APPENDIX
B
DEVELOPMENT
OF
THE
METEOROLOGICAL
MODEL
SAN
ANTONIO
EAC
REGION
ATTAINMENT
DEMONSTRATION
MARCH
2004
Appendix
B
Table
of
Contents
Page
Background                               
B­
1
Meteorological
Run
5d                        .  
B­
3
Diurnal
Temperatures                      . 
B­
4
Wind
Speeds                         . ..
B­
4
Soil
Moisture                           
B­
4
Vertical
Layers                          .
B­
4
Weather
Patterns                         
B­
5
Evaluation
of
the
Best
Performing
Simulation             
B­
5
Meteorological
Run
6f                          ..
B­
6
Diurnal
Temperatures                       .
B­
8
Wind
Speed                           ..
B­
8
Weather
Patterns                         .
B­
9
Evaluation
of
the
Best
Performing
Simulation       .     ...
B­
9
Final
MM5
Configuration                         
B­
11
Meteorological
Run
5g                      ...   
B­
15
Statistical
Evaluation
of
MM5
Run
5g                .
B­
20
Processing
of
MM5
Meteorological
Fields
for
CAMx         ...
B­
21
Appendix
B
List
of
Tables
Page
Table
B­
1
Summary
of
Meteorological
Sensitivity
Tests
(
eight
runs)     
B­
3
Table
B­
2
Summary
of
Revised
MM5
Applications            ..
B­
8
Table
B­
3
Comparison
of
Mean
Daily
Statistics
Against
Statistical
Benchmark
for
the
4­
km
Grid                ...
B­
21
Table
B­
4
Meteorological
Data
Requirements
for
CAMx         ...
B­
22
Table
B­
5
Vertical
Layer
Structure
for
MM5
and
CAMx
for
Sept.
13­
20,
1999
Episode                  
B­
23
Appendix
B
List
of
Figures
Page
Figure
B­
1
Comparison
of
Ozone
Levels
Measured
during
Baylor
University
Airborne
Sampling
Project
on
September
17th
And
18th
with
Ozone
Levels
Predicted
by
Original
1999
Model
Simulation
Met
3b
for
Time
of
Day
and
Altitude
of
Flights.
(
Dotted
lines
represent
collected
Ozone
data
in
parts
per
billion)                ..
B­
2
Figure
B­
2
Observed
(
black)
and
Predicted
(
red)
Values
for
Run
5d
For
San
Antonio/
Austin
Region                .
B­
6
Figure
B­
3
Comparison
of
the
Mixing
Height
Between
the
Original
Model
(
Run
4c),
5d,
and
6f                  .
B­
3
Figure
B­
4
Comparison
of
Wind
Speed
and
Direction,
Temperature,
Humidity
Statistics
for
Original
Meteorological
Model
Run
(
black),
5d
(
red),
and
6f
(
blue)
for
San
Antonio
 
Austin     ..
B­
11
Figure
B­
5
Wind
Statistics
for
Original
Meteorological
Model
Run
(
black),
5d
(
red),
and
6f
(
blue)
for
San
Antonio­
Austin
Region                          .
B­
12
Figure
B­
6
Temperature
Statistics
for
Meteorological
Model
Run
(
black),
5d
(
red),
and
6f
(
blue)
for
San
Antonio­
Austin
Region                          .
B­
13
Figure
B­
7
Humidity
Statistics
for
Original
Meteorological
Model
Run
(
black),
5d
(
red),
and
6f
(
blue)
for
San
Antonio­
Austin
Region                         . 
B­
14
Figure
B­
8
Hourly
Wind
Speeds
for
Runs
5d
and
5g
in
the
4­
km
San
Antonio
Area
Domain                    
B­
16
Figure
B­
9
Hourly
Temperature
for
Runs
5d
and
5g
in
the
4­
km
San
Antonio
Area
Domain                    
B­
16
Figure
B­
10
Hourly
Wind
Direction
for
Runs
5d
and
5g
in
the
4­
km
San
Antonio
Area
Domain                    
B­
17
Figure
B­
11
Hourly
Humidity
for
Runs
5d
and
5g
in
the
4­
km
San
Antonio
Area
Domain                    
B­
17
Figure
B­
12
Comparison
of
Ozone
Levels
Measured
During
Baylor
University
Airborne
Sampling
Project
on
September
17th
With
Ozone
Levels
Predicted
by
5d
and
5g
Model
Simulation
For
Time
of
Day
and
Altitude
of
Flights.
(
Dotted
line
Represents
collected
ozone
data
in
parts
per
billion
(
ppb))    
B­
19
B­
1
BACKGROUND
The
meteorological
inputs
used
to
create
the
original
1999
episode
were
developed
by
ENVIRON
using
the
Fifth
Generation
Mesoscale
Model,
referred
to
as
MM5.
In
their
report
on
development
of
the
1999
episode
simulation,
ENVIRON
acknowledged
certain
placement,
and
precipitation
rather
well
for
the
September
1999
episode
for
the
entire
4­
km
domain
(
South
Texas).
However,
the
model
only
performed
marginally
when
predicting
humidity
and
pressure.
Two
persistent
problems
with
the
original
MM5
model
simulation,
referred
to
as
Met
3b,
included
wind
speed
 
over
predicting
of
wind
speed
at
night
and
under
predicting
during
the
daytime
 
and
over
predicting
of
early
morning
temperatures.

The
most
significant
problem
in
the
San
Antonio
area
was
aloft
wind
direction1.
On
September
17
and
18,
1999,
air
quality
sampling
was
conducted
in
the
San
Antonio
 
Austin
area
as
part
of
the
Baylor
University
Airborne
Sampling
Project
conducted
by
TCEQ.
The
data
collected
from
the
Baylor
aircraft
flights
were
compared
to
predictions
from
the
1999
model
simulation
for
the
same
days.
While
the
model
performed
well
in
replicating
peak
ozone
aloft
in
the
urban
plumes
for
the
17th
and
18th,
the
spatial
distribution
was
poor,
indicating
a
problem
with
wind
direction
at
those
altitudes.
Figure
B­
1
provides
a
comparison
between
the
aircraft
data
and
ozone
levels
predicted
by
the
model
for
the
correct
time
period
and
altitude
of
the
flights.
On
September
17th
the
aircraft
took
measurements
between
600
 
800
meters
beginning
at
1400
CDT.
The
flight
on
September
18th
began
at
1700
CDT
and
data
were
collected
at
about
700
meters
during
most
of
the
flight.
(
Emery,
et.
al.,
2002)
As
shown
in
the
figure,
the
peak
ozone
plumes
predicted
by
the
model
are
south
of
observed
plumes
for
both
days
that
data
were
collected.

In
addition
to
wind
direction
issues,
simulated
ozone
levels
between
the
plumes
were
under
predicted
by
10
 
20
ppb
when
compared
to
observed
data.
This
problem
suggests
that
the
model
was
generating
insufficient
regional
background
ozone
levels.
According
to
ENVIRON,
insufficient
background
ozone
indicates
problems
with
the
regional
emissions
inputs
to
the
model,
such
as
too
few
VOC
emissions
from
biogenic
sources.
(
Emery,
et.
al.,
2002)
As
part
of
the
effort
to
improve
the
accuracy
of
the
model,
the
South
Texas
Near
Non
Attainment
areas
contracted
with
the
ENVIRON
and
with
the
University
of
Texas'
Center
for
Energy
and
Environmental
Resources
(
UT­
CEER)
to
improve
accuracy
of
meteorological
input
data
for
the
1999
episode
model.

1
Typically,
surface
winds
transport
pollutants
locally,
while
upper
(
aloft)
winds
have
the
potential
to
transport
ozone
and
its
precursors
much
greater
distances,
often
hundreds
of
kilometers
(
APTI).
B­
2
Figure
B­
1.
Comparison
of
Ozone
Levels
Measured
during
Baylor
University
Airborne
Sampling
Project
on
September
17th
and
18th
with
Ozone
Levels
Predicted
by
Original
1999
Model
Simulation
Met
3b
for
Time
of
Day
and
Altitude
of
Flights.
(
Dotted
lines
represent
collected
ozone
data
in
parts
per
billion.)

September
17,
1999
2:
00
P.
M.
September
18,
1999
6:
00
P.
M.
B­
3
ENVIRON,
in
conjunction
with
UT­
CEER,
tested
alternative
MM5
configurations
and
parameter
algorithms
to
find
the
combination
that
best
replicate
actual
meteorological
conditions
during
the
1999
high­
ozone
episode.
The
best
of
these
runs
were
presented
to
the
air
quality
planners
at
TCEQ,
and
the
four
NNA
partners
for
further
analysis.
The
runs,
labeled
5d,
5g,
and
6f,
each
have
unique
strengths
and
weaknesses
and
impact
the
photochemical
model
in
various
ways,
as
described
below.

METEOROLOGICAL
RUN
5d
The
ENVIRON/
UT
team
conducted
eight
meteorological
runs,
labeled
5a,
b,
c,
d,
e,
f,
h,
and
I,
using
version
3.4
of
the
MM5
model
while
incorporating
new
databases
and
model
configurations
that
proved
to
be
successful
in
other
applications
throughout
the
country.

 
Change
to
an
alternative
boundary
layer
scheme
(
Blackadar
or
MRF)
to
investigate
sensitivity
to
boundary
layer
mixing;
 
Change
to
an
alternative
radiation
scheme
(
RRTM)
that
is
known
to
perform
better
in
the
humid
Texas
climate
and
may
reduce
the
morning
over­
predicted
surface
temperatures;
 
Utilize
interactive
multi­
layer
soil
moisture
schemes
now
available
with
the
latest
release
of
MM5
(
v3.5)
that
would
provide
a
more
realistic
feedback
between
soil
and
atmosphere;
and
 
Test
the
effects
of
alternative
observational
analyses
and
FDDA
techniques
that
may
better
characterize
conditions
in
the
south
central
U.
S.

Table
B­
1.
Summary
of
Meteorological
Sensitivity
Tests
(
eight
runs)
Run
ID
Configuration
Run
5c
Identical
to
Run
4c
(
the
best
performing
of
the
original
runs
reported
by
Emery
and
Tai,
2002),
except
that
the
Blackadar
PBL
scheme
was
replaced
by
the
Gayno­
Seaman
PBL
scheme.

Run
5
Identical
to
Run
5c
except
that
the
Dudhia
Cloud
radiation
scheme
was
replaced
by
the
RRTM
radiation
scheme
Run
5b
Identical
to
Run
5
except
that
data
from
the
Texas
Coastal
Ocean
Observation
Network
(
TCOON)
and
NOAA
National
Buoy
Center
were
added
to
the
original
observational
FDDA
input
data
set.

Run
5d
Identical
to
Run
5b,
except
that
the
MRF
PBL
scheme
replaced
the
Blackadar
PBL
scheme.

Run
5e
Identical
to
Run
5d,
except
that
the
standard
5­
layer
soil
model
was
augmented
by
the
bucket
soil
moisture
option,
and
Run
5e
used
the
standard
climatological
default
soil
moisture
to
define
the
initial
soil
conditions
by
land
use
category
(
up
to
this
point,
soil
moisture
was
reduced
25%
from
standard
values
as
in
the
original
Run
4c).

Run
5f
Identical
to
Run
5e,
except
that
the
reduced
soil
moisture
was
used
similarly
to
Runs
4c
and
5­
5d.

Run
5i
Identical
to
Run
5d
except
that
the
number
of
vertical
layers
was
increased
from
28
to
41,
resulting
in
about
twice
the
vertical
resolution
between
approximately
250
and
4600
meters
above
the
surface.

Run
5h
Identical
to
Run
5e
(
bucket
soil
moisture
with
standard
default
initial
soil
moisture
values)
except
that
the
number
of
vertical
layers
was
increased
from
28
to
41.
B­
4
When
conducting
the
eight
runs,
which
are
listed
on
table
B­
1,
modelers
tested
two
methodologies
for
determining
the
depth
of
the
planetary
boundary
layer
(
the
Blackadar
and
Medium
Range
Forecast
schemes),
two
radiation
schemes2
(
the
Dudhia­
Cloud
and
Rapid
Radiation
Transfer
Model),
three
versions
of
the
four
dimensional
data
assimilation
(
FDDA)
model3
(
versions
11,
12,
and
13),
three
soil
moisture
schemes,
and
two
vertical­
layer4
resolutions
(
28
layers
versus
41
layers).

Diurnal
Temperatures
The
Dudhia
Cloud
scheme
overestimated
the
amount
of
radiation
absorbed
and
reradiated
by
the
atmosphere.
This
resulted
in
very
warm
nighttime
minimum
temperatures.
When
the
RRTM
radiation
scheme
was
used
rather
than
the
Dudhia
Cloud
scheme,
the
simulated
temperature
range
compared
very
closely
with
the
observed
diurnal
temperature
ranges.
The
simulated
maximum
temperatures
were
too
cool
compared
to
the
observed
maximum
temperatures.

Wind
Speeds
Wind
speeds
were
analyzed
with
the
Blackadar
PBL
scheme
and
the
MRF
PBL
scheme.
The
Blackadar
PBL
produced
high
daytime
and
nighttime
winds,
which
suggested
that
the
Blackadar
approach
produced
an
overly
aggressive
vertical
transfer
of
momentum.
It
was
also
noted
that
higher
winds
near
the
top
of
the
boundary
layer
may
mix
to
the
surface
too
rapidly.
These
occurrences
can
be
related
to
the
under
prediction
of
maximum
temperatures.
The
MRF
PBL
scheme
did
result
in
an
improved
prediction
of
both
daytime
and
nighttime
wind
speeds.
Nighttime
winds
were
noted
to
be
high
however.
With
the
MRF
PBL
scheme,
the
daytime
maximum
temperatures
warmed
by
1
to
2
K
but
remained
below
the
observed
temperatures.

Soil
Moisture
The
use
of
the
bucket
soil
moisture
option
in
the
model,
rather
than
the
five­
layer
soil
model,
produced
a
consistent
increase
in
wind
speed.
Maximum
temperatures
also
improved
slightly
during
the
final
days
of
the
episode.

Vertical
Layers
In
the
modified
run,
the
model
was
configured
to
run
with
41
layers
rather
than
28
in
order
to
investigate
the
effects
of
increased
vertical
resolution.
Higher
resolution
(
usually
more
than
30
layers)
in
the
vertical
direction
is
recommended
and
widely
adopted.
The
run
did
not
indicate
improvement
in
model
performance
at
the
boundary
layer
or
at
the
surface.

2
Shortwave
radiation
from
the
sun
and
longwave
radiation
from
the
earth
impact
atmospheric
processes
by
producing
heat,
moisture,
and
momentum
exchanges,
which
drive
the
PBL.
3
Four
dimensional
data
assimilation
(
FDDA)
refers
to
a
sophisticated
method
of
initializing
a
predictive
model
such
as
MM5.
MM5
and
similar
models
estimate
meteorological
processes
using
numerical
equations.
To
make
numerical
forecasts,
the
models
must
have
a
starting
point
in
which
initial
conditions
are
provided
in
the
form
of
gridded
data.
FDDA
combines
numerical
predictions
with
observations
to
provide
a
4­
dimensional
estimate
of
initial
meteorological
parameters.
The
FDDA
technique
utilized
to
develop
the
refined
1999
episode
was
the
"
nudging"
technique
in
which
the
model
is
gently
pushed
toward
observed
values
using
numeric
equations.
4
The
modeling
team
employed
the
"
Sigma
Coordinate
System"
in
which
the
lowest
vertical
coordinate
follows
a
smoothed
version
of
the
actual
terrain.
The
higher
sigma
surfaces
parallel
the
lowest
coordinate
but
gradually
transition
to
being
nearly
horizontal
at
the
top
of
the
coordinate
system,
typically
above
the
tropopause.
When
28
sigma
levels
are
used,
the
8
lowest
layers
make
up
the
planetary
boundary
layer.
By
increasing
the
vertical
layers
to
41,
additional
vertical
layers
are
focused
in
the
boundary
layer
and
jet
stream,
which
provides
a
higher
resolution
than
the
28­
layer
system.
B­
5
Weather
Patterns
Surface
pressure
patterns
predicted
by
both
the
Blackadar
and
MRF
runs
compared
well
with
observed
pressure
patterns.
Cloud
type
and
coverage
across
the
domain
were
similar,
however
the
Blackadar
runs
produced
greater
amounts
of
low­
level
cloudiness
over
South
Texas
and
Gulf
of
Mexico
on
some
of
the
modeling
days.
Rainfall
amounts
was
not
predicted
after
the
14th
of
September,
which
was
in
concurrence
with
the
observed
rainfall
levels.
A
deeper
mixed
layer
was
produced
by
the
MRF
PBL
scheme
than
the
Blackadar
runs
with
heights
greater
by
25%
­
35%.

Evaluation
of
the
Best
Performing
Simulation
Overall,
run
5d
produced
the
best
results
of
the
eight
sensitivity
runs.
Wind
speed
and
wind
direction
improved
when
comparing
run
5d
results
with
the
meteorological
inputs
used
in
the
original
1999
simulation.
However,
5d
maintained
a
northerly
wind
bias
throughout
much
of
the
episode.
Run
5d
predicted
cooler
daily
temperatures
than
observed;
but
daily
minimum
temperatures
were
much
closer
to
observed
values
than
those
predicted
by
other
runs.
Humidity
resulted
as
erroneous
predictions
and
may
be
a
cause
for
concern
not
on
model
performance
but
rather
errors
in
the
simulation
of
spatial
and
temporal
evolution
of
the
boundary
layer
by
the
PBL
scheme,
particularly
along
the
coast
line.
Figure
B­
2
provides
a
comparison
between
observed
and
predicted
(
run
5d)
wind
speeds,
wind
direction,
temperature,
and
humidity
in
the
San
Antonio
 
Austin
region
during
the
September
1999
episode.
As
shown,
there
was
a
high
degree
of
correlation
between
observed
and
predicted
values
for
these
four
meteorological
variables.
B­
6
Figure
B­
2.
Observed
(
black)
and
Predicted
(
red)
Values
for
Run
5d
for
San
Antonio/
Austin
Region
METEOROLOGICAL
RUN
6f
The
ENVIRON/
UT­
CEER
team
also
developed
a
series
of
meteorological
runs
which
took
advantage
of
new
/
additional
input
data
from
EPA
as
well
as
the
expanded
capabilities
of
MM5
version
3.5.
Improvements
within
version
3.5
included
improvements
within
known
deficiencies
and
to
access
additional
modeling
capabilities.
These
modifications
included:

1.
The
same
four­
domain
nested
mesh
with
108/
36/
12/
4­
km
resolution,
but
with
an
expanded
36
km
grid
in
order
to
move
possible
108/
36
boundary
artifacts
away
from
the
area
of
interest
and
to
better
simulate
the
dominant
regional­
scale
Observed/
Predicted
Windspeed
0
2
4
6
9/
13
9/
14
9/
15
9/
16
9/
17
9/
18
9/
19
9/
20
m/
s
ObsWndSpd
PrdWndSpd
Observed/
Predicted
Wind
Direction
0
60
120
180
240
300
360
9/
13
9/
14
9/
15
9/
16
9/
17
9/
18
9/
19
9/
20
deg
ObsWndDir
PrdWndDir
Observed/
Predicted
Temperature
290
295
300
305
310
9/
13
9/
14
9/
15
9/
16
9/
17
9/
18
9/
19
9/
20
K
ObsTemp
PrdTemp
Predicted/
Observed
Humidity
0
5
10
15
20
25
9/
13
9/
14
9/
15
9/
16
9/
17
9/
18
9/
19
9/
20
g/
kg
ObsHum
PrdHum
B­
7
meteorology
over
the
entire
central
U.
S.
that
dictated
flow
and
pressure
patterns
in
Texas
during
the
episode.

2.
The
coupled
Pleim­
Xiu
Land
Surface
Model
and
boundary
layer
model,
which
required
additional
datasets
such
as
soil
type,
vegetation
categories,
deep
soil
temperature,
and
vegetation
fraction
archived
at
the
National
Center
for
Atmospheric
Research.

3.
Three­
hourly
observational
"
analysis"
fields
from
the
Eta
Data
Assimilation
System,
(
EDAS)
as
opposed
to
EDAS
"
initialization"
data
used
in
previous
modeling
to
establish
initial/
boundary
conditions
and
inputs
to
the
MM5
Four
Dimensional
Data
Assimilation
(
FDDA)
package.

4.
Incorporation
of
routine
surface
and
upper­
air
observation
data
obtained
from
NCAR
archives
into
the
EDAS
fields
processed
for
each
MM5
modeling
grid.
This
modification
was
made
to
ensure
that
the
mesoscale
and
local
meteorological
features
in
the
south­
central
U.
S.
were
faithfully
characterized
in
the
EDAS
analysis
dataset.
This
preprocessing
step
was
skipped
in
the
original
application
because
it
was
believed
that
the
relatively
high
spatial
and
temporal
resolution
of
the
EDAS
fields
was
sufficient
to
capture
these
details.

5.
Use
of
the
RRTM
radiation
scheme
for
all
grids,
based
on
the
favorable
results
from
the
sensitivity
tests.

6.
Use
of
two­
way
interactive
nesting
for
all
grids.
The
4­
km
grid
was
run
as
an
independent
one­
way
nest
in
the
original
application.

7.
Modifications
to
the
FDDA
nudging
technique
to
include
two­
dimensional
surface
analysis
nudging,
altered
nudging
strengths,
and
recommendations
of
Dr.
Nelson
Seaman
at
the
Pennsylvania
State
University.
The
TCOON
and
NOAA
buoy
data
were
also
added
to
the
observation
FDDA
nudging
inputs.

These
capabilities
were
not
available
in
version
3.4
of
MM5
used
to
develop
the
original
1999
simulation,
nor
were
they
addressed
in
runs
5a
 
5i.
The
new
runs,
labeled
6c
 
6f,
were
each
modified
to
reflect
different
parameters
applied
to
each
run.

Runs
5a
 
5i
contained
a
36
 
km
grid
system
that
was
arranged
in
55
X
55
grid
cells.
The
"
run
6"
series
used
an
expanded
36­
km
grid
system
that
covered
85
X
61
grid
cells.
All
nested
grids
within
the
108­,
36­,
12­,
and
4­
km
grid
system,
were
run
in
two­
way
interactive
mode.
5
In
contrast,
the
run
5
series
incorporated
one­
way
interaction6
on
the
4­
km
grid,
with
2­
way
interaction
on
other
nested
grids.
All
"
6
series"
runs
were
conducted
with
28
vertical
layers,
since
the
run
5
series
results
indicated
this
resolution
performed
best.

5
Finer
resolution
grids
are
nested
inside
of
coarser­
resolution
grids.
The
information
for
the
outermost
grid
is
supplied
from
an
outside
source
using
one­
way
interaction.
The
coarse
grids
provide
boundary
conditions
on
the
mesh
interfaces
between
coarse/
fine
grids.
Forecast
variables
developed
in
the
fine
grids
are
used
to
update
the
coarse
grids
that
they
cover,
resulting
in
two­
way
interaction
since
information
flows
from
coarse
to
fine
grid
as
well
as
from
fine
grid
to
coarse
grid.
6
In
one­
way
interaction,
information
flows
in
one
direction:
from
the
coarser
grid
system
to
the
finer
grid
system.
Computations
within
the
fine
model
do
not
affect
the
larger
grid
system.
B­
8
Cloud
options
remained
consistent
with
runs
4
and
5,
which
mainly
consisted
of
treating
cloud
microphysics
with
the
"
simple
ice"
mechanism.

The
run
6
series
also
varied
by
the
grids
in
which
the
settings
were
modified.
For
example,
the
sole
difference
between
runs
6c
and
6e
was
that
6e
included
initial
/
boundary
condition
modifications
applied
to
all
grids,
whereas
these
modifications
were
only
applied
to
the
three
largest
grids
in
6c.

ENVIRON/
UT
also
employed
a
Land
Surface
Model
to
improve
handling
of
surfaceatmospheric
interactions
for
the
run
6
series.
The
more
sophisticated
Land
Surface
Models
(
LSMs)
would
provide
advantages
for
mesoscale
modeling
than
did
the
simple
"
five­
layer"
soil
model.
Surface­
atmosphere
processes
affect
the
magnitude
and
direction
of
sensible
and
latent
heat
transfer
which
then
defines
boundary
layer
development,
surface
temperature,
and
humidity
which
are
important
for
successful
air
pollution
modeling.
The
Pleim­
Xiu
approach
was
reputable
for
outstanding
results
in
air
quality
planning
in
other
parts
of
the
county
therefore
was
utilized
in
correlation
with
the
MM5
application.

In
addition,
the
modeling
team
incorporated
supplemental
data
sets,
such
as
soil
type,
vegetation
categories,
and
deep
soil
temperature.
The
following
information
describes
the
revised
MM5
applications.

Table
B­
2.
Summary
of
Revised
MM5
Applications
Run
ID
Configuration
Run
6c
Includes
2­
D
surface
nudging
toward
wind,
temperature,
and
humidity
analyses,
and
soil
moisture
nudging
toward
surface
humidity
Run
6d
Identical
to
Run
6c,
except
soil
moisture
nudging
was
turned
off
Run
6e
Identical
to
Run
6c,
except
2­
D
analysis
and
soil
moisture
nudging
was
applied
to
the
4
km
domain
Run
6f
Identical
to
Run
6e,
except
with
additional
surface
observations
from
TCOON
and
NBDC
buoy
sites
in
the
observation
nudging
database.

Diurnal
Temperatures
The
diurnal
temperature
range
was
suppressed,
resulting
in
cooler
afternoon
maxima
and
warmer
morning
minima.
In
run
6d,
temperatures
improved
significantly
when
the
soil
moisture
nudging
was
turned
off
yet
still
produced
worse
results
than
in
run
4c.
Run
6e
had
comparable
diurnal
temperatures
to
the
4c
original
run.

Wind
Speed
Wind
speed
trends
in
run
6c
were
much
better
simulated
than
in
the
4c
original
run.
The
speeds
were
stronger
in
the
afternoon
and
lighter
at
night.
The
first
four
days
of
the
simulation
had
wind
speeds
that
were
overpredicted.
However,
wind
speeds
during
September
17th
through
September
20th
corresponded
well
to
the
observed
wind
speeds.
Run
6d
resulted
in
stronger
wind
speeds.
Run
6e
predicted
wind
speeds
and
direction
which
were
compatible
to
observed
levels.
B­
9
Weather
Patterns
Diurnal
trends
of
moisture
were
not
predicted
well
in
run
6c
as
compared
to
the
run
4c.
Run
6d
had
poor
moisture
performance
on
all
modeling
days.
Humidity
in
run
6e
was
generally
higher
than
the
other
runs
but
most
closely
matched
observed
values.

Evaluation
of
the
Best
Performing
Simulation
The
final
run,
6f,
incorporated
additional
surface
observation
data
from
the
Texas
Coastal
Ocean
Observation
Network
and
the
National
Buoy
Data
Center.
Of
this
series,
run
6f
was
considered
the
best
performing
simulation.
Run
6f
predicted
temperature
and
humidity
more
accurately
than
5d
and
demonstrated
improved
wind
speed
and
wind
direction
when
compared
to
the
original
1999
simulation.
However,
the
improved
wind
statistics
for
6f
were
inferior
to
the
improvements
demonstrated
by
5d.
Furthermore,
the
boundary
layer
patterns
from
6f
were
considered
questionable.
(
ENVIRON,
2003)
This
is
shown
graphically
in
figure
B­
3,
which
compares
the
mixing
height
between
the
original
model,
5d,
and
6f.
The
graph
for
6f
displays
an
area
of
suppressed
mixing
that
appears
to
track
a
swath
of
sandy
soil
from
southern
San
Antonio
to
Bryan,
Texas.
According
to
ENVIRON,
this
indicates
the
run
6f
configuration
was
excessively
sensitive
to
sandy
soil.
B­
10
Figure
B­
3.
Comparison
of
the
Mixing
Height
Between
the
Original
Model
(
Run
4c),
5d,
and
6f
B­
11
FINAL
MM5
CONFIGURATION
Overall,
the
ENVIRON/
UT­
CEER
team
considered
run
5d
to
outperform
all
other
5
and
6
series
runs.
Their
conclusions
were
based
on
measurements
as
to
how
accurately
the
runs
simulated
observed
conditions
as
well
as
other
performance
statistics.
These
performance
results
are
shown
in
figures
B­
4
through
B­
7.
Figure
B­
4
provides
a
comparison
of
observed
wind
speed,
wind
direction,
temperature,
and
humidity
with
values
predicted
by
the
original
meteorological
model,
run
5d,
and
6f.
Figures
B­
5
through
B­
7
provide
statistical
measures
of
performance
for
wind
speed
/
direction,
temperature
and
humidity,
respectively.

Figure
B­
4.
Comparison
of
Wind
Speed
and
Direction,
Temperature,
Humidity
Statistics
for
Original
Meteorological
Model
Run
(
black),
5d
(
red),
and
6f
(
blue)
for
San
Antonio
 
Austin
B­
12
Figure
B­
5.
Wind
Statistics
for
Original
Meteorological
Model
Run
(
black),
5d
(
red),
and
6f
(
blue)
for
San
Antonio
 
Austin
Region
B­
13
Figure
B­
6.
Temperature
Statistics
for
Meteorological
Model
Run
(
black),
5d
(
red),
and
6f
(
blue)
for
San
Antonio
 
Austin
Region
B­
14
Figure
B­
7.
Humidity
Statistics
for
Original
Meteorological
Model
Run
(
black),
5d
(
red),
and
6f
(
blue)
for
San
Antonio
 
Austin
Region
B­
15
METEOROLOGICAL
RUN
5g
The
ENVIRON/
UT­
CEER
team
undertook
one
additional
run
to
merge
the
best
configurations
of
the
5
and
6
series
runs.
The
team
recommended
that
the
MM5
Run
5d
set
of
meteorological
fields
for
the
photochemical
model
be
used
in
combination
with
important
FDDA
and
input
database
changes
adopted
in
MM5
Run
6f,
but
the
MRF
PBL
scheme
and
five­
layer
soil
model
of
Run
5d
be
maintained.
This
configuration
included:
 
28
sigma
levels
 
Expanded
36­
km
domain
used
in
Run
6f
 
Two­
way
interactive
108/
36/
12/
4­
km
grids
 
FDDA
analysis
nudging
on
the
108/
36/
12­
km
grids:
­
3­
D
analysis
nudging:
MM5
was
lightly
nudged
toward
3­
hourly
gridded
EDAS
analysis
of
winds
(
in
the
boundary
layer
and
aloft)
and
temperature
and
humidity
(
only
above
the
boundary
layer),
which
were
improved
by
the
blending
of
routine
surface
and
upper­
air
observational
data
­
2­
D
surface
analysis
nudging:
MM5
was
lightly
nudged
toward
3­
hourly
gridded
surface
analyses
of
winds,
temperature,
and
humidity
generated
by
the
RAWINS
program.
 
Observation
nudging
on
the
12/
4­
km
grids:
MM5
was
strongly
nudged
toward
discrete
hourly
wind
observations
from
routine
and
special
measurement
networks
operating
in
Texas
during
the
episode.
 
MRF
PBL
 
Simple
ice
cloud
microphysics
 
Kain­
Fritsch
cumulus
parameterization
except
on
4­
km
grid
 
Five­
layer
soil
model
 
RRTM
radiation
scheme
 
Reduced
soil
moisture
and
thermal
inertia
to
account
for
drier
conditions
The
results
of
run
5d
and
5g
are
very
comparable;
for
example,
both
runs
predicted
similar
PBL
heights.
However,
there
were
strengths
/
weaknesses
found
in
both
models.
Run
5g
out­
performed
5d
in
terms
of
predicting
temperature
and
humidity.
Conversely,
run
5d
out­
performed
5g
in
terms
of
wind
speed
and
ground­
level
wind
direction.
Run
5g
wind
speed
and
direction
predictions
for
Central
Texas
during
the
first
half
of
the
episode
were
slightly
degraded
compared
to
Run
5d;
however,
along
the
coast,
Run
5g
showed
enhanced
onshore
afternoon
flow
that
was
in
better
agreement
with
observations.
Figures
B­
8
through
B­
11
provide
comparisons
of
wind
speed,
wind
direction,
temperature,
and
humidity,
between
observed
values
and
values
predicted
by
runs
5d
and
5g.
B­
16
Figure
B­
8.
Hourly
Wind
Speed
for
Runs
5d
and
5g
in
the
4­
km
San
Antonio
Area
Domain
Figure
B­
9.
Hourly
Temperature
for
Runs
5d
and
5g
in
the
4­
km
San
Antonio
Area
Domain.
B­
17
Figure
B­
10.
Hourly
Wind
Direction
for
Runs
5d
and
5g
in
the
4­
km
San
Antonio
Area
Domain
Figure
B­
11.
Hourly
Humidity
for
Runs
5d
and
5g
in
the
4­
km
San
Antonio
Area
Domain.

Since
the
results
of
the
5d
and
5g
wind
speed/
direction,
temperature,
and
humidity
comparisons
were
inconclusive
in
terms
of
which
one
was
the
superior
meteorological
B­
18
run,
additional
analyses
were
conducted.
These
included
running
the
two
meteorological
simulations
through
the
CAMx
photochemical
model
to
determine
the
impact
each
run
had
on
predicting
ozone
levels;
and
further,
comparing
the
photochemical
results
to
the
Baylor
aircraft
sampling
data
for
the
time
and
altitude
of
the
flights.
The
two
photochemical
runs,
labeled
run
14
(
using
the
5d
meteorological
run)
and
run
16
(
using
the
5g
meteorological
run)
were
identical
with
the
exception
of
the
meteorological
inputs.

A
problem
that
was
significant
in
the
San
Antonio
area
was
aloft
wind
direction.
While
the
model
performed
well
in
replicating
peak
ozone
aloft
in
the
urban
plumes
with
the
original
Met
3b
run
for
the
17th
and
18th,
the
spatial
distribution
was
poor,
indicating
a
problem
with
wind
direction
at
those
altitudes.
The
data
collected
from
the
Baylor
aircraft
flights
were
compared
to
predictions
from
the
1999
model
simulation
for
the
same
days
for
Met
Run
5d
and
5g.
Figure
B­
12
provides
a
comparison
between
the
aircraft
data
collected
by
Baylor
University
and
ozone
levels
predicted
by
the
model
for
the
correct
time
period
and
altitude
of
the
flights.
As
shown
in
the
figure,
the
peak
ozone
plumes
predicted
by
the
model
are
vastly
improved.
B­
19
Figure
B­
12.
Comparison
of
Ozone
Levels
Measured
during
Baylor
University
Airborne
Sampling
Project
on
September
17th
with
Ozone
Levels
Predicted
by
Met
5d
and
5g
Model
Simulation
for
Time
of
Day
and
Altitude
of
Flights.
(
Dotted
line
represents
collected
ozone
data
in
parts
per
billion
(
ppb).)

CAMx
Run
with
Met
5d
CAMx
Run
with
Met
5g
B­
20
STATISTICAL
EVALUATION
OF
MM5
RUN
5g
The
University
of
Texas
at
Austin
provided
the
following
information
regarding
the
statistical
evaluation
of
the
MM5
model
performance.
Winds,
temperature,
and
humidity
were
quantitatively
assessed
to
analyze
MM5
performance
at
all
available
surface
observation
stations
across
the
4­
km
domain.
The
METSTAT
program
developed
by
ENVIRON
(
2001)
was
utilized
for
the
evaluation.
The
METSTAT
program
generates
pairing
of
observations
and
predictions
and
calculates
statistical
measures
for
wind
speed,
wind
direction,
temperature,
and
humidity.
The
following
statistical
metrics
were
examined:

 
Bias
error
 
mean
difference
between
pairings
of
predicted
and
observed
data
over
a
region.
 
Gross
error
 
mean
absolute
value
of
difference
between
pairings
of
predicted
and
observed
data
over
a
region.
 
Root
Mean
Square
Error
(
RMSE)
 
the
square
root
of
the
mean
of
the
squared
difference
between
pairings
of
predicted
and
observed
data
over
a
region.
 
Index
of
agreement
(
IOA)
 
at
each
monitoring
site,
calculate
the
sum
of
the
absolute
value
of
the
difference
between
the
prediction
and
the
mean
of
the
observations
and
the
absolute
value
of
the
difference
between
the
observation
and
the
mean
of
the
observations.
These
sums
are
added
over
all
monitoring
sites
and
divided
into
the
square
of
the
RMSE.
This
value
is
then
subtracted
from
one.

Performance
goals
for
the
above
parameters
were
established
from
a
comparison
of
statistical
summaries
of
the
results
of
nearly
thirty
regional
meteorological
model
simulations
used
to
drive
photochemical
models
throughout
the
country.
Performance
goals
were
chosen
to
establish
a
level
of
performance
that
most
past
modeling
has
achieved
and
to
filter
out
those
applications
that
exhibit
particularly
poor
performance.
It
should
be
stressed
that
these
goals
are
guided
by
the
results
of
meteorological
models
that
have
been
accepted
and
used
in
support
of
historical
regulatory
photochemical
air
quality
modeling
efforts.
The
performance
goals
will
require
refinement
as
the
state
of
the
science
of
meteorological
modeling
improves.

Comparisons
of
mean
daily
statistics
on
the
4­
km
grid
against
the
statistical
benchmarks
are
summarized
in
table
B­
3
for
the
San
Antonio/
Austin,
Corpus
Christi/
Victoria,
and
Houston/
Galveston
sub­
domains.
The
importance
of
the
various
meteorological
input
fields
on
CAMx
air
quality
modeling
can
be
ranked
as
follows
(
in
descending
order):

1.
Surface
and
vertical
profiles
of
wind
speed/
direction;
2.
Boundary
layer
mixing
depth
and
intensity;
3.
Temperature
(
primarily
the
extent
to
which
it
influences
boundary
layer
characterization,
but
secondarily,
the
extent
to
which
it
affects
chemical
reaction
rates);
4.
Humidity
and
clouds
(
assuming
cloud
cover
was
insignificant,
which
was
the
case
during
this
episode).

The
table
B­
3
reveals
that
excellent
performance
for
wind
speed
and
direction
and
good
performance
for
temperature
and
humidity
was
achieved
by
the
Run
5g
simulation
on
the
4­
km
domain.
The
reader
is
cautioned
that
these
results
are
based
on
comparisons
to
observations
obtained
from
ground­
level
monitoring
stations.
Upper
air
observations
B­
21
in
the
4­
km
domain
were
limited
to
two
locations.
Vertical
profiles
of
observed
wind,
temperature,
and
humidity
were
available
from
the
Corpus
Christi
National
Weather
Service
rawinsonde
station.
Vertical
profiles
of
boundary
layer
winds
were
available
from
a
special
air
quality
study
(
Big
Bend
Regional
Aerosol
and
Visibility
Observational)
profiler
located
in
Llano,
Texas.
Run
5g
and
Run5d
achieved
the
best
performance
at
these
two
monitoring
stations.
7
(
EMERY,
C.
A.,
et.
al.,
2003)

Table
B­
3.
Comparison
of
Mean
Daily
Statistics
Against
Statistical
Benchmark
for
the
4­
km
grid.
8
Episode
Mean
Parameter
Benchmark
Austin/
San
Antonio
Corpus
Christi/
Victoria
Houston/
Galveston/
Beaumont/
Port
Arthur
Wind
Speed
RMSE*
<
2.0
m/
s
1.2
1.3
1.3
Wind
Speed
Bias
±
0.5
m/
s
0.0
0.5
0.4
Wind
Speed
IOA**
>
0.60
0.68
0.81
0.63
Wind
Direction
Gross
Error
<
30
deg
36
23
30
Wind
Direction
Bias
±
10
deg
­
6
­
5
2
Temperature
Gross
Error
<
2.0
K
2.1
1.3
1.5
Temperature
Bias
±
0.5
K
­
1.3
0.4
­
0.6
Temperature
IOA**
>
0.80
0.92
0.92
0.95
Humidity
Gross
Error
<
2.0
g/
kg
1.4
2.4
1.1
Humidity
Bias
±
1.0
g/
kg
­
0.3
­
1.6
­
0.3
Humidity
IOA**
>
0.60
0.47
0.53
0.61
*
RMSE:
root
mean
square
error
**
IOA:
index
of
agreement
Processing
of
MM5
Meteorological
Fields
for
CAMx
Meteorological
data
from
the
Run5g
simulation
were
used
to
generate
the
required
three­
dimensional
gridded
meteorological
fields
shown
in
table
B­
4
for
the
September
13­
20,
1999
CAMx
model.
The
MM5
output
fields
were
translated
to
CAMx­
ready
inputs
using
ENVIRON's
MM5CAMx
translation
software.
This
program
performs
several
functions:

 
Extracts
wind,
temperature,
pressure,
humidity,
cloud,
and
rain
fields
from
each
MM5
grid
that
matches
the
corresponding
CAMx
grid;
 
Performs
mass­
weighted
vertical
aggregation
of
data
for
CAMx
layers
that
span
multiple
MM5
layers;
 
Diagnoses
fields
of
vertical
diffusion
coefficient
(
Kv),
which
are
not
directly
output
by
MM5
(
Kv
was
diagnosed
using
the
O'Brien
1970
method);

7
Emery,
C.
A.,
E.
Tai,
and
G.
McGaughey.
2003b.
"
Revised
Meteorological
Modeling
of
the
September
13­
20,
1999
Texas
Ozone
Episode
 
Final
Report."
Prepared
for
The
Texas
Joint
Near
Nonattainment
Areas
and
Texas
Commission
on
Environmental
Quality,
by
ENVIRON
International
Corporation,
101
Rowland
Way,
Novato,
CA
94945
and
the
University
of
Texas
at
Austin,
Center
for
Energy
and
Environmental
Resources,
10100
Burnet
Road,
MS
R7100,
Austin,
TX
78758.
31
March
2003
8
Value
in
red
denotes
statistics
outside
the
benchmark.
B­
22
Table
B­
4.
Meteorological
Data
Requirements
for
CAMx.

CAMx
Input
Parameter
Description
Layer
interface
height
(
m)
3­
D
time­
varying
layer
heights
for
the
start
and
end
of
each
hour
Winds
(
m/
s)
3­
D
wind
vectors
(
u,
v)
for
the
start
and
end
of
each
hour
Temperature
(
K)
3­
D
temperature
and
2­
D
gridded
surface
temperature
for
the
start
and
end
of
each
hour
Pressure
(
mb)
3­
D
pressure
for
the
start
and
end
of
each
hour
Vertical
Diffusivity
(
m^
2/
s)
3­
D
vertical
exchange
coefficients
for
each
hour
Water
Vapor
(
ppm)
3­
D
water
vapor
mixing
ratio
for
each
hour
Cloud
Cover
3­
D
cloud
cover
for
each
hour
Rainfall
Rate
(
in/
hr)
2­
D
rainfall
rate
for
each
hour
The
MM5CAMx
program
has
been
written
to
carefully
preserve
the
consistency
of
the
predicted
wind,
temperature,
and
pressure
fields
output
by
MM5.
This
is
important
for
preparing
mass­
consistent
inputs,
and
consequently,
for
obtaining
high
quality
performance
from
CAMx.

Most
data
prepared
by
MM5CAMx
were
directly
input
to
CAMx.
A
single
40­
meter
deep
CAMx
surface
layer
was
extracted
from
aggregation
of
the
lowest
two
20­
meter
MM5
layers.
The
structures
for
vertical
layers
for
MM5
and
CAMx
are
shown
in
table
B­
5.
The
CAMx
vertical
layer
structure
is
consistent
with
recommendations
described
in
the
EPA's
8­
hour
modeling
guidance
(
EPA,
1999)
which
specifies
that
the
surface
layer
should
be
no
more
than
50
meters
deep,
no
layer
beneath
the
mixing
height
should
be
greater
than
about
300
meters
thick,
7­
9
vertical
layers
with
the
planetary
boundary
layer
and
1­
2
layers
above
it.
(
Emery,
C.
A.,
et.
al.,
2002)

The
horizontal
extent
of
the
MM5
4­
km
domain
was
defined
to
be
much
larger
than
the
CAMx
4­
km
domain
(
the
MM5
domain
reached
to
the
Texas­
Louisiana
border).
The
differences
in
spatial
extents
of
the
domains
could
lead
to
inconsistencies
in
the
flow
and
hydrodynamic
fields
just
inside
and
along
the
eastern
boundary
of
the
CAMx
4­
km
grid
and
12­
km
grids,
if
meteorological
fields
for
the
12­
km
CAMx
grid
were
derived
only
from
the
12­
km
MM5
output.
To
ensure
consistency
for
this
portion
of
the
CAMx
grids,
an
alternative
approach
was
designed.
The
4­
km
meteorological
output
fields
were
extracted
for
the
entire
MM5
4­
km
grid
coverage
using
MM5CAMx,
averaged
to
12­
km
resolution,
then
used
to
replace
the
meteorological
fields
on
that
portion
of
the
CAMx
12­
km
grid.
B­
23
Table
B­
5.
Vertical
Layer
Structure
for
MM5
and
CAMx
for
Sept.
13­
20,
1999
Episode
An
alternative
set
of
vertical
diffusivity
input
fields
were
developed
for
CAMx.
Vertical
diffusivities
(
Kv)
are
important
inputs
to
the
CAMx
model
because
they
determine
the
rate
and
depth
of
mixing
in
the
PBL
and
above.
Original
diffusivity
fields
derived
by
MM5CAMx
were
passed
through
an
additional
algorithm
that
sets
minimum
Kv
values
between
layers
1
and
2
to
ensure
that
nocturnal
stability
near
the
surface
is
not
overstated
The
minimum
value
is
tied
to
the
land
use
(
e.
g.,
urban,
forest,
agricultural,
water,
etc.)
to
represent
different
impacts
of
mechanical
mixing
and
surface
heat
input
(
e.
g.,
urban
heat
island
effect).

References:

Emery,
C.
A,
E.
Tai.,
G.
M.
Wilson,
and
G.
Yarwood.
2002.
"
Development
of
a
Joint
CAMx
Photochemical
Modeling
Episode
for
the
Four
Southern
Texsa
Near
Non­
Attainment
Area."
Prepared
for
the
Texas
Near
Non­
Attainment
Areas
through
the
Alamo
Area
Council
of
Governments,
by
ENVIRON
International
Corporation,
101
Rowland
Way,
Novato,
CA
94945.
3
April
2002.

Emery,
C.
A.,
E.
Tai,
and
G.
McGaughey.
2003.
"
Revised
Meteorological
Modeling
of
the
September
13­
20,
1999
Texas
Ozone
Episode
 
Final
Report."
Prepared
for
The
Texas
Joint
Near
Nonattainment
Areas
and
Texas
Commission
on
Environmental
Quality,
by
ENVIRON
International
Corporation,
101
Rowland
Way,
Novato,
CA
94945
and
the
University
of
Texas
at
Austin,
Center
for
Energy
and
B­
24
Environmental
Resources,
10100
Burnet
Road,
MS
R7100,
Austin,
TX
78758.
31
March
2003
ENVIRON,
UT/
CEER
2003
presentation
"
Revised
MM5
Modeling
for
the
Texas
Joint
NNAs",
slide
50
United
States
Environmental
Protection
Agency
(
EPA),
1999.
"
Draft
Guidance
on
the
Use
of
Models
and
Other
Analyses
in
Attainment
Demonstrations
for
the
8­
Hour
Ozone
NAAQS.
EPA
454/
R­
99­
004,
U.
S.
Environmental
Protection
Agency,
Research
Triangle
Park,
NC.
May
1999.
