PM
2.5
Nonattainment
Boundaries
in
Triad,
North
Carolina
1
8/
25/
04
Appendix
C
 
PM2.5
/
Population
Density
Correlation
Analysis
*
The
following
analysis
is
the
work
of
Hoy
Bohanon,
with
edits
made
by
Wyat
Appel.

Hoy
Bohanon,
PE
August
25,
2004
Bohanon
Engineering
PLLC
PO
Box
448
Clemmons,
NC
27012
bohanoneng@
triad.
rr.
com
The
information
about
the
methodology
is
contained
in
two
spreadsheets
containing
multiple
sheets
and
some
sheets
containing
large
amounts
of
data.
The
sheets
are
generally
labeled
"
PM2.5
DESIGNATIONS
DATA
SPREADSHEET
­
6/
14/
04."

The
author
assumes
that
data
in
these
spreadsheets
contains
the
best
available
data.
It
appears
that
EPA
prejudged
the
data
with
a
conclusion
that
nonattainment
monitors
must
be
influenced
by
80%
of
the
total
emissions
in
immediately
surrounding
counties
and
in
some
cases
such
as
the
Triad
counties
that
may
be
two
or
three
counties
distant.
There
was
some
attempt
to
make
a
correction
for
background
by
an
"
urban
excess"
calculation
that
applied
weighting
factors
to
emissions
data.
Some
unknown
algorithm
created
values
for
carbon
and
crustal
which
were
then
weighted
and
combined
with
SO2
and
NOx
to
determine
the
weighted
emissions
score.
Wind
patterns
are
not
considered.

The
measured
data
that
exist
for
the
Triad
and
surrounding
counties
(
as
selected
by
EPA)
is
shown
in
the
table
below.
Counties
with
missing
data
are
omitted.

Design
Values
Total
Emissions,
2001
(
tons)

Sequence
COU
'
01­'
03
Pop
Density
PM
SO2
NOX
VOC
530
Guilford
14.1
663
2,418
2,833
19,068
34,464
531
Davidson
15.8
274
1,951
1,398
11,281
14,970
532
Forsyth
14.6
768
1,559
5,885
14,552
20,679
534
Alamance
13.7
315
1,181
749
5,618
8,967
538
Chatham
12.2
79
1,714
11,605
5,823
4,734
545
Orange
13.1
301
857
756
6,264
6,751
548
Montgomery
12.1
56
516
484
1,631
4,175
549
Caswell
13.3
55
483
199
1,071
1,622
An
initial
basic
analysis
of
the
measured
data
is
to
determine
the
independent
and
dependent
variables,
(
or
the
potential
causes
and
the
effect)
and
then
investigate
whether
there
is
a
relationship.
The
effect
(
dependent
variable)
is
PM2.5
which
is
shown
as
Design
PM
2.5
Nonattainment
Boundaries
in
Triad,
North
Carolina
2
8/
25/
04
Values
(
DV).
It
may
be
influenced
(
independent
variables)
by
population
density
(
which
is
a
surrogate
for
many
emissions),
PM,
SO2,
NOx,
or
VOC.

Examining
the
Population
Density
The
data
for
the
immediate
Triad
area
is:

Triad
Density
Design
Value
Guilford
663
14.1
Forsyth
768
14.6
Alamance
315
13.7
Davidson
274
15.8
Where
density
is
people
/
square
mile
and
design
value
is
PM2.5
in
µ
g/
m3
Plotting
the
data
gives
PM2.5
Pop
Density
v.
Design
Value
R
2
=
0.075
13.5
14.0
14.5
15.0
15.5
16.0
0
200
400
600
800
1,000
Population
/
sm
Design
Value
The
added
trend
line
shows
that
the
correlation
is
negative.
This
means
that
higher
population
results
in
lower
PM.
This
doesn't
make
sense.
The
correlation
coefficient
is
very
low.
PM
2.5
Nonattainment
Boundaries
in
Triad,
North
Carolina
3
8/
25/
04
A
trend
line
that
excludes
the
Davidson
County
data
point
results
in
the
following:

R
2
=
0.873
13.0
13.5
14.0
14.5
15.0
15.5
16.0
0
200
400
600
800
1,000
Series1
Davidson
Linear
(
Series1)

This
is
a
better
fit
and
the
line
shows
that
increasing
population
results
in
increased
PM2.5.
This
look
at
the
data
leads
to
an
obvious
question.
Is
Davidson
County
different
from
the
other
counties
in
the
area?
To
make
a
more
informed
analysis,
the
other
surrounding
counties
listed
in
the
EPA
spreadsheet
should
be
added.

Triad+
Density
Design
Value
Guilford
663
14.1
Forsyth
768
14.6
Alamance
315
13.7
Davidson
274
15.8
Chatham
79
12.2
Orange
301
13.1
Montgomery
56
12.1
Caswell
55
13.3
PM
2.5
Nonattainment
Boundaries
in
Triad,
North
Carolina
4
8/
25/
04
Triad
PM2.5
DV
Guilford
Davidson
Forsyth
Alamance
Chatham
Orange
Montgomery
Caswell
12.0
12.5
13.0
13.5
14.0
14.5
15.0
15.5
16.0
0
200
400
600
800
1,000
Pop
Density
DV
Design
Value
vs
Population
R
2
=
0.3256
12.0
12.5
13.0
13.5
14.0
14.5
15.0
15.5
16.0
0
200
400
600
800
1,000
Pop
/
sq
mi
DV
With
the
additional
data,
the
regression
line
now
shows
increasing
PM2.5
with
increasing
population,
but
the
fit
isn't
very
good
(
r2
=
0.3).
PM
2.5
Nonattainment
Boundaries
in
Triad,
North
Carolina
5
8/
25/
04
Excluding
Davidson
County
from
the
regression
results
in
the
following:

Design
Value
vs
Population
R
2
=
0.7752
12.0
12.5
13.0
13.5
14.0
14.5
15.0
15.5
16.0
0
200
400
600
800
1,000
Pop
/
sq
mi
DV
Triad+

Davidson
Linear
(
Triad+)

The
fit
is
much
better.
The
intercept
corresponds
closely
to
what
EPA
used
as
the
background
value
in
its
urban
excess
calculation.
The
intercept
of
the
regression
line
(
value
where
the
population
equals
zero)
is
12.4.
In
the
urban
excess
calculation
EPA
appears
to
have
used
a
value
of
12.2
from
James
River
Face,
VA
for
a
background
(
zero
population)
number
for
the
Triad.
The
two
numbers
are
quite
close.

The
question
arises,
is
there
something
different
about
Davidson
County?
Is
it
an
outlier
in
terms
of
the
surrounding
area?
An
initial
determination
can
be
made
by
determining
the
95%
confidence
interval
around
the
Triad+
(
excluding
Davidson)
data
points.

Design
Value
vs
Population
11.0
12.0
13.0
14.0
15.0
16.0
17.0
18.0
0
200
400
600
800
1,000
Pop
/
sq
mi
DV
Triad+
Davidson
LCI
UCI
Linear
(
Triad+)
Linear
(
LCI)
Linear
(
UCI)
PM
2.5
Nonattainment
Boundaries
in
Triad,
North
Carolina
6
8/
25/
04
The
Triad+
data
set
(
not
including
Davidson)
is
bounded
by
a
95%
upper
confidence
interval
and
a
95%
lower
confidence
interval.
Points
within
these
intervals
are
statistically
likely
(
with
95%
confidence)
to
be
a
part
of
the
same
data
set.
Davidson
County
clearly
falls
outside
of
this
data
set.

One
may
logically
conclude
based
upon
the
data,
that
the
other
counties
exhibit
an
increased
design
value
due
to
increased
population
(
which
appears
to
be
a
decent
surrogate
for
increased
cars,
commerce,
jobs,
etc.).

One
may
also
conclude
that
Davidson
County
is
different
from
all
other
counties.

It
is
therefore
appropriate
to
"
single
out"
Davidson
County
and
begin
the
process
of
determining
what
makes
its
PM2.5
concentration
higher.
It
is
inappropriate
to
include
any
of
the
other
counties
adjacent
to
or
surrounding
Davidson
in
a
nonattainment
designation.
Their
design
values
all
attain
the
current
standard
and
there
is
no
evidence
that
the
contribution
to
each
other
is
significant.
Their
differences
can
be
explained
by
the
"
local
effect"
of
population.
Davidson
has
some
unexplained
"
local
effect"
making
it
unique
among
the
counties
in
the
Triad
area
of
North
Carolina.

Additional
Analysis
One
may
ask
whether
or
not
the
other
possible
input
factors
such
as
PM
give
similar
results.
Is
there
another
measured
component
that
closely
correlates
to
the
design
values?
The
answer
is
no.
The
following
table
lists
the
R
square
values
for
the
data
set
excluding
Davidson
County.
Higher
values
indicate
better
correlation.
Lower
values
indicate
lack
of
correlation.

Component
R­
square
Population
density
0.77
PM
0.20
SO2
0.02
NOx
0.52
VOC
0.50
Could
two
factors
be
involved?
The
stepwise
procedure
for
multiple
linear
regression
is
to
add
the
next
factor
highest
r­
square
factor
to
the
analysis.
A
multiple
linear
regression
using
population
density
combined
with
NOx
shows
NOx
to
not
be
a
significant
factor.
(
p=
0.32).
