1
Final
Draft
November
2004
EVALUATION
AND
CHARACTERISATION
OF
REACTIVITY
METRICS
Report
to
the
United
States
Environmental
Protection
Agency
Research
Triangle
Park,
NC27711
Order
No.
4D­
5844­
NATX
By
R.
G.
(
Dick)
Derwent
rdscientific
Newbury,
United
Kingdom
2
EXECUTIVE
SUMMARY
This
report
describes
the
evaluation
and
characterisation
of
reactivity
metrics
to
support
the
United
States
Environmental
Protection
Agency's
efforts
to
update
its
reactivity
policy
for
the
attainment
of
ozone
air
quality
standards.
The
review
is
based
on
three
detailed
computer
modelling
studies
which
taken
together
have
made
a
significant
contribution
to
basic
understanding
of
the
likely
role
that
reactivity
considerations
could
play
in
ozone
control
strategies.
Reactivity­
based
policies
should
work
efficiently
on
both
the
urban
and
regional
scales
by
reducing
episodic
peak
ozone
levels
and
by
reducing
exceedances
of
the
ozone
air
quality
standards.
Two
preferred
metrics
have
been
selected
as
the
most
promising
for
policy
development
and
these
are
the
EKMA
and
Regional
or
3­
D
Maximum
Incremental
Reactivity
metrics.
3­
D
Eulerian
regional
air
quality
models
have
been
demonstrated
to
be
ideal
tools
for
the
visualisation
of
the
ozone
impacts
of
reactivity­
based
VOC
substitution
strategies
across
central
California,
the
south­
western
and
south­
eastern
United
States
of
America.
The
choice
of
chemical
mechanism,
together
with
its
adequacy
and
completeness,
are
crucial
to
the
estimation
of
reactivity
metrics
and
the
compilation
of
reactivity
scales.

If
the
potential
value
of
reactivity­
based
VOC
control
strategies
is
to
become
accepted
by
policy­
makers
for
the
attainment
of
ozone
air
quality
standards,
then
two
basic
tools
will
be
required.
The
first
is
an
evaluated
reactivity
scale
that
expresses
our
best
understanding
of
the
relative
importance
of
each
of
upwards
of
100
to
200
VOCs
in
photochemical
ozone
formation
under
given
environmental
conditions.
The
second
is
a
3­
D
Eulerian
regional
air
quality
model
to
provide
the
best
possible
visualisation
of
the
urban
and
regional
scale
ozone
impacts
of
candidate
reactivity­
based
VOC
control
strategies.
3
1.
Introduction
Sunlight­
driven
photochemical
reactions
in
the
atmospheric
boundary
layer
involving
volatile
organic
compounds
(
VOCs)
and
nitrogen
oxides
(
NOx)
result
in
the
formation
of
ozone
and
fine
particles.
Each
VOC
has
widely
differing
reaction
rates
and
shows
differing
propensities
to
form
ozone.
Each
VOC
will
therefore
make
a
different
contribution
to
the
elevated
ozone
concentrations
found
in
each
pollution
episode
in
each
situation.
Basic
understanding
of
these
different
reaction
rates
and
reaction
pathways
for
each
VOC
has
increased
steadily
in
recent
years
to
the
extent
that,
the
United
States
Environmental
Protection
Agency
is
considering
whether
it
is
possible
to
use
this
increased
understanding
to
develop
more
efficient,
accurate
and
cost­
effective
control
strategies
for
meeting
the
ozone
air
quality
standards.

The
focus
of
controls
on
the
emissions
of
ozone
precursors
has
been
largely
on
the
reductions
in
mass
emissions
of
VOCs
without
regard
to
their
different
contributions
to
photochemical
ozone
production
(
Dimitriades,
1999).
However,
included
in
the
policy
framework
has
been
the
provision
for
the
exclusion
from
the
emission
control
regulations
of
those
compounds
that
are
considered
to
be
of
negligible
reactivity.
Inherent
in
this
approach
is
the
view
that
selective
control
of
VOC
emissions
is
more
cost­
effective
than
the
indiscriminate
mass
control
approach.

The
contribution
made
by
any
individual
VOC
to
photochemical
ozone
formation
depends
crucially
on
environmental
conditions.
However,
the
contributions
made
by
one
VOC
relative
to
another,
or
pairs
or
groups
of
VOCs
relative
to
each
other,
are
much
less
variable.
Ethane
is
thus
relatively
unreactive
compared
with
most
VOCs
and
isoprene
is
relatively
more
highly
reactive,
under
most
conditions.
The
relative
rankings
of
VOCs
in
a
quantitative
table
is
termed
a
reactivity
scale
and
the
most
frequently
quoted
such
scale
is
the
Maximum
Incremental
Reactivity
(
MIR)
scale
(
Carter,
1994).
The
MIR
scale
is
the
most
appropriate
scale
to
quantify
VOC
reactivity
under
photochemical
ozone
formation
conditions
most
sensitive
to
VOC
emissions
under
the
highest
NOx
conditions
associated
with
intense
ozone
source
regions
in
the
Los
Angeles
basin.
It
has
been
adopted
in
the
state
of
California
for
the
purpose
of
implementing
reactivity­
based
regulations
(
CARB,
1993).

The
Reactivity
Research
Working
Group
(
RRWG,
1999)
coordinates
policy­
relevant
research
related
to
VOC
reactivity
and
has
overseen
three
studies
performed
by
Carter
et
al.
(
2003),
Arunachalam
et
al.
(
2003)
and
Hakami
et
al.
(
2003,
2004).
Taken
together,
these
three
studies
have
made
a
significant
contribution
to
basic
understanding
of
the
likely
role
that
reactivity
considerations
could
play
in
ozone
control
strategies.
These
studies
have
shown
that
reactivity­
based
policies
should
work
efficiently
on
both
the
urban
and
regional
scales
by
reducing
episodic
peak
ozone
levels
and
by
reducing
the
exceedances
of
the
ozone
air
quality
standards.

This
report
describes
the
evaluation
and
characterisation
of
reactivity
metrics
to
support
the
United
States
Environmental
Protection
Agency's
efforts
to
update
its
reactivity
4
policy
for
the
attainment
of
ozone
air
quality
standards.
The
review
is
based
on
the
above
three
studies
performed
by
Carter
et
al.
(
2003),
Arunachalam
et
al.
(
2003)
and
Hakami
et
al.
(
2003,
2004)
and
on
the
existing
VOC
reactivity
literature,
as
appropriate.

2.
Review
and
Assessment
of
VOC
Reactivity
Metric
Studies
The
incremental
reactivity
of
a
VOC
species
is
estimated
from
the
incremental
increases
in
ozone
exposure
levels
caused
by
the
incremental
increases
in
the
emissions
of
particular
VOCs.
A
ranked
table
of
incremental
reactivities
for
a
series
of
VOCs
comprises
a
reactivity
scale.
Because
ozone
exposure
levels
can
be
represented
in
a
number
of
ways,
these
different
representations
or
metrics
can
be
used
to
define
a
number
of
reactivity
scales.

Since
reactivity­
based
policies
are
ultimately
about
achieving
ambient
air
quality
standards
for
ozone,
the
most
appropriate
quantification
of
ozone
exposure
levels
is
to
use
the
1­
hour
and
8­
hour
air
quality
standards.
In
addition,
however,
there
may
be
many
alternative
methodologies
for
the
integration
of
the
ozone
impacts
across
the
model
domains
and
over
each
episode
day.
Only
a
subset
of
the
many
possibilities
have
been
considered
here
in
the
studies
by
Carter
et
al.
(
2003)
and
Hakami
et
al.
(
2003,
2004).

Before
reactivities
and
reactivity
scales
can
be
quantified,
it
is
necessary
to
specify
the
time
period
being
studied,
the
model
domains
that
are
being
analysed,
the
methods
used
to
quantify
ozone
exposure
levels
in
a
given
cell
and
the
methodologies
used
to
derive
a
single
reactivity
metric
for
the
entire
model
domain.

Question
1.
What
are
the
attributes
of
the
reactivity
metrics
that
influence
their
suitability
for
use
in
large­
scale
policy
applications
?

The
main
attribute
of
the
reactivity
metrics
that
influences
their
suitability
for
policy
applications
is
robustness
across
model
domains
and
between
episode
days.
It
is
crucial
that
the
relative
positions
of
the
VOCs
or
groups
of
VOCs
in
a
reactivity
scale,
estimated
for
a
given
reactivity
metric,
do
not
shift
significantly
or
change
or
reverse
between
different
domains
and
episode
days.
An
important
issue
is
whether
a
reactivity
scale
that
works
in
New
York,
with
its
high
population
density
will
also
work
in
Atlanta
with
its
large
biogenic
VOC
emissions
(
Hakami
et
al.
2004).

A
further
important
characteristic
of
a
reactivity
scale
and
hence
an
important
attribute
of
a
reactivity
metric,
from
a
policy
perspective
is
its
effective
range
(
Carter
et
al.
2003).
The
effective
range
of
a
reactivity
scale
is
the
range
of
reactivity
values
generated
between
the
most
reactive
VOC
and
an
unreactive
VOC
such
as
ethane.
If
a
reactivity
scale
is
used
with
too
low
an
effective
range
then
there
would
be
insufficient
incentive
for
the
consideration
of
reactivity­
based
policies.
A
reactivity
scale
with
an
unreasonably
5
high
an
effective
range
may
lead
to
unacceptable
ozone
impacts
with
some
large
substitutions
of
low
reactivity
VOCs.

Hakami
et
al.
(
2004)
consider
six
properties
of
reactivity
scales
and
metrics
that
are
relevant
for
air
quality
management
purposes.
These
include
spatial
robustness,
uncertainty,
risk
of
being
inaccurate,
population
weighting,
fairness
and
regionally
representivity.
Ease
of
calculation
is
not
considered
to
be
an
important
attribute
for
a
reactivity
metric.
Whilst
spatial
robustness
is
the
most
important
property
of
a
reactivity
scale,
the
other
five
properties
are
best
evaluated
when
candidate
reactivity­
based
strategies
have
been
visualized
in
a
3­
D
Eulerian
air
quality
model.

Question
2.
What
are
the
most
suitable
reactivity
metrics
from
both
scientific
and
practical
application
perspectives
for
potential
development
of
VOC
control
strategies
based
on
VOC
reactivity
?

Carter
et
al.
(
2003)
describe
how
they
have
employed
the
CAMx
model
to
study
reactivity
metrics
across
the
eastern
United
States.
The
study
period
chosen
was
the
NARSTO­
NE
episode
which
covered
four
days
during
July
1995.
The
model
employed
the
Carbon
Bond
4
(
CB4)
mechanism
and
was
run
at
4
km
x
4
km,
12
km
x
12
km
and
36
km
x
36
km
scales.
Reactivity
metrics
were
estimated
using
the
direct
decoupled
method
(
DDM)
for
7
methods,
4
episode
days,
3
grid
resolutions
and
two
ozone
standards.
In
each
case,
a
reactivity
scale
was
produced
comprising
the
following
nine
VOC
species:
OLE,
formaldehyde,
acetaldehyde,
ethane,
xylenes,
ethanol,
PAR,
ethane
and
TOL.

Table
1
summarises
the
evaluation
of
each
of
the
reactivity
metrics
over
the
different
episode
days,
model
domains
and
grid
resolutions
based
on
Carter
et
al.
(
2003).
Carter
et
al.
(
2003)
comment
that
the
differences
between
the
different
reactivity
metrics
for
the
same
species
were
less
than
the
differences
between
the
species
for
the
same
metric.
Hence,
in
each
case,
a
usable
reactivity
scale
could
be
constructed
from
the
available
results.
In
general,
the
use
of
the
8­
hour
or
1­
hour
ozone
standards
made
little
difference
to
the
reactivity
metrics
and
the
use
of
cut­
offs
in
the
ozone
levels
of
up
to
80
ppb
also
did
not
significantly
change
the
reactivity
metrics.
Significant
differences
were
seen
between
the
4
km
x
4
km
grid
resolutions
and
the
12
km
x
12
km
and
36
km
x
36
km
resolutions
because
of
the
better
representation
at
the
4
km
x
4
km
resolution
of
urban
source
regions
which
are
more
VOC­
sensitive.

In
summarizing
their
results,
Carter
et
al.
(
2003)
noted
that
for
most
model
species
the
EKMA
reactivity
metrics
were
surprisingly
close
to
the
regional
model
relative
reactivity
metrics.
The
three
3­
D
grid
model
reactivity
metrics
that
were
favoured
were:
Regional
MIR,
Regional
MIR
to
MOIR
and
Minimum
Substitution
Error
 
Method
1.
The
EKMA
and
Regional
MIR
reactivity
scales
showed
the
largest
effective
ranges,
whilst
those
of
the
Regional
MIR
to
MOIR
and
Minimum
Substitution
Error
 
Method
1,
showed
reduced
effective
ranges.
6
Table
1.
The
evaluation
of
the
reactivity
metrics
over
the
different
episode
days,
model
domains
and
grid
resolutions
described
by
Carter
et
al.
(
2003).

Reactivity
metric
Evaluation
Regional
Average
Ozonea
Disadvantage
is
that
this
metric
includes
many
low
ozone
cells
that
are
the
least
VOC­
sensitive
Regional
Maximum
Ozoneb
Disadvantage
is
that
it
is
highly
variable
between
episode
days
and
model
domains
and
sometimes
the
absolute
peak
ozone
is
insensitive
to
VOC
emissions
Regional
Average
Ozone
Over
Standardc
This
is
a
robust
metric
that
shows
little
variability
and
gives
the
best
estimates
for
regions
in
exceedance
of
standards
Minimum
Substitution
Error
 
Method
1d
Disadvantage
that
this
method
does
not
work
well
for
low
and
negative
reactivity
species
but
weights
heavily
VOC­
sensitive
cells
Minimum
Substitution
Error
 
Method
2e
Disadvantage
that
this
method
is
sometimes
unreliable
for
some
species
Regional
Maximum
Incremental
Reactivityf
This
is
a
robust
metric
because
all
the
cells
used
have
similar
VOC­
sensitivities
and
negative
NOx
sensitivities.
Disadvantage
is
that
this
is
not
a
truly
global
metric
Regional
MIR
to
MOIRg
This
is
a
robust
metric
that
shows
little
variability
because
all
the
cells
used
have
similar
VOC
sensitivities
and
are
where
NOx
controls
would
be
counter­
productive
Notes:
a.
the
average
incremental
reactivity
over
the
entire
model
domain.
b.
the
incremental
reactivity
in
the
cell
with
the
domain­
wide
peak
concentration.
c.
the
average
incremental
reactivity
for
all
grid
cells
exceeding
the
standard.
d.
the
incremental
reactivity
that
minimizes
the
sum
of
squares
of
ozone
changes
in
all
cells
caused
by
substituting
a
VOC
by
the
base
ROG.
e.
the
incremental
reactivity
that
minimizes
the
sum
of
squares
of
ozone
changes
in
all
cells
caused
by
substituting
the
base
ROG
by
a
VOC.
f.
The
incremental
reactivities
in
the
cells
with
the
highest
base
ROG
sensitivity.
g.
The
incremental
reactivities
for
all
grid
cells
with
negative
NOx­
sensitivity.
7
Table
2.
The
evaluation
of
the
reactivity
metrics
over
the
different
episode
days,
model
domains
and
grid
resolutions
described
by
Hakami
et
al.
(
2004).

Reactivity
metric
Evaluation
3­
D
Maximum
Incremental
Reactivity
MIR­
3D
Consistent
across
model
domains
and
days
and
protective
of
urban
populations
with
high
NOx
levels
3­
D
Peak
Ozone
Incremental
Reactivitya
POIR­
3D
or
3­
D
Maximum
Ozone
Incremental
Reactivity
MOIR­
3D
Based
on
the
relative
reactivity
at
one
point
and
perhaps
relatively
rare
conditions
of
most
NOx­
limited,
less
typical
of
population
centres,
least
VOC­
sensitive
Least
Square
Relative
Reactivity
LS
­
RR
Consistent
across
model
domains
and
episode
days
Regional
Average
Ozone
AVG
Showed
considerable
inter­
domain
variability
Regional
Average
Ozone
Above
Standard
AVS
Showed
considerable
inter­
domain
variability
3­
D
Maximum
Incremental
Reactivity
to
Maximum
Ozone
Incremental
Reactivity
M2M
Showed
greatest
inter­
domain
consistency,
little
variability
between
episode
days
Notes:
a.
for
the
conditions
where
the
sensitivity
of
the
maximum
ozone
to
the
initial
NOx
is
zero.

Hakami
et
al.
(
2003,
2004)
describe
how
they
have
used
the
Urban­
to­
Regional
Multiscale
(
URM)
and
MAQSIP
3­
D
models
to
estimate
reactivity
metrics
over
the
Eastern
United
States
and
in
central
California.
The
NARSTO­
NE
episodes
during
May
and
July
1995
were
used
for
the
former
domain
and
the
SARMAP
episode
for
the
latter.
The
models
were
applied
at
a
spatial
resolution
of
24
km
x
24
km
with
other
grid
sizes
up
to
48
km,
96
km
and
192
km
and
used
the
SAPRC
99
chemical
mechanism.
This
mechanism
represented
42
explicit
VOC
species
of
which
9
were
lumped
VOC
species.
Reactivity
metrics
using
6
different
methodologies
were
calculated
using
the
direct
decoupled
method.

In
summarizing
their
results,
Hakami
et
al.
(
2003)
observed
that
the
Peak
or
Maximum
Ozone
Incremental
Reactivity
POIR­
3D
or
MOIR­
3D
reactivity
metric
was
not
adequately
robust
across
the
episode
days
and
model
domains.
The
Maximum
Incremental
Reactivity
MIR­
3D
metric
gave
the
best
agreement
with
the
box
model
reactivity
scales
for
all
episodes.
The
Least
Square
 
Relative
Reactivity
LS
 
RR
metric
showed
the
best
spatial
consistency
over
the
model
domains.
Reactivity
metrics
based
on
the
1­
hour
and
8­
hour
ozone
standards
behaved
similarly
and
there
was
little
difference
between
the
metrics
calculated
with
80
ppb
and
60
ppb
ozone
cut­
offs.
There
were
also
8
little
differences
between
the
reactivity
metrics
calculated
with
the
different
model
emission
inventories.
However,
the
reactivities
of
aldehydes
in
the
3­
D
model
were
generally
lower
than
those
reported
from
box
model
studies
due
to
the
more
efficient
carry
over
of
radical
sources
in
the
3­
D
model.
Reactivities
of
 ­
pinene
and
the
trimethylbenzenes
were
lower
than
box
model
values
and
that
of
methanol
was
higher
in
the
3­
D
model.

Table
2
summarises
the
evaluation
of
the
different
reactivity
metrics
put
forward
by
Hakami
et
al.
(
2004)
covering
the
different
models,
episodes
and
domains.
The
three
3­
D
reactivity
metrics
that
show
the
required
consistency
between
the
different
model
domains
and
episode
days
were
reported
as:
Least
Square
Relative
Reactivity
LS
 
RR,
Regional
Maximum
Incremental
Reactivity
to
Maximum
Ozone
Incremental
Reactivity
M2M
and
Maximum
Incremental
Reactivity
MIR­
3D.

Question
3.
What
are
the
two
or
three
most
promising
metrics
from
the
science
and
policy
perspective
for
use
in
developing
reactivity­
based
control
policies
?

Taking
together
the
Carter
et
al.
(
2003)
and
the
Hakami
et
al.
(
2003,
2004)
studies,
the
most
promising
metrics
for
the
development
of
reactivity
policies
are:
 
EKMA­
MIR,
Regional
MIR
or
MIR­
3D,
 
Regional
MIR
to
MOIR
or
M2M
and
 
Minimum
Substitution
Error
 
Method
1
or
Least
Square
Relative
Reactivity.

However,
there
is
an
important
requirement
to
develop
reactivity
metrics
that
generate
reactivity
scales
with
the
widest
effective
ranges
consistent
with
our
understanding
of
the
atmospheric
chemistry
of
VOCs.
On
this
basis,
the
most
promising
reactivity
metrics
are
EKMA­
MIR
and
Regional
MIR
or
MIR­
3D.

3.
Review
and
assessment
of
the
ozone
impacts
of
reactivity­
based
policies
for
VOC
controls
Based
on
the
detailed
studies
performed
by
Arunachalam
et
al.
(
2003),
Carter
et
al.
(
2003)
and
Hakami
et
al.
(
2003,
2004)
a
number
of
basic
questions
can
be
answered
about
the
likely
efficacy
of
reactivity­
based
strategies
for
VOC
emission
control
policies.

Arunachalam
et
al.
(
2003)
have
used
the
Multi­
scale
Air
Quality
Simulation
Platform
(
MAQSIP)
model
to
investigate
the
impact
of
reactivity­
based
VOC
emission
control
policies
on
the
attainment
of
ozone
air
quality
standards.
The
MAQSIP
model
was
run
for
the
August­
September
2000
period
of
the
Texas
AQS
campaign
and
for
a
period
during
June
1996
over
the
eastern
United
States.
For
the
Texas
study,
the
model
resolutions
were
5
km
x
5
km,
15
km
x
15
km
and
45
km
x
45
km
whereas
for
the
eastern
United
States
study,
they
were
set
at
12
km
x
12
km
and
36
km
x
36
km.
Two
different
9
chemical
mechanisms
were
employed,
namely
Carbon
Bond
CB4
and
RADM­
2
(
Stockwell
et
al.
1990)
and
7
different
ozone
exposure
metrics
were
calculated.
The
reactivity­
based
substitution
scenarios
evaluated
were:
a
15%
reduction
in
man­
made
VOC
emissions,
substitution
of
15%
of
all
man­
made
VOC
emissions
with
butanone
and
the
substitution
of
non­
mobile
xylene
emissions
with
2­
butoxyethanol.

The
key
finding
presented
by
Arunachalam
et
al.
(
2003)
was
that
reactivity­
based
VOC
substitution
strategies
resulted
in
reduced
ozone
levels,
particularly
in
source
regions
and
in
areas
downwind
of
major
urban
and
industrial
centres.
Noticeable
changes
in
ozone
were
seen
with
the
15%
substitution
by
butanone
in
a
region
dominated
by
natural
biogenic
emissions
and
not
usually
thought
to
be
VOC­
sensitive.
The
substitution
of
the
highly
reactive
xylene
with
2­
butoxyethanol
was
beneficial
in
reducing
exceedances
of
the
ozone
standards
but
small
increases
were
noted
in
some
ozone
metrics,
in
some
model
domains
and
on
some
days.
VOC
substitution
strategies
appear
to
work
better
on
peak
ozone
levels.
Finally,
gram­
based
VOC
substitutions
seem
to
produce
more
favourable
ozone
outcomes
compared
with
mole
or
mol­
C
based
strategies.

In
addition
to
the
studies
of
reactivity
metrics
discussed
above,
Carter
et
al.
(
2003)
also
performed
some
VOC
substitution
experiments
with
a
view
to
evaluating
the
impact
of
the
substitution
of
current
VOC
emissions
with
low
reactivity
or
borderline
exempt
VOCs.
Ethane
was
the
chosen
surrogate
for
these
low
reactivity
VOCs.
Varying
amounts
of
ethane
were
added
back
to
replace
the
entire
man­
made
VOC
emission
inventory.
This
type
of
substitution
led
to
almost
as
much
ozone
reduction
as
the
complete
curtailment
of
man­
made
VOC
emissions.
If
enough
of
a
low
but
positive
reactivity
VOC
such
as
ethane
was
added,
then
ozone
levels
would
eventually
equal
or
exceed
current
levels.
This
level
of
substitution
was
found
to
involve
replacing
all
man­
made
VOC
emissions
with
about
five
times
the
mass
of
ethane.

Question
4.
Is
it
possible
to
use
incremental
reactivities
to
develop
more
efficient
control
strategies
for
meeting
the
ozone
air
quality
standards
?

Taken
together,
the
studies
by
Carter
et
al.
(
2003),
Arunachalam
et
al.
(
2003)
and
Hakami
et
al.
(
2003,
2004)
have
made
a
significant
contribution
to
basic
understanding
of
the
likely
role
that
reactivity
considerations
could
play
in
ozone
control
policies.
These
studies
have
shown
that
reactivity­
based
policies
should
work
efficiently
on
both
the
urban
and
regional
scales
by
reducing
episodic
peak
ozone
levels
and
by
reducing
the
exceedances
of
the
ozone
air
quality
standards.
Regional
air
quality
3­
D
models
have
been
shown
to
be
ideal
platforms
with
which
to
visualize
the
likely
impacts
of
VOCsubstitution
strategies
on
the
attainment
of
the
ozone
standards.
10
Question
5.
What
view
should
be
taken
on
the
increases
in
ozone
found
in
some
highly
NOx­
sensitive
regions
with
large
substitutions
of
low
reactivity
VOC
species
?

In
principle,
if
the
ozone
increases
in
highly
NOx­
sensitive
regions,
following
large
substitutions
of
low
reactive
VOC
species,
do
not
take
ozone
levels
above
the
ozone
standards
then
they
are
probably
of
little
policy
significance.
However,
they
may
still
be
of
some
scientific
interest.
Some
form
of
process­
analysis
should
be
conducted
to
understand
the
mechanism
of
any
such
ozone
increases
found
in
3­
D
Eulerian
regional
air
quality
models.
An
important
issue
is
whether
such
ozone
increases
are
likely
to
be
a
real
feature
of
the
atmospheric
chemistry
of
the
VOC
substitution
or
an
artifact
of
the
cutdown
representation
of
the
oxidation
of
those
VOCs
with
NOx­
sinks,
together
with
free
radical
sources
and
sinks,
in
the
chemical
mechanisms
adopted
in
3­
D
models.

Question
6.
Are
reactivity­
based
controls
ineffective
or
counter­
productive
in
multiday
transport
or
stagnation
scenarios
?

The
studies
by
Carter
et
al.
(
2003),
Arunachalam
et
al.
(
2003)
and
Hakami
et
al.
(
2003,
2004)
certainly
addressed
multi­
day
ozone
episodes
but
this
is
a
long
way
from
addressing
multi­
day
transport
or
stagnation
scenarios.
This
question
therefore
remains
largely
unanswered
in
the
current
studies
evaluated
above.

Further
work
will
be
required
to
address
these
issues
in
the
North
American
context.
This
will
entail
developing
an
improved
set
of
EKMA
or
trajectory
scenarios
so
that
box
model
reactivity
scales
can
be
estimated
for
multi­
day
and
stagnation
conditions.
Although
box
models
cannot
represent
the
meteorological
aspects
of
multi­
day
effects,
they
can
represent
the
relevant
issues
of
atmospheric
chemistry.
Stockwell
et
al.
(
2001)
have
shown
that
the
EKMA­
type
models
can
be
readily
adapted
to
calculate
multi­
day
effects
and
that
the
resulting
reactivity
scales
generated
are
more
representative
of
nonurban
areas.
3­
D
Eulerian
regional
air
quality
models,
in
contrast,
can
fully
represent
the
meteorological
processes
driving
multi­
day
and
stagnation
effects.
However,
caution
must
be
exercised
in
the
representation
of
the
oxidation
of
those
VOCs
with
NOx­
sinks,
together
with
free
radical
sources
and
sinks,
in
the
cut­
down
chemical
mechanisms
employed
in
3­
D
Eulerian
models.

In
Europe,
long­
range
transboundary
transport
and
multi­
day
formation
are
paramount
and
reactivity
scales
already
fully
take
these
issues
into
account
(
Derwent
et
al.
1998).
Multi­
day
reactivity
refers
to
the
process
by
which
emitted
VOCs
contribute
to
ozone
formation
on
time
scales
beyond
the
day
during
which
they
were
emitted.
All
VOCs,
including
the
most
reactive
alkenes
and
carbonyls,
have
a
capacity
to
produce
ozone
during
the
day
following
their
emission
and
multi­
day
ozone
formation
is
driven
by
the
carry­
over
of
unreacted
parent
VOC,
of
organic
nitrogen
compounds
and
by
the
step­
wise
formation
and
destruction
of
carbonyl
compounds
(
Derwent
et
al.
2004).
The
cut­
down
chemical
mechanisms
used
in
3­
D
Eulerian
air
quality
models
may
not
fully
represent
all
11
these
processes
and
so
they
consequently
underestimate
the
extent
of
multi­
day
ozone
formation
with
the
reactive
alkenes
and
carbonyls.

4.
Discussion
of
the
major
issues
involved
in
estimating
reactivity
metrics
Question
7.
How
important
is
the
choice
of
chemical
mechanism
in
estimating
reactivity
metrics
?

The
choice
of
chemical
mechanism,
together
with
its
adequacy
and
completeness,
are
crucial
to
the
estimation
of
reactivity
metrics
and
the
compilation
of
reactivity
scales.
The
main
task
in
the
development
of
a
reactivity
scale
is
the
distillation
of
all
that
is
known
concerning
the
kinetics
and
mechanism
of
the
oxidation
of
individual
VOCs
into
the
one
single
number
for
each
VOC
in
the
reactivity
scale.

The
three
reactivity
studies
employed
three
distinctly
different
chemical
mechanisms:
CB4,
SAPRC
99
and
RADM­
2.
These
are
all
widely
used
mechanisms,
each
with
its
own
strengths
and
weaknesses.
It
is
a
relatively
straightforward
task
to
evaluate
them
against
each
other
under
particular
environmental
conditions.
Such
an
evaluation
would
not
significantly
help
with
the
evaluation
of
reactivity
metrics
because
the
above
three
mechanisms
have
been
designed
largely
to
represent
photochemical
ozone
formation
in
the
base
case
environment.
To
develop
a
reactivity
scale,
we
need
chemical
mechanisms
for
each
of
the
many
hundreds
of
VOCs
that
may
not
necessarily
be
explicitly
represented
in
the
base
case
but
are
an
essential
component
of
the
perturbed
cases
in
which
incremental
reactivities
are
being
estimated.

To
illustrate
this
point,
it
is
useful
to
examine
the
three
reactivity
studies
considered
above.
An
explicit
mechanism
for
ethane
photo­
oxidation
had
to
be
added
to
the
CB4
mechanism
to
represent
the
contribution
to
photochemical
ozone
formation
from
a
VOC
which
is
exempt
because
of
its
low
reactivity.
Explicit
mechanisms
for
2­
butoxyethanol
and
butanone
were
also
added
to
CB4.
RADM­
2
has
no
explicit
representation
of
butanone
(
Stockwell
et
al.
1990)
and
so
it
was
assumed
to
behave
in
the
same
manner
as
the
lumped
species,
KET,
which
represents
all
ketones
of
all
carbon
chain
lengths.
An
explicit
mechanism
for
2­
butoxyethanol
was
also
added
to
RADM­
2.
How
are
such
extensions
and
assumptions
to
be
adequately
and
robustly
evaluated
?

For
a
reactivity
scale
to
be
a
practical
and
policy­
relevant
tool,
it
will
need
to
address
upwards
of
100
to
200
VOCs.
Detailed
explicit
chemical
mechanisms
will
be
required
that
can
address
accurately,
the
fate
and
behaviour
of
a
large
number
of
VOCs.
Without
access
to
these
detailed
and
explicit
chemical
mechanisms,
it
will
not
be
possible
to
check
out
the
simplifications,
assumptions
and
extensions
made
when
additional
VOCs
are
added
to
the
simplified
chemical
mechanisms
that
are
employed
in
3­
D
Eulerian
regional
air
quality
models.
12
Additional
laboratory,
field
and
smog
chamber
experimental
data
may
be
required
to
construct
and
validate
the
detailed
and
explicit
chemical
mechanisms
for
individual
VOCs.
There
are
still
many
uncertainties
in
the
explicit
chemical
mechanisms
for
the
aromatics
and
other
important
classes
of
VOCs.

Question
8.
What
is
the
appropriate
modelling
framework,
3­
D
or
box
model,
for
the
development
of
reactivity
scales
?

Calculating
reactivity
scales
using
3­
D
Eulerian
air
quality
models
is
a
significant
undertaking,
as
the
three
studies
discussed
above
have
shown.
It
has
been
possible
to
calculate
reactivities
for
up
to
42
VOCs
in
the
URM
model
(
Hakami
et
al.
2003,
2004)
but
this
is
a
long
way
off
the
number
that
would
be
required
for
the
development
of
a
practical
and
policy­
relevant
reactivity
scale.
A
significant
effort
will
be
required
in
the
preparation
of
accurate
but
simplified
representations
of
the
fates
and
behaviour
of
each
VOC
so
that
computer
run
times
do
not
become
inordinately
long.
It
is
not
clear
whether
such
simplifications
will
necessarily
be
available
for
all
candidate
VOCs.

A
more
practical
alternative
is
the
development
of
an
improved
set
of
EKMA­
type
box
model
or
trajectory
model
scenarios
that
can
be
used
to
evaluate
reactivities.
Although
box
models
cannot
represent
specific
episodes
in
detail,
they
permit
the
use
of
fully
detailed
and
explicit
chemical
mechanisms.
They
can
represent
the
major
conditions
that
control
VOC
reactivity
(
Carter,
1994)
and
they
can
represent
the
atmospheric
chemistry
issues
relevant
to
multi­
day
and
stagnation
conditions
(
Derwent
et
al.
2004).

To
exploit
fully
the
potential
of
reactivity­
based
strategies
in
the
future,
both
box
and
3­
D
Eulerian
regional
air
quality
models
will
be
required.
Indeed,
one
of
the
important
conclusions
of
the
Carter
et
al.
(
2003)
study
addressed
the
concern
as
to
whether
reactivity
scales
generated
using
EKMA­
type
box
models
could
represent
reactivity
issues
over
the
regional
scale.
The
close
similarity
between
the
EKMA­
type
box
model
scales
and
the
corresponding
3­
D
regional
MIR
scale
on
the
4
km
x
4
km
model
domain,
shows
that,
indeed,
both
box
models
and
3­
D
Eulerian
models
are
capturing
the
same
important
features
of
regional
scale
photochemical
ozone
formation
(
Carter
et
al.
2003).

This
viewpoint
necessarily
represents,
at
first
sight,
a
departure
from
that
set
out
by
the
Reactivity
Research
Working
Group
(
RRWG,
1999)
which
considered
that
3­
D
Eulerian
regional
scale
air
quality
models
are
the
only
tools
that
can
realistically
capture
the
longrange
impacts
of
VOCs.

Hakami
et
al.
(
2003,
2004)
describe
how
3­
D
Eulerian
models
account
for
the
carry­
over
of
species
that
provide
a
source
of
free
radicals
into
the
next
day's
chemistry
and
how
this
feature
accounts
for
the
lower
reactivities
for
aldehydes
compared
with
box
models.
It
may
well
also
account
for
the
lower
reactivity
found
for
 ­
pinene
and
the
higher
reactivity
found
for
methanol.
This
is
an
important
observation
and
it
strongly
supports
the
viewpoint
that
only
3­
D
Eulerian
models
can
realistically
simulate
the
long­
range
13
impacts
of
VOCs.
However,
there
is
a
potential
danger
here
if
the
carry­
over
of
radical
source
species
were
not
a
real
feature
of
VOC
chemistry
but
an
artifact
of
the
simplifications
and
non­
stoichiometric
factors
introduced
in
the
cut­
down
representations
of
VOC
chemistry
in
the
3­
D
models.
If
the
carry­
over
of
radical
sources
were
an
artifact
of
the
chemical
mechanism
then
its
importance
would
have
been
considerably
overstated
by
the
3­
D
Eulerian
models
and
this
would
have
led
to
the
estimation
of
biased
reactivities.

Whatever
results
are
generated
by
3­
D
Eulerian
models,
they
are
only
as
reliable
as
the
simplified
chemical
mechanisms
employed
in
them.
Without
access
to
box
models,
there
will
be
no
route
by
which
the
situation
envisaged
by
the
Reactivity
Research
Working
Group
can
be
reached
and
3­
D
Eulerian
regional
scale
air
quality
models
become
the
only
reliable
tools
that
can
realistically
simulate
the
long
range
impacts
of
VOCs.

5.
Next
steps
in
the
development
of
reactivity­
based
policies
for
VOC
controls
If
the
potential
value
of
reactivity­
based
VOC
control
strategies
become
accepted
by
policymakers
for
the
attainment
of
ozone
air
quality
standards,
then
a
number
of
steps
will
need
to
be
taken
to
underpin
their
development.
These
steps
are
laid
out
in
the
following
paragraphs.

Two
sets
of
basic
tools
will
be
required,
as
follows:

 
An
evaluated
reactivity
scale
containing
evaluated
reactivity
values
for
upwards
of
100
to
200
VOCs,
and
 
3­
D
Eulerian
regional
air
quality
models
to
visualise
the
impacts
of
potential
reactivity­
based
strategies
on
the
attainment
of
ozone
air
quality
standards.

The
basic
principles
being
that
the
evaluated
reactivity
scale
expresses
our
best
understanding
of
the
relative
importance
of
each
VOC
in
photochemical
ozone
formation
under
given
environmental
conditions
and
that
the
3­
D
Eulerian
regional
air
quality
models
provide
the
best
possible
visualization
of
the
urban
and
regional
scale
impacts
of
potential
reactivity­
based
strategies.

Step
1:
Development
of
an
interim
reactivity
scale
There
are
currently
two
independent
methodologies
for
estimating
reactivity
values
for
upwards
of
a
100
VOCs
under
North
American
conditions.
The
first
methodology
has
led
to
the
production
of
the
MIR
scale
and
is
based
on
the
SAPRC
family
of
chemical
mechanisms
(
Carter,
1994).
The
second
has
led
to
the
POCP
scale
and
is
based
on
the
Master
Chemical
Mechanism
(
Derwent
et
al.,
2001).
Both
adopt
EKMA­
like
box
models
and
both
employ
the
same
input
data
based
on
Baugues
(
1990).
14
The
general
level
of
agreement
between
the
MIR
and
POCP
values
is
excellent
for
a
wide
range
of
VOCs
and
VOC
classes.
The
remaining
differences
need
to
be
resolved
where
they
exist.
The
two
reactivity
scales
should
then
be
reviewed
and
a
combined
interim
reactivity
scale
should
be
constructed.

Step
2:
Development
of
candidate
reactivity­
based
strategies
Using
the
interim
reactivity
scale,
all
stake­
holders
should
be
encouraged
to
develop
candidate
reactivity­
based
strategies.
In
this
way
policy­
makers
can
obtain
some
understanding
of
the
types
of
substitutions
and
their
geographic
scales,
that
are
being
contemplated.

Step
3:
Development
of
the
SMOKE
emissions
processing
system
It
is
likely
that
some
changes
will
be
required
to
the
SMOKE
emissions
processing
system
in
CMAQ
so
that
the
candidate
reactivity­
based
strategies
can
be
visualized
in
CMAQ
and
other
3­
D
Eulerian
regional
air
quality
models.

Step
4:
Development
of
EKMA­
type
scenarios
The
input
data
employed
in
the
construction
of
the
MIR
and
POCP
scales
were
taken
from
Baugues
(
1990)
and
represent
39
urban
areas
at
some
point
in
time
in
the
past
when
peak
ozone
concentrations
were
far
greater
than
currently
observed
(
Carter
et
al.,
2003).
There
is
a
case
for
the
development
of
an
updated
set
of
EKMA­
type
scenarios
for
intense
urban
ozone
episodes
and
for
those
conditions
responsible
for
multi­
day
transport
and
stagnation.
Once
these
scenarios
are
available,
it
would
be
a
straightforward
task
to
generate
MIR
and
POCP
scales
and
to
produce
an
updated
and
evaluated
reactivity
scale.
The
proposal
would
be
to
generate
up
to
100
of
these
updated
scenarios
from
3­
D
Eulerian
regional
air
quality
model
simulations
for
a
range
of
different
air
basins
and
ozone
episodes,
using
present
and
future
emission
levels.

Step
5:
Development
of
an
evaluated
reactivity
scale
Whilst
reactivity
scales
and
the
reactivity
values
that
populate
them
are
not
basic
geophysical
constants,
they
are
derived
from
them.
In
this
context,
basic
geophysical
constants
include
reaction
rate
coefficients,
quantum
yields
and
absorption
cross­
sections.
There
is
a
case
for
extending
the
work
of
the
JPL
and
IUPAC
chemical
kinetic
data
15
evaluation
panels
to
the
review
of
the
chemical
kinetic
data
for
the
reactions
of
OH,
NO3
and
O3
with
VOCs
and
of
the
reactions
of
the
alkyl,
oxy,
peroxy
and
peroxyacyl
radicals
produced
by
them.
Such
a
panel
could
also
review
quantum
yield
and
absorption
crosssection
data
for
the
VOCs
and
their
degradation
products.

Evaluated
chemical
kinetic
data
could
then
be
fed
into
the
SAPRC
mechanisms
and
Master
Chemical
Mechanism
as
they
become
available.
These
mechanisms
would
then
be
used
with
the
updated
EKMA­
scenarios
to
develop
an
evaluated
reactivity
scale
which
would
replace
the
interim
scale
for
the
development
of
candidate
reactivity­
based
strategies.

Step
6:
Assemble
a
3­
D
Eulerian
regional
air
quality
model
with
an
explicit
mechanism
There
is
a
case
for
taking
a
3­
D
regional
air
quality
model
and
adding
an
explicit
chemical
mechanism
for
each
candidate
VOC
species
and
performing
a
`
gold
standard'
visualisation
of
one
reactivity­
based
strategy
in
one
air
basin.
This
model
run
would
then
be
used
to
bench­
mark
cut­
down
chemical
mechanisms
for
that
VOC
species.

Step
7:
Establish
cut­
down
representations
of
VOC
oxidation
for
3­
D
Eulerian
regional
air
quality
models
The
SAPRC
and
Master
Chemical
Mechanisms
provide
a
framework
within
which
cutdown
representations
of
VOC
oxidation
could
be
developed
for
3­
D
Eulerian
regional
air
quality
models.
Protocol
would
need
to
be
laid
out,
setting
out
the
basic
principles
to
be
adhered
to
in
mechanism
simplification
and
the
accuracy
to
be
achieved
in
the
representation
of
the
production
of
ozone,
temporary
NOx
reservoirs,
free
radical
sources
and
sinks
and
major
oxidation
products.
Cut­
down
representations
could
then
be
tested
against
the
`
gold
standard'
visualisation
produced
in
step
6
above,
on
a
VOC­
by­
VOC
basis.

Step
8:
Visualisation
of
the
impacts
of
reactivity­
based
VOC
controls
on
the
attainment
of
ozone
standards
Using
the
evaluated
reactivity
scale,
potential
reactivity­
based
VOC
control
strategies
could
be
generated
for
ozone
non­
attainment
areas.
Cut­
down
representations
of
the
oxidation
mechanisms
of
the
VOCs
involved
would
then
be
added
to
the
3­
D
Eulerian
regional
air
quality
models,
together
with
the
required
speciated
VOC
emission
inventories
generated
using
SMOKE.
The
3­
D
Eulerian
regional
air
quality
models
would
16
then
be
able
to
visualize
the
impacts
of
the
reactivity­
based
strategies
on
ozone
nonattainment
The
model
results
would
also
show
up
any
areas
where
ozone
concentrations
had
increased
and
any
impacts
of
meteorological
and
transport
processes
that
cannot
be
allowed
for
in
simple
box
models.

In
this
context,
it
is
useful
if
the
evaluated
reactivity
scale
has
the
largest
effective
range
that
is
consistent
with
our
understanding
of
the
atmospheric
chemistry
of
VOCs.
In
this
way,
there
would
be
maximum
incentive
given
to
the
consideration
of
reactivity­
based
policies.
The
reactivity
scale
would
then
give
an
upper
limit
to
the
magnitude
of
the
substitutions
of
low
reactivity
VOCs
for
highly
reactive
VOCs.
Visualisation
with
3­
D
Eulerian
models
would
then
provide
a
vital
check
that
there
were
no
unacceptable
ozone
increases,
exactly
as
envisaged
by
Carter
et
al.
(
2003),
Hakami
et
al.
(
2003,
2004)
and
Arunachalam
et
al.
(
2003).

6.
References
Arunachalam,
S.,
Mathur,
R.,
Holland,
A.,
Lee,
M.
R.,
Olerud,
D.,
Jeffries,
H.
(
2003).
Final
report:
Investigation
of
VOC
reactivity.
Assessment
with
comprehensive
air
quality
modelling.
Carolina
Environmental
Program,
University
of
North
Carolina
at
Chapel
Hill,
Chapel
Hill,
North
Carolina.

Baugues,
K.
(
1990).
Preliminary
planning
information
for
updating
the
ozone
regulatory
impact
analysis
version
of
EKMA,
U.
S.
Environmental
Protection
Agency,
Research
Triangle
Park,
North
Carolina.

CARB.
(
1993).
Proposed
regulations
for
low­
emission
vehicles
and
clean
fuels.
California
Air
Resources
Board,
Sacramento,
California.

Carter,
W.
P.
L.
(
1994).
Development
of
ozone
reactivity
scales
for
volatile
organic
compounds.
Journal
of
Air
and
Waste
Management
Association
44,
881­
889.

Carter,
W.
P.
L.,
Tonnesen,
G.,
Yarwood,
G.
(
2003).
Investigation
of
VOC
reactivity
effects
using
existing
regional
air
quality
models.
Center
for
Environmental
Research
and
Technology,
University
of
California,
Riverside,
California.

Derwent,
R.
G.,
Jenkin,
M.
E.,
Saunders,
S.
M.,
Pilling,
M.
J.
(
1998).
Photochemical
ozone
creation
potentials
for
organic
compounds
in
north
west
Europe
calculated
with
a
Master
Chemical
Mechanism.
Atmospheric
Environment
32,
2429­
2441,

Derwent,
R.
G.,
Jenkin,
M.
E.,
Saunders,
S.
M.,
Pilling,
M.
J.
(
2001).
Characterization
of
the
reactivites
of
volatile
organic
compounds
using
a
Master
Chemical
Mechanism.
Journal
of
Air
and
Waste
Management
Association
51,
699­
707.
17
Derwent,
R.
G.,
Jenkin,
M.
E.,
Saunders,
S.
M.,
Pilling,
M.
J.,
Passant,
N.
R.
(
2004).
Multiday
ozone
formation
for
alkenes
and
carbonyls
investigated
with
a
Master
Chemical
Mechanism
under
European
conditions.
Atmospheric
Environment,
in
press.

Dimitriades,
B.
(
1999).
Scientific
basis
of
an
improved
EPA
policy
on
control
of
organic
emissions
for
ambient
ozone
reduction.
Journal
of
Air
and
Waste
Management
Association
49,
831­
838.

Hakami,
A.,
Bergin,
M.
S.,
Russell,
A.
S.
(
2003).
Assessment
of
the
ozone
and
aerosol
formation
potentials
(
reactivities)
of
organic
compounds
over
the
eastern
United
States.
Atlanta,
Georgia.

Hakami,
A.,
Arhami,
M.,
Russell,
A.
G.
(
2004).
Further
analysis
of
VOC
reactivity
metrics
and
scales.
Atlanta,
Georgia.

RRWG
(
1999).
VOC
reactivity
science
assessment:
Executive
Summary.
Reactivity
Research
Working
Group.
http://
www.
cgenv.
com/
narsto/
reactinfo.
html.

Stockwell,
W.
R.,
Middleton,
P.,
Chang,
J.
S.,
Tang,
X.
(
1990).
The
second
generation
regional
acid
deposition
model
chemical
mechanism
for
regional
air
quality
modelling.
J.
Geophys.
Res.
95,
16,343­
16,367.

Stockwell,
W.
R.,
Geiger,
H.,
Becker,
K.­
H.
(
2001).
Estimation
of
incremental
reactivities
for
multiple
day
scenarios:
An
application
to
ethane
and
dimethoxymethane.
Atmospheric
Environment
35,
929­
939.
