UNITED
STATES
ENVIRONMENTAL
PROTECTION
AGENCY
Office
of
Air
Quality
Planning
and
Standards
Research
Triangle
Park,
North
Carolina
27711
August
29,
2003
MEMORANDUM
SUBJECT:
Analyses
of
2000­
2002
PM
Data
for
the
PM
NAAQS
Review
FROM:
Mark
Schmidt,
OAQPS
David
Mintz,
OAQPS
Tesh
Rao,
OAQPS
Lance
McCluney,
OAQPS
(
Intern)

TO:
file
The
purpose
of
this
memorandum
is
to
describe
and
summarize
multiple
sets
of
analyses
conducted
for
the
review
of
the
Particulate
Matter
(
PM)
national
ambient
air
quality
standards
(
NAAQS).
Both
PM10
and
PM2.5
data
were
analyzed,
as
well
as
the
calculated
differences
of
the
two
particle
size
cuts
(
PM10­
2.5),
and
PM2.5
composition
data.
Most
PM10
and
PM2.5
data,
and
corresponding
meteorological
information,
were
extracted
from
EPA's
Air
Quality
System
(
AQS)
data
base
on
various
dates
in
May,
2003.
PM2.5
composition
data
from
urban
sites
in
the
EPA
Speciation
Network
(
ESpN)
were
pulled
from
AQS
in
July,
2003.
PM
mass
and
PM2.5
composition
data
from
rural
sites
in
the
Interagency
Monitoring
of
PROtected
Visual
Environmental
(
IMPROVE)
aerosol
monitoring
network
were
acquired
from
the
National
Park
Service
in
July,
2003.
Additional
meteorological
data
were
obtained
from
the
National
Weather
Service.
Meteorological
data
were
necessary
to
convert
AQS
PM10
samples
reported
at
`
standard
conditions'
(
25
°
C,
760
mm
Hg)
to
`
local
conditions'
(
actual
temperature
and
pressure).
The
conversion
was
necessary
to
facilitate
size
cut
comparisons
of
PM10
to
PM2.5
and
calculate
an
accurate
difference;
PM10
data
are
generally
reported
to
AQS
at
standard
conditions
and
PM2.5
data
are
reported
at
local
conditions.
There
are
three
attachments
to
this
memo,
each
corresponding
to
the
different
type
of
data
analyzed:
Attachment
A
describes
the
AQS­
based,
24­
hour
duration
analyses;
Attachment
B
describes
the
AQS­
based
hourly
analyses;
and
Attachment
C
describes
the
ESpN
and
IMPROVE
data
analyses.
Each
attachment
contains
a
discussion
of
the
methods
and
assumptions
used
to
generate
results.
All
AQS­
based
24­
hour
duration
PM
(
10
and
2.5
micron
size
cuts)
data
and
hourly
PM10
data
used
in
the
analyses
were
sampled
with
Federal
Reference
Methods
(
FRM)
or
Federal
Equivalent
Methods
(
FEM).
Hourly
AQS
PM2.5
data
and
particle
data
collected
in
the
ESpN
and
IMPROVE
networks
(
Attachment
C)
utilized
non­
FRM/
FEM
techniques.

States
are
required
to
certify
their
data
submitted
to
AQS
on
an
annual
basis
for
each
calender
year;
this
certification
must
be
done
by
July
1st
of
the
following
year.
Since
the
2002
data
used
for
these
analyses
were
queried
from
AQS
prior
to
the
certification
deadline,
it
should
Page
­
2­
be
noted
that
the
2002
data
are
subject
to
change
and
that
additional
2002
data
may
have
been
reported
after
the
retrievals
used
in
these
analyses.

Some
analysis
results
are
summarized
at
a
broad
regional
level
using
the
geographic
regions
specified
below.
The
regional
definitions
correspond
to
the
regions
identified
by
the
Health
Effects
Institute
(
HEI)
in
a
recent
PM
study.
[
See
Figure
1,
page
8,
in
Samet,
J.
M.,
et
al.,
"
The
National
Morbidity,
Mortality,
and
Air
Pollution
Study
Part
II:
Morbidity,
Mortality,
and
Air
Pollution
in
the
United
States,"
Health
Effects
Institute,
Research
Report
Number
94,
Part
II,
June
2000.]
The
origin
of
the
HEI
region
definitions
can
be
traced
back
to
Figure
6­
30
of
EPA's
1996
PM
Criteria
Document,
which
identified
regions
on
the
basis
of
"
uniqueness
in
aerosol
trends,
seasonality,
size
distribution,
or
chemical
composition."
Some
sites
(
e.
g.,
ones
in
Alaska,
Hawaii,
Puerto
Rico,
and
the
Virgin
Islands,
)
were
not
assigned
to
an
HEI
region.
For
these
analyses,
these
sites
were
placed
in
Region
0,
`
Not
in
PM
Region'.
Data
for
these
sites
are
excluded
from
charts
shown
`
by
region'
but
are
included
elsewhere.

PM
REGION
CODE
PM
REGION
DESCRIPTION
HOW
DEFINED
1
Northeast
ME,
NH,
VT,
MA,
RI,
CT,
NJ,
DE,
MD*,
PA*,
NY*,
VA*,
WV*
(*
east
of
­
78.50
°
W
longitude)

2
Southeast
NC,
SC,
TN,
GA,
FL,
AL,
MS,
LA,
AR,
OK*,
TX*
(*
east
of
­
97.70
°
W
longitude)

3
Industrial
Midwest
NY*,
PA*,
WV*,
VA*,
KY,
OH,
MI,
IN,
IL,
WI#,
MN#,
IA#,
MO#
(*
west
of
­
78.50
°
W
longitude,
#
east
of
­
91.50
°
W
longitude)

4
Upper
Midwest
MN*,
WI*,
IA*,
MO*,
ND,
SD,
NE,
KS,
CO#
(*
west
of
­
91.50
°
W
longitude,
#
east
of
­
104.05
°
W
longitude)

5
Southwest
OK*,
TX*,
NM,
AZ,
NV#,
CA#
(*
west
of
­
97.70
°
W
longitude,
#
south
of
37.00
°
N
latitude
and
east
of
­
115.50
°
W
longitude)

6
Northwest
WA,
ID,
MT,
WY,
UT,
OR,
CO*,
CA#,
NV#
(*
west
of
­
104.05
°
W
longitude,
#
north
of
37.00
°
N
latitude)

7
Southern
California
CA*,
NV*
(*
west
of
­
115.50
°
W
longitude
and
south
of
37.00
°
N
latitude)

For
additional
information
on
the
analyses
documented
in
the
attachments,
please
contact
Mark
Schmidt
at
(
919)
541­
2416.

3
Attachments
Page
­
A1­
ATTACHMENT
A
Processing
Details
for
AQS
24­
hour
Sample
Duration
Files
and
Figures
This
attachment
describes
the
data
(
2000­
2002)
and
processing
procedures
used
to
generate
the
following
AQS­
based,
24­
hour
sample
duration
files
and
figures:
C
File
PM25_
sitemon_
info.
xls:
PM2.5
Site­
Monitor
Information
C
File
PM10_
sitemon_
info.
xls:
PM10
Site­
Monitor
Information
C
File
PMC_
sitemon_
info.
xls:
PM10­
2.5
Site­
Monitor
Information
C
File
PM25_
sitemon_
summary.
xls:
PM2.5
Monitor
Data
Summary
C
File
PM10_
sitemon_
summary.
xls:
PM10
Monitor
Data
Summary
C
File
PMC_
sitemon_
summary.
xls:
PM10­
2.5
Monitor
Data
Summary
C
File
PM25ctymax.
xls:
PM2.5
County
Max
Data
Summary
C
File
PM10ctymax.
xls:
PM10
County
Max
Data
Summary
C
File
PMCctymax.
xls:
PM10­
2.5
County
Max
Data
Summary
C
Figure
2­
4.
Distribution
of
annual
mean
PM2.5
and
estimated
annual
mean
PM10­
2.5
concentrations
by
region,
2000­
2002.
C
Figure
2­
5.
Distribution
of
98th
percentile
24­
hour
average
PM2.5
and
estimated
PM10­
2.5
concentrations
by
region,
2000­
2002.
C
Figure
2­
6.
County­
level
maximum
annual
mean
PM2.5
concentrations,
2000­
2002.
C
Figure
2­
7.
County­
level
maximum
98th
percentile
24­
hour
average
PM2.5
concentrations,
2000­
2002.
C
Figure
2­
9.
County­
level
maximum
annual
mean
PM10
concentrations,
2000­
2002.
C
Figure
2­
10.
County­
level
maximum
98th
percentile
24­
hour
average
PM10
concentrations,
2000­
2002.
C
Figure
2­
11.
Estimated
county­
level
maximum
annual
mean
PM10­
2.5
concentrations,
2000­
2002.
C
Figure
2­
12.
Estimated
county­
level
maximum
98th
percentile
24­
hour
average
PM10­
2.5
concentrations,
2000­
2002.
C
Figure
2­
16.
Distribution
of
ratios
of
annual
mean
PM2.5
to
PM10
by
region,
2000­
2002.
C
Figure
2­
17.
Regional
average
correlation
of
24­
hour
average
PM
by
size
fraction.
C
Figure
2­
18.
Urban
24­
hour
average
PM2.5
concentration
distributions
by
region
and
month,
2000­
2002.
C
Figure
2­
19.
Urban
24­
hour
average
PM10­
2.5
concentration
distributions
by
region
and
month,
2000­
2002.
CC
Figure
2­
20.
Distribution
of
annual
mean
vs.
98th
percentile
24­
hour
average
PM2.5
concentrations,
2000­
2002.
C
Figure
2­
21.
Distribution
of
estimated
annual
mean
vs.
98th
percentile
24­
hour
average
PM10­
2.5
concentrations,
2000­
2002.

General
Data
Description
All
data,
except
for
supplemental
meteorological
data
obtained
from
NWS,
were
extracted
from
EPA's
Air
Quality
System
database
(
AQS).
After
downloading
and
necessary
preprocessing,
data
for
PM10,
PM2.5,
and
calculated
PM10­
2.5,
sites
were
subjected
to
data
completeness
criteria.
The
data
selection
criteria
for
all
PM
size
cuts
(
applied
independently
to
PM10,
PM2.5,
and
calculated
PM10­
2.5)
was
(
by
site)
the
most
recent
4,
8,
or
12
consecutive
quarters
of
11
or
more
samples.
A
simple
example
is
shown
below.
For
this
example
site,
the
quarters
that
would
have
been
utilized
are
shaded.
Since
the
selection
criterion
evaluates
available
data
in
increments
of
4
quarters,
previous
quarters
could
not
be
used
due
to
the
shortfall
in
2001,
quarter
1.
An
additional
increment
of
4
consecutive
quarters
meets
the
11
minimum
sample
threshold
(
1999,
quarters
1­
4),
but
would
not
have
been
used
since
the
more
recent
band
of
data
(
shaded)
were
available.
Although
the
utilized
selection
criteria
do
not
guarantee
a
calendar
year(
s)
of
data,
it
does
provide
at
least
one
full
year
consisting
of
four
quarters,
thus
reducing
seasonal
bias.
Data
present
in
quarters
not
part
of
the
4­,
8­,
or
12­
quarter
period
of
interest
were
deleted
and
thus,
not
included
in
a
site
summaries
`
00
Q1
`
00
Q2
`
00
Q3
`
00
Q4
`
01
Q1
`
01
Q2
`
01
Q3
`
00
Q4
`
00
Q4
`
02
Q1
`
02
Q3
`
02
Q4
N=
12
13
14
15
10
15
16
14
15
13
11
9
Page
­
A2­
Means
and
percentiles
were
calculated
for
each
site
that
met
completeness
criteria.
Weighted
`
annual'
means
(
referenced
as
`
ANNMEAN'
in
summary
data
files)
were
computed
for
each
site
as
follows:
quarterly
averages
were
calculated
for
each
kept
quarter;
4­
quarter
averages
were
then
computed
from
the
applicable
one,
two,
or
three
sets
of
quarterly
averages
(
e.
g.,
in
the
example
above,
from
the
`
00
Q4,
`
01
Q1,
`
01
Q2,
and
`
01
Q3
averages);
then,
the
4­
quarter
means
were
averaged
together.
Percentiles
of
interest
(
minimum,
maximum,
median,
5th,
25th,
75th,
and
95th)
were
computed
on
the
entire
4­,
8­,
or
12­
quarter
distribution
of
data.

Although
all
data
submitted
to
AQS
are
considered
valid,
some
data
are
tagged
with
quality
assurance
qualifiers,
natural
event
flags,
and/
or
exceptional
event
flags.
All
data,
regardless
of
flags,
were
included
in
these
analyses.

All
concentrations
presented
in
the
analyses
outputs
are
shown
in
units
of
micrograms
per
cubic
meter
(
µ
g/
m3)
at
local
conditions.
Since
most
regulatory
PM10
data
are
reported
in
units
of
µ
g/
m3
at
standard
temperature
and
pressure
conditions
(
25
°
C,
760
mm
Hg),
the
data
had
to
be
converted
to
local
temperature
and
pressure
conditions
using
meteorological
information
(
see
next
section).
The
standard
conditions
PM10
data
were
converted
to
local
conditions
before
the
site
completeness
criteria
(
most
recent
4,
8,
or
12
consecutive
quarters
of
11+
samples)
were
applied.

PM10
Data
PM10
data
from
Federal
Reference
Methods
(
FRM)
and
Federal
Equivalent
Methods
(
FEM)
monitors
were
extracted
from
AQS
on
May
28,
2003.
Four
separate
queries
were
made:
1)
raw
daily
(
24­
hour)
for
parameter
81102
[
PM10,
standard
temperature
and
pressure
conditions
(
STP)];
2)
raw
daily
for
parameter
85101
[
PM10,
local
temperature
and
pressure
conditions
(
LTP)];
3)
summary
daily
(
hourly
reported
measurements
aggregated
within
AQS
to
a
24­
hour
period)
for
parameter
81102;
and
4)
summary
daily
for
parameter
85101.
The
`
daily'
monitors
collected
24­
hour
average
samples
on
a
filter
for
each
successful
day
of
monitoring.
The
monitors
are
typically
scheduled
to
collect
PM10
samples
once
every
6
days,
though
some
collected
samples
more
frequently.
The
PM10
filter
samples
are
weighed
in
a
laboratory
environment
to
obtain
mass
concentrations
expressed
in
micrograms
per
cubic
meter
(
µ
g/
m3).
The
`
hourly'
monitors
are
generally
operated
continuously
almost
every
hour
of
the
year
(
with
occasional
down
time
for
calibrations
and
audits).
AQS
maintains
the
raw
hourly
data
and
also
aggregates
the
hourly
information
into
summary
daily
records.
A
summary
record
is
only
created
if
75%
or
more
of
the
hourly
data
($
16)
are
present.

Parameter
81102
data,
both
summary
and
daily,
were
converted
to
local
conditions
using
collocated
temperature
and
pressure
information.
If
collocated
temperature
and/
or
pressure
data
were
not
available,
meteorological
data
from
the
nearest
NWS
station
were
used.
If
collocated
data
were
not
available
and
the
NWS
data
were
missing,
the
STP
data
were
not
converted
to
LTP
and
not
used
in
the
analyses.
Parameter
85101
and
converted
81102
data
were
merged
by
site
day.
If
more
than
one
type
of
PM10
data
were
present
on
a
given
day,
LTP
reported
data
were
favored
over
converted
(
STP
to
LTP)
data,
and
lower
Pollutant
Occurrence
Codes
(
POC's)
were
favored
over
higher
POC
numbers.
Typically,
monitors
with
lower
POC
numbers
will
have
more
data
(
sample
more
frequently)
than
monitors
with
higher
POC
numbers
(
for
the
same
duration
data).
`
Primary'
sampler
POC
numbers
are
generally
lower
than
quality
assurance
(
QA)
monitor
POC's;
thus
primary
data
were
favored
over
QA
data.
Daily
sampling
monitors
Page
­
A3­
also
usually
have
lower
POC
numbers
than
hourly
sampling
instruments,
so
the
daily
data
were
generally
selected
over
the
summary
daily
data.
Hence,
the
following
selection
priority
was
implemented:
85101
daily
>
85101
summary
>
81102
daily
>
81102
hourly.

Site
data
were
evaluated
on
the
11
sample
per
quarter
threshold
and
a
determination
made
as
to
which
4,
8,
or
12
(
all)
quarters
to
keep.
1235
sites
had
at
least
4
consecutive
quarters
of
11+
samples:
707
sites
had
12
quarters
available,
246
sites
had
8
quarters
available,
and
282
sites
had
4
quarters
to
use.
The
1235
sites
are
mapped
in
Figure
B­
1.
Metadata
for
the
1235
sites
are
provided
in
file,
`
PM10_
sitemon_
info.
xls'.
Summary
data
for
the
sites
(
means
and
moments)
are
provided
in
file,
`
PM10_
sitemon_
summary.
xls'.
County
maxima
of
site
`
ANNMEAN'
and
98th
percentile
are
provided
in
`
PM10ctymax.
xls'.

PM2.5
Data
PM2.5
data
from
non­
continuous
FRM
monitors
were
extracted
from
AQS
on
May
28,
2003.
Only
one
AQS
query
was
necessary:
24­
hour
(
daily)
data
for
parameter
88101
(
PM2.5,
local
temperature
and
pressure
conditions).
Hourly
data
were
not
extracted
for
these
analyses
since
are
there
are
no
continuous
FRM/
FEM
methods
for
PM2.5.
PM2.5
monitors
are
typically
scheduled
to
collect
PM2.5
samples
once
every
3
days,
though
some
collected
samples
more
or
less
frequently.
The
PM2.5
filter
samples
are
weighed
in
a
laboratory
environment
to
obtain
mass
concentrations
(
µ
g/
m3).
PM2.5
data,
reported
as
parameter
88101,
are
in
local
conditions.
Only
primary
monitors
were
used
in
the
analyses;
primary
monitors
are
the
first
occurring
POC,
generally
`
1'.
Site
data
were
evaluated
on
the
11
sample
per
quarter
threshold
and
a
determination
made
as
to
which
4,
8,
or
12
(
all)
quarters
to
keep.
1152
sites
had
at
least
4
consecutive
quarters
of
11+
samples:
789
sites
had
all
12
quarters
available,
193
sites
had
8
quarters
available,
and
170
sites
had
4
quarters
to
use.
The
1152
sites
are
mapped
in
`
pm25­
map.
gif'.
Metadata
for
the
1152
sites
are
provided
in
file,
`
PM25_
sitemon_
info.
xls'.
Summary
data
for
the
sites
(
means
and
moments)
are
provided
in
file,
`
PM25_
sitemon_
summary.
xls'.
County
maxima
(
of
site
`
ANNMEAN'
and
98th
percentile)
are
provided
in
`
PM25ctymax.
xls'.

PM(
10­
2.5)
Estimates
In
order
to
characterize
a
PM
coarse
fraction
(
i.
e.,
PM
less
than
10
micrometers
but
greater
than
2.5
micrometers),
a
simplistic
difference
method
was
utilized.
At
locations
where
both
PM10
and
PM2.5
were
recorded,
PM2.5
daily
averages
are
subtracted
from
PM10
daily
averages.
Currently
there
are
no
federal
PM
coarse
fraction
monitoring
requirements.
Although
there
are
no
federal
reference
or
equivalent
methods
stipulated
for
the
PM
coarse
fraction,
only
FRM/
FEM
PM10
and
PM2.5
daily
averages
were
used
to
construct
the
PM10­
2.5
estimates.
No
effort
was
made
to
account
for
differences
in
sampling
instruments
or
protocols
between
the
colocated
PM10
and
PM2.5
monitors.
Because
of
these
differences
(
and
other
factors),
occasionally
the
calculated
PM10­
2.5
values
were
negative;
this
is
not
unexpected
for
two
independent
observations
and
negative
PM10­
2.5
concentrations
were
not
censored
from
the
analyses.
Both
the
PM10
and
PM2.5
data
used
in
the
difference
calculation
were
in
units
of
µ
g/
m3
at
local
conditions,
thus
the
calculated
PM10­
2.5
values
also
are
in
those
units.
All
available
FRM/
FEM
PM10
and
PM2.5
data
were
used
to
construct
the
PM10­
2.5
estimates;
the
completeness
criteria
(
most
recent
4,
8,
or
12
consecutive
quarters
of
11+
samples)
were
evaluated
after
the
daily
difference
estimates
were
calculated.
488
sites
met
the
completeness
criteria
of
at
least
4
consecutive
quarters
of
11+
Page
­
A4­
samples;
219
sites
had
12
quarters
available,
129
sites
had
8
quarters
available,
and
140
sites
had
4
quarters
to
use.
The
488
sites
are
mapped
in
Figure
B­
3.
Metadata
for
the
488
sites
are
provided
in
file,
`
PMC_
sitemon_
info.
xls'.
Summary
data
for
the
sites
(
means
and
moments)
are
provided
in
file,
`
PMC_
sitemon_
summary.
xls'.
County
maxima
(
of
site
`
ANNMEAN'
and
98th
percentile)
are
provided
in
`
PMCctymax.
xls'.

Boxplot
Figures
Many
of
the
analyses
figures
are
boxplots.
Unless
otherwise
noted,
in
all
of
the
AQSbased
24­
hour
average
duration
boxplots,
the
following
definitions
apply:
C
The
bottom
of
the
box
depicts
the
25th
percentile
of
the
plotted
distribution
C
The
top
of
the
box
depicts
the
75th
percentile
of
the
plotted
distribution
C
The
line
through
the
box
identifies
the
distribution
median
C
The
top
whisker
cap
identifies
the
95th
percentile
of
the
plotted
distribution
C
The
bottom
whisker
cap
identifies
the
5th
percentile
of
the
plotted
distribution
C
The
distribution
maximum
and
minimum
are
shown
as
asterisks
Data
Files
and
Processing
Code
The
graphics
generated
for
the
PM
NAAQS
review
and
the
3
size
faction
site
maps
are
included
in
the
attached
file,
`
Attach­
A­
Graphics'.

The
nine
Microsoft
Excel
spreadsheets
mentioned
above
are
contained
in
the
attached
file,
`
Attach­
B­
spreadsheets.
zip'
along
with
a
data
dictionary.
A
spreadsheet
of
boxplot
plotting
points
is
also
included.

Raw,
intermediate,
and
final
SAS
data
files
generated
for
these
analyses
are
included
in
the
attached
file,
`
Attach­
B­
SASdata.
zip':

Data
were
processed
with
SAS
software.
The
SAS
programs
are
included
in
the
attached
file,
`
Attach­
B­
SAS.
zip':

Comments
on
Specific
Figures:

Figure
2­
4.
Distribution
of
annual
mean
PM2.5
and
estimated
annual
mean
PM10­
2.5
concentrations
by
region,
2000­
2002:
C
Shows
distribution
of
site
level
ANNMEAN
by
size
fraction
and
region.
C
Region
0
data
(
23
sites
for
PM2.5,
13
sites
for
PM10­
2.5)
were
excluded
from
chart.
Figure
2­
5.
Distribution
of
98th
percentile
24­
hour
average
PM2.5
and
estimated
PM10­
2.5
concentrations
by
region,
2000­
2002:
C
Shows
distribution
of
site
level
PCT98
by
size
fraction
and
region.
C
Region
0
data
(
23
sites
for
PM2.5,
13
sites
for
PM10­
2.5)
were
excluded
from
chart.
Figure
2­
6.
County­
level
maximum
annual
mean
PM2.5
concentrations,
2000­
2002:
C
Shows
county
maximum
of
PM2.5
site
mean
(
referenced
as
ANNMEANMAX
in
Page
­
A5­
PM25ctymax.
xls)
C
Concentration
ranges
provided
by
AQSSD,
HEEG.
Figure
2­
7.
County­
level
maximum
98th
percentile
24­
hour
average
PM2.5
concentrations,
2000­
2002:
C
Shows
county
maximum
of
PM2.5
site
98th
percentile
(
referenced
as
PCT98MAX
in
PM25ctymax.
xls)
C
Concentration
ranges
provided
by
AQSSD,
HEEG.
Figure
2­
9.
County­
level
maximum
annual
mean
PM10
concentrations,
2000­
2002:
C
Shows
county
maximum
of
PM10
site
mean
(
referenced
as
ANNMEANMAX
in
PM10ctymax.
xls)
C
Concentration
ranges
provided
by
AQSSD,
HEEG.
Figure
2­
10.
County­
level
maximum
98th
percentile
24­
hour
average
PM10
concentrations,
2000­
2002:
C
Shows
county
maximum
of
PM10
site
98th
percentile
(
referenced
as
PCT98MAX
in
PM10ctymax.
xls)
C
Concentration
ranges
provided
by
AQSSD,
HEEG.
Figure
2­
11.
Estimated
county­
level
maximum
annual
mean
PM10­
2.5
concentrations,
2000­
2002:
C
Shows
county
maximum
of
PM10­
2.5
site
mean
(
referenced
as
ANNMEANMAX
in
PMCctymax.
xls)
C
Concentration
ranges
provided
by
AQSSD,
HEEG.
Figure
2­
12.
Estimated
county­
level
maximum
98th
percentile
24­
hour
average
PM10­
2.5
concentrations,
2000­
2002:
C
Shows
county
maximum
of
PM10­
2.5
site
98th
percentile
(
referenced
as
PCT98MAX
in
PMCctymax.
xls)
C
Concentration
ranges
provided
by
AQSSD,
HEEG.
Figure
2­
16.
Distribution
of
ratios
of
24­
hour
average
PM2.5
to
PM10
by
region,
2000­
2002:
C
The
ratio
of
PM2.5
to
PM10
was
first
calculated
for
each
site­
day.
Because
the
parameter
selection
criteria
(
most
recent
4,
8,
or
12
consecutive
quarters
of
11+
samples)
were
applied
separately
for
PM10
and
PM2.5,
the
selected
time
periods
did
not
necessarily
match.
If
the
common
time
periods
of
constituent
raw
data
(
for
the
PM10
and
PM2.5
sites
that
met
the
selection
criteria)
were
used
for
this
analysis,
some
sites
common
to
both
parameters
would
not
have
any
matches
(
by
site­
day)
and
others
would
have
a
seasonal
bias
(
only
have
matches
in
certain
quarters).
To
avoid
this
situation,
the
raw
data
used
in
this
analysis
were
culled
from
the
PM10­
2.5
database
(
pmc_
raw_
meetscomp.
sas7bdat).
This
insured
an
equal
number
of
each
quarter
for
each
site
and
also
insured
a
minimum
of
44
samples
for
each
site
(
4
quarters
*
11
samples
each).
C
The
site­
day
ratios
of
PM2.5
to
PM10
were
averaged
by
site
and
the
distribution
of
the
site
ratios
plotted
by
region.
Figure
2­
17.
Regional
average
correlation
of
24­
hour
average
PM
by
size
fraction:
C
For
the
same
reason
noted
in
the
first
bullet
above,
all
data
used
in
this
analysis
were
culled
from
the
PM10­
2.5
database
(
PMC_
raw_
meetscomp.
sas7bdat).
C
A
Pearson
correlation
coefficient
was
calculated
for
each
site
fraction
pair
(
PM10
versus
PM2.5,
PM2.5
versus
PM10­
2.5,
and
PM10
versus
PM10­
2.5.
C
The
site
correlation
coefficients
for
each
fraction
were
averaged
by
region.
C
Region
0
data
(
23
sites
for
PM2.5,
13
sites
for
PM10­
2.5)
were
excluded
from
the
chart.
Figure
2­
18.
Urban
24­
hour
average
PM2.5
concentration
distributions
by
region
and
month,
Page
­
A6­
2000­
2002.
C
Only
data
from
monitors
with
location
setting
(
referenced
`
LOCATION"
in
`
PM25_
sitemon_
info.
xls')
=
`
URBAN
AND
CENTER
CITY'
or
`
SUBURBAN'
were
used.
C
All
24­
hour
average
values
from
kept
quarters
at
above
noted
monitors
were
averaged
together
by
region­
month.
C
In
the
plots,
the
boxes
represent
the
interquartile
range
(
25th
to
75th
percentiles)
of
each
monthly
distribution
and
the
line
inside
the
box
is
the
median
of
the
distribution.
The
trend
line
represents
the
mean,
and
the
number
above
each
box
represents
the
number
of
24­
hour
average
observations
that
were
used
to
generate
each
box
plot.
C
Only
valid
regions
(
1­
7)
were
plotted.
°
Seven
separate
graphics
were
produced
for
the
PM
NAAQS
review
memo;
a
2­
character
HEI
region
name
abreviation
differentiates
the
attached
plots.
Figure
2­
19.
Urban
24­
hour
average
PM10­
2.5
concentration
distributions
by
region
and
month,
2000­
2002.
C
Only
data
from
monitors
with
location
setting
(
referenced
`
LOCATION"
in
`
PMC_
sitemon_
info.
xls')
=
`
URBAN
AND
CENTER
CITY'
or
`
SUBURBAN'
were
used.
C
All
24­
hour
average
values
from
kept
quarters
at
above
noted
monitors
were
averaged
together
by
region­
month.
C
In
the
plots,
the
boxes
represent
the
interquartile
range
(
25th
to
75th
percentiles)
of
each
monthly
distribution
and
the
line
inside
the
box
is
the
median
of
the
distribution.
The
trend
line
represents
the
mean,
and
the
number
above
each
box
represents
the
number
of
24­
hour
average
observations
that
were
used
to
generate
each
box
plot.
C
Only
valid
regions
(
1­
7)
were
plotted.
°
Seven
separate
graphics
were
produced
for
the
PM
NAAQS
review
memo;
a
2­
character
HEI
region
name
abbreviation
differentiates
the
attached
plots.
Figure
2­
20.
Distribution
of
annual
mean
vs.
98th
percentile
24­
hour
average
PM2.5
concentrations,
2000­
2002:
C
Shows
distribution
of
site
level
98th
percentile
by
annual
mean
range.
C
Concentration
ranges
provided
by
AQSSD,
HEEG.
Figure
2­
21.
Distribution
of
estimated
annual
mean
vs.
98th
percentile
24­
hour
average
PM10­
2.5
concentrations,
2000­
2002:
C
Shows
distribution
of
site
level
98th
percentile
by
annual
mean
range.
C
Concentration
ranges
provided
by
AQSSD,
HEEG.
Page
­
B1­
ATTACHMENT
B
Processing
Details
for
AQS
Hourly
Sample
Duration
Figures
This
attachment
describes
the
data
(
2000­
2002)
and
processing
procedures
used
to
generate
the
following
AQS
based,
hourly
sample
duration
figures:
C
Figure
2­
22.
Hourly
average
PM2.5
distributions
(
upper
panel)
and
seasonal
average
PM2.5
concentrations
(
lower
panel)
at
a
Cleveland,
OH
monitoring
site,
2000­
2002.
C
Figure
2­
23.
Hourly
average
PM10­
2.5
distributions
at
a
Cleveland,
OH
monitoring
site,
2000­
2002.
The
graphics
are
included
in
the
attached
file,
`
Attach­
B­
graphics'.

Data
Description
Hourly
(
AQS
duration='
1')
PM10
and
PM2.5
data
for
calendar
years
2000­
2002
were
extracted
from
AQS
on
May
28,
2003.
Two
queries
were
made
for
PM10:
one
for
parameter
81102
[
PM10,
standard
temperature
and
pressure
conditions
(
STP)]
and
one
for
parameter
85101
[
PM10,
local
temperature
and
pressure
conditions
(
LTP)].
Only
one
query
was
necessary
for
PM2.5,
for
parameter
88101
(
PM2.5
LTP).
Hourly
meteorological
data
were
obtained
from
the
National
Weather
Service
(
NWS).
Meteorological
data,
specifically
temperature
and
pressure,
were
necessary
to
convert
PM10
data
reported
at
STP
to
LTP.
To
convert
the
PM10
STP
data
to
an
LTP
basis,
the
following
formula
was
used:

PM10­
LTP=
PM10­
STP
*
[
298
/
temperature
(
degrees
C)]
*
[
pressure
(
mm
Hg)
/
760].

Hourly
PM10­
2.5
records
were
created
by
subtracting
the
PM2.5
from
the
LTP
(
reported
or
converted)
PM10
values.
Since
data
are
reported
to
AQS
in
Standard
Time,
hours
were
adjusted
to
reflect
daylight
savings
time
to
be
consistent
with
human
activity.
The
seasons
were
derived
in
the
following
manner:
°
December,
January,
February
(
quarter
1)
=
Winter
°
March,
April,
May
(
quarter
2)
=
Spring
°
June,
July,
August
(
quarter
3)
=
Summer
°
September,
October,
November
(
quarter
4)
=
Fall
Outputs
Description
Two
types
of
plots
are
shown
in
the
analyses
figures,
boxplots
and
a
diurnal
line
plots.
In
the
boxplots,
the
boxes
represent
the
interquartile
range
(
25th
to
75th
percentiles)
of
each
hourly
distribution,
the
line
inside
each
box
shows
the
median
of
the
distribution,
the
dot
inside
each
box
represents
the
distribution
mean
(
the
distribution
means
are
connected
by
lines),
and
the
box
whiskers
represent
the
5th
and
95th
percentiles.
The
diurnal
line
graph
plots
hourly
averages
by
season.
The
hourly
averages
(
by
season)
are
joined
by
lines;
plotting
symbols
are
not
shown.
Although
the
outputs
were
produced
for
multiple
sites,
only
the
graphics
for
the
Cleveland,
OH
Page
­
B2­
site
(
AQS
ID
=
390350060)
were
used
in
the
analyses
memorandum.

Methodology
The
statistical
software
package
SAS
was
used
to
process
the
data
and
create
the
associated
graphics.
The
SAS
procedure
PROC
UNIVARIATE
was
used
to
calculate
the
summary
statistics
and
the
procedure
PROC
GPLOT
was
used
to
procedure
the
graphics.
The
code
used
to
generate
the
plots
are
contained
in
Attach­
B­
SAScode.
Page
­
C1­
ATTACHMENT
C
Processing
Details
for
IMPROVE
and
ESpN
Figures
This
attachment
describes
the
data
and
processing
procedures
used
to
generate
the
following
ESpN
and
IMPROVE
figures:
C
Figure
2­
8.
Average
annual
mean
PM2.5
concentration
trend
at
IMPROVE
sites,
1999­
2001.
C
Figure
2­
13.
Average
annual
mean
PM10­
2.5
concentration
trend
at
IMPROVE
sites,
1999­
2001.
C
Figure
2­
14.
Annual
average
composition
of
PM2.5
by
Region.
Upper
panel
rural
sites,
lower
panel
urban
sites.
C
Figure
2­
15.
Average
annual
mean
sulfate,
total
carbon,
and
crustal
material
concentration
trend
at
IMPROVE
sites,
1999­
2001.
The
graphics
are
included
in
the
attached
file,
`
Attach­
C­
Graphics'.

Data
Description
Two
types
of
data
were
used
in
these
analyses:
Rural
particle
data
were
derived
from
daily
measurements
from
the
Interagency
Monitoring
of
PROtected
Visual
Environmental
(
IMPROVE)
aerosol
monitoring
network.
Urban
aerosol
particle
data
were
derived
from
daily
measurements
from
the
EPA
Speciation
Monitoring
Network
(
ESpN).
Data
from
ESpN
were
extracted
from
AQS
on
July
3,
2003.
Data
from
IMPROVE
were
obtained
in
two
manners.
Debbie
Miller
of
the
National
Park
Service
provided
1992­
2001
site­
level
annual
summary
data
on
July
10,
2003.
Additional
IMPROVE
summary
data
for
the
time
period
September,
2001
through
August,
2002
were
downloaded
from
the
IMPROVE
website
on
July
20,
2003.
All
summary
data
used
in
these
analyses
are
included
in
the
attached
file,
`
Attach­
C­
Data'.

The
IMPROVE
trends
analyses
(
Figures
2­
8,
2­
13
and
2­
15)
show
10­
year
trends
for
PM10­
2.5
mass;
PM2.5
mass;
and
the
three
PM2.5
mass
components:
sulfate
(
SO4),
total
carbon
(
TC,
sum
of
organic
carbon
and
elemental
carbon),
and
crustal
material.
All
five
major
components
of
PM2.5
mass
are
reported
in
the
IMPROVE
/
ESpN
bar
graphs
of
Figure
2­
14.
These
components
include:
ammonium
(
NH4),
nitrate
(
NO3),
total
carbonaceous
mass
(
TCM),
and
crustal
material.
TCM
is
calculated
as
organic
carbon
mass
(
OCM)
+
elemental
carbon
(
EC).
OCM
is
estimated
as
measured
and
blank­
corrected
organic
carbon
(
OC)
multiplied
by
1.40
(
to
convert
to
mass).
Crustal
material
concentrations
are
estimated
using
the
`
IMPROVE
equation:
2.2[
Aluminum]+
2.49[
Silicon]+
1.63[
Calcium]+
2.42[
Iron]+
1.94[
Titanium].

Outputs
Description
The
IMPROVE
trends
plots
(
Figures
2­
8,
2­
13
and
2­
15)
show
the
annual
average
concentrations
for
PM10­
2.5,
PM2.5,
sulfate,
total
carbon,
and
crustal
material
averaged
across
9
eastern
sites
and
23
western
sites.
There
is
also
a
separate
plot
series
for
the
Washington,
DC
site
because
it
is
not
a
rural
site
like
the
others.
A
list
of
sites
by
area
classification
is
provided
in
the
file
`
Trends
Region
X
Site.
xls'
(
which
is
located
in
`
Attach­
C­
Data.
zip'.)
Page
­
C2­
The
bar
graphs
contrast
rural
(
top
panel,
data
from
IMPROVE)
versus
urban
(
bottom
panel,
data
from
ESpN)
average
concentrations
of
PM2.5
constituents
by
geographic
region.
For
this
particular
analysis,
the
following
geographic
regions
were
used
(
in
lieu
of
HEI
region
definitions):
South
East,
Mid
West,
East
Coast/
North
East,
East
Texas/
South,
Far
North
East,
North
Plains,
California,
Desert
West,
and
North
West.
See
spreadsheet
`
Bar
Charts
Region
X
Site.
xls'
(
located
in
`
Attach­
C­
data.
zip')
for
the
regional
site
assignments.

Methodology
For
the
trends
analyses,
sites
were
required
to
have
8
of
10
valid
years
of
data.
Missing
years
were
interpolated
using
surrounding
years.
The
lines
plotted
are
the
averages
across
the
trend
sites
in
each
region.
The
Washington
plot
is
based
on
a
single
site.
PROC
UNIVARIATE
in
the
SAS
statistical
software
package
was
used
to
calculate
the
averages
in
each
plot.
The
SAS
code
is
included
in
the
attached
file,
`
Attach­
C­
SAScode'.
Plotting
points
for
each
of
the
three
plots
are
presented
in
file,
`
Trends
Plotting
Points.
xls'
which
is
located
in
`
Attach­
C­
Data.
zip'.

For
the
bar
graphs,
only
data
for
`
complete'
sites
were
used.
For
both
the
urban
and
rural
data,
`
complete'
data
consisted
of
having
50%
or
more
observations
(
of
the
major
chemical
components
of
PM2.5
mass:
sulfate,
nitrate,
organic
carbon
mass,
elemental
carbon,
aluminum,
calcium,
titanium,
iron,
and
silicon)
per
quarter
for
the
year
analyzed
Use
of
Outputs
The
data
as
presented
in
the
bar
charts
should
be
used
only
to
gauge
relative
levels
of
urban
versus
rural
concentrations
of
the
chemical
species
in
each
of
the
regions.
Because
urban
and
rural
sites
were
not
specifically
matched
(
in
terms
of
separation
distance,
representative
upwind
locations,
etc.),
no
inferences
should
be
drawn
about
`
urban
increments'
based
on
these
data
and
graphics.
The
two
bar
charts
simply
illustrate
how
PM2.5
constituents
vary
for
a
group
of
urban
and
rural
sites
that
are
roughly
located
in
the
same
geographic
region.
For
further
details
on
`
urban
increment'
analyses,
refer
to:
V.
Rao,
N.
Frank,
A.
Rush,
F.
Dimmick,
"
Chemical
Speciation
of
PM2.5
in
Urban
and
Rural
Areas",
In
the
Proceedings
of
the
Air
&
Waste
Management
Association
Symposium
on
Air
Quality
Measurement
Methods
and
Technology,
San
Francisco,
November
13­
15,
2002.
