_
20% -
V
y
0%-
40
24-hr NAAQS
mean 4.659 %
median 4.167%
60
boot 50th percentile uncert, all sites
15-
>.
'to "
-c.
-------
Annual NAAQS, rel uncert (%), all sites
50-
45-
40-
35-
30-
25-
5%
4%
3%
2%
1%
0%
-120
-100
-80
long
24-hr NAAQS, rel uncert (%), all sites
50"
45-
40-
35-
30-
25-
10.0%
7.5%
5.0%
2.5%
0.0%
-120
-100
-80
long
Figure 45: Spatial distribution of the relative difference between the C'J and the actual DY from the 50%
CIs for the 20f3 PMa.s DVs.
68
-------
2010-2012 PM2.5 bootstrap results
69
-------
CO
E
~3)
D
>
Q
CL
CO
1
-t-J
ฃ
o
o
JD
Annual bootstrap results
Cl
15-
10-
5-
16
CO
E
100-
O)
13
s'
75-
>
Q
50-
Q.
CD
i_
25-
to
o
o
.a
Annual DV (ug/m3)
24-hr bootstrap results
ci
24-hr DV (ug/m3)
Difference between DVs and boot results, Annual NAAQ2?i
Annual DV (ug/m3)
Difference between DV and boot results, Daily NAAQS ci
VP
0s
0)
O
c
0
0
4
-a
>
a
60-
40-
20-
o-
Daily DV (ug/m3)
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
Pigure 46: Bootstrap results for the 2012 I'DVs, showing the 50%, 68%, 75%, and 95SS GIs, along with
the mean and median bootstrap DVs. The top two panels show the values; for the DVs at the various GIs,
while the bottom two panels show the relative difference between the Gl and the actual DV.
70
-------
Annual NAAQS bootstrap summary
25-
20-
0
O 15-
c
0
0
^ 10-
T3
>
Q
5-
o-
ฅ
B
1
1
1
CD
q3
0
0
CL
CL
5
0
3
-0
-0
o~-
o~-
LO
LO
h-
LO
h-
0>
CD CD
CD --jj
E
E S
ฆo'
4'
C/)
lo
a>
bootstrap metric
24-hr NAAQS bootstrap summary
0
O
C
0
1
0
T3
>
Q
25-
20-
10-
CD
ง"
LO
CM
LO
CM
O
LO
CD
CL
Q_
O
LO
CD
CL
Q.
0 (D
Q- >
ง"
CO
CD
CO
CD
LO
r^
LO
r--
LO
CD
0)
CL
Q.
LO
o>
X
CD
E
ฃ=
CD
CD
E
ฃ=
CD
in
CD
E
"O
bootstrap metric
Figure 47: Bootstrap results for the 2012 PM2.5 DVs, showing distribution of the relative differences between
the bootstrap DVs and the actual DV at the 50%, 68%, 75%, and 95% CIs, along with the mean, median,
maximum, minimum, standard deviations of the relative differences.
71
-------
5% -
boot 50th percentile uncert, all sites
mean 1.757 %
rftedian 1.685%
4% -
c 3% -
0.5-
4 8 12 16
annual NAAQS
boot 50th percentile uncert, all sites
mean 0.163
median 0.150
0.4-
O 0.3 -
0.1 -
mm mm
i i i
4 8 12
annual NAAQS
16
boot 50th percentile uncert, all sites
80% -
60% -
-l'
c
iS
!_
CD
O
c 40%-
(D
>
0
20% -
mean 4.698 %
median 4.167%
0%-
20 40
24-hr NAAQS
60
boot 50th percentile uncert, all sites
15-
>.
c
'TO 10-
-------
Annual NAAQS, rel uncert (%), all sites
50-
45-
40-
35-
30-
25-
5%
4%
3%
2%
1%
0%
-120
-100
-80
long
24-hr NAAQS, rel uncert (%), all sites
50"
45-
40-
35-
30-
25-
10.0%
7.5%
5.0%
2.5%
0.0%
-120
-100
-80
long
Figure 49: Spatial distribution of the relative difference between the C'J and the actual DY from the 50%
CIs for the 2012 PMa.s DVs.
7:',
-------
2009-2011 PM2.5 bootstrap results
74
-------
Annual bootstrap results
Cl
CO
E
100-
O)
75-
>
Q
50-
Q.
CD
i_
25-
to
o
o
.a
a)
o
c
2
0
'u
>
Q
Annual DV (ug/m3)
24-hr bootstrap results
ci
20 30 40
24-hr DV (ug/m3)
60
Difference between DVs and boot results, Annual NAAQ2?i
40-
30-
20-
10-
0-
Annual DV (ug/m3)
Difference between DV and boot results, Daily NAAQS ci
VP
0s
0)
O
c
0
0
4
-a
>
a
60-
40"
20-
o-
30 40
Daily DV (ug/m3)
60
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
Figure 50: Bootstrap results for the 2011 I'DVs, showing the 50%, 68%, 75%, and 95งS GIs, along with
the mean and median bootstrap DVs. The top two panels show the values; for the DVs at the various GIs,
while the bottom two panels show the relative difference between the Gl and the actual DV.
7.5
-------
Annual NAAQS bootstrap summary
25-
20-
0
O 15-
c
0
0
^ 10-
T3
>
Q
5-
o-
ฅ
1
d
i
1 1
4
1
CD CD
CD --jj
E
E S
ฆo'
4'
C/)
bootstrap metric
24-hr NAAQS bootstrap summary
0
O
c
0
0
u_
"O
>
Q
25-
20-
LO
CM
CD
ง"
LO
CM
O
LO
CD
CL
Q_
O
LO
CD
CL
Q.
0
ง"
CO
CD
CO
CD
m
r^
LO
r--
io
CD
CD
CL
Q.
lO
cr>
X
CD
E
ฃ=
CD
CD
E
ฃ=
CD
TD
CD
E
"D
C/)
bootstrap metric
Figure 51: Bootstrap results lor the 2011 PM2.5 DVs, showing distribution ol the relative differences between
the bootstrap DVs and the actual DV at the 50%, 68%, 75%, and 95% CIs, along with the mean, median,
maximum, minimum, standard deviations ol the relative differences.
76
-------
boot 50th percentile uncert, all sites
5% -
>, 4% -
-I'
c
"iS
i
CD
o
ง 3% -
(D
>
jc
CD
2% -
1%-
mean 1.798 %
median 1.720
%
A
%
ซ
' %
\
\
*
>.
ป
%
\s\
ซ
ป
ป
1
I
i
1
w
I 1
1
10
15
0.5-
0.4-
>*
c
'03
0 n o
o 0.3 -
c
3
w
-Q
03
0.2-
annual NAAQS
boot 50th percentile uncert, all sites
mean 0.171
median 0.150
m
mm
10
15
annual NAAQS
boot 50th percentile uncert, all sites
75% -
-I'
c
"iS
g 50%-
c
=3
(D
>
jo
0
25% -
0%~
mean 4.879 %
median 4.200 %
>
.ซฆ
10
20
30 40
24-hr NAAQS
50
60
boot 50th percentile uncert, all sites
15-
c
ฆe 10-
-------
Annual NAAQS, rel uncert (%), all sites
50-
45-
40-
35-
30-
25-
5%
4%
3%
2%
1%
0%
-120
-100
-80
long
24-hr NAAQS, rel uncert (%), all sites
50"
45-
40-
35-
30-
25-
10.0%
7.5%
5.0%
2.5%
0.0%
-120
-100
-80
long
Figure 53: Spatial distribution of the relative difference between the C'J and the actual DY from the 50%
CIs for the 2011 I'M-;..-. D\ >.
78
-------
2008-2010 PM2.5 bootstrap results
79
-------
CO
E
20-
~3)
>
15-
O
CL
10-
CO
1
2
5-
0
0
.Q
Annual bootstrap results
Cl
a)
o
c
2
0
'u
>
Q
CO
E
"o>
90-
N-^
>
60-
Q
Q.
CD
i_
30-
to
O
O
-Q
Annual DV (ug/m3)
24-hr bootstrap results
ci
24-hr DV (ug/m3)
Difference between DVs and boot results, Annual NAAQ2?i
40-
30-
20-
10-
0-
A W
J,-
4
JA|AV
Mil
lid
W
/\ /\
10 15
Annual DV (ug/m3)
Difference between DV and boot results, Daily NAAQS ci
VP
0s
0)
O
c
0
0
4
-a
>
a
60-
40"
20-
o-
\\
Daily DV (ug/m3)
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
Figure 54: Bootstrap results for the 2010 PM2.5 DVs, showing the 50%, 68%, 75%, and 95งS GIs, along with
the mean and median bootstrap DVs. The top two panels show the values; for the DVs at the various GIs,
while the bottom two panels show the relative difference between the Gl and the actual DV.
80
-------
Annual NAAQS bootstrap summary
25-
20-
0
O 15-
c
Q
5-
o-
3.
0
งฆ o
LD
CN
LD
CN
O
ID
CD
CL
Q_
O
ID
CO
CD
CD
CL
CL
CO
CD
0
ง- o
ID
r^
LD
LD
0>
0
CL
CL
ID
0>
X
03
E
i i
d ฃ=
CD CD
0 ^
E 0
E
c >
E S
ฆo'
4'
C/)
bootstrap metric
24-hr NAAQS bootstrap summary
25-
20-
1
< .
15-
<1)
o
c
Q
5-
I
0 0
CL >
ง"
LD
CN
LD
CN
O
ID
0
CL
CL
O
ID
0
CL
CL
0 0
CL >
ง"
CO
CD
CO
CD
ID
r^
LD
r--
LD
CD
0
CL
CL
LD
cr>
X
CD
E
ฃ=
CD
0
E
ฃ=
CD
TD
0
E
"D
C/)
bootstrap metric
Figure 55: Bootstrap results for the 2010 PM2.5 DVs, showing distribution of the relative differences between
the bootstrap DVs and the actual DV at the 50%, 68%, 75%, and 95% CIs, along with the mean, median,
maximum, minimum, standard deviations of the relative differences.
81
-------
boot 50th percentile uncert, all sites
boot 50th percentile uncert, all sites
50% -
7.5% -
40% -
-I'
c
3
g 5.0%-
c
=3
(D
>
jc
0
meln 1.853 %
median 1.740 %
2.5% -
* ซซ
10 15
annual NAAQS
boot 50th percentile uncert, all sites
0.6-
>.
c
'to
-c
a>
o
c
13
20%-
10%-
0%-
20 40
24-hr NAAQS
60
boot 50th percentile uncert, all sites
mean 0.181
medten 0.15
an ซ
i
o
ป
mm
mm
i
m
i
i
i i
i
10-
mean M .242
median 1.000
20 40
24-hr NAAQS
60
Figure 56: Bootstrap results from the 50% CIs for PM2.5 DVs. The top two panels show the relative difference
between the CI and the actual DV and the bottom two panels show the absolute difference between the values
for the DVs at each site and the CI.
82
-------
Annual NAAQS, rel uncert (%), all sites
50-
45-
40-
35-
30-
25-
5%
4%
3%
2%
1%
0%
-120
-100
long
24-hr NAAQS, rel uncert (%), all sites
-80
50-
45-
40-
35-
30-
25-
10.0%
7.5%
5.0%
2.5%
0.0%
-120
-100
long
-80
Figure 57: Spatial distribution of the relative difference between the C'J and the actual DY from the 50%
CIs for the 20f0 PMa.s DVs.
83
-------
2007-2009 PM2.5 bootstrap results
84
-------
Annual bootstrap results
Cl
CO
E
a>
60-
>
a
40-
Q.
CD
i_
20-
to
o
o
.a
Annual DV (ug/m3)
24-hr bootstrap results
ci
24-hr DV (ug/m3)
Difference between DVs and boot results, Annual NAAQ2?i
Annual DV (ug/m3)
Difference between DV and boot results, Daily NAAQS ci
VP
0s
0)
O
c
0
0
4
-a
>
a
60-
40"
20-
o-
Daily DV (ug/m3)
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
Figure 58: Bootstrap results for the 2009 PMjDVs, showing the 50%, 68%, 75%, and 95งS GIs, along with
the mean and median bootstrap DVs. The top two panels show the values; for the DVs at the various GIs,
while the bottom two panels show the relative difference between the Gl and the actual DV.
85
-------
Annual NAAQS bootstrap summary
25-
20-
0
o 15-
c
Q
0-
S
!
>
i
I
>
ป
1
I
-
H *
, ;
1
i
ป
ป
r
Li. J
M
I
r=H
4
_i_ _L
0
งฆ o
LD
CN
LD
CN
O
ID
CD
CL
Q_
O
ID
CO
CD
CD
CL
CL
CO
CD
0
งฆ o
ID
r^
LD
LD
C7)
0
CL
CL
ID
C7)
X
03
E
CD CD
0 --jj
E 0
E
c >
E S
ฆo'
4'
C/)
bootstrap metric
24-hr NAAQS bootstrap summary
25-
20-
0
O 15-
c
0
1
0)
^ 10-
T3
>
Q
5-
! ' I
i 1
1
1
0 0
CL >
ง"
LD
CN
LD
CN
O
ID
0
CL
CL
O
ID
0
CL
CL
0 0
CL >
ง"
CO
CD
CO
CD
ID
r^
LD
r--
LD
CD
0
CL
CL
LD
cr>
X
CD
E
JZ c
CD CD
0 --jj
E 0
E
"O
c/)
bootstrap metric
Figure 59: Bootstrap results for the 2009 PM2.5 DVs, showing distribution of the relative differences between
the bootstrap DVs and the actual DV at the 50%, 68%, 75%, and 95% CIs, along with the mean, median,
maximum, minimum, standard deviations of the relative differences.
86
-------
boot 50th percentile uncert, all sites
boot 50th percentile uncert, all sites
8% -
40% -
mean* 1.942 %
median 1.830%
ฃ 6% -
S> 4% -
**ป
12 16
annual NAAQS
boot 50th percentile uncert, all sites
0.6-
>*
c
03
-E 04.
CD
O
c
13
C/)
-Q
03
12
annual NAAQS
16
c 30%-
i_
0
O
c
=3
g 20%-
jo
0
10%-
mean 5.086 %
median 4.200 %
V
_ ซSLa
0% -
20
40
24-hr NAAQS
60
boot 50th percentile uncert, all sites
m
mean 0.203
10
>
median 0.2)0 c
03
mm*
l
1 * <
1 i
mm
0
0
c
13
* W
03 5
1 Miai <
mmmmmm
0
mm m
M
mean 1.353
median 1.000
<
M MM
mm m m mm m
m mm m
m m
20
40
24-hr NAAQS
60
Figure 60: Bootstrap results from the 50% CIs for PM2.5 DVs. The top two panels show the relative difference
between the CI and the actual DV and the bottom two panels show the absolute difference between the values
for the DVs at each site and the CI.
87
-------
Annual NAAQS, rel uncert (%), all sites
50-
45-
40-
35-
30-
25-
5%
4%
3%
2%
1%
0%
-120
-100
long
24-hr NAAQS, rel uncert (%), all sites
-80
50-
45-
40-
35-
30-
25-
10.0%
7.5%
5.0%
2.5%
0.0%
-120
-100
-80
long
Figure 61: Spatial distribution of the relative difference between the C'J and the actual DY from the 50%
CIs for the 2009 PMa.s DVs.
88
-------
2006-2008 PM2.5 bootstrap results
89
-------
CO
E
~3)
20-
>
15-
o
CL
CO
10-
1
w
-4'
5-
o
o
.Q
Annual bootstrap results
Cl
a)
o
c
2
0
'u
>
Q
10 15
Annual DV (ug/m3)
CO
E
a>
60-
>
a
40-
Q.
CD
i_
20-
to
o
o
24-hr bootstrap results
ci
24-hr DV (ug/m3)
Difference between DVs and boot results, Annual NAAQ2?i
40-
30-
20-
10-
0-
10 15
Annual DV (ug/m3)
Difference between DV and boot results, Daily NAAQS ci
VP
0s
0)
O
c
0
0
4
-a
>
a
60-
40"
20-
o-
Daily DV (ug/m3)
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
Figure 62: Bootstrap results for the 2008 I'DVs, showing the 50%, 68%, 75%, and ง5% GIs, along with
the mean and median bootstrap DVs. The top two panels show the values; for the DVs at the various GIs,
while the bottom two panels show the relative difference between the Gl and the actual DV.
90
-------
Annual NAAQS bootstrap summary
25-
20-
0
O 15-
c
0
0
^ 10-
T3
>
Q
5-
o-
9
B 8
0
งฆ o
LD
CN
LD
CN
O
ID
CD
CL
CL
O
ID
CD
CL
CL
0
งฆ o
CO
CD
CO
CD
ID
r^
ID
LD
0>
24-hr NAAQS bootstrap summary
0
CL
CL
ID
0>
X
03
E
i i
d ฃ=
CD CD
0) --jj
E 0
E
c >
E S
ฆo'
4'
C/)
bootstrap metric
0
o
c
0
1
0
'~o
>
Q
25-
20-
0 0
CL >
ง"
LD
CN
LD
CN
O
ID
0
CL
CL
O
ID
0
CL
CL
CO
CD
CO
CD
0 0
CL >
ง"
ID
r^
LD
r--
LD
o>
0
CL
CL
LD
cr>
X
CD
E
ฃ=
CD
0
E
ฃ=
CD
TD
0
E
"D
C/)
bootstrap metric
Figure 63: Bootstrap results for the 2008 PM2.5 DVs, showing distribution of the relative differences between
the bootstrap DVs and the actual DV at the 50%, 68%, 75%, and 95% CIs, along with the mean, median,
maximum, minimum, standard deviations of the relative differences.
91
-------
boot 50th percentile uncert, all sites
6% -
-I'
c
"ro
i
0
O
c
=3
0
>
jc
0
4% -
2% -
mean
1^48
%
m
r
1
nedian
i 1.870%
<
*
m
^
ซ
.
10 15
annual NAAQS
30% -
'c
.2 20%-
(D
O
c
=3
0
>
"co
0
10%-
o%-
boot 50th percentile uncert, all sites
mean 4.979 %
median 4.000 %
"
20
boot 50th percentile uncert, all sites
40
24-hr NAAQS
boot 50th percentile uncert, all sites
10.0-
0.6-
>*
C
'03
0 A ,
o 0.4 -
c
3
w
-Q
03
0.2-
ป mean 0.216
median 0.20
mmm mmmmmmmmm
mm
7.5-
>*
c
C\3
o 5.0 -
c
3
C/)
_Q
n3
10 15
annual NAAQS
20
mean 1.394
median 1.000
2.5-
m m
MM*
0.0-
20
40
24-hr NAAQS
60
Figure 64: Bootstrap results from the 50% CIs for PM2.5 DVs. The top two panels show the relative difference
between the CI and the actual DV and the bottom two panels show the absolute difference between the values
for the DVs at each site and the CI.
92
-------
Annual NAAQS, rel uncert (%), all sites
50-
45-
40-
35-
30-
25-
5%
4%
3%
2%
1%
0%
-120
-100
long
24-hr NAAQS, rel uncert (%), all sites
-80
50-
45-
40-
35-
30-
25-
10.0%
7.5%
5.0%
2.5%
0.0%
-120
-100
long
-80
Figure 65: Spatial distribution of the relative difference between the C'J and the actual DY from the 50%
CIs for the 2008 PMa.s DVs.
93
-------
2005-2007 PM2.5 bootstrap results
94
-------
CO
E
~3)
1
O
CM
>
15-
O
CL
10-
CO
1
2
5-
0
0
.Q
Annual bootstrap results
Cl
a)
o
c
2
0
'u
>
Q
10 15
Annual DV (ug/m3)
20
CO
E
80-
O)
13
*
60-
>
Q
40-
Q.
CD
i_
20-
W
O
O
24-hr bootstrap results
ci
24-hr DV (ug/m3)
Difference between DVs and boot results, Annual NAAQ2?i
40-
30-
20-
10-
0-
10 15
Annual DV (ug/m3)
20
Difference between DV and boot results, Daily NAAQS ci
VP
0s
0)
O
c
0
0
4
-a
>
a
60-
40"
20-
o-
Daily DV (ug/m3)
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
Figure 66: Bootstrap results for the 2007 PM2.5 DVs, showing the 50%, 68%, 75%, and 95SS GIs, along with
the mean and median bootstrap DVs. The top two panels show the values; for the DVs at the various GIs,
while the bottom two panels show the relative difference between the Gl and the actual DV.
95
-------
Annual NAAQS bootstrap summary
25-
20-
0
O 15-
c
Q
5-
o-
I I I
0
ง- o
ID
CN
ID
CN
O
ID
CD
CL
Q_
O
ID
CD
CL
CL
0 (D
CL >
งฆ o
CO
CD
CO
CD
ID
r^
ID
ID
0>
CD
CL
CL
ID
0>
X
03
E
i i
d ฃ=
CD CD
E S
ฆo'
4'
C/)
bootstrap metric
24-hr NAAQS bootstrap summary
25-
O 15-
bootstrap metric
Figure 67: Bootstrap results for the 2007 PM2.5 DVs, showing distribution of the relative differences between
the bootstrap DVs and the actual DV at the 50%, 68%, 75%, and 95% CIs, along with the mean, median,
maximum, minimum, standard deviations of the relative differences.
96
-------
5% -
boot 50th percentile uncert, all sites
mean 1.959 %
medicfh 1.890 %
boot 50th percentile uncert, all sites
30% -
mean 4.770 %
> 4% -
=i 3% -
20% -
(D
o
c
=3
0
>
"co
0 10%-
median 4.300 %
* % #
A A A
0.5-
5 10 15
annual NAAQS
boot 50th percentile uncert, all sites
mem 0.233
median 0>850
20
_
I
m
w
ฆ 1
0%-
20
24-hr NAAQS
boot 50th percentile uncert, all sites
6-
0.4-
c
'03
V no
o 0.3 -
c
3
C/)
-Q
03
0.2-
0.1 -
m mm
>*
c
C\3
* 0
o
c
=3
C/)
_Q
n3
4-
mean 1.457
median 1.500
M (MM
MWI
ป
ซ
2-
<
ซ
4
0-
10
15
20
20
annual NAAQS
40
24-hr NAAQS
60
Figure 68: Bootstrap results from the 50% CIs for PM2.5 DVs. The top two panels show the relative difference
between the CI and the actual DV and the bottom two panels show the absolute difference between the values
for the DVs at each site and the CI.
97
-------
Annual NAAQS, rel uncert (%), all sites
50-
45-
40-
35-
30-
25-
5%
4%
3%
2%
1%
0%
-120
-100
long
24-hr NAAQS, rel uncert (%), all sites
-80
50-
45-
40-
35-
30-
25-
10.0%
7.5%
5.0%
2.5%
0.0%
-120
-100
long
-80
Figure 69: Spatial distribution of the relative difference between the C'J and the actual DY from the 50%
CIs for the 2007 PMa.s DVs.
98
-------
3.10 2004-2006 PM2.5 bootstrap results
99
-------
CO
E
~3)
D
>
Q
CL
CO
1
-t-J
ฃ
o
o
JD
CO
E
O)
>
Q
Q.
CD
i_
to
-4'
o
o
X!
Annual bootstrap results
Cl
a)
o
c
2
0
'u
>
Q
20-
15-
10-
5-
10 15
Annual DV (ug/m3)
24-hr bootstrap results
ci
80-
60-
40-
20-
24-hr DV (ug/m3)
Difference between DVs and boot results, Annual NAAQ2?i
40-
30-
20-
10-
0-
~AlW^
10 15
Annual DV (ug/m3)
Difference between DV and boot results, Daily NAAQS ci
VP
0s
0)
O
c
0
0
M
-a
>
a
60-
40-
20-
o-
Daily DV (ug/m3)
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
Figure 70: Bootstrap results for the 2006 I'DVs,:showing the 50%, 68%, 75%, and 95% GIs, along with
the mean and median bootstrap DVs. The top two panels show the values; for the DVs at the various GIs,
while the bottom two panels show the relative difference between the Gl and the actual DV.
100
-------
Annual NAAQS bootstrap summary
25-
Q 15-
B
1
1
1
X
ฃ=
ฃ=
03
03
03
E
CD
TD
E
CD
E
c >
E S
ฆo'
4'
C/3
bootstrap metric
24-hr NAAQS bootstrap summary
Q
25-
20-
CD
ง"
LO
CM
LO
CM
O
LO
CD
CL
Q_
O
LO
CD
CL
Q.
CD
ง"
CO
CD
CO
CD
LO
r^
LO
I--
LO
CD
0)
CL
CL
LO
cr>
X
03
E
ฃ=
ro
CD
E
ฃ=
in
CD
E
~o
bootstrap metric
Figure 71: Bootstrap results lor the 2006 PM2.5 DVs, showing distribution of the relative differences between
the bootstrap DVs and the actual DV at the 50%, 68%, 75%, and 95% CIs, along with the mean, median,
maximum, minimum, standard deviations of the relative differences.
101
-------
boot 50th percentile uncert, all sites
mean 1.939 %
boot 50th percentile uncert, all sites
30% -
mean 4.854 %
5% -
>- 4% -
_c
'03
i
0
o
ง 3% -
0
>
median 1*880 %
.v\V
20% -
median 4.400 %
0
O
c
=3
0
>
'
_ro
0
2% -
1%-
V X. 0 10%- . # V
n.JsS.'. *
* ""***.. 0%- *T "
10 15
annual NAAQS
20
20 40
24-hr NAAQS
60
0.6-
0.4-
boot 50th percentile uncert, all sites
mean 0.231
median 0.360
boot 50th percentile uncert, all sites
6-
mean 1.475
median 1.500
CO
t
CD
O
c
=5
C/)
-Q
03
0.2-
>*
c
"03
0
O
c
=5
C/)
_Q
03
4-
2-
0-
10
15
20
mm*
20
It
annual NAAQS
40
24-hr NAAQS
60
Figure 72: Bootstrap results from the 50% CIs for PM2.5 DVs. The top two panels show the relative difference
between the CI and the actual DV and the bottom two panels show the absolute difference between the values
for the DVs at each site and the CI.
102
-------
Annual NAAQS, rel uncert (%), all sites
50-
45-
40-
35-
30-
25-
5%
4%
3%
2%
1%
0%
-120
-100
long
24-hr NAAQS, rel uncert (%), all sites
-80
50-
45-
40-
35-
30-
25-
10.0%
7.5%
5.0%
2.5%
0.0%
-120
-100
long
-80
Figure 73: Spatial distribution of the relative difference between the C'J and the actual DY from the 50%
C Is for the 2006 I'Mo - DVs .
103
-------
3.11 2003-2005 PM2.5 bootstrap results
104
-------
Annual bootstrap results
Cl
a)
o
c
2
0
'u
>
Q
10 15
Annual DV (ug/m3)
CO
E
o>
>
Q
Q.
CD
L_
-4'
o
o
X!
24-hr bootstrap results
ci
100-
75-
50-
25-
J
k
20
40
60
24-hr DV (ug/m3)
Difference between DVs and boot results, Annual NAAQ2?i
40-
30-
20-
10-
0-
HA
10 15
Annual DV (ug/m3)
Difference between DV and boot results, Daily NAAQS ci
VP
0s
0)
O
c
0
0
4
-a
>
a
60-
40-
20-
o-
Daily DV (ug/m3)
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
Figure 74: Bootstrap results for the 2005 PMjDVs, showing the 50%, 68%, 75%, and 95งS GIs, along with
the mean and median bootstrap DVs. The top two panels show the values; for the DVs at the various GIs,
while the bottom two panels show the relative difference between the Gl and the actual DV.
105
-------
Annual NAAQS bootstrap summary
25-
20-
0
O 15-
c
0
0
^ 10-
T3
>
Q
5-
o-
4=-Li
I I I
0
งฆ o
LD
CN
LD
CN
O
ID
CD
CL
Q_
O
ID
CD
CL
CL
0
งฆ o
CO
CD
CO
CD
ID
r^
ID
LD
0>
0
CL
CL
ID
0>
X
03
E
03
0
E
03
in
CD
E
>
0
TD
-o'
-t<
U)
bootstrap metric
24-hr NAAQS bootstrap summary
0
O
C
0
0
>
o
25-
20-
10-
0 0
CL >
ง"
LD
CN
LD
CN
O
ID
0
CL
CL
O
ID
CO
CD
0
CL
CL
0 0
CL >
ง"
CO
CD
ID
r^
LD
r--
LD
o>
0
CL
CL
LD
cr>
X
03
E
ฃ=
03
0
E
ฃ=
ro
TD
CD
E
"D
C/)
bootstrap metric
Figure 75: Bootstrap results for the 2005 PM2.5 DVs, showing distribution of the relative differences between
the bootstrap DVs and the actual DV at the 50%, 68%, 75%, and 95% CIs, along with the mean, median,
maximum, minimum, standard deviations of the relative differences.
106
-------
5% -
boot 50th percentile uncert, all sites
mean 1.940 %
median 1.870%
= 3% -
V*
10 15 20
annual NAAQS
boot 50th percentile uncert, all sites
0.5-
0.4-
>.
C
"co
tl
0 A O
o 0.3 -
c
3
w
-Q
03
0.2-
0.1 -
1 0.236
&
mear
90 <
a
r
1 lUUIdl 1
A AAA
mi
1
(
w
ft A
w W W
mm <
m m
a
WBVW W
mm m
a a a
www
1
1
W
1
1
1
10 15
annual NAAQS
20
boot 50th percentile uncert, all sites
30% - mean 5.008 %
c 20%-
ro
!_
0
O
c
=3
0
>
jo
0
median 4.700 %
10%-
o%-
1
w
V
\\
* %
^
H
*.
>
m
20 40
24-hr NAAQS
60
boot 50th percentile uncert, all sites
6- mean 1.547
median 1.500
4-
m
03
0
O
c
=5
C/)
_Q
03
m
mm
mm
ซ
mmm m
2-
20 40 60
24-hr NAAQS
Figure 76: Bootstrap results from the 50% CIs for PM2.5 DVs. The top two panels show the relative difference
between the CI and the actual DV and the bottom two panels show the absolute difference between the values
for the DVs at each site and the CI.
107
-------
Annual NAAQS, rel uncert (%), all sites
50-
45-
40-
35-
30-
25-
5%
4%
3%
2%
1%
0%
-120
-100
long
24-hr NAAQS, rel uncert (%), all sites
-80
50-
45-
40-
35-
30-
25-
10.0%
7.5%
5.0%
2.5%
0.0%
-120
-100
long
-80
Figure 77: Spatial distribution of the relative difference between the C'J and the actual DY from the 50%
C Is for the 2005 PMaDVs.
108
-------
3.12 2002-2004 PM2.5 bootstrap results
109
-------
CO
E
~3)
D
>
Q
CL
CO
I
ฃ
o
o
JD
Annual bootstrap results
Cl
a)
o
c
2
0
'u
>
Q
20-
10-
10 15
Annual DV (ug/m3)
25
CO
E
O)
>
Q
Q.
CD
L_
cn
-4'
o
o
X!
24-hr bootstrap results
ci
100-
75-
50-
25-
24-hr DV (ug/m3)
Difference between DVs and boot results, Annual NAAQ2?i
40-
30-
20-
10-
0-
10 15 20
Annual DV (ug/m3)
Difference between DV and boot results, Daily NAAQS ci
VP
0s
0)
O
c
0
0
M
-a
>
a
60-
40-
20-
o-
t
Daily DV (ug/m3)
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
Pigure 78: Bootstrap results for the 2004 I'DVs, showing the 50%, 68%, 75%, and 95SS GIs, along with
the mean and median bootstrap DVs. The top two panels show the values; for the DVs at the various GIs,
while the bottom two panels show the relative difference between the Gl and the actual DV.
110
-------
Annual NAAQS bootstrap summary
25-
Q 15-
!t= 10
CD CD
CD --jj
E
E S
-a'
4'
C/)
bootstrap metric
24-hr NAAQS bootstrap summary
Q
25-
20-
10-
CD
ง"
LO
CM
LO
CM
O
LO
CD
CL
Q_
O
LO
CD
CL
Q.
0 (D
Q- >
ง"
CO
CD
CO
CD
LO
r^
LO
r--
LO
CD
0)
CL
Q.
LO
o>
X
CD
E
ฃ=
CD
CD
E
ฃ=
CD
in
CD
E
"O
bootstrap metric
Figure 79: Bootstrap results for the 2004 PM2.5 DVs, showing distribution of the relative differences between
the bootstrap DVs and the actual DV at the 50%, 68%, 75%, and 95% CIs, along with the mean, median,
maximum, minimum, standard deviations of the relative differences.
Ill
-------
boot 50th percentile uncert, all sites
boot 50th percentile uncert, all sites
6% -
ฆ! 4% -
i_
0
O
c
=3
0
>
CD
mean
1.903 %
median
1.840 %
m
30% -
2% -
-i'
c
'ro
i
CD
O
c
=3
0
>
jo
0
20% -
10%-
0%-
5.233 7,
r
nean ฃ
)
a
median
4.700 %
-
# # ^
P'.i -
s
w
I 1
1
1
0.5-
0.4-
c
CT3
0
5 10 15 20
annual NAAQS
boot 50th percentile uncert, all sites
mean i.227
median 0.250
m m m
mmmmm
25
20 40
24-hr NAAQS
60
boot 50th percentile uncert, all sites
7.5-
O 0.3- Ml Ml
c
w
-Q
03
0.2-
0.1 -
c
5.0-
* 0
o
C
13
(/)
-Q
CD
2.5-
0.0-
mean 1.606
median 1.500
m
mm
WW w
ป
10
15
20
25
20
annual NAAQS
40
24-hr NAAQS
60
Figure 80: Bootstrap results from the 50% CIs for PM2.5 DVs. The top two panels show the relative difference
between the CI and the actual DV and the bottom two panels show the absolute difference between the values
for the DVs at each site and the CI.
112
-------
Annual NAAQS, rel uncert (%), all sites
50-
45-
40-
35-
30-
25-
5%
4%
3%
2%
1%
0%
-120
-100
long
24-hr NAAQS, rel uncert (%), all sites
-80
50-
45-
40-
35-
30-
25-
10.0%
7.5%
5.0%
2.5%
0.0%
-120
-100
long
-80
Figure 81: Spatial distribution of the relative difference between the C'J and the actual DY from the 50%
CIs for the 2004 PMa.s DVs.
113
-------
3.13 2001-2003 PM2.5 bootstrap results
114
-------
CO
E
~3)
D
>
Q
CL
CO
1
ฃ
o
o
JD
CO
E
a>
>
Q
Q.
CD
L_
-*-
C0
-4'
o
o
.a
CI limits
50% CI
68% CI
75% CI
95% CI
max/min
median
CI limits
50% CI
68% CI
75% CI
95% CI
max/min
median
Annual bootstrap results
20-
10 20
Annual DV (ug/m3)
24-hr bootstrap results
60-
20 40 60
24-hr DV (ug/m3)
Difference between DVs and boot results, Annual NAAQ2?i limits
40 - 50% CI
30- j 68% CI
^ i i
10 20 median
Annual DV (ug/m3)
Difference between DV and boot results, Daily NAAQS ci limits
VP
CD 60"
O
0 40"
0
1 20"
Q ฐ"
Daily DV (ug/m3)
50% CI
68% CI
75% CI
95% CI
max/min
median
Pigure 82: Bootstrap results for the 2003 I'DVs, showing the 50%, 68%, 75%, and 95% GIs, along with
the mean and median bootstrap DVs. The top two panels show the values; for the DVs at the various GIs,
while the bottom two panels show the relative difference between the Gl and the actual DV.
115
-------
Annual NAAQS bootstrap summary
25-
20-
0
O 15-
c
0
0
^ 10-
T3
>
Q
5-
o-
i i
0
งฆ o
LD
CN
LD
CN
O
ID
CD
CL
Q_
O
ID
CD
CL
CL
0
งฆ o
CO
CD
CO
CD
ID
r^
ID
LD
0>
0
CL
CL
ID
0>
X
03
E
03
0
E
03
in
CD
E
>
0
TD
-o'
-t<
U)
bootstrap metric
24-hr NAAQS bootstrap summary
0
O
c
0
i
O
25-
20-
10-
0 0
CL >
ง"
LD
CN
LD
CN
O
ID
0
CL
CL
O
ID
0
CL
CL
0 0
CL >
ง"
CO
CD
CO
CD
ID
r^
LD
r--
LD
o>
0
CL
CL
LD
cr>
X
03
E
ฃ=
03
0
E
ฃ=
ro
TD
CD
E
"D
C/)
bootstrap metric
Figure 83: Bootstrap results for the 2003 PM2.5 DVs, showing distribution of the relative differences between
the bootstrap DVs and the actual DV at the 50%, 68%, 75%, and 95% CIs, along with the mean, median,
maximum, minimum, standard deviations of the relative differences.
116
-------
boot 50th percentile uncert, all sites
6% -
mean 1.901 %
boot 50th percentile uncert, all sites
30%- mean 5.276%
-i<
I 4% ฆ
i_
0
O
c
=3
0
>
jc
0
median 1.840%
2% -
0.6-
0.4-
10 20
annual NAAQS
boot 50th percentile uncert, all sites
meanฎ0.233
m
medics 0.290
mm
03
CD
O
c
=5
C/)
-Q
03
-i'
c
'ro
0
O
c
=3
0
>
0
20% -
median 4.400 %
10%-
o%-
40 60
24-hr NAAQS
boot 50th percentile uncert, all sites
6-
ซ
0.2-
>.
c
03
0 .
o 4-
c
=3
C/)
_Q
03
mean 1.70C
i*
median ^.500
2-
!
0-
10 20
annual NAAQS
20 40 60
24-hr NAAQS
80
Figure 84: Bootstrap results from the 50% CIs for PM2.5 DVs. The top two panels show the relative difference
between the CI and the actual DV and the bottom two panels show the absolute difference between the values
for the DVs at each site and the CI.
117
-------
Annual NAAQS, rel uncert (%), all sites
50-
45-
40-
35-
30-
25-
5%
4%
3%
2%
1%
0%
-120
-100
long
24-hr NAAQS, rel uncert (%), all sites
-80
50-
45-
40-
35-
30-
25-
10.0%
7.5%
5.0%
2.5%
0.0%
-120
-100
long
-80
Figure 85: Spatial distribution of the relative difference between the C'J and the actual DY from the 50%
C Is for the 2003 PMaDVs.
118
-------
3.14 2000-2002 PM2.5 bootstrap results
119
-------
CO
E
~3)
D
>
Q
CL
CO
1
-t-J
ฃ
o
o
JD
Annual bootstrap results
Cl
a)
o
c
CD
i_
0)
'u
>
Q
30-
20-
10-
30
CO
E
a>
90-
Nฆ*'
>
a
60-
Q.
CD
i_
30-
to
o
o
.a
Annual DV (ug/m3)
24-hr bootstrap results
ci
24-hr DV (ug/m3)
Difference between DVs and boot results, Annual NAAQ2?i
40-
30-
20-
10-
0-
30
Annual DV (ug/m3)
Difference between DV and boot results, Daily NAAQS ci
VP
0s
0)
O
c
0
0
4
-a
>
a
60-
40-
20-
o-
W
25
Daily DV (ug/m3)
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
limits
- 50% CI
- 68% CI
75% CI
- 95% CI
- max/min
- median
Pigure 86: Bootstrap results for the 2002 I'DVs, showing the 50%, 68%, 75%, and 95SS GIs, along with
the mean and median bootstrap DVs. The top two panels show the values; for the DVs at the various GIs,
while the bottom two panels show the relative difference between the Gl and the actual DV.
120
-------
Annual NAAQS bootstrap summary
25-
20-
0
O 15-
c
0
1
0)
^ 10-
T3
>
Q
5-
o-
=T= 4
I I
0
งฆ o
LD
CN
LD
CN
O
ID
CD
CL
Q_
O
ID
CD
CL
CL
0
งฆ o
CO
CD
CO
CD
ID
r^
ID
LD
0>
0
CL
CL
ID
0>
X
03
E
03
0
E
03
TD
CD
E
>
0
TD
-o'
-t<
U)
bootstrap metric
24-hr NAAQS bootstrap summary
0
O
c
0
i
O
25-
20-
10-
0 0
CL >
ง"
LD
CN
LD
CN
O
ID
0
CL
CL
O
ID
0
CL
CL
0 0
CL >
ง"
CO
CD
CO
CD
ID
r^
LD
r--
LD
o>
0
CL
CL
LD
cr>
X
03
E
ฃ=
03
0
E
ฃ=
ro
TD
CD
E
"D
C/)
bootstrap metric
Figure 87: Bootstrap results for the 2002 PM2.5 DVs, showing distribution of the relative differences between
the bootstrap DVs and the actual DV at the 50%, 68%, 75%, and 95% CIs, along with the mean, median,
maximum, minimum, standard deviations of the relative differences.
121
-------
boot 50th percentile uncert, all sites
6% -
mean 1.895 %
boot 50th percentile uncert, all sites
40% -
median 1.840 %
jg 4% "
i
0
o
c
=3
0
>
JO
CD
30% -
-i'
c
3
!_
(D
C 20%-
=3
0
>
jo
0
L_
10%-
mean 4.962 %
median 4.500 %
#
.
2555? .
10 20
annual NAAQS
30
boot 50th percentile uncert, all sites
0.6-
>.
c
03
ฆC. 04-
(D
O
c
13
C/)
-Q
03
0.2-
25 50 75
24-hr NAAQS
boot 50th percentile uncert, all sites
m
8-
mean 0.234
m
medisW 9ป250*
m a a a m m t
m
m
ป
ป
>*
c
03
6-
M MM
rnrnmm ซ
0
O
c
4-
mm
iMMM
wmm m
13
C/)
_Q
03
mmmmm
2 -
mmmmm
m
0-
10 20
annual NAAQS
30
mean 1.640
mi&flian 1.500
m mmmmm m m
mmmmm m m
ป mm
m
25 50 75
24-hr NAAQS
Figure 88: Bootstrap results from the 50% CIs for PM2.5 DVs. The top two panels show the relative difference
between the CI and the actual DV and the bottom two panels show the absolute difference between the values
for the DVs at each site and the CI.
122
-------
Annual NAAQS, rel uncert (%), all sites
50-
45-
40-
35-
30-
25-
5%
4%
3%
2%
1%
0%
-120
-100
long
24-hr NAAQS, rel uncert (%), all sites
-80
50-
45-
40-
35-
30-
25-
10.0%
7.5%
5.0%
2.5%
0.0%
-120
-100
long
-80
Figure 89: Spatial distribution of the relative difference between the CI and the actual DV from the 50%
C Is for the 2002 P.Yla DVs.
123
-------
4 Comparison plots of nearby sites
Comparison of PM2.5 data for paired, nearby sites for the spatial analysis conducted in Section 3.1.2.
124
-------
County 1: Honolulu State 1: Hawaii
Sites: 150031001 & 150031004
20-
15-
10-
SiteJD
~ 150031001
~ 150031004
i i i i i i i
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
150031001
Figure 90: Comparison of I data, for sites 150031001 and 150031001. Top panel shows time Series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r2)j witlt2feta points are colored by month.
-------
County 1: Clark State 1: Indiana
Sites: 180190006 & 211110067
~ 180190006
~ 211110067
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
20
180190006
Figure 91: Comparison of I data, for sites 180190006 and 180190006. Top panel shows time Series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (rS), wit Jrilata points are colored by month.
-------
County 1: Marion State 1: Indiana
Sites: 180970078 & 180970081
~ 180970078
~ 180970081
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
20
180970078
Figure 92: Comparison of I data, for sites 180970078 and 180970078. Top panel shows time Series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (rS), wit llrjTata points are colored by month.
-------
40-
30-
20-
10-
County 1: Clinton State 1: Iowa
Sites: 190450019 & 190450021
SiteJD
~ 190450019
~ 190450021
01/01/12 07/01/12 01/01/13
07/01/13
DATE
01/01/14 07/01/14 01/01/15
40-
30-
8 20 -
10-
delta 6.425 km
m = 0.89; r2 = 0.909
10
i
20
190450019
30
40
seas
~ DJF
~ JJA
~ MAM
~ SON
Figure 93: Comparison of I data, for sites 190450019 and 190450019. Top panel shows time Series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r2)j witlt2Sata points are colored by month.
-------
40-
30-
ง20-
CL
10-
County 1: East Baton Rouge State 1: Louisiana
Sites: 220330009 & 221210001
* ~
~ ~ *
ซ* ~
.... :
1 \V.JU"
v*i.
& . i*
SiteJD
~ 220330009
~ 221210001
i i i i i i i
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
40-
30-
2 20-
10-
delta 5.419 km
m = 0.76: r2 = 0.637
seas
~ DJF
~ JJA
~ MAM
~ SON
10
20
220330009
30
40
Figure 94: Comparison of I data, for sites 220330009 and 220330009. Top panel shows time Series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r2)j wit SriSata points are colored by month.
-------
County 1: Washington State 1: Minnesota
Sites: 271630447 & 271630448
30-
20-
10-
* -f
At ^ f ~ *$!/~;*< *t *b
SiteJD
~ 271630447
~ 271630448
i i i i i i i
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
30
: 20
10
delta 1.04 km
m = 0.94; r:
2 = 0.857
~
/ /<
// 4
~ . ~
~ ~
~ ~ Iwft
i.W51'
i ififlni *
ป
~
1
*:*
it>n
.JgP
\iri ~*
cr*
i i
l%iป ~
p* ~
ป
i
i
1
10
20
271630447
30
seas
~ DJF
~ JJA
~ MAM
~ SON
Figure 95: Comparison of I data, for sites 271630447 and 271630447. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r&), wit Btjjfl&ta points are colored by month.
-------
County 1: Washoe State 1: Nevada
Sites: 320310016 & 320311005
100
75-
50-
25
o-
~
~
~
~
~
~
X
%
~
~
~
~ i
k
i i
L
~
i
ป
~
<
Sl.
f A*
K JL
X mil
M9a ^ ttlR
V
un
"pH*
n
~
&
SiteJD
~ 320310016
~ 320311005
i i i i i
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
delta 5.522 km
100
75-
50-
25-
o-
m = 0.95;
r2 = 0.779
~
// ~
*
*
yr ~
#
~
~ *
***~
dHULปซ
~
c
i i i i
) 25 50 75 100
seas
~ DJF
~ JJA
~ MAM
~ SON
320310016
Figure 96: Comparison of I data, for sites 320310016 and 320310016. Top panel shows time Series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r2)j witliM&ta points are colored by month.
-------
80-
60-
County 1: Bernalillo State 1: New Mexico
Sites: 350010023 & 350010024
:40-
SiteJD
~ 350010023
~ 350010024
20-
I
~ ~
I I I I I I
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
delta 7.947 km
80-
60-
! 40 -
20-
m = 0.43; r2 = 0.209
20
40
350010023
60
80
seas
DJF
JJA
MAM
SON
Figure 97: Comparison of PMjjs data, for sites 350010023 and 350010023. Top panel shows time Series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (rS), witltiSJata points are colored by month.
-------
County 1: Northampton State 1: Pennsylvania
Sites: 420950025 & 420950027
40-
20-
~ ~ xป tn %
fVl .vk.
01/01/12 07/01/12
01/01/13 07/01/13 01/01/14
DATE
SiteJD
~ 420950025
~ 420950027
i i
07/01/14 01/01/15
delta 5.702 km
40-
20
m = 0.88; r2 = 0.835
4
~
~ ~
~
~ i
WW****
*. .
J|^
t> -:
~
~
~ ~
i i
i i
20
40
seas
~ DJF
~ JJA
~ MAM
~ SON
420950025
Figure 98: Comparison of I data, for sites 420950025 and 420950025. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r&), wit l|i:i;fc>t;ซ points are colored by month.
-------
County 1: Philadelphia State 1: Pennsylvania
Sites: 421010047 & 421010057
~ 421010047
~ 421010057
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
20
421010047
Figure 99: Comparison of I data, for sites 421010047 and 421010047. Top panel shows time Series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r2)j wit points are colored by month.
-------
County 1: Philadelphia State 1: Pennsylvania
Sites: 421010055 & 421010047
40-
30-
20-
10-
~
~
~~
~
~ t
~
w w
**
i * 1*1 ~ ~
* v., 'diii? /;ปซif.
fjr* : ~ SSiT*A1
SiteJD
~ 421010047
~ 421010055
01/01/12 07/01/12 01/01/13
07/01/13
DATE
01/01/14 07/01/14 01/01/15
421010055
Figure 100: Comparison of I'M- - data for sites 421010055 and 421010055. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r2)j with3$ata points are colored by month.
-------
County 1: Providence State 1: Rhode Island
Sites: 440070022 & 440071010
40-
30-
ฃ20-
10-
SiteJD
~ 440070022
~ 440071010
i i i i i i i
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
20
440070022
Figure 101: Comparison of I'M- - data for sites 440070022 and 440070022. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r2)j wit a points are colored by month.
-------
County 1: Salt Lake State 1: Utah
Sites: 490353006 & 490353010
60-
40-
20-
~
<
r
~
~
<
*
<
V
2
ฆฆ
jr
V
J
(
ฆ
SiteJD
~ 490353006
~ 490353010
01/01/12 07/01/12 01/01/13
07/01/13
DATE
01/01/14 07/01/14 01/01/15
60-
40-
20-
delta 7.274 km
m = 0.92; r2 = 0
~
.888
~
~
~
~
20
40
60
seas
~ DJF
~ JJA
~ MAM
~ SON
490353006
Figure 102: Comparison of I'M- - data for sites 490353006 and 490353006. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (rS), witlliSjata points are colored by month.
-------
County 1: New Castle State 1: Delaware
Sites: 100032004 & 420450002
~ 100032004
~ 420450002
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
20
100032004
Figure 103: Comparison of I'M- - data for sites 100032004 and 100032004. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r&), with3iata points are colored by month.
-------
County 1: District of Columbia State 1: District Of Columbia
Sites: 110010043 & 510590030
40-
30-
20-
10-
~
K *: SI IBS!
~ ~
~ ซ
~
~
1
A
A
ฆc
i i i i i i i
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
SiteJD
~ 110010043
~ 510590030
20
110010043
Figure 104: Comparison of I'M- - data for sites 11001004.3 and 110010043, Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r$), witlti3งata points are colored by month.
-------
County 1: Cobb State 1: Georgia
Sites: 130670003 & 130890002
~ 130670003
~ 130890002
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
130670003
Figure 105; Comparison of I'M- - data for sites 130670003 and 130670003, Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r2)j wit l|i ij&ta points are colored by month.
-------
30-
County 1: Hawaii State 1: Hawaii
Sites: 150011006 & 150012023
20-
10-
u
.ป
<
~
%
~
~
. ~
~
t
. *<*ฃ. .* t.
.V* : 4-,
/.'S> ป $ *.ซป
>ฃ&
f
~
SiteJD
~ 150011006
~ 150012023
01/01/12
01/01/13
07/01/13
DATE
01/01/14 07/01/14 01/01/15
150011006
Figure 106: Comparison of FM2.5 data for sites 150011006 and 1S0011006. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r2)j wit l|i klata points are colored by month.
-------
County 1: Hawaii State 1: Hawaii
Sites: 150011012 & 150012020
~ 150011012
~ 150012020
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
150011012
Figure 107: Comparison of FM^.s data for sites 150011012 and 1S0011012. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (rS), wit l|i iiata points are colored by month.
-------
County 1: Hawaii State 1: Hawaii
Sites: 150012016 & 150012020
100 -
75-
50-
SiteJD
~ 150012016
~ 150012020
25-
i i i lii
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
100-
75-
50-
25-
delta 32.771 km
m = 0.28; r2 = 0.0749
o-
seas
~ DJF
~ JJA
~ MAM
~ SON
25
50
150012016
75
100
Figure 108: Comparison of FM^.s data for sites 150012016 and 1S0012014, Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r&), wit^4Jata points are colored by month.
-------
County 1: Honolulu State 1: Hawaii
Sites: 150031001 & 150031004
20-
15-
10-
SiteJD
~ 150031001
~ 150031004
i i i i i i i
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
150031001
Figure 109: Comparison of FM^.s data for sites 150031001 and 1S0031001. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r2)j wit l|i Hat a points are colored by month.
-------
County 1: Honolulu State 1: Hawaii
Sites: 150032004 & 150031004
~ 150031004
~ 150032004
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
150032004
Figure 110: Comparison of FM^.s data for sites 150032004 and 1S0032004. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r2)j with4
-------
County 1: Clark State 1: Indiana
Sites: 180190006 & 211110067
~ 180190006
~ 211110067
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
20
180190006
Figure 111: Comparison of I'M- - data for sites 180190006 and 180190006. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (rS), with4lata points are colored by month.
-------
County 1: Marion State 1: Indiana
Sites: 180970078 & 180970081
~ 180970078
~ 180970081
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
20
180970078
Figure 112: Comparison of I'M- - data for sites 180970078 and 180970078. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r2)j with4fata points are colored by month.
-------
40-
30-
20-
10-
County 1: Clinton State 1: Iowa
Sites: 190450019 & 190450021
SiteJD
~ 190450019
~ 190450021
01/01/12 07/01/12 01/01/13
07/01/13
DATE
01/01/14 07/01/14 01/01/15
40-
30-
8 20 -
10-
delta 6.425 km
m = 0.89; r2 = 0.909
10
i
20
190450019
30
40
seas
~ DJF
~ JJA
~ MAM
~ SON
Figure 113: Comparison of I'M;...- data for sites 190450019 and 190450019. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (rS), with4Sata points are colored by month.
-------
40
30
20
10
County 1: Johnson State 1: Iowa
Sites: 191032001 & 191130040
~
~
~
~
~
~
~
:
~
~
X .
i
i*
~
~
' t! 1
~
~
~
~
~
P
k4
Ms?*
~
SiteJD
~ 191032001
~ 191130040
i i i i i i i
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
delta 38.713 km
40-
30-
ฃ20-
10-
m = 0.97; r2 = 0.898
seas
~ DJF
~ JJA
~ MAM
~ SON
10
i
20
191032001
I
30
40
Figure 114: Comparison of I'M;...- data for sites 191032001 and 191032001. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (rS), witlti4งata points are colored by month.
-------
60-
40-
20-
County 1: Muscatine State 1: Iowa
Sites: 191390015 & 191630015
~
ป
:
~
t
ฆy:
? _
I ~ , ~. V* ~ ซ ~
.. Sv j
ป * r#ป*\
V ป*; ; Vj~ si j
{ m
SiteJD
~ 191390015
~ 191630015
01/01/12 07/01/12 01/01/13
07/01/13
DATE
01/01/14
i i
07/01/14 01/01/15
191390015
Figure 115; Comparison of I'M- - data for sites 191.390015 and 191390015. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r2)j with5fl)ata points are colored by month.
-------
County 1: Jefferson State 1: Kentucky
Sites: 211110051 & 180190006
~ 180190006
~ 211110051
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
20
211110051
Figure 116: Comparison of FM^.s data for sites 211110051 and 211110051. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r&), witfelata points are colored by month.
-------
40-
30-
ง20-
CL
10-
County 1: East Baton Rouge State 1: Louisiana
Sites: 220330009 & 221210001
* ~
~ ~ *
ซ* ~
.... :
1 \V.JU"
v*i.
& . i*
SiteJD
~ 220330009
~ 221210001
i i i i i i i
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
40-
30-
2 20-
10-
delta 5.419 km
m = 0.76: r2 = 0.637
seas
~ DJF
~ JJA
~ MAM
~ SON
10
20
220330009
30
40
Figure 117: Comparison of I'M;...- data for sites 220330009 and 220330009. Top panel shows time series for
both sites for yeaฃS 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r2)j witltgjata points are colored by month.
-------
County 1: Cecil State 1: Maryland
Sites: 240150003 & 420290100
40-
30-
20
10-
SiteJD
~ 240150003
~ 420290100
01/01/12 07/01/12 01/01/13
07/01/13
DATE
01/01/14 07/01/14 01/01/15
20
240150003
Figure 118: Comparison of FM2.5 data for sites 240150003 and 240150003, Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r&), witBiSjata points are colored by month.
-------
County 1: Harford State 1: Maryland
Sites: 240251001 & 245100040
30-
20
10-
: * \* ป
\
SiteJD
~ 240251001
~ 245100040
i i i i i i i
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
delta 29.292 km
30-
20-
10-
m = 0.89; r2 = 0.794
seas
~ DJF
~ JJA
~ MAM
~ SON
10
20
30
240251001
Figure 119: Comparison of FM^.s data for sites 240251001 and 240211001. Top panel shows time series for
both sites for yeaฃS 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r2)j wit l|i."i|ata points are colored by month.
-------
40-
County 1: Kent State 1: Maryland
Sites: 240290002 & 240150003
30-
|20-
ci-
IO-
Site J D
~ 240150003
~ 240290002
01/01/12 07/01/12 01/01/13
07/01/13
DATE
01/01/14 07/01/14 01/01/15
delta 44.457 km
40-
30-
ฐ20-
10-
m = 0.92; r2 = 0.727
seas
~ DJF
~ JJA
~ MAM
~ SON
i
10
20
240290002
30
40
Figure 120: Comparison of I'M;...- data for sites 240290002 and 240290002. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r2)j witBiSfota points are colored by month.
-------
County 1: Montgomery State 1: Maryland
Sites: 240313001 & 240330030
30-
20-
10-
01/01/12 07/01/12 01/01/13 07/01/13
DATE
I I
01/01/14 07/01/14
SiteJD
~ 240313001
~ 240330030
01/01/15
30
,20
10
delta 20.812
km
m = 0.87; r2 =
0.764
#
~
~
$
~~ ~ ซ!
~ /
~ v/
~
*
~
~ UP * .
y ~
~
~
~
~ ~~
. vfw #
'^r* * *
pi
i i i i
0 10 20 30
seas
~ DJF
~ JJA
~ MAM
~ SON
240313001
Figure 121: Comparison of I'M- - data for sites 240313001 and 240313001. Top panel shows time series for
both sites for yeaฃS 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r2)j witlL5งata points are colored by month.
-------
County 1: Prince George's State 1: Maryland
Sites: 240330030 & 110010043
30-
20-
10-
i i i i i i i
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
SiteJD
~ 110010043
~ 240330030
240330030
Figure 122: Comparison of I'M;...- data for sites 240330030 and 240330030. Top panel shows time series for
both sites for yeaฃS 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r2)j withsfata points are colored by month.
-------
County 1: Wayne State 1: Michigan
Sites: 261630001 & 261630039
~ 261630001
~ 261630039
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
20
261630001
Figure 123: Comparison of I'M- - data for sites 261630001 and 261630001. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r2)j witlLSSata points are colored by month.
-------
County 1: Anoka State 1: Minnesota
Sites: 270031002 & 270530963
30-
20-
10-
* .~ : I
* * |
, . V * v
t I ป # ซ*
*" . 4ป*: <. 2.4
.*>vป . .ป . >
tv '!': 'ซ. ;'i -a!A iป . t > u
mli ฃ x?
Site J D
~ 270031002
~ 270530963
i i i i i i
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
30-
ฆ 20-
10-
delta 20.867 km
m = 0.97; r2 = 0.827
10
20
30
seas
~ DJF
~ JJA
~ MAM
~ SON
270031002
Figure 124: Comparison of I'M;...- data for sites 270031002 and 270031002. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r&), witlL5งata points are colored by month.
-------
County 1: Hennepin State 1: Minnesota
Sites: 270530963 & 271230871
~ 270530963
~ 271230871
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
delta 17.526 km
30-
.20-
10-
m = 0.85; r2 =
0.806
? * V-
* *
seas
~ DJF
~ JJA
~ MAM
~ SON
10
20
30
270530963
Figure 125; Comparison of I'M- - data for sites 270S3096.3 and 270530963, Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r&), with6fl)ata points are colored by month.
-------
County 1: Washington State 1: Minnesota
Sites: 271630447 & 271630448
30-
20-
10-
* -f
At ^ f ~ *$!/~;*< *t *b
SiteJD
~ 271630447
~ 271630448
i i i i i i i
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
30
: 20
10
delta 1.04 km
m = 0.94; r:
2 = 0.857
/ /<
y<-'
// 4
ซ
~ . ~
~ ~
~ ~ Iwft
i.W51'
i ififlni *
ป
~
1
*:*
it>n
.JgP
\iri ~*
cr*
i i
l%iป ~
p* ~
ป
i
i
1
10
20
271630447
30
seas
~ DJF
~ JJA
~ MAM
~ SON
Figure 126: Comparison of I'M- - data for sites 271630447 and 271630447. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r&), wit lliolata points are colored by month.
-------
County 1: Cass State 1: Missouri
Sites: 290370003 & 290950034
~ 290370003
~ 290950034
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
290370003
Figure 127: Comparison of I'M;...- data for sites 290370003 and 290370003, Top panel shows time series for
both sites for yeaฃS 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r2)j wit Btfijata points are colored by month.
-------
County 1: Clay State 1: Missouri
Sites: 290470005 & 290950034
~ 290470005
~ 290950034
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
delta 27.712 km
30-
8 20 ฆ
10-
m = 0.88; r2 = 0.758
# # -A
seas
~ DJF
~ JJA
~ MAM
~ SON
10
20
290470005
I
30
Figure 128: Comparison of I'M;...- data for sites 290470005 and 290470005, Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r&), witltiffilata points are colored by month.
-------
County 1: Jefferson State 1: Missouri
Sites: 290990019 & 295100007
j
:~
: .~
KAgJj
pp
\
t ~ *
L * 1 * $*t ~ *
i ~ \f ~
s* - JSav:* u m ฃ
R 4 *i*_~*
*. ~ .tf yv
*~ A #
~
~
ซ
1
: :?
>
tซ ~
fit
1
~
w
SiteJD
~ 290990019
~ 295100007
i i i i i i i
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
delta 15.725 km
40
20
m = 0.93; r2 = 0.819
yS,'''
~
~ . \
~ 1A 1
y
L
~ .
jpp
~
1 1 1
0 20 40
seas
~ DJF
~ JJA
~ MAM
~ SON
290990019
Figure 129: Comparison of I'M- - data for sites 290990019 and 290990019. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r&), wit l|i(i|ata points are colored by month.
-------
County 1: Saint Louis State 1: Missouri
Sites: 291893001 & 295100085
t
i i i i iii
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
SiteJD
~ 291893001
~ 295100085
delta 13.216 km
20 40
291893001
seas
~ DJF
~ JJA
~ MAM
~ SON
Figure 130: Comparison of I'M- - data for sites 291893001 and 291893001. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r2)j with6$ata points are colored by month.
-------
County 1: St. Louis City State 1: Missouri
Sites: 295100007 & 295100085
CL
on
SiteJD
~ 295100007
~ 295100085
0
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
delta 13.89 km
40-
20-
m = 0.96; r2 = 0.842
ซ z*
***s
20
295100007
40
seas
~ DJF
~ JJA
~ MAM
~ SON
Figure 131: Comparison of I'M;...- data for sites 291100007 and 295100007. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r2)j with6lata points are colored by month.
-------
County 1: Lewis and Clark State 1: Montana
Sites: 300490004 & 300490026
40-
20
~
:~
~
n ~
vซ
~*
t
01/01/12 07/01/12 01/01/13 07/01/13
DATE
I I I
01/01/14 07/01/14 01/01/15
SiteJD
~ 300490004
~ 300490026
300490004
Figure 132: Comparison of I'M- - data for sites 300490004 and 300490004. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r&), with6fata points are colored by month.
-------
100 ~
County 1: Missoula State 1: Montana
Sites: 300630024 & 300630037
75-
50
25-
%
~
~
~
SiteJD
~ 300630024
~ 300630037
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
delta 24.47 km
100-
75
50
25
/ 4
m = 0.8;
r2 = 0.837
~
' *
ป
ซ
V ~ ~
^ a* flr(k ' a
~
ป 'w
wgr~z
~
seas
~ DJF
~ JJA
~ MAM
~ SON
25
50
300630024
75
100
Figure 133: Comparison of I'M- - data for sites 300630024 and 300630024. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (rS), witheSata points are colored by month.
-------
County 1: Douglas State 1: Nebraska
Sites: 310550019 & 311530007
40-
30-
20-
10-
~
~
~
~ ~
~
~ ~ ~ <
~
v
4, A. ~
* , jft *
>* -/V ปปf 4 jljl* ~ *
~
I ~
~
~
I
~
I I I I I I I
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
SiteJD
~ 310550019
~ 311530007
310550019
Figure 134: Comparison of I'M- - data for sites 310550019 and 310550019. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (rS), witlti6งata points are colored by month.
-------
County 1: Washoe State 1: Nevada
Sites: 320310016 & 320311005
100
75-
50-
25
o-
~
~
~
~
~
~
X
%
~
~
~
~ i
k
i i
L
~
i
ป
~
<
Sl.
f A*
K JL
~if c i#
M9a
V
un
"pH*
n
~
&
SiteJD
~ 320310016
~ 320311005
i i i i i
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
delta 5.522 km
100
75-
50-
25-
o-
m = 0.95;
r2 = 0.779
~
// ~
~
*
yr ~
#
~
~ *
***~
dHULปซ
~
c
i i i i
) 25 50 75 100
seas
~ DJF
~ JJA
~ MAM
~ SON
320310016
Figure 135; Comparison of I'M- - data for sites 320310016 and 320310016. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (rS), with7<(lata points are colored by month.
-------
40-
30-
(Q 20"
5
CL
10-
County 1: Cheshire State 1: New Hampshire
Sites: 330050007 & 330115001
~
~
~ ~
#
I I I I I I I
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
SiteJD
~ 330050007
~ 330115001
40
30
ฐ20
10-
delta 33.029 km
m = 0.26; r
2 = 0.286
~ v
'*&-. . :
. KjSOmw* r c ซ
~ ซ/~~~
seas
~ DJF
~ JJA
~ MAM
~ SON
10
l
20
330050007
30
40
Figure 136: Comparison of FM^.s data for sites 330050007 and 330050007. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r&), wit I|i7<|ata points are colored by month.
-------
County 1: Rockingham State 1: New Hampshire
Sites: 330150018 & 330115001
~ 330115001
~ 330150018
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
delta 40.652 km
330150018
Figure 137: Comparison of FM^.s data for sites 330150018 and 330150018. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r2)j with7
-------
County 1: Hudson State 1: New Jersey
Sites: 340171003 & 340390004
50
40
30
20
10
~
~ ~
~
~I
~ ,
~
1
ป.
1 ll
r V
%
~ ~
x i :
~ ป
I
~
;
~
~
i . .. 1 ^- T
1; .
~
ป iflv
* ~
SiteJD
~ 340171003
~ 340390004
01/01/12 07/01/12 01/01/13
07/01/13
DATE
01/01/14 07/01/14 01/01/15
20
340171003
Figure 138: Comparison of I'M- - data for sites 340171003 and 340171003, Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (rS), wit SiTEfota points are colored by month.
-------
County 1: Mercer State 1: New Jersey
Sites: 340210008 & 420170012
40-
30-
'20-
10-
~
%
~ A .
:*
.~ #
~
~ *
if
~ ~
~
$
?.i6i
. ~
;*:y.
SiteJD
~ 340210008
~ 420170012
01/01/12 07/01/12 01/01/13
07/01/13
DATE
01/01/14
i i
07/01/14 01/01/15
20
340210008
Figure 139: Comparison of I'M- - data for sites 340210008 and 340210008. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r&), witlh7data points are colored by month.
-------
80-
60-
County 1: Bernalillo State 1: New Mexico
Sites: 350010023 & 350010024
:40-
SiteJD
~ 350010023
~ 350010024
20-
I
~ ~
I I I I I I
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
delta 7.947 km
80-
60-
! 40 -
20-
m = 0.43; r2 = 0.209
20
40
350010023
60
80
seas
DJF
JJA
MAM
SON
Figure 140: Comparison of I'M;...- data for sites 35001002.3 and 350010023. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (rS), with7(Sata points are colored by month.
-------
County 1: Queens State 1: New York
Sites: 360810124 & 340171003
~ 340171003
~ 360810124
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
360810124
Figure 141: Comparison of I'M- - data for sites 360810124 and 360810124. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r&), with7
-------
County 1: Mercer State 1: North Dakota
Sites: 380570004 & 380650002
~ *
40-
30-
20-
10-
SiteJD
~ 380570004
~ 380650002
i i i i i i i
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
delta 28.522 km
40
30
20
10
m = 0.68; r2 = 0.405
ซ
~
~
c-
\
i .*4M
~ v/
1$..^
Ihv i
~
~ *+
~ X*
f"<
1 1 1 1 1
0 10 20 30 40
seas
~ DJF
~ JJA
~ MAM
~ SON
380570004
Figure 142: Comparison of I'M- - data for sites 3-80570004 and 380570004. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r2)j wit l|i7
-------
50
40
30
20
10
County 1: Adams State 1: Pennsylvania
Sites: 420010001 & 420410101
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~ ~
~ ~
IB m II
4
4
_ 4
: i
vt.
~ ~ ~
~ I
~
~
<
4
Ff ~ #vr
~
i
i
>
SiteJD
~ 420010001
~ 420410101
01/01/12 07/01/12 01/01/13
07/01/13
DATE
01/01/14 07/01/14 01/01/15
420010001
Figure 143: Comparison of I'M- - data for sites 420010001 and 420010001. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r2)j with7
-------
60
40
20
County 1: Allegheny State 1: Pennsylvania
Sites: 420030008 & 420030064
~
~
~
~
~
(
ป
~
~
~
~
~
~
4
~
ป
~
~
1 1
*
^
*j* V * *1.
%VJL ~ i
~
~
/
t X
fi^y,
i i i i i i i
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
SiteJD
~ 420030008
~ 420030064
delta 17.618 km
80
60
: 40
20
,
*
y
s
*
**
m = 1.3; r2 = 0.538
~ ~
<ป
.. .--v/
. \ * * .'
*lseHF
~
~
c
i i i i
) 20 40 60
seas
~ DJF
~ JJA
~ MAM
~ SON
420030008
Figure 144: Comparison of I'M- - data for sites 420030008 and 420030008. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r2)j witlti7<งata points are colored by month.
-------
County 1: Beaver State 1: Pennsylvania
Sites: 420070014 & 421255001
~ 420070014
~ 421255001
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
delta 34.812 km
420070014
Figure 145; Comparison of I'M- - data for sites 420070014 and 420070014. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r&), witltiBdata points are colored by month.
-------
County 1: Berks State 1: Pennsylvania
Sites: 420110011 & 420750100
~ 420750100
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
Figure 146: Comparison of I'M- - data for sites 420110011 and 420110011. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r2)j wit l|is|ata points are colored by month.
-------
50-
40-
30-
20-
10-
County 1: Cumberland State 1: Pennsylvania
Sites: 420410101 & 420430401
*
h
Av
~
~ ~
~
:
:*
* t*
#
~
~
~
~ ~
~ ~ #
* #
s " *
,Sv*4 -F? ! ]$ *
'U.V v}!X;Vwซ> t
MBi>yrai! '~~iu.'iitSw,ป^ ซF
-------
County 1: Delaware State 1: Pennsylvania
Sites: 420450002 & 421010055
40
30
20
10
l
~
J
~
~
t .
~
~
~
~
~
~
~
~
t
~
a
J
_ <
s
i
*
r .*V
\ J ~442*'
ฆ " * II
C s: ..
-ft?- f
n ~ ~
~
w?
~
SiteJD
~ 420450002
~ 421010055
01/01/12 07/01/12 01/01/13
07/01/13
DATE
01/01/14 07/01/14 01/01/15
420450002
Figure 148: Comparison of I'M- - data for sites 420450002 and 420450002. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r2)j witlMJl&ta points are colored by month.
-------
County 1: Lancaster State 1: Pennsylvania
Sites: 420710007 & 420750100
~ 420710007
~ 420750100
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
420710007
Figure 149: Comparison of I'M- - data for sites 420710007 and 420710007. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r&), witlhSiata points are colored by month.
-------
County 1: Montgomery State 1: Pennsylvania
Sites: 420910013 & 421010057
40-
30-
20-
*
~~
~
~
~
; .
SiteJD
~ 420910013
~ 421010057
I I 1" I I I I
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
delta 22.121 km
40
30
5 20
10
<
ป ^
m = u.yj; = u./^o
~
~
~
~ ~
~ ~
1
~
~
~ ~~
. %
. ~ 'kij.
~ /S'
*
> x.
x
~
~ * ~
V * *
jaiflj
A A 'ISL-., . >
*
HPPEj ~
Ifฎ?*
w * ~
- *
iSMResa ' *
* *
~
~
1
1
1
i i
i
i
seas
~ DJF
~ JJA
~ MAM
~ SON
10
20 30
420910013
40
Figure 150: Comparison of I'M- - data for sites 420910013 and 420910013, Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r2)j withSfeta points are colored by month.
-------
County 1: Northampton State 1: Pennsylvania
Sites: 420950025 & 420950027
40-
20-
~ ~ xป tn %
fVl .vk.
01/01/12 07/01/12
01/01/13 07/01/13 01/01/14
DATE
SiteJD
~ 420950025
~ 420950027
i i
07/01/14 01/01/15
delta 5.702 km
40-
20
m = 0.88; r2 = 0.835
4
~
~ ~
~
~ i
*. .
J|^
t> -:
~
~
~ ~
i i
i i
20
40
seas
~ DJF
~ JJA
~ MAM
~ SON
420950025
Figure LSI: Comparison of I'M- - data for sites 420950025 and 420950025. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r&), with8lata points are colored by month.
-------
County 1: Philadelphia State 1: Pennsylvania
Sites: 421010047 & 421010057
~ 421010047
~ 421010057
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
20
421010047
Figure 132: Comparison of I'M- - data for sites 421010047 and 421010047. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r2)j witligfata points are colored by month.
-------
County 1: Philadelphia State 1: Pennsylvania
Sites: 421010055 & 421010047
40-
30-
20-
10-
~
~
~~
~
~ t
~
w w
**
i * 1*1 ~ ~
* v., 'diii? /;ปซif.
fjr* : ~ SSiT*A1
SiteJD
~ 421010047
~ 421010055
01/01/12 07/01/12 01/01/13
07/01/13
DATE
01/01/14 07/01/14 01/01/15
421010055
Figure 153: Comparison of I'M- - data for sites 421010055 and 421010055. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r2)j withSSata points are colored by month.
-------
60
40
20
County 1: Washington State 1: Pennsylvania
Sites: 421250005 & 420030064
~
~
~
~
~
~
~
~
~
ป
ป
~
ป
~
~
~ ~. ~,
<*ซ
i i
$8r
He
1 T
;
^ ' mm
%
i i
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
SiteJD
~ 420030064
~ 421250005
delta 19.927 km
60-
40-
20
m = 0.94; r2 = 0.381
v
seas
~ DJF
~ JJA
~ MAM
~ SON
20
i
40
421250005
60
Figure 184: Comparison of I'M;...- data for sites 421250005 and 421250005, Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r&), witli8ง&ta points are colored by month.
-------
County 1: Washington State 1: Pennsylvania
Sites: 421250200 & 421250005
~ 421250005
~ 421250200
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
421250200
Figure 185; Comparison of I'M- - data for sites 421250200 and 421250200. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (rS), with9fl)ata points are colored by month.
-------
County 1: Washington State 1: Pennsylvania
Sites: 421255001 & 421250200
~ 421250200
~ 421255001
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
421255001
Figure 156: Comparison of I'M- - data for sites 421255001 and 421255001. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (rS), witliSi&ta points are colored by month.
-------
60
40
20
County 1: Westmoreland State 1: Pennsylvania
Sites: 421290008 & 420030064
~
~
~
~
~
~
~
~
~
1
ป
i
i
<
1
aM.,: \
mirg jag J rT1 &IL2RiSSl *
9 IF #*Tv _ kVJ]
** '*! ^ J j ซ m MW1& fป5* '
f ' 1 | 1 r **ซ
~
~
~
4
|ฃg
Wr* f
i i i i i i i
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
SiteJD
~ 420030064
~ 421290008
delta 30.834 km
60
40
20
~
m = 0.9; r2 = 0.
~
* ~
329
~
~ V
~ *
~ ซ
!~* /
:m3&
MS0
'SbSkPS
JSPr*
~ /
>\y*Z
vTป
~ ~
ซ
till
0 20 40 60
seas
~ DJF
~ JJA
~ MAM
~ SON
421290008
Figure 157: Comparison of I'M- - data for sites 421290008 and 421290008. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (rS), wit I|i9i;iia points are colored by month.
-------
County 1: York State 1: Pennsylvania
Sites: 421330008 & 420430401
50-
t
40-
30-
20-
10-
: 9
~~t J
ป
~ ~.
~ vJj
* >
.
>
ปgya * at/ ~ V
/>ซVSKtS AfDH*& x&i
SiteJD
~ 420430401
~ 421330008
01/01/12 07/01/12 01/01/13
07/01/13
DATE
01/01/14 07/01/14 01/01/15
421330008
Figure 158: Comparison of I'M- - data for sites 421.330008 and 421330008. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r&), witltiQiata points are colored by month.
-------
County 1: Kent State 1: Rhode Island
Sites: 440030002 & 440070022
40-
30-
20-
10-
>\
SiteJD
~ 440030002
~ 440070022
i i i i i i i
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
delta 33.186 km
40-
m = 0.93: r2 = 0.532
8 20 -
seas
~ DJF
~ JJA
~ MAM
~ SON
10 20
440030002
Figure 159: Comparison of I'M;...- data for sites 440030002 and 440030002. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r&), wit points are colored by month.
-------
County 1: Providence State 1: Rhode Island
Sites: 440070022 & 440071010
40-
30-
ฃ20-
10-
SiteJD
~ 440070022
~ 440071010
i i i i i i i
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
20
440070022
Figure 160: Comparison of I'M- - data for sites 440070022 and 440070022. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r2)j withOfeta points are colored by month.
-------
County 1: Charleston State 1: South Carolina
40
30
20
10
Sites: 450190048 & 450190049
~
~
1
A
i
4
4
~
# 1 ^ ' 5 ฃ I
it{# .tv
uk'iu
SiteJD
~ 450190048
~ 450190049
i i i i i i i
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
delta 23.295 km
m = 0.67; r2 = 0.626
40-
30-
20-
10-
seas
~ DJF
~ JJA
~ MAM
~ SON
10
i i
20 30
450190048
I
40
Figure 161: Comparison of I'M;...- data for sites 450190048 and 450190048. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (rS), with9lata points are colored by month.
-------
County 1: Greenville State 1: South Carolina
Sites: 450450015 & 450830011
~ 450450015
~ 450830011
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
delta 46.412 km
30
20
10
~
m = 0.9; r2 = 0.866
#
ป<
//'''
XV''*
* *SS
r
, ป&&&*'
MWrfc
*
V, ~
ป
1 jW
5*1
~ *
Br t
i i i i
0 10 20 30
seas
~ DJF
~ JJA
~ MAM
~ SON
450450015
Figure 162: Comparison of I'M- - data for sites 450450015 and 450450015. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r2)j wit l|i!ปtata points are colored by month.
-------
County 1: Lexington State 1: South Carolina
Sites: 450630008 & 450790019
30-
-------
County 1: Harris State 1: Texas
Sites: 482011035 & 482011039
30-
20-
10-
SiteJD
~ 482011035
~ 482011039
t .
i i i i i i i
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
482011035
Figure 164: Comparison of I'M- - data for sites 482011035 and 482011085, Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r2)j witlti9งata points are colored by month.
-------
County 1: Salt Lake State 1: Utah
Sites: 490353006 & 490353010
60-
40-
20-
~
<
r
~
~
<
*
<
V
2
ฆฆ
jr
V
J
(
ฆ
SiteJD
~ 490353006
~ 490353010
01/01/12 07/01/12 01/01/13
07/01/13
DATE
01/01/14 07/01/14 01/01/15
60-
40-
20-
delta 7.274 km
m = 0.92; r2 = 0
~
.888
~
~
~
~
20
40
60
seas
~ DJF
~ JJA
~ MAM
~ SON
490353006
Figure 165; Comparison of I'M- - data for sites 490353006 and 490353006. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (rS), wit&Odata points are colored by month.
-------
120 -
80-
40
County 1: Utah State 1: Utah
Sites: 490490002 & 490494001
t
i
~
t
~
~
K
1
<
<
4
*
<
h
>**
3
n:
ur ~ jjf*
\h:A i
t Pit* Lr*Jk . il
l .
K If jฃ
~
SiteJD
~ 490490002
~ 490494001
01/01/12 07/01/12 01/01/13
07/01/13
DATE
01/01/14 07/01/14 01/01/15
delta 10.672 km
m = 1; r2 = 0.934
100-
50-
o-
seas
~ DJF
~ JJA
~ MAM
~ SON
40
80
120
490490002
Figure 166: Comparison of I'M- r, data for sites 490490002 and 490490002. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r2), wit Liu lata points are colored by month.
-------
80
60
^ 40
20
County 1: Weber State 1: Utah
Sites: 490570002 & 490353010
4
<
>
t.
*
ป
i
(
~:
%
ฆฆ
~t*
If
t ~ i
ซ ft v
> i jdtw #
ซK . ;|CJt ซ*ซ
~rf # w * ^
% ฃ.
ป V
t, '
. \ *:&< .
~ r ฅ ~
~
~
%
y
~
SiteJD
~ 490353010
~ 490570002
01/01/12 07/01/12 01/01/13
07/01/13
DATE
01/01/14 07/01/14 01/01/15
490570002
Figure 167: Comparison of I'M- - data for sites 490570002 and 490570002. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r2)j witSOfeta points are colored by month.
-------
County 1: Pierce State 1: Washington
Sites: 530530029 & 530332004
40-
30-
20-
10-
~
~
~
~
~
n
~
~
~
~
~
~ S* it\t8
i1' ! -i* ^ ฆ 'Sl^j. * iii it* ^ปr * . ^ ~ ฆ i >jH
SiteJD
~ 530332004
~ 530530029
01/01/12 07/01/12 01/01/13
07/01/13
DATE
01/01/14 07/01/14 01/01/15
20
530530029
Figure 168: Comparison of I'M- - data for sites 530530029 and 530530029. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r2), witSOiata points are colored by month.
-------
County 1: Snohomish State 1: Washington
Sites: 530610005 & 530330080
30-
20-
10-
<
~
~
#
i
~
(
>
~
ป tT
~
1
~~
i
~
~ ~
~
4
~
t
f {/ :
% tfp
% Si
' '
it
* l* ~ It ซ
ป ~
*
.
~~
~ . ปt
SiteJD
~ 530330080
~ 530610005
01/01/12 07/01/12 01/01/13
07/01/13
DATE
01/01/14 07/01/14 01/01/15
530610005
Figure 169: Comparison of I'M- - data for sites 530610005 and ง30610005, Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (rS), wit Lfulata points are colored by month.
-------
County 1: Snohomish State 1: Washington
Sites: 530610020 & 530611007
60-
~
40-
20-
#~
SiteJD
~ 530610020
~ 530611007
01/01/12 07/01/12 01/01/13
07/01/13
DATE
01/01/14 07/01/14 01/01/15
delta 47.348 km
60-
m = 0.58: r2 = 0.368
40"
20
* .
~ .. ... \ *
v y?\
.
~V 4* f ' ~**
~, -1 * t ~
*9* *
V/
'~f* ~
20
i
40
60
530610020
seas
~ DJF
~ JJA
~ MAM
~ SON
Figure 170: Comparison of I'M;...- data for sites 530610020 and ง30610020. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (x2), witSoSata points are colored by month.
-------
County 1: Snohomish State 1: Washington
Sites: 530611007 & 530610005
60
40
lO
rsi
5
CL
SiteJD
~ 530610005
~ 530611007
20
0
01/01/12 07/01/12 01/01/13 07/01/13 01/01/14 07/01/14 01/01/15
DATE
530611007
Figure 171: Comparison of I'M- - data for sites 530611007 and ง30611007. Top panel shows time series for
both sites for years 2012-2014. Bottom panel shows scatter plot of paired data, along with slope for the
linear regression and correlation coefficient (r2), witSOiata points are colored by month.
-------
County 1: Brown State 1: Wisconsin
Sites: 550090005 & 550870009
40-
30-
20-
10-
j:
~
V#
~
# ~
~
~
ป
t
# *~ ~~~
~
~
~ ~
~t
I*
Si ~ ; i 4 I / f #t i ~ i j V
I |l*\ .~ f 4 h % 2^*e #}v a #ji
~ *
-------
5 Comparison of air quality variability for ozone sensitivity tests
Results from the ozone sensitivity analysis discussed in Section 2.2.3.
5.1 All available data, no quarterly subsets
208
-------
2013_files_no_QT_all_sens_test bootstrap 50th percentile uncert
6% ฆ
c
'ro
-e
ฉ
^ 4% ฆ
=}
CD
>
J9 2% ฆ
0
0%-
mean 1.512%
itiedian 1.427%
t
* M
% *V
ฆฆ
40
60 80
4th 8-hr high NAAQS
100
LONG
1-80
-100
-120
-140
2013_files_no_QT_all_sens_test bootstrap 50th percentile uncert
>
-ป-
c
A-
t 4.
0
o
c
ฉ
3 2-
O
w
CD
CD
50"
45-
40-
35-
30-
25-
mean 1.075
median 1.000
. :
40
60
80
100
4th 8-hr high NAAQS
4th MDA8, rel uncert (%)
2.0%
1.5%
1.0%
0.5%
0.0%
-120
-100
-80
long
Figure 173: Bootstrap results lor the ozone 2013 DVs, showing the 50%, 68%, 75%, and 95% CIs, along with
the mean and median bootstrap DVs. The top panel shows the DVs at the various CIs, the middle panel
shows the relative difference between the CI and the actual DV, and the bottom panel shows the distribution
ol the relative differences between the CI and the actual DV.
209
-------
All available data, with quarterly subsets
210
-------
5% -
.E 4% ฆ
ro
ฉ
o 3% ฆ
a
> 2% ฆ
>
-i'
.E 3-
ro
ฆc
0
o
ฉ
ฐ 1
.Q
CD
ro
50-
45-
40-
35-
30-
25-
2013_files_QT_sens_test bootstrap 50th percentile uncert
mean 1.517% rrfkdian 1.423%
1%-
LONG
1-80
-100
-120
-140
30
50 70
4th 8-hr high NAAQS
90
110
2013_files_QT_sens_test bootstrap 50th percentile uncert
mean 1.073 median 1.000
. :_<&
30
50
70
90
110
4th 8-hr high NAAQS
4th MDA8, rel uncert (%)
2.0%
1.5%
1.0%
0.5%
0.0%
-120
-100
-80
long
Figure 174: Bootstrap results lor the ozone 2013 DVs, showing the 50%, 65%, 75%, and 95% CIs, along with
the mean and median bootstrap DVs. The top panel shows the DVs at the various CIs, the middle panel
shows the relative difference between the CI and the actual DV, and the bottom panel shows the distribution
ol the relative differences between the CI and the actual DV.
211
-------
6 Analysis of temporal lag on ozone data and results from a
blocked bootstrap sensitivity analysis
This section presents results from an analysis to examine the temporal correlation of air quality levels, i.e.,
the tendency of high concentration days to occur after other days with high concentrations. Such behavior,
if present, would be a function of both emission trends (e.g., weekday traffic versus weekend traffic) and
meteorology (e.g., high pressure systems often hinder the transport of pollutants and also accompany higher
temperatures, which tend to increase the formation of ozone). The primary motivation for this assessment
is to determine whether the implementation of a block bootstrap procedure is needed for the bootstrapping
analysis described in Section 2.2.3 in order to account for possible temporal correlation, and if so, what is the
appropriate block size. If not properly accounted for, correlation can affect the assessment of uncertainty (i.e.,
the standard errors used to calculate confidence intervals). While in this analysis confidence intervals were
constructed using empirical percentiles, it is important to consider whether autocorrelation may be affecting
the distributional characteristics of the bootstrapped data. Thus a sensitivity analysis is considered.
A block bootstrap method can be used in the presence of autocorrelation to replicate the correlation
structure in the data. Blocks are designed such that the dependence between adjacent or closely spaced
measurements is contained within a block, and there is induced independence between measurements in
adjacent blocks. Block size selection can be tricky, as the blocks should be large enough to induce inde-
pendence but small enough to retain important characteristics of the data, including natural variation and
overall trends (i.e., the variance-bias trade-off for avoiding over-smoothing). There is no one agreed upon
method for selection of block size for bootstrapping procedures. Many considerations can come into play,
including practical issues and subject-matter scientific expertise. The analysis presented here first attempts
to determine the "length of lag" in the ambient ozone data (i.e., how long do correlations of concentrations
between MDA8 values persist). Based on the lag analysis, a secondary "blocked" bootstrap analysis was
completed which sampled blocks of days corresponding to the lag found in the initial analysis. Ultimately,
a 7-day lag was selected from the lag analysis. The resultant bootstrap results were similar to the original
non-parametric bootstrap, which sampled individual days rather than blocks of 7-days.
6.1 Analysis procedure and results
The R software package [R Core Team, 2017] was used to conduct the lag analysis. The acf (autocorrelation
function) and pacf (partial autocorrelation function) were used to determine the autocorrelation of the time
series of MDA8 values at each measurement site for all data available from 2016. The results from the
network-wide correlations were summarized in Figures 175 and 176.
The results from the acf analysis suggest that autocorrelations drop off after lag 3, as the mean and
median correlation coefficients at 4, 5 and 6 days lag are equivalent (top panel of Figure 175). While the
correlations are still within the 95% confidence interval returned from acf out to the 6-day lag, the fact that
the distribution of the differences between the individual correlation and the confidence intervals (middle
and bottom panel of Figure 175) are virtually identical starting at 3 days of lag suggest that correlations
at this level would be found at any lag period. The pacf analysis accounts for the autocorrelation found
in the previous lag periods (i.e., the correlation found for the 2-day lag removes the correlation found from
the 1-day lag). The results from this analysis suggest that the autocorrelation is only significant to one day.
Taking these results into account, a 3-day lag should appropriately account for any autocorrelation in the
ozone data. This is implemented in the bootstrapping analysis via a 7-day block size to account for +- 3 days
surrounding the sampled daily value. Thus, prior to bootstrapping, the data is grouped into fixed blocks of
size n 7 and the sampline with replacement is performed on the blocks. This is also consistent with block
sizes used in Inoue and Shintani [2006] and Hall and Horowitz [1996]. A 7-day block size also addresses the
consideration of weekly (7-day) pollution patterns across weekday to weekend that may exist.
A second bootstrap analysis was completed for the 2016 ozone data using a block sampling method,
with the 7-day block sample size, in order to determine the effect of possible temporal autocorrelation on
the bootstrap confidence intervals. The analysis was conducted with the R "boot" package with the tsboot
(time seties bootstrap) package with block resampling with fixed block lengths. Simple blocking, rather
than overlapping blocks of randomly varying widths, should suffice for initial consideration of possible effect
of dependence on bootstrapped confidence intervals (Lahiri [1999] and Andrews [2002]). The results from
212
-------
this bootstrap approach are shown in Figures 177 and 178 (which can be compared to the results from the
non-paremetric bootstrap in Sections 3.1 and 4.2 in the main document). The results detailed in Figure 178
indicate slightly greater variability in the blocked bootstrap result, with the mean variability from the blocked
bootstrap was 1.62%, versus 1.42% from the non-parametric bootstrap, and the median was 1.55%, versus
1.47% from the non-parametric bootstrap, while there are a lew sites with notable larger variability, as with
the non-parametric bootstrap, there is no large-scale trend in the variability. The only location ol note is
perhaps the Uinta Basin in Utah, where a cluster ol sites are grayed out in the map, indicating variability
greater than the color scale. These sites have the highest variability from the blocked bootstrap. The Uinta
Basin is known to have a unique patter in high-ozone days, with the maximum concentrations occurring in
the winter during unique meteorological events, such that the high days are always clustered together. As a
result, this highly unique ozone pattern has distinctly different results in the blocked bootstrap as compared
to the non-parametric bootstrap.
213
-------
median and mean correlation coefficient using acf
1.0
0.50-
-0.25-
lag_0_delt lag_1_delt lag_2_delt lag_3_delt lag_4_delt lag_5_delt lag_6_delt
variable
lag
median and mean difference between cc and 95% CI using acf
0.75-
d)
"O
0.50-
lag
box plot of delta between corr and 95% CI at each lag
0.75-
Figure 175: Mean (red lines) and median (black lines) correlations from the acl analysis for ozone data from
2016. 214
-------
median and mean correlation coefficient using pacf
0.6
o
o
0.2-
2
6
4
lag
median and mean difference between cc and 95% CI using pacf
0.4
-------
25-
20-
0)
O
c
E
15-
5
10-
>
Q
5-
o-
rLi
bootstrap metric
9
CI limits
50% CI
68% CI
75% CI
95% CI
max/min
median
40 60 80 100
Daily DV (ppb)
CI limits
50% CI
68% CI
75% CI
95% CI
max/min
median
X!
CL
Q.
>
Q
Q.
CO
100-
40 60 80 100
Daily DV (ppb)
Figure 177: Blocked-bootstrap results for the ozone 2016 DVs, showing the 50%, 68%f 75%, and 95% CIs,
along with the mean and median bootstrap DVs, The top panel shows the DVs at the various CIs, the
middle panel shows the relative difference between the CI and the actual DV, and the bottom panel shows
the distribution of the relative differences between the CI and the actual DV.
216
-------
2016 bootstrap 50th percentile uncert
>
-i'
c
'ro
ฉ
o
a
CD
>
_ro
0
9% -
6% -
3% -
0%-
mean 1.620%
wedian
^5
a1-
CO
IT)
~
ซ
"* ซ
1
:!
A
w
:
LONG
-80
-100
-120
-140
40
60 80
4th 8-hr high NAAQS (ppb)
100
2016 bootstrap 50th percentile uncert
.Q
Q.
CL
-i'
C
'co
ฆc
0
o
c
ฉ
o
w
.Q
CD
mean 1.074
1 median 1.000
40
60 80
4th 8-hr high NAAQS (ppb)
100
4th MDA8, rel uncert (%)
CD
50-
45-
40-
35-
30-
25-
4%
3%
2%
1%
0%
-120
-100
-80
long
Figure 178: Blocked-bootstrap results from the 50% CIs for the 2016 ozone DVs. The top panel shows the
relative difference between the CI and the actual DV, the middle panel shows the absolute difference between
the values for the DVs at each site and the CI, and the bottom panel shows the spatial distribution ol the
relative difference between the 50% CIs for the 2015 ozone DV at each site.
217
-------
7 Results from cluster analyses and other spatial groupings
This section presents results from several cluster analyses and other analyses conducted to examine the
presence of spatial groupings or trends. If strong correlation in the variability can be found in natural
spatial groupings, there may be reason to consider the variability at a regional, rather than national, level.
The primary purpose of this analysis is to attempt to identify natural spatial groupings and determine if
there strong correlations in the variability within these spatial groupings and if the variability between spatial
groupings are significantly different. Since there is no clear pathway to determine the spatial correlations,
the analysis presented here consists of several iterations of cluster analysis as well as an analysis of variability
based on well-established climate regions to explore this issue from various perspectives.
7.1 Cluster analyses
Cluster analysis is an analysis technique that attempts to group data by similar characteristics of the data
in question. This is generally done by assigning quantitative values to each characteristic and measuring
and minimizing the "distance" between the existing clusters. The "distance" parameter can be calculated in
a variety of ways, but the most common (and the one used here) is simply the Euclidian distance between
the input variables. Two types of clustering algorithms are applied, a K-means algorithm and a hierarchical
algorithm. The K-mean algorithm uses a pre-determined number of clusters and initially randomly assigns all
items to clusters. The distance between cluster centers and all individuals are calculated, then individuals
are reassigned to their closest cluster. The algorithm repeats a set number of times or until a minimum
convergence threshold is reached. Hierarchical algorithms do not use a predetermined a number of clusters,
but instead start with each individual as part of their own cluster. The first step in a hierarchical analysis
combines the two closest clusters (which are just the two closest members at the first step). Each subsequent
step combines the next closest clusters, until only 2 clusters are left. The R software package [R Core Team,
2017] was used to conduct the cluster analysis, using the kmeans and hclust functions. The analysis was
performed on the results from the 2014-2016 PM variability results, as described in the following sections.
7.1.1 Cluster analysis with latitude, longitude, and variability values
This cluster analysis used the latitude, longitude (both in degrees eastwest and northsouth), and the relative
variability (as a percentage of the site's DV). Thus, the distance between individuals and clusters is defined
as the difference between the latitude, longitude, and relative variability. Since the longitudes and latitude
varies on a much larger scale between sites (longitude ranges from -64 to -160 degrees, latitude ranges from
17 to 64 degrees) than the relative variability (0-5 percent for the annual and 0-75 percent for the daily
DVs), the spatial input component will have a greater impact on the resulting than the site-level variability
(clusters for the annual and daily DVs were computed separately). That is, the spatial closeness will be the
primary factor in forming these clusters, but the analysis will then try to group nearby sites with similar
levels of variability. Hierarchical and K-mean clustering were applied independently.
The clusters formed from this analysis is shown in Figures 179 and 180 and statistics are summarized in
Tables 1- 4. The K-means analysis used 10 clusters, which was picked based on the number of EPA Regions.
The figure also shows the hierarchical cluster results at 10 clusters for comparison. The clusters from the
hierarchical analysis have relatively little recognizable geographic correlation. For example, cluster 1 (orange
circles in both the annual and 24-hr figures) consists of a group of sites over California and Arizona and a
group over the south eastern US (Florida, Georgia, Alabama), with a major discontinuity in this grouping,
with no data points in New Mexico, Texas, Louisiana, and Mississippi. Table 1 and 2 show the statistics
from the hierarchical clusters for the annual and 24-hr standards. The table includes a comparison of the
mean variability from each cluster to the mean from the entire dataset using a Welch Modified Two-Sample
t-Test (determined from the tsum.test function from the BSDA package in R) to determine if the means
are significantly different. For the annual standard, the p values are all fairly high, with the smallest value
just over 0.1, which is well above the nominal p value of 0.5 typically identified as an indicator that the
means may be different. For the 24-hr standard, there are 2 clusters with p values less than 0.05. Cluster
7 has a p value of 0.02, which may be different than the annual mean, but the sites in this cluster (light
blue squares with an "x") are spread across the country, i.e., they are not spatially distinct. Cluster 8 has
the smallest p value (0.006) and has the smallest mean variability. For the most part, this cluster is in the
218
-------
same region (purple asterisk), in the eastern US, from North Carolina up to New York. However, this cluster
is interspersed with several other clusters. Thus, while it has distinct variability values and is spatially
correlated, it is not spatially distinct.
The results from the K-mean cluster analysis are starkly different from the hierarchical analysis. The
clusters are all geographically distinct and the results of the t-test indicate that several of the clusters are
distinctly different from the mean dataset. For the annual standard, half of the clusters have p-values less
than 0.05 (2, 3, 4, 8, and 10) while 7 clusters have p-values less than 0.05 for the daily standard (1, 2, 3,
4, 5, 6, and 9, though cluster 1 and 9 only have a few members and cluster 2 is close enough to 0.05 to
discount as significantly different, leaving only 4 clusters of note). On the surface, this suggests there are
regional differences in the variability. However, the differences between the results from the annual and daily
standards suggest the result is less certain. For example, cluster 4 in the annual analysis stands out as having
the largest mean variability and a very small p-value, suggesting the variability in this subset is significantly
different from the mean dataset. However, these sites are part of a larger cluster in the daily results (cluster
8), which include California sites, and has lower mean variability than the mean from the dataset (though
not significantly different, with a p-value of 0.31). Another example of inconsistency between the annual and
daily results is cluster 3 in the annual results, which roughly correlates to cluster 6 in the daily results. In
this case, the clusters represent approximately the same geographic region. However, for the annual result,
cluster 3 has mean variability that appears to be significantly higher than the mean dataset's (p-value of
0.016), but significantly lower mean variability for the daily standard than the mean dataset's (p-value of
0.018). Thus, these particular geographic areas have higher than average variability in the long-term, but
lower than average variability in the short-term. The inconsistent results from the K-means analysis make it
difficult draw specific conclusions about the geographic nature of the variability as estimated by this analysis.
Table 1: Comparison of hierarchical clusters for lat-long-annual variability
n.sites
grp
mean
median
sd
ann.pval
107
1
1.628
1.538
0.7296
0.1969
44
2
1.917
1.786
0.7377
0.1052
43
3
1.759
1.724
0.7835
0.7972
57
4
1.759
1.744
0.5889
0.7024
43
5
1.659
1.630
0.5025
0.4108
44
6
1.745
1.765
0.4914
0.8207
73
7
1.736
1.754
0.5604
0.8929
26
8
1.674
1.714
0.6251
0.6762
49
9
1.816
1.744
0.7380
0.4171
38
10
1.660
1.453
0.8483
0.6384
524
All
1.727
1.705
0.6700
1.0000
Table 2: Comparison of hierarchical clusters for lat-long-24-hr variability
n.sites
grp
mean
median
sd
TF.pval
107
1
5.985
3.846
8.882
0.718242
44
2
6.396
5.013
5.046
0.365465
43
3
6.339
5.556
6.332
0.501141
57
4
5.150
4.054
3.758
0.370173
43
5
6.192
4.762
6.745
0.618803
44
6
5.098
4.692
3.403
0.336031
73
7
4.720
4.167
2.598
0.021686
26
8
4.338
4.006
1.917
0.005972
49
9
5.989
4.545
8.297
0.787386
38
10
6.217
4.583
6.185
0.594067
524
all
5.659
4.447
6.163
1.000000
219
-------
Table 3: Comparison of K-means clusters for lat-long-annual variability
n.sites
grp
mean
median
sd
ann.pval
61
1
1.634
1.613
0.5826
2.508e-01
26
2
1.425
1.291
0.5835
1.618e-02
96
3
1.863
1.796
0.4657
1.599e-02
26
4
2.487
2.395
0.7843
4.550e-05
58
5
1.661
1.658
0.5383
3.941e-01
38
6
1.619
1.695
0.6162
3.044e-01
57
7
1.573
1.471
0.5912
7.068e-02
66
8
1.423
1.373
0.4362
2.550e-06
57
9
1.663
1.471
0.8888
6.006e-01
39
10
2.266
2.174
0.7903
1.556e-04
524
all
1.727
1.705
0.6700
1.000e+00
Table 4: Comparison of K-means clusters for lat-long-24-hr variability
n. sites
grp
mean
median
sd
TF.pval
3
1
61.111
58.333
12.729
1.701e-02
59
2
4.794
4.762
2.367
3.604e-02
44
3
9.083
8.477
3.979
2.320e-06
104
4
3.807
3.333
2.131
8.750e-08
33
5
4.041
2.941
3.060
9.146e-03
109
6
4.751
4.545
2.849
1.831e-02
38
7
4.994
4.762
2.825
2.154e-01
68
8
5.165
4.202
3.342
3.111e-01
7
9
30.768
30.952
9.002
3.049e-04
59
10
5.020
3.846
3.136
1.938e-01
524
all
5.659
4.447
6.163
1.000e+00
220
-------
Annual NAAQS, Hierarchical clusters
-120 -100 -80
long
Annual NAAQS, kmeans clusters
-120 -100 -80
long
Figure 179: Hierarchical and K-means clusters for the annual variability for 2016.
1
2
3
4
5
6
7
8
9
10
1
2
3
4
5
6
7
8
9
10
221
-------
1
2
3
4
5
6
7
8
9
V
1
2
3
4
5
6
7
8
9
24-hr NAAQS, Hierarchical clusters
50-
45-
40-
35-
30-
25-
-120
-100
long
24-hr NAAQS, kmeans clusters
-80
50-
45-
40-
35-
30-
25-
-120
-100
long
-80
Figure 180: Hierarchical and K-means clusters for the 24-hr variability for 2016.
222
-------
7.1.2 Cluster analysis with time series of variability values (2014-2016)
This cluster analysis used the relative variability (as a percentage of the site's DV) from each site over 3
DV periods (2014-2016). Thus, the distance between individuals and clusters is defined as the difference
between each year's variability values (i.e., the variability from 2014 data, the variability from 2015 data,
the variability from 2016 data) for a particular standard. Unlike the previous analysis, all input variables
are on the same scale, such that no one parameter is driving the cluster formation. Therefore, this analysis
attempts to group sites with similar levels of variability over time in order to see if those variability trends have
spatial correlation. Since this approach incorporates the variability over time, it reflects the final composite
variability value determined in the main analysis, which is the average over 3 DV periods. Hierarchical and
K-mean clustering were applied independently.
The clusters formed from this analysis is shown in Figures 181 and 182 and statistics are summarized
in Tables 5- 8. The K-means analysis used 10 clusters, which was picked based on the number of EPA
Regions. The figure also shows the hierarchical cluster results at 10 clusters for comparison. As with the
latlongvariability analysis presented in the previous section, the clusters from the hierarchical analysis have
relatively little recognizable geographic correlation. However, most of the clusters have mean variability levels
that are distinctly different from the mean dataset (note that the mean values presented here represent the
mean from all years), though this approach also resulted in more clusters with very few members, such that
the annual results only had 5 clusters with 20 or more members and the 24-hr results only had 3 clusters
with 20 or more members. The spatial distribution of the results from the K-means analysis was similar to
the hierarchical results, in that relatively little recognizable geographic correlation. However, the K-mean
algorithm resulted in more meaningful clusters in terms of number of members and the statistical significance.
Thus, while the cluster analysis conducted with the 3-year variability trends resulted in groups that were
distinct with respect to their variability levels, it showed essentially no spatial correlation, suggesting that
geographic differences in variability do not need to be taken into account.
Table 5: Comparison of hierarchical clusters for 2014-2016 variability, annual
n'sites
grp
mean
median
sd
ann'pval
86
1
1.027
1.053
0.2387
1.418e-49
84
2
1.475
1.449
0.2964
9.408e-08
30
3
2.422
2.381
0.2443
2.756e-19
147
4
1.877
1.852
0.2730
1.498e-05
20
5
1.488
1.559
0.4272
3.258e-02
7
6
2.185
2.439
0.8897
2.129e-01
12
7
3.017
3.008
0.3734
3.841e-08
9
8
3.099
3.125
0.4706
1.585e-05
1
9
4.843
5.128
0.8896
NA
2
10
3.987
4.139
0.5126
9.888e-02
398
11
1.716
1.666
0.5855
1.000e+00
Table 6: Comparison of hierarchical clusters for 2014-2016 variability, 24-hr
n'sites
grp
mean
median
sd
TF'pval
272
1
3.630
3.333
1.662
1.499e-08
49
2
8.034
7.895
2.365
1.311e-ll
58
3
4.848
4.545
2.134
7.556e-01
9
4
11.821
11.765
2.320
1.106e-05
4
5
9.097
5.409
7.870
3.701e-01
1
6
18.480
20.312
4.399
NA
1
7
55.556
75.000
41.107
NA
1
8
25.966
29.630
7.354
NA
2
9
13.220
14.460
6.831
3.363e-01
1
10
41.250
46.667
20.340
NA
223
-------
n'sites
grp
mean
median
sd
TF'pval
398
11
4.957
4.129
4.143
1.000e+00
Table 7: Comparison of K-means clusters for 2014-2016 variability, annual
n'sites
grp
mean
median
sd
ann'pval
23
1
1.027
1.053
0.2387
1.068e-14
3
2
1.475
1.449
0.2964
2.939e-01
61
3
2.422
2.381
0.2443
1.781e-38
66
4
1.877
1.852
0.2730
3.753e-04
48
5
1.488
1.559
0.4272
1.382e-03
5
6
2.185
2.439
0.8897
3.038e-01
68
7
3.017
3.008
0.3734
4.822e-50
28
8
3.099
3.125
0.4706
3.779e-16
27
9
4.843
5.128
0.8896
9.066e-17
69
10
3.987
4.139
0.5126
2.528e-56
398
11
1.716
1.666
0.5855
1.000e+00
Table 8: Comparison of K-means clusters for 2014-2016 variability, 24-hr
n'sites
grp
mean
median
sd
TF'pval
2
1
3.630
3.333
1.662
4.577e-01
59
2
8.034
7.895
2.365
2.016e-13
44
3
4.848
4.545
2.134
7.770e-01
104
4
11.821
11.765
2.320
4.609e-65
37
5
9.097
5.409
7.870
3.098e-03
46
6
18.480
20.312
4.399
1.172e-26
2
7
55.556
75.000
41.107
3.319e-01
22
8
25.966
29.630
7.354
6.561e-12
73
9
13.220
14.460
6.831
7.360e-16
9
10
41.250
46.667
20.340
6.812e-04
398
11
4.957
4.129
4.143
1.000e+00
224
-------
1
2
3
4
5
6
7
8
9
1l
1
2
3
4
5
6
7
8
9
Annual NAAQS, Hierarchical clusters
-120 -100 -80
long
Annual NAAQS, kmeans clusters
-120 -100 -80
long
Figure 181: Hierarchical and K-means clusters lor the annual variability lor 2014-2016.
225
-------
24-hr NAAQS, Hierarchical clusters
-120 -100 -80
long
24-hr NAAQS, kmeans clusters
-120 -100 -80
long
Figure 182: Hierarchical and K-means clusters for the 24-hr variability for 2014-2016.
1
2
3
4
5
6
7
8
9
10
1
2
3
4
5
6
7
8
9
10
226
-------
7.1.3 Cluster analysis with time series of variability values (2012-2016)
This cluster analysis used the relative variability (as a percentage of the site's DV) from each site over 5
DV periods (2012-2016). Thus, the distance between individuals and clusters is defined as the difference
between each year's variability value (i.e., the variability from 2012 data, ..., the variability from 2016 data)
for a particular standard. Unlike the previous analysis, all input variables are on the same scale, such that
no one parameter is driving the cluster formation. Therefore, this analysis attempts to group sites with
similar levels of variability over time in order to see if those variability trends have spatial correlation. Since
this approach incorporates the variability over time, it partly reflects the final composite variability value
determined in the main analysis, which is the average over 3 DV periods. The extended period is evaluated
in addition to the 3 DV periods presented above in order to improve correlations that may exist with a longer
data record. Hierarchical and K-mean clustering were applied independently.
The clusters formed from this analysis is shown in Figures 183 and 184 and statistics are summarized in
Tables 9- 12. The K-means analysis used 10 clusters, which was picked based on the number of EPA Regions.
The figure also shows the hierarchical cluster results at 10 clusters for comparison. The results from this
analysis are fairly similar to the results using the 3 DV periods. The clusters are not spatially distinct;
many clusters have few members, though most have distinct variability levels. Thus, while the cluster
analysis conducted with the 5-year variability trends resulted in groups that were distinct with respect to
their variability levels, it showed essentially no spatial correlation, suggesting that geographic differences in
variability do not need to be taken into account.
Table 9: Comparison of hierarchical clusters for 2012-2016 variability, annual
n'sites
grp
mean
median
sd
ann'pval
76
1
1.138
1.099
0.2967
1.272e-27
30
2
2.292
2.273
0.3028
6.841e-12
55
3
1.533
1.531
0.3474
9.338e-04
105
4
1.809
1.802
0.3136
6.702e-02
1
5
3.295
2.632
1.2706
NA
16
6
2.929
2.871
0.3333
1.926e-ll
2
7
3.691
3.562
1.0601
2.315e-01
1
8
3.117
3.670
0.8943
NA
1
9
4.096
4.348
0.5319
NA
3
10
2.185
2.232
0.7266
3.884e-01
290
11
1.727
1.678
0.5550
1.000e+00
Table 10: Comparison of hierarchical clusters for 2012-2016 variability, 24-hr
n'sites
grp
mean
median
sd
TF'pval
193
1
3.460
3.226
1.469
6.142e-09
25
2
7.603
7.143
2.917
1.610e-04
32
3
5.251
4.762
2.776
5.508e-01
33
4
7.187
7.143
2.787
1.074e-04
1
5
36.947
10.000
37.861
NA
1
6
13.357
7.692
10.734
NA
2
7
14.391
12.812
5.456
2.455e-01
1
8
36.444
10.000
39.143
NA
1
9
23.033
28.571
9.620
NA
1
10
27.333
18.750
23.875
NA
290
11
4.927
4.132
3.802
1.000e+00
227
-------
Table 11: Comparison of K-means clusters for 2012-2016 variability, annual
n'sites
grp
mean
median
sd
ann'pval
46
1
1.138
1.099
0.2967
1.113e-18
39
2
2.292
2.273
0.3028
5.306e-15
56
3
1.533
1.531
0.3474
8.645e-04
51
4
1.809
1.802
0.3136
1.358e-01
34
5
3.295
2.632
1.2706
2.922e-08
3
6
2.929
2.871
0.3333
2.210e-02
3
7
3.691
3.562
1.0601
8.448e-02
21
8
3.117
3.670
0.8943
5.988e-07
21
9
4.096
4.348
0.5319
5.424e-16
16
10
2.185
2.232
0.7266
2.442e-02
290
11
1.727
1.678
0.5550
1.000e+00
Table 12: Comparison of K-means clusters for 2012-2016 variability, 24-hr
n'sites
grp
mean
median
sd
TF'pval
1
1
3.460
3.226
1.469
NA
72
2
7.603
7.143
2.917
1.189e-09
67
3
5.251
4.762
2.776
4.262e-01
20
4
7.187
7.143
2.787
2.264e-03
26
5
36.947
10.000
37.861
2.218e-04
75
6
13.357
7.692
10.734
2.885e-09
2
7
14.391
12.812
5.456
2.455e-01
3
8
36.444
10.000
39.143
2.978e-01
2
9
23.033
28.571
9.620
2.285e-01
22
10
27.333
18.750
23.875
2.491e-04
290
11
4.927
4.132
3.802
1.000e+00
228
-------
1
2
3
4
5
6
7
8
9
1l
1
2
3
4
5
6
7
8
9
Annual NAAQS, Hierarchical clusters
-120 -100 -80
long
Annual NAAQS, kmeans clusters
-120 -100 -80
long
Figure 183: Hierarchical and K-means clusters for the annual variability for 2012-2016.
229
-------
24-hr NAAQS, Hierarchical clusters
-120 -100 -80
long
24-hr NAAQS, kmeans clusters
-120 -100 -80
long
Figure 184: Hierarchical and K-means clusters for the 24-hr variability for 2012-2016.
1
2
3
4
5
6
7
8
9
10
1
2
3
4
5
6
7
8
9
10
230
-------
7.2 Spatial analysis using NOAA Climate Regions
The National Oceanic and Atmospheric Administration (NOAA) has identified 9 "climate regions" [Thomas
and Koss, 1984], which have been identified to have distinct climatologically characteristics (more information
available at the NOAA website). This spatial grouping thus represents an independent spatial grouping with
which to evaluate regional variability characteristics. The mean annual and 24-hr variability values from sites
within these regions are compared in Figure 185 and detailed in Tables 13 and 14. The Pacific Northwest
(region 4) and the Central Northwest (region 9) stand out as having higher variability, which was seen in the
first K-means cluster analysis (using latitude, longitude, and the variability). The p-values for the annual
results are less than 0.05 for these two regions (though the p-value for region 9 is just barely less than 0.05
and the p-value for region 4 is still relatively large). The p-values for these two regions from the 24-hr results
are well above the nominal value of 0.05 and so are not significantly different from the mean dataset. Thus,
the results again make it difficult draw specific conclusions about the geographic nature of the variability
as estimated by this analysis, though the overall interpretation of these results is that most regions are not
significantly different from the mean dataset.
Table 13: Comparison of variability within NOAA climate regions, annual
Region Number
n sites
ann mean
ann median
ann sd
ann pval
1
96
1.724119413
1.724137931
0.538832598
0.986301656
2
69
1.864963958
1.875
0.501458469
0.547340281
3
94
1.457900874
1.388888889
0.545905166
0.245378188
4
22
2.473011166
2.405978785
0.845285667
0.013781951
5
39
1.719076105
1.704545455
0.622791499
0.970268689
6
70
1.563384611
1.583124478
0.459051349
0.469019357
7
25
1.623411728
1.595744681
0.621819606
0.676247393
8
66
1.610690042
1.405159932
0.84597489
0.627212461
9
42
2.255001105
2.198067633
0.808280411
0.047892639
All
523
1.727970703
1.727970703
0.670071968
1
Table 14: Comparison of variability within NOAA climate regions, 24-hr
Region Number
n sites
TF mean
TF median
TF sd
TF pval
1
96
5.861069073
4.545454545
6.966363904
0.924845491
2
69
4.690627066
4.545454545
1.965280928
0.632945983
3
94
3.628720758
3.125
2.092141946
0.326832786
4
22
10.13095985
7.417582418
11.55838373
0.165608839
5
39
5.280129178
5
3.051945505
0.853550834
6
70
5.893319944
4.653679654
5.146873277
0.911597545
7
25
5.383605813
4
3.836720273
0.896970478
8
66
5.93939837
3.923076923
10.5914199
0.90680014
9
42
8.69767221
7.692307692
3.979459168
0.166016078
All
523
5.660892679
4.347826087
6.168440212
1
231
-------
CD
O 4"
C\J
Q_
I o -
O J
10
I
c
c 2-
co
60-
CD
O
CN
a. 40-
o
LO
20-
50-
45-
40-
35-
30-
25-
X
/\
V
X
T*
l
3
-
-9-
4 5
regnum
I
regnum
x
-e-
a
/
\
\
T
regnum
^2
^3
*
ฃ3 5
^9
regnum
+ 1
^2
^3
04
ฃ3 5
regnum
1
2
3
4
5
-120
-100
-80
Figure 185: Comparison of variability within NOAA climate regions for 2016.
232
-------
References
D. W. Andrews. The Block-block Bootstrap: Improved Asymptotic Refinements. Cowles Foundation Dis-
cussion Papers 1370, Cowles Foundation for Research in Economics, Yale University, May 2002. URL
https://ideas.repec.org/p/cwl/cwldpp/1370.html.
P. Hall and J. L. Horowitz. Bootstrap critical values for tests based on generalized-method-of-moments
estimators. Econometrica, 64(4):891-916, 1996. ISSN 0165-0270. doi: 10.2307/2171849. URL http:
//www.j stor.org/stable/2171849.
A. Inoue and M. Shintani. Bootstrapping gmm estimators for time series. Journal of Econometrics,
133(2):531555, 2006. URL http://EconPapers.repec.org/RePEc:eee:econom:v:133:y:2006:i:2:p:
531-555.
S. N. Lahiri. Theoretical comparisons of block bootstrap methods. The Annals of Statistics, 27(l):386-404,
1999.
R Core Team. R: A Language and Environment for Statistical Computing. R Foundation for Statistical
Computing, Vienna, Austria, 2017.
R. K. Thomas and W. J. Koss. Regional and national monthly, seasonal, and annual temperature weighted
by area, 1895-1983. Historical Climatology Series 4-3, 1984.
233
-------
-------
United States Office of Air Quality Planning and Standards Publication No. EPA-454/R-18-001
Environmental Protection Air Quality Assessment Division April, 2018
Agency Research Triangle Park, NC
-------
|