1
Preparing
Highway
Emissions
Inventories
for
Urban
Scale
Modeling:

A
Case
Study
in
Philadelphia
R.
Cook
a,
A.
Beidler
b,
J.
S.
Touma
c,
M.
Strum
d
a
US
Environmental
Protection
Agency,
Office
of
Transportation
and
Air
Quality,
National
Vehicle
and
Fuel
Emissions
Laboratory,
2000
Traverwood
Drive,
Ann
Arbor,
MI
48105,
USA
b
Computer
Sciences
Corporation
P.
O
Box
12804
Research
Triangle
Park,
NC
27709,
USA
c
NOAA/
Atmospheric
Sciences
Modeling
Division
(
In
Partnership
with
the
U.
S.

Environmental
Protection
Agency),
Research
Triangle
Park,
NC
d
US
Environmental
Protection
Agency
MD
243­
01
Research
Triangle
Park,
NC
27707,
USA
Abstract
Use
of
local
data
to
develop
air
toxic
pollutant
inventories,
rather
than
estimates
based
on
national
default
data
and
"
top­
down"
allocation
schemes,
can
have
a
very
large
impact
on
the
magnitude
and
distribution
of
air
toxic
emissions.
Such
data
can
be
used
to
significantly
improve
the
accuracy
of
local
scale
assessment.
This
paper
provides
a
comparison
of
"
top
down"
and
"
bottom
up"
approaches
for
one
urban
area,
Philadelphia,
in
calendar
year
1999.
Under
the
"
top
down"

approach,
emissions
are
estimated
at
the
county
level,
typically
starting
from
more
aggregated
information,
e.
g.,
metropolitan
statistical
area
or
state
data
on
activities,
and
average
or
default
inputs
to
the
highway
vehicle
emissions
factor
model.
The
"
bottom
up"
approach
utilizes
Travel
Demand
Model
(
TDM)
data
in
conjunction
with
a
highway
vehicle
emissions
factor
model
to
2
better
estimate
levels
and
spatial
distribution
of
onroad
motor
vehicle
emissions.
TDM
data
can
include
information
on
the
spatial
distribution
of
vehicle
activity,
speeds
along
those
roads
(
which
can
have
a
large
impact
on
emissions),
and
the
distribution
of
the
vehicle
miles
traveled
(
VMT)

among
vehicle
classes
for
different
speed
ranges.
As
expected
these
two
approaches
yield
different
spatial
distribution
of
emissions;
however,
they
also
result
in
different
levels
of
emissions
at
the
county
level.

Introduction
Air
quality
models
are
useful
tools
for
identifying
sources
and
pollutants
of
greatest
concern,

evaluating
the
effectiveness
of
existing
or
potential
reduction
strategies,
and
helping
to
locate
monitors.
Emission
sources
that
are
input
into
these
models
include
major,
area
and
mobile
sources.
The
mobile
sources
include
the
highway
or
"
on­
road"
vehicles
category,
which
are
motorized
vehicles
that
are
normally
operated
on
public
roadways
(
e.
g.,
passenger
cars,

motorcycles,
minivans,
sport­
utility
vehicles,
light­
duty
trucks,
heavy­
duty
trucks,
and
buses)
and
the
"
nonroad"
vehicles
category
(
e.
g.,
commercial
marine
vessels,
locomotives,
aircraft,
etc.).
In
this
paper
we
will
focus
on
air
toxics
emissions
from
highway
vehicles
only.

Highway
vehicle
emissions
for
regional
and
national
scale
air
toxics
assessments
are
usually
developed
using
a
"
top
down"
approach.
Under
this
approach,
emissions
are
estimated
at
the
county­
level
using
activity
data
which
itself
has
been
allocated
from
a
larger
geographic
scale
3
(
State,
or
urban
area)
using
spatial
surrogates
such
as
population.
Default
or
average
inputs
are
then
used
in
a
highway
vehicle
emissions
factor
model
to
compute
county
level
emissions.
Other
spatial
surrogates
are
used
to
allocate
county
level
emissions
to
grid
cells
or
census
tracts
as
required
by
the
air
quality
model.
This
approach
is
taken
due
to
the
difficulty
of
gathering
and
incorporating
more
precise
local
data
for
such
a
large
domain.
An
example
of
an
assessment
which
uses
this
approach
is
EPA's
National­
Scale
Air
Toxics
Assessment
(
U.
S.
Environmental
Protection
Agency,
2002).
The
National
Scale
Air
Toxics
Assessment
relies
on
data
from
the
National
Emissions
Inventory
(
NEI),
a
national
database
of
air
pollutant
emissions
for
major
and
area
stationary
sources,
highway
vehicles,
and
nonroad
equipment
(
U.
S.
Environmental
Protection
Agency,
2004a).
The
issue
with
this
approach
is
that
default
inputs
used
may
not
reflect
local
scale
conditions.
This
can
result
in
mischaracterization
of
emissions,
not
only
at
the
sub­
county
(
local)
scale
but
also
at
the
county
scale.

Large­
scale
air
toxics
assessments,
such
as
the
National­
Scale
Air
Toxics
Assessment,
are
very
useful
as
screening
tools,
and
for
identifying
priorities
for
further
analysis.
However,
both
the
level
and
distribution
of
toxic
emissions
vary
greatly
with
estimates
produced
based
on
localized
data.
This
can
in
turn
affect
the
modeled
concentrations
and
estimated
risk.
A
recent
study
conducted
in
the
Minneapolis­
St.
Paul
area
of
Minnesota,
which
used
spatial
surrogates
to
allocate
mobile
source
emissions
to
census
tracts,
found
that
the
Industrial
Source
Complex
(
ISC)

dispersion
model
used
tended
to
overpredict
at
low
monitored
concentrations,
and
undepredict
at
levels
of
high
monitored
concentrations
(
Pratt
et
al.,
2004).
The
investigators
attributed
this
4
pattern
to
the
failure
to
represent
roadway
emissions
properly.
Recent
studies
(
Kinnee
et
al.,

2004;
Cohen
et
al,
2005)
show
that
using
more
refined
local
activity
data
to
allocate
emissions
can
improve
model
performance
and
have
a
significant
impact
on
how
toxic
emissions
are
distributed
in
an
urban
area.
These
studies
also
show
that
there
are
strong
spatial
gradients
of
toxic
emissions
associated
with
roads,
and
that
the
high
levels
of
air
toxics
associated
with
roads
may
increase
risks
for
a
number
of
adverse
health
effects.
Thus,
more
accurate
emissions
data
at
the
local
scale
is
important
in
developing
strategies
to
address
air
toxics
at
the
local
level.

While
the
previous
studies
focused
on
the
distribution
of
emissions
and
resulting
toxics
concentrations
at
the
census
tract
level
and
below,
this
study
also
looks
at
differences
at
the
broader,
county­
level.
Even
for
a
national
assessment,
incorporation
of
more
local
data
through
the
current
NEI
development
process
can
significantly
improve
modeling
results.

Travel
demand
model
data
used
for
transportation
planning
can
provide
much
more
detailed
information
on
the
spatial
distribution
of
roadway
types,
vehicle
activity,
and
speeds
along
those
roads
(
which
can
have
a
large
impact
on
emissions).
These
data
can
be
used
to
more
accurately
estimate
the
magnitude
of
toxic
emissions
at
the
local
scale
and
where
they
occur.
5
In
this
study,
activity
and
speed
data
from
travel
demand
modeling
(
the
TRANPLAN
TRANsportation
PLANning
integrated
transportation
planning
model)
were
used
in
conjunction
with
emission
factor
data
from
EPA's
MOBILE6.2
model
to
develop
a
link­
level
highway
vehicle
air
toxics
inventory
and
to
characterize
the
distribution
of
air
toxic
emissions
for
the
Philadelphia
metropolitan
area.
These
results
were
then
compared
with
the
inventory
and
spatial
distribution
of
emissions
developed
for
the
1999
National
Air
Toxics
Study
(
EPA,
2002).

Methodology
Development
of
a
highway
emission
inventory
for
national
scale
assessments:
the
"
top
down"

approach
Development
of
Emission
Factors
­­
Regional
and
national
scale
Assessments,
such
as
the
National
Scale
Air
Toxics
Assessment
(
U.
S.
EPA,
2002)
typically
rely
on
data
from
EPA's
NEI.

For
the
NEI,
EPA
calculated
air
toxic
emission
factors
for
highway
vehicles
using
EPA's
MOBILE6.2
emission
factor
model
(
EPA,
2004b;
Cook
and
Glover,
2002).
In
addition
to
modeling
criteria
pollutant
emission
factors,
MOBILE6
estimates
emission
factors
for
six
air
toxics
(
benzene,
formaldehyde,
acetaldehyde,
1,3
butadiene,
acrolein,
and
methyl
tertiary
butyl
ether).
Modeling
these
pollutants
requires
detailed
information
on
fuel
parameters
within
the
MOBILE6.2
scenario
descriptions,
along
with
the
standard
inputs,
such
as
roadway
type,
average
temperature,
altitude,
and
inspection
maintenance
program,
required
to
model
criteria
pollutant
6
emissions.
Emission
factors
for
other
HAPs
can
be
estimated
using
a
command
(
ADDITIONAL
HAPS)
which
allows
the
user
to
enter
emission
factors
or
air
toxic
ratios
for
additional
air
toxic
pollutants.

In
the
development
of
the
1999
NEI,
default
values
were
used
for
certain
key
modeling
inputs.

For
instance,
MOBILE6.2
requires
vehicle
registration
distributions
by
age
as
an
input.
Another
required
input
is
the
fraction
of
VMT
for
each
class
of
vehicles.
The
1999
NEI
utilized
the
national
defaults
contained
within
the
MOBILE6.2
model.
Also,
national
default
speeds
for
different
roadway
types
(
e.
g.,
freeways,
arterials,
local
roads,
connector
ramps)
were
used.
For
fuel
inputs
to
MOBILE6.2,
the
NEI
utilized
data
from
fuel
surveys
done
at
service
stations
(
E.
H.

Pechan
and
Associates,
2005;
Cook
and
Glover,
2002).
These
parameters
can
significantly
impact
emission
factors,
and
in
more
localized
assessments,
inventory
accuracy
can
be
greatly
improved
by
replacing
National
defaults
with
local
input
data.

Emission
factors
are
calculated
by
season
or
month,
road
type,
and
vehicle
type,
for
exhaust
and
evaporative
emissions
(
E.
H.
Pechan
and
Associates,
2005).
For
the
1999
National
Emissions
Inventory,
air
toxic
emission
factors
are
estimated
for
each
season
rather
than
each
month.
For
the
NEI,
state,
local
and
tribal
agencies
can
replace
EPA­
generated
emission
estimates
with
their
own
emission
estimates
by
submitting
the
data
to
EPA
during
the
NEI
development
or
review
stages.

The
preferred
method,
however,
is
for
the
state,
local
and
tribal
agencies
to
provide
inputs
to
the
National
Mobile
Inventory
Model
(
EPA,
2005),
the
framework
housing
MOBILE6.2.
This
7
allows
more
local
data
to
be
used
in
a
more
transparent
way.

Development
of
County
Level
VMT
­­
To
develop
county
level
VMT
for
large
urban
areas
such
as
Philadelphia
for
the
NEI,
U.
S.
EPA
relies
on
data
supplied
by
the
Federal
Highway
Administration
(
FHWA)
and
publicly
available
data
from
FHWA's
Highway
Statistics
series
(
DOT,
2002).
Highway
Statistics
is
the
result
of
a
cooperative
effort
between
the
FHWA
and
the
States.
Data
from
this
report
are
used
to
apportion
VMT
into
various
vehicle
categories.
In
addition,
other
data
provided
by
FHWA
are
used
to
apportion
urban
area
VMT
to
different
roadway
types.
Large
urban
area
VMT
are
allocated
by
roadway
type
to
the
county/
roadway
level
using
Bureau
of
the
Census
1990
Number
of
Inhabitants
(
CNOI)
data
at
the
county
level
(
U.
S.

Bureau
of
the
Census,
1992).
The
following
equation
is
used:

A
UX
C
UX
A
UX
C
UX
POP
POP
VMT
VMT
,
,
,
,
×
=
(
1)

where:

VMT
UX,
C
=
Urban
area
VMT
on
roadway
type
X
in
county
C
(
calculated)

VMT
UX,
A
=
Urban
area
VMT
on
roadway
type
X
for
total
urban
area
A
contained
in
state
(
FHWA)

POP
UX,
C
=
Urban
area
population
in
county
C
with
nonzero
mileage
from
urban
roadway
type
X
POP
UX,
A
=
Urban
area
population
for
total
urban
area
A
contained
in
state
totaled
for
all
counties
8
with
nonzero
mileage
from
urban
roadway
type
X
Using
this
methodology,
VMT
could
be
overestimated
for
areas
where
average
VMT
per
capita
is
lower
than
average,
such
as
urban
centers
with
mass
transit,
and
underestimated
in
areas
where
people
tend
to
travel
more,
such
as
outskirts
of
an
urban
area.
Finally,
county­
level
annual
VMT
is
temporally
allocated
to
different
seasons
using
National
seasonal
allocation
factors
(
E.
H.

Pechan
and
Associates,
2005).

Allocation
of
Highway
Vehicle
Emissions
to
Census
Tracts
 
Highway
vehicle
county
level
air
toxic
emissions
from
the
NEI
can
be
allocated
to
census
tracts
or
grid
cells
for
national
or
regional
scale
modeling
using
the
Emission
Modeling
System
for
Hazardous
Air
Pollutants
(

EMSHAP
(
Strum
et
al.,
2004).
EMS­
HAP
allocates
highway
vehicle
emissions
using
the
following
equation:

J
county
tract
J
county
J
county
tract
S
E
E
,
,
,
,
,
×
=
(
2)

where:

E
tract
,
county,
j
=
census
tract
or
grid
cell
emissions
from
source
category
j
in
a
county
E
county,
j
=
emissions
from
category
j
in
county
that
contains
census
tract
or
grid
cell.

S
tract,
county
j
=
spatial
allocation
factor
for
tract
or
grid
cell
in
county
that
corresponds
to
spatial
9
surrogate
assigned
to
source
category
j.

Six
different
spatial
surrogates
are
available
for
allocation
to
census
tracts
nationwide.
Table
1
shows
how
urban
area
VMT
is
allocated
to
counties
and
census
tracts
in
the
NEI.
Onroad
mobile
source
categories
are
mapped
to
these
six
spatial
surrogates.
Roadway
miles
are
used
as
a
surrogate
for
all
categories
but
local
roads,
and
population
is
the
surrogate
for
local
roads.
The
surrogates
were
developed
using
data
from
the
TIGER
and
the
Census
Bureau
as
described
in
Table
2.1
Development
of
a
highway
emissions
inventory
for
local
assessments:
the
"
bottom
up"
approach
Metropolitan
Planning
Organizations
(
MPOs)
are
a
transportation
policy­
making
organizations
made
up
of
representatives
from
local
government
and
transportation
authorities.
MPOs
have
responsibility
for
planning,
programming
and
coordination
of
federal
highway
and
transit
investments
in
urbanized
areas
with
a
population
greater
than
50,000.
One
of
an
MPO's
functions
is
to
develop
long­
range
transportation
plans
for
the
urban
area.
Travel
Demand
Models
(
TDM's)
are
commonly
used
to
predict
the
demand
for
transportation
services
such
as
roads
and
to
assist
in
the
development
of
alternative
plans.
These
models
use
a
link­
node
network
tied
to
geographic
coordinates
to
characterize
travel
patterns
in
the
urban
area.
Associated
with
this
network
are
data
attributes
such
as
number
of
lanes,
roadway
type,
volume,
speed,
and
capacity.

1
More
details
on
surrogates
are
posted
on
http://
www.
epa.
gov/
ttn/
chief/
emch/
spatial/
newsurrogate.
html
10
These
data
can
be
used
with
the
MOBILE6.2
emissions
model
to
create
detailed
emissions
concentrations
and
spatial
distributions.
2
The
geographic
database
and
associated
attribute
data
of
the
Philadelphia
area
roadway
network
were
obtained
from
the
MPO
technical
staff
of
the
Delaware
Valley
Regional
Planning
Commission.
This
database
contains
information
on
all
roads,
except
local
roads,
based
on
1999
information.
Each
road
is
divided
into
links
and
nodes
which
are
road
segments
and
endpoints.

Information
provided
includes
coordinates
of
the
nodes,
roadway
type,
directional
average
annual
daily
traffic
(
AADT)
values,
and
number
of
lanes.
Other
urban
area
MPOs
use
regional
planning
organizations,
local
governments,
state
Departments
of
Transportation,
or
contractors
to
perform
the
technical
aspects
of
TDMs.

This
network
data
was
first
projected
into
Universal
Transverse
Mercator
(
UTM)
coordinates
which
are
required
by
the
air
quality
model.
Each
roadway
segment,
or
link,
in
the
database
consisted
of
x
and
y
coordinates,
a
roadway
type
indicator,
directional
AADT
values,
and
number
of
lanes.
Directional
AADT
values
were
summed
and
multiplied
by
the
roadway
segment
length
to
obtain
daily
Vehicle
Miles
Traveled
(
VMT)
values.
VMT
by
season
and
hour
of
the
day
was
2
(
See
http://
www.
planning.
dot.
gov/
Documents/
BriefingBook/
BBook.
htm#
2BB
for
additional
information
on
the
Metropolitan
Transportation
Planning
Process)
11
then
estimated
using
temporal
distributions
from
the
Delaware
Valley
Regional
Planning
Commission
and
allocated
to
vehicle
classes.

EPA's
MOBILE6.2
model
was
used
to
create
seasonal,
fleet
average
emission
factors
(
in
grams/
mile)
that
were
applied
to
the
hourly
VMT
for
each
season,
producing
vehicle
emissions
for
air
toxics
The
basic
input
files
used
by
EPA
for
the
development
of
their
national
emissions
inventory
were
used
as
the
initial
template
for
this
study.
These
were
then
modified
using
Linux
shell
scripts
to
replace
generalized
inputs
with
inputs
representing
local
conditions.

The
seasonal
fleet
average
emission
factors
from
MOBILE
6.2
varied
by
facility
type
(
freeway,

arterial,
local,
ramp)
and
emission
type
(
exhaust
running,
exhaust
start,
evaporative
hot
soak,

evaporative
diurnal,
evaporative
resting
loss,
evaporative
running
loss,
and
crankcase).
The
same
speed
distribution
and
VMT
fractions
for
individual
vehicle
classes
were
used
for
all
links
of
a
given
facility
type.
Emission
factors
for
running
emissions
(
emissions
produced
from
engine
exhaust
and
evaporation
as
the
engine
is
running)
were
extracted
from
the
output
by
season
and
facility
(
arterial,
interstate),
and
matched
to
the
VMT.
This
yields
the
hourly
emissions
for
each
of
the
air
toxics
for
individual
links.

 
×
=
J
county
J
county
J
county
VMT
FAC
E
,
,
,
(
3)
12
where:

E
county,
j
=
seasonal,
hourly
link
emissions
from
facility
j
in
a
county
FAC
county,
j
=
seasonal
fleet
average
emission
factor
for
facility
type
j
within
each
county
VMT
county,
j
=
seasonal,
hourly
link
VMT
for
facility
type
j
within
each
county
Nonrunning
(
evaporative
emissions
produced
while
the
motor
vehicle
is
not
running
or
exhaust
emissions
from
engine
starting)
emissions
were
derived
from
applying
the
MOBILE6.2
factors
to
total
county
VMT
by
hour,
and
were
then
spatially
distributed
to
1km
x
1km
grids
using
gridded
interstate
and
other
roads
as
surrogates.
Gridding
surrogates
were
developed
by
calculating
the
ratio
between
the
length
of
roadway
in
a
1km
square
modeling
grid
and
the
total
length
of
roadway
in
a
county.
These
ratios
were
then
used
as
surrogates
to
spatially
distribute
the
total
county
emissions
to
the
individual
grid
cells.

While
many
air
toxics
were
included
in
the
Philadelphia
assessment,
this
paper
focuses
primarily
on
benzene
and
four
others
generated
directly
from
MOBILE6.2.
These
four
are
1,3­
butadiene,

formaldehyde,
acetaldehyde,
and
acrolein.
Benzene
is
a
product
of
both
exhaust
and
evaporation
while
the
other
air
toxics
are
only
a
product
of
exhaust.
13
Results
Both
the
spatial
distribution
of
emissions
and
the
total
county
emissions
differ
significantly
between
the
top­
down
and
the
bottom­
up
methodologies
described
above.
Figure
1
illustrates
the
spatial
distribution
of
benzene
emissions
using
the
local
approach.
Higher
emissions
are
distributed
along
major
roadways
such
as
interstates
and
are
concentrated
at
major
highway
intersections
reflecting
a
more
accurate
spatial
distribution
the
emissions
when
compared
to
the
national
methodology,
where
emissions
are
treated
as
"
pseudo"
point
sources
within
a
grid
cell
or
census
tract.

Figure
2
displays
benzene
emission
distribution
by
census
tract
as
used
in
the
1999
NATA.

Although
the
greater
emissions
density
tracts
generally
follow
the
interstates
and
freeways,
the
distribution
within
a
tract
is
uniform.
Differences
in
the
spatial
distribution
of
emissions
between
the
two
methodologies
is
shown
in
Figure
3.
Link
based
emissions
were
aggregated
to
tracts
for
the
purpose
of
this
comparison.

Using
the
two
approaches,
the
total
inventories
for
air
toxics
vary
significantly
for
the
modeling
domain.
We
compared
the
two
approaches
for
counties
entirely
within
the
modeling
domain.

Table
3
shows
that
total
county
benzene
emissions
vary
even
more
significantly
between
the
approaches.
Philadelphia
county
benzene
emissions,
for
example,
using
the
local
approach,
are
about
half
of
the
emissions
derived
from
using
the
national
approach.
A
similar
pattern
is
seen
for
14
other
air
toxics
(
Table
4).

Most
differences
in
emissions
among
counties
between
the
two
approaches
can
be
explained
by
examining
the
VMT,
summarized
in
Table
5.
While
county
level
VMT
varies
greatly,
the
total
VMT
for
the
counties
in
the
Philadelphia
metropolitan
area
is
close.
The
differences
are
a
direct
result
of
the
approaches
used.
The
national
approach
allocates
VMT
to
the
individual
counties
in
an
urban
area
based
on
population
ratios
while
the
local
approach
uses
actual
traffic
count
data
(
e.
g.,
VMT)
for
each
roadway
segment
in
a
county
to
arrive
at
total
county
emissions.

However,
despite
the
fact
that
total
VMT
for
the
Philadelphia
metropolitan
area
is
similar
between
the
two
approaches,
the
total
benzene
inventory
for
counties
entirely
within
the
modeling
domain
is
about
13%
lower
when
the
local
approach
is
used.
This
indicates
that
the
MOBILE6.2
inputs
used
in
the
local
approach
are
also
playing
a
role.
In
this
assessment,
the
differences
are
attributable
to
exhaust
emissions,
rather
than
evaporative.
This
is
because
they
are
even
more
pronounced
for
the
air
toxics
in
Table
4,
which
do
not
have
an
evaporative
component,
than
they
are
for
benzene.
A
sensitivity
analysis
of
criteria
pollutant
estimates
done
for
MOBILE6
indicated
that
the
following
inputs
had
a
large
impact
on
emission
factors:
speed,
registration
distribution,

VMT
fractions
for
individual
vehicle
classes,
temperature,
and
RVP
(
Gianelli
et
al.,
2002).

Furthermore,
other
fuel
properties
have
a
large
impact
on
toxic
emission
rates
as
well
(
U.
S.
EPA,

1999).
Thus,
we
explored
the
impacts
these
key
inventory
inputs
had
on
benzene
emission
results
using
the
two
approaches.
15
Small
increases
in
the
fraction
of
vehicles
traveling
at
low
speeds
can
have
a
large
impact
on
emissions.
For
default
MOBILE6.2
runs,
such
as
those
done
for
the
NEI,
an
AVERAGE
SPEED
command
is
used.
With
this
command,
the
VMT
are
assigned
to
two
speed
ranges
for
freeways
and
one
speed
range
for
arterials.
In
the
Philadelphia
modeling,
the
MOBILE6
SPEED
VMT
command
was
used,
which
assigns
VMT
to
14
speed
ranges
for
both
freeways
and
arterials.
A
comparison
of
emission
rates
using
the
SPEED
VMT
command
versus
the
AVERAGE
SPEED
command
at
various
average
speeds
found
very
little
difference
in
results
for
benzene.

The
registration
distributions
used
in
the
local
approach
can
also
have
a
large
impact.
A
20%

shift
in
vehicle
fractions
among
age
classes
can
lead
to
an
increase
in
emissions
of
up
to
50%

(
Gianelli,
et
al.,
2002).
A
comparison
of
benzene
emission
rates
for
Philadelphia
County,
using
local
distributions
versus
the
MOBILE6
defaults,
with
all
other
inputs
unchanged,
had
a
significant
impact.
Emission
factors
were
more
than
10%
lower
using
the
local
distributions
(
Table
6).

Another
input
which
varied
between
the
local
approach
and
the
NEI
was
the
fraction
of
the
total
VMT
assigned
to
individual
vehicle
classes.
However,
a
comparison
of
the
impact
of
using
local
VMT
fractions
on
benzene
emission
factors
in
Philadelphia
County,
with
all
other
inputs
unchanged,
found
this
had
little
effect
(
about
0.6%).
Other
inputs
which
have
been
found
to
significantly
impact
toxic
emission
factors,
such
as
temperature
and
RVP,
did
not
vary
between
16
the
two
modeling
approaches.
Thus,
differences
in
registration
distributions
between
the
two
approaches
explain
a
significant
portion
of
the
total
inventory
difference
for
the
modeling
domain.

Conclusion
Use
of
local
level
inputs
can
have
a
significant
impact
on
the
magnitude
and
distribution
of
highway
vehicle
air
toxic
emissions.
This
study
shows
that
estimating
vehicle
running
emissions
for
individual
road
links
to
develop
an
inventory
results
in
a
much
different
spatial
distribution
of
emissions
than
allocating
emissions
using
spatial
surrogates.
Furthermore,
the
population
ratio
methodology
used
to
allocate
VMT
to
counties
in
the
NEI
is
inconsistent
with
travel
demand
model
estimates
of
where
this
activity
is
occurring.
Moreover,
use
of
local
data
rather
than
default
national
inputs
can
also
result
in
significant
differences
in
air
toxics
emissions
estimates.
In
the
case
of
Philadelphia,
using
local
registration
distribution
data
results
in
significantly
lower
air
toxics
emission
factors
and
resultant
emissions.
In
addition
to
improving
the
quality
of
local
scale
assessment,
using
these
local
data
can
improve
the
quality
of
National­
scale
assessment
through
the
current
NEI
development
process.
Use
of
local
inputs
and
TDM
developed
VMT
with
the
"
top
down"
approach
used
in
the
NEI
results
in
county­
wide
inventories
which
closely
approximate
county­
wide
inventories
developed
using
a
more
disaggregate
"
bottom
up"

approach.

Acknowledgements
17
We
would
like
to
acknowledge
the
contribution
of
Ray
Chalmers,
Dr.
James
D.
Smith,
Brian
Rehn
and
Alan
Cimorelli
of
EPA
Region
3,
and
Harvey
Michaels,
David
Brzezinski,
Megan
Beardsley,
and
Chad
Bailey
of
EPA's
Office
of
Transportation
and
Air
Quality.
We
also
appreciate
the
comments
of
Laurel
Driver,
George
Pouliot,
and
William
Benjey.

Disclaimer
This
paper
has
been
reviewed
in
accordance
with
the
US
Environmental
Protection
Agency's
peer
and
administrative
review
policies
and
approved
for
publication.
Mention
of
trade
names
or
commercial
products
does
not
constitute
endorsement
or
recommendation
of
their
use.
18
References
Cohen,
J.,
Cook,
R.,
Bailey,
C.,
Carr,
E.
(
2005)
Relationship
between
motor
vehicle
emissions
of
hazardous
air
pollutants,
roadway
proximity,
and
ambient
concentrations
in
Portland,
Oregon.

Environ.
Modelling
and
Software
20,
7­
12.

Cook,
R.;
Glover,
E.,
(
2002)
Technical
Description
of
the
Toxics
Module
for
MOBILE6.2
and
Guidance
on
Its
Use
for
Emission
Inventory
Preparation.
U.
S.
EPA,
Office
of
Transportation
and
Air
Quality,
Ann
Arbor,
MI,
Report
No.,
EPA420­
R­
02­
011.
Located
at:

<
http://
www.
epa.
gov/
otaq/
m6.
htm>

E.
H.
Pechan
and
Associates,
(
2005)
Documentation
for
the
Onroad
National
Emissions
Inventory
(
NEI)
for
Base
Years
1970­
2002.
Prepared
for
U.
S.
EPA,
Office
of
Air
Quality
Planning
and
Standards,
Emission
Factor
and
Inventory
Group,
Research
Triangle
Park,
NC.

Located
at:
<
http://
www.
epa.
gov/
ttn/
chief/
nei
Gianelli,
R.,
Gillmore,
J.,
Landman,
L.,
Srivastava,
S.,
Beardsley,
M.,
Brzezinski,
D.,
Dolce,
G.,

Koupal,
J.,
Pedelty,
J.,
Shyu,
G.
(
2002)
Sensitivity
Analysis
of
MOBILE6.0.
EPA
Report
No.

EPA420­
R­
02­
035.
http://
www.
epa.
gov/
OMSWWW/
models/
mobile6/
m6tech.
htm
19
Kinnee,
E.,
Touma,
J.,
Mason,
R,
Thurman,
J.,
Beidler,
A.,
Bailey,
C.,
Cook,
R.
(
2004).

Allocation
of
Onroad
Mobile
Emissions
to
Road
Segments
for
Air
Toxics
Modeling
in
an
Urban
Area,
Transportation
Research
Part
D
9,
139­
150.

Pratt,
G.,
Wu,
C.,
Bock,
D.
(
2004).
Comparing
air
dispersion
model
predictions
with
measured
concentrations
of
VOCs
in
urban
areas.
Environmental
Scientific
Technology.
38:
423­
430.

Strum,
M.;
Mason,
R,;
Thurman,
J.
(
2004)
User's
Guide
for
the
Emissions
Modeling
System
for
Hazardous
Air
Pollutants
(
EMS­
HAP)
Version
3.0.
U.
S.
Environmental
Protection
Agency,

Office
of
Air
Quality
Planning
and
Standards,
Emissions,
Monitoring
and
Analysis
Division,

Research
Triangle
Park,
NC.
Report
No.
EPA­
454/
B­
03­
006.
Located
at:

<
http://
www.
epa.
gov/
scram001/
userg/
other/
emshapv3ug.
pdf>

DOT
(
U.
S.
Department
of
Transportation),
(
2002).
Highway
Statistics
1999:
May
2002
Update.

Office
of
Highway
Information
Management,
Office
of
Policy
Development,
Federal
Highway
Administration.
U.
S.
Government
Printing
Office.
Washington,
D.
C.
Located
at:

<
http://
www.
fhwa.
dot.
gov/
ohim/
hs99/
index.
html>

U.
S.
Bureau
of
the
Census,
(
1992)
1990
Census
of
Population,
Volume
I
Characteristics
of
Population,
Chapter
B
Number
of
Inhabitants.
Located
at:
<

http://
www.
census.
gov/
prod/
www/
abs/
decennial.
html>
20
U.
S.
EPA.
(
1999).
Analysis
of
the
Impacts
of
Control
Programs
on
Motor
Vehicle
Toxics
Emissions
and
Exposure
in
Urban
Areas
and
Nationwide.
Prepared
by
Sierra
Research,
Inc.
and
Eastern
Research
Group,
Inc.
for
Office
of
Transportation
and
Air
Quality,
Ann
Arbor,
MI.

Report
No.
EPA420­
R­
99­
029.
http://
www.
epa.
gov/
otaq/
regs/
toxics/
tox_
archive.
htm#
1
EPA
(
U.
S.
Environmental
Protection
Agency),
(
2002)
National
Air
Toxics
Assessment
for
1996.

Located
at:
<
http://
www.
epa.
gov/
ttn/
atw/
nata>

EPA
(
U.
S.
Environmental
Protection
Agency),
(
2004b).
1999
National
Emissions
Inventory,

Final
Version
3.0.
Located
at:
<
http://
www.
epa.
gov/
ttn/
chief/
net/
1999inventory.
html>

EPA
(
U.
S.
Environmental
Protection
Agency),
(
2004b)
MOBILE6
Vehicle
Emissions
Model.

Located
at:
<
http://
www.
epa.
gov/
otaq/
m6.
htm>

EPA
(
U.
S.
Environmental
Protection
Agency),
(
2005).
EPA's
National
Mobile
Inventory
Model
(
NMIM),
A
Consolidated
Emissions
Modeling
System
for
MOBILE6
and
NONROAD.

http://
www.
epa.
gov/
otaq/
nmim.
htm
21
Table
1
Mapping
of
Mobile
Source
Emission
Roadway
Categories
to
Spatial
Surrogates
Highway
Mobile
Source
(
FHWA
category)
Spatial
Surrogate
Assignment
(
Miles)

rural
interstate
rural
primary
roads
rural
other
principal
arterial
rural
primary
roads
rural
minor
arterial
rural
primary
roads
rural
major
collector
rural
secondary
roads
rural
minor
collector
rural
secondary
roads
rural
local
rural
population
urban
interstate
urban
primary
roads
urban
other
freeways
and
expressways
urban
primary
roads
urban
other
principal
arterial
urban
primary
roads
urban
minor
arterial
urban
primary
roads
urban
collector
urban
secondary
roads
urban
local
urban
population
22
Table
2
Data
Used
to
Develop
Surrogates
for
Mobile
Sources
for
National/
Regional
Scale
Modeling
Surrogate
Description
of
how
surrogate
was
developed
Urban
Primary
Roads
Road
Miles
of
Urban
Primary
Roads.
Overlayed
US
Census
Bureau's
urban
areas
with
TIGER
roads.
Urban
primary
roads
are
roads
with
CFCC
A11
through
CFCC
A28
and
A63
that
fall
within
census
designated
urban
areas.

Rural
Primary
Roads
Road
Miles
of
Rural
Primary
Roads.
Overlayed
US
Census
Bureau's
urban
areas
with
TIGER
roads.
Rural
primary
roads
are
roads
with
CFCC
A11
through
CFCC
A28
and
A63
that
fall
within
census
designated
rural
areas.

Urban
Secondary
Roads
Road
Miles
of
Urban
Secondary
Roads.
Overlayed
US
Census
Bureau's
urban
areas
with
TIGER
roads.
Urban
secondary
roads
are
roads
with
CFCC
A31
through
CFCC
A38
that
fall
within
census
designated
urban
areas.

Rural
Secondary
Roads
Road
Miles
of
Rural
Secondary
Roads.
Overlayed
US
Census
Bureau's
urban
areas
with
TIGER
roads.
Rural
secondary
roads
are
roads
with
CFCC
A31
through
CFCC
A38
that
fall
within
census
designated
rural
areas.

Table
3
Comparison
of
Annual
1999
Benzene
Emissions
(
tons/
year)
from
Two
Approaches
County
Local
(
TDM)
Based
National
(
NEI)
Percent
Difference
Camden
165
210
­
27%
Delaware
162
160
1%
Gloucester
110
104
6%
Montgomery
333
209
59%
Philadelphia
255
467
­
45%
Total
1,025
1,150
­
12%
23
Table
4
Comparison
of
Annual
1999
Emissions
(
tons/
year)
for
Selected
HAPs
Using
the
Two
Approaches
County
Butadiene,
1,3­
Formaldehyde
Acetaldehyde
Acrolein
TDM
NEI
TDM
NEI
TDM
NEI
TDM
NEI
Camden
25
34
93
128
29
37
4
6
Delaware
21
26
79
105
24
30
3
4
Gloucester
17
17
62
66
19
19
3
3
Montgomery
44
35
165
139
51
40
7
6
Philadelphia
34
77
127
305
39
86
5
13
Total
141
189
526
744
162
211
22
32
24
Table
5
Comparison
of
Annual
VMT
From
Two
Approaches
Annual
VMT
(
Vehicle
Miles
Traveled)
1999
County
VMT
Derived
from
TDM
(
Travel
Demand
Model
)
VMT
From
Highway
Statistics
(
used
in
NEI)
Percent
Difference
Bucks
4,878,636,990
3,710,307,400
31%
Burlington
4,187,820,178
3,567,877,700
17%
Camden
3,682,826,156
4,199,354,700
­
12%
Chester
4,491,670,453
3,046,058,600
47%
Delaware
3,155,112,296
3,373,141,300
­
6%
Glouster
2,524,701,655
2,195,654,200
15%
Mercer
3,136,953,593
3,555,648,300
­
12%
Montgomery
6,381,619,077
4,508,702,800
42%
Philadelphia
4,864,568,590
9,804,935,300
­
50%
Total
37,303,908,988
37,961,680,300
­
2%

Table
6.
Impact
of
Local
Registration
Distribution
on
MOBILE6
Benzene
Emission
Factors
(
All
Other
Inputs
Unchanged)

Emission
Factor
(
gm/
mi)
Facility
Speed
Local
M6
Default
%
Difference
Arterial
10
64
74
­
13
Arterial
15
51
59
­
13
Arterial
20
44
51
­
14
Arterial
25
41
47
­
14
Arterial
30
38
44
­
13
Arterial
35
37
42
­
13
Arterial
40
36
42
­
13
Arterial
45
36
41
­
13
Freeway
10
61
71
­
14
Freeway
15
47
55
­
14
Freeway
20
43
49
­
14
Freeway
25
40
47
­
13
Freeway
30
39
45
­
13
Freeway
35
38
43
­
13
Freeway
40
37
43
­
13
Freeway
45
37
42
­
12
Freeway
50
37
42
­
12
Freeway
55
37
42
­
12
Freeway
60
37
41
­
12
Freeway
65
36
40
­
12
25
Figure
1
Philadelphia
Link
Based
Benzene
Emissions
Distribution
0­
0.03
0.03
­
0.10
0.10
­
0.25
0.25
­
0.5
Tons/
Year
26
Figure
2
Philadelphia
Tract
Emissions
Density
from
1999
NATA
0.087532
­
0.136542
0.050421
­
0.087531
0.028344
­
0.050420
0.012306
­
0.028343
0.000000
­
0.012305
Tons/
Year/
Sq.
Mile
27
Figure
3
Difference
in
Emissions
Density
(
1999
NATA
Versus
Local
Approach
Using
Link
Based
Emissions)

Tons/
Year/
Sq.
Mile
­
0.034
­­
0
0
­­
0.120
