►
Description
2020 CESM Workshop Day 2 - Tuesday, 16 June 2020 - 08:30-12:00 - Atmosphere Model Working Group
The 25th Annual CESM Workshop will be a virtual workshop with a modified schedule on its already scheduled date. Specifically, the virtual Workshop will begin with a full-day schedule on 15 June 2020 with presentations on the state of the CESM. On 16-17 June 2020, working groups and cross working groups have half-day sessions, some with presentations and some that are discussion only.
More information about the CESM Atmosphere Model Working Group:
http://www.cesm.ucar.edu/working_groups/Atmosphere/
More information about the workshop:
http://www.cesm.ucar.edu/events/workshops/2020/
Talks included:
Parameterizing Unified Microphysics Across Scales (PUMAS): an evolution - Andrew Gettelman
Ice Microphysical Properties Below -40C based on Seven NSF Flight Campaigns and NCAR CAM6 Model Simulations - Minghui Diao
CAM6 Boundary Layer Evolution Over An Idealized Diurnal Cycle - Richard Neale
A new unified boundary layer and convection parameterization CPT project: The multi-plume Eddy-Diffusivity/Mass-Flux (EDMF) approach - Joao Teixeira
Merging the CLUBB and EDMF frameworks in CAM6: Approach: SCM Results and Next Steps - Mikael Witte
Turbulent dissipation is not pressure damping: Moving beyond Mellor's single master length scale - Zhun Guo
Reference Climate of CESM Coupled Aqua and Ridge Planets - Xiaoning Wu
Simulation of Convectively Coupled Kelvin Waves Using the Single Column Atmosphere Model Coupled to the Damped Gravity Wave Method - I-Kuan Hu
The Shifting Scales of Western US Landfalling Atmospheric Rivers Under Climate Change - Alan Rhoades
Tropical Cyclones in Variable-Resolution CAM: Impacts of High-Resolution Grid Extent and Climate Change Forcing - Alyssa Stansfield
Impact of horizontal resolution on the meteorology and climate of Greenland - Adam Herrington
Investigations into the vertical resolution for the next "workhorse" version of CESM - Isla Simpson
A
C
B
So
today,
I'm
going
to
talk
about
the
evolution
of
club
micro
physics
and
CES
M.
This
is
called.
We've
now
talked
a
little
bit
about
this
in
the
winter.
This
is
sort
of
a
further
development
of
this
and
I'm
gonna
touch
on
some
other
things.
That
should
be
a
general
interest
to
people
across
the
institution
that
will
across
across
the
model,
not
just
those
interested
in
cloud
microphysics
per
se.
B
So
this
is
work,
that's
actually
been
funded
by
both
NASA
and
NSF
and
the
Department
of
Energy,
and
it's
got
a
huge
cast
of
characters
from
a
number
of
different
locations
across
NCAR
and
also
outside
of
NCAR.
So
that
starts
to
speak
to
the
fact
that
we're
now
sort
of
introducing
a
new
name
for
the
micro
physics.
That's
in
Cam,
the
parameterisation
have
unified
microphysics
across
scales,
pumas.
B
B
So
it's
sort
of
critical
as
part
of
that,
and
just
to
note
that,
regardless
of
scale
of
the
model,
you
still
need
micro,
physics,
so
micro
physics,
by
definition,
is
representing
statistically
things
at
10
to
the
minus
6
scales,
so
we're
nowhere
near
being
able
to
resolve
any
of
this
in
something
like
a
large
scale
model.
If
you
run
kam
at
7,
kilometers
or
1
kilometer,
you
could
start
to
get
rid
of
different
parameterizations,
but
you'd
still
need
a
micro,
physical
parameterization.
It
doesn't
solve
the
problem.
B
The
advantage,
of
course,
is
that
the
micro
physical
parameterization
is
still
very
well
scale
separated
from
the
grid
scale,
unlike
something
like
deep
convection,
for
example.
So
we
don't
have
to
worry
so
much
about
whether
our
assumptions
are
valid
across
scales.
We
have
an
advantage
there.
There
are.
Let
me
use
this
one.
There
are
several
types
of
micro
physics
schemes.
B
Basically,
there
are
either
sort
of
defined
as
being
bulk
bin
and
then
Lagrangian
particle
based-
and
this
goes
in
reverse
order
of
complexity,
so
Lagrangian
particle
based
schemes
basically
represent
a
distribution
by
sampled
super
articles,
with
a
multiplicity
factor.
To
give
you
a
size
distribution,
you
can
explicitly
calculate
the
interaction
of
any
of
these
pop
any
of
these
drops
and
scale
it
up
by
the
population.
B
You
can
have
discrete
bins
where
each
of
these
is
a
variable
in
the
model,
it's
a
particle
size
or
particle
mass,
and
then
you
count
the
number
to
get
a
distribution
and
again
each
size
can
interact
with
each
other
size
for
a
lot
of
explicit
calculations,
and
then
these
are
typically
defined
as
a
set
of
bins
for
each
class
or
a
class
is
something
like
liquid
or
ice.
It
usually
defines
phase.
Sometimes
it's
defined
where
you
define
precipitation
as
being
a
different
class,
because
there's
a
very
different
size.
B
You
have
rain
and
snow
and
you
can
also
have
rhymed
ice
or
something
like
groppler
hail.
So
these
bin
schemes
are
often
used
in
large
Eddy
simulations
they're,
also
computationally
expensive,
but
you
can
do
things
much
more
explicitly
because
you
have
a
population
of
drops
that
can
interact
with
the
different
sizes
of
the
population
of
drops,
but
because
this
obviously
is
statistical
and
has
a
definite
shape.
You
we
often
use
a
bulk
scheme
and
a
bulk
scheme
as
a
to
moment
scheme
again
just
going
downward
in
complexity.
B
A
to
moment
scheme
basically
defines
this
distribution
from
a
bin
scheme
with
the
functional
form
and
when
you
do
that,
you
can
integrate
to
get
a
total
mass
and
you
can
get
a
representative
number
the
second
moment
of
that
distribution
or
another
moment
of
that
distribution.
If
you
define
its
form
and
then
you
can
analytically
do
a
lot
of
the
same
calculations,
you
would
do
on
a
bin
scheme.
You
can
also,
if
you
reduce
that
to
a
one
moment
scheme.
B
You
just
have
the
prognostic
mass
and
you
make
an
assumption
about
this
size
and
they
make
it
diagnostic
and
it
sort
of
reduces
to
this
with
some
representation
of
a
functional
form.
So
in
the
current
version
of
kam
we're
obviously
using
a
we're
using
a
two
moment
scheme
to
moment
bulk
scheme
that
has
a
number
of
different
classes.
The
evolution
looks
like
this.
We
started
in
kam,
for
which
that
was
a
one
moment,
bulk
scheme
in
kam
5.
B
We
introduced
the
two
moment,
Morrison
gentlemen
scheme
that
evolved
to
version
2,
which
had
prognostic
precipitation
in
cam
six.
We
have
a
version.
That's
a
there
is
a
version
that
you
can
turn
on
a
version
of
this
inside
of
the
convective
microphysics
recently
in
cam
6.2,
and
this
will
actually
be
released
in
the
next
version,
we're
releasing
a
slightly
updated
version
mg
3.
B
Some
other
efforts
that
have
gone
on
are
this
unified
ice
and
I'll
talk
a
little
bit
more
about
that
where
we
put
rain
and
we
put
snow
and
ice
together
and
then
we
there's
been
some
work:
optimizing
the
micro
physics
code
and
adding
GPU
directives
to
get
it
to
run
on
GPUs
it'll
also
still
run
on
CPUs.
So
that's
getting
ported
back
into
the
main
code.
B
John
dennis
is
doing
some
work
on
that
and
sizzle
right
now
so
mg
2
mg,
3
versions
are
currently
being
used
or
available
in
cam,
5
and
cam
6,
and
also
the
derivative
that
should
say
models.
Sorry
so
things
like
e3,
sm
and
nori
sm,
there's
a
version
of
it
that
andy
acraman
is
coded
and
the
gifts
model,
and
it's
used
in
some
of
the
higher
resolution
versions
of
the
gos
model
that
don't
offend
our
Ohana's
worked
on.
B
So,
as
I
showed
on
my
title
slide,
there's
a
lot
more
people
involved
in
this
than
just
myself
and
Hugh
Morrison's,
so
M
and
G
are
not
really
appropriate
anymore.
So
we've
renamed
this
as
Pumas
parameterisation,
beautified
microphysics
across
scales.
It's
again
it's
basically
it's
based
on
mg3,
but
we're
evolving
it
forward.
B
It's
we're
designing
it
to
be
scalable
and
efficient
for
different
kinds
of
models
and
we're
pulling
it
from
an
external
repository,
we're
adding
unit
tests
to
that
interfaces
for
different
models
and
trying
to
make
this
easier
to
contribute
for
the
whole
community
and
also
make
sure
it
performs
as
well
as
we
can
the
current
status.
You
can
take
a
look
at
this
on
github
we're
pulling
the
utilities.
Actually.
B
This
is
now
called
Pumas
utils,
and
this
will
be
called
Puma
soon
into
cam
as
an
external
and
developing
unit
tests,
and
this
new
workflow
is
going
to
facilitate
development
across
the
model
systems.
We're
gonna
start
to
move
over
to
incorporate
the
CCPP
common
community
physics
package,
port
of
that
in
the
next
six
months.
The
average
cam
user
doesn't
have
to
worry
about
this,
but
it's
happening
behind
the
scenes
and
it's
gonna
hopefully
make
it
more
robust
and
traceable.
So
other
science
efforts
again
unified
ice.
B
We've
actually
done
some
work
on
machine
learning
which
I
could
have
talked
about
here,
but
that
paper
is
going
to
be
submitted
where
we've
replaced
the
warm
rain
process
in
in
cam
with
a
stochastic
collection
kernel
and
done
an
emulator
of
that
again.
This
efficiency
work
and
there
are
we're
working
on
lightweight
aerosol
formulations
that
don't
require
full
aerosol
model
for
different
models.
There's
been
work,
there's
work
on
ice
nucleation
and
a
CCPP
converge
so.
B
Skip
the
unified
ice
stuff
and
just
talk
about
this
perturb
parameter
ensemble.
So
the
goal
is
to
understand
the
sensitivity
of
cam
six
to
parameter
changes.
We've
never
really
done
a
full-blown
version
of
this.
We
also
want
to
help
tune
the
current
and
new
parameterization
so
like
our
developments
within
the
micro
physics,
so
we're
developing
a
list
of
parameters
and
ranges
to
do
a
Latin,
hypercube
sampling
of
the
parameters.
B
This
will
probably
take
about
three
to
five
times
as
many
sets
of
parameters
or
model
runs
as
parameters
and
we're
gonna
try
to
run
a
series
of
experiments
and
develop
an
emulator
for
the
output.
We
have
some
CPU
time
to
do
this,
we're
starting
now
with
the
with
getting
the
code
ready
and
doing
some
single
column
model
tests
as
a
validation,
the
method
the
goal
is
to
make
all
this
output
available
to
the
community
and
the
intent
is
to
modify
cam
physics
to
pull
all
this
stuff
up
to
nameless
parameters.
B
So
we
can
easily
do
automated
runs
with
this.
We
actually
have
a
list
of
these
parameters
for
the
different
parameters:
ations
things
in
club
things,
an
mg
things
that
are
only
in
some
of
the
more
advanced
versions,
pumas
things
in
the
aerosol
scheme
and
things
in
the
deep
convection.
We
hope
to
start
scam
testing
in
a
month
or
so,
and
what
we're
gonna
do
is
actually
standard
one
degree
camp,
six,
a
mep
simulations
we
may
do
them
one
year,
nudged,
because
we're
focused
on
clouds.
B
We
may
also
look
at
it:
do
them
with
pre
industrial
aerosols
to
look
at
how
this
affects
aerosol,
forcing
an
SST
+
4k
to
understand
how
it
affects
cloud,
feedbacks
and
climate
feedbacks
if
you're
interested.
Let
me
know,
we've
got
an
active
group,
that's
helping
to
develop
this.
If
you
want
to
do
some
of
the
help
doing
the
analysis
or
you're
interested
in
particular
perturbations,
let
me
know
just
to
show
that
we
put
mg
3
into
the
IFS,
so
this
is
actually
running
in
the
IFS
single
column
model.
B
This
right
here
is
the
standard
control
ifs
case
in
orange
and
mg
3
in
in
blue.
We're
actually
doing
a
pretty
good
job
reproducing
things
like
clouds
or
the
total
supercooled
liquid
on
average,
in
the
simulations-
and
this
is
a
relatively
long
single-column
case
over
Macquarie
Island.
Looking
at
mix
phase
clouds
in
the
southern
ocean
I'll
skip
to
the
summary
and
stop
there
so
mg
2
and
mg
3
with
grapple
are
currently
available
in
cam
6.
The
micro
physics
is
now
pulled
from
an
external
repository.
D
B
B
Michael
says
we
in
the
winter
meeting
we
mentioned
switching
warm
process
rates
from
KK
deciphered
by
hang.
Is
that
still
the
plan
I'd
like
it
to
be
the
plan
we
can
do
that
whenever
we
want
just
has
to
be
in
the
tuning
I
think
we
should
also
be
changing
up
some
of
the
parameters
for
our
deep
convection
in
default,
and
maybe
we
want
to
do
this
all
at
once
and
only
tune
it.
A
D
E
E
Okay,
so
I'm
going
to
talk
about
the
ice
micro,
physical
properties
based
on
seven
and
sfy
campaigns
and
the
encar
chem
six
model
simulations
most
of
this
work
is
down
by
my
graduate
student
Ryan
pet
nog,
and
we're
also
collaborating
with
Romeo
and
Suzanne
show.
So
I
want
to
thank
all
the
flight
and
Grunk
crew
from
encar,
encar
RAF
and
all
the
science
teams
at
NCAR.
E
It
is
very
important
to
represent
serious
clout
correctly
in
climbing
model,
so
they
cover
up
to
about
30
to
40
percent
of
Earth's
surface,
and
if
you
misrepresent
the
clear
sky
supersaturation
and
as
create
cirrus
cloud,
they
will
cause
about
2.5
watt
per
meter
square
average
radiative
biases
at
the
top
of
atmosphere.
As
you
know,
cirrus
cloud
can
block
some
of
the
shortwave
radiation
by
the
clear
sky
supersaturation,
which
is
just
water.
Vapor,
cannot
block
some
of
these
four
wave
radiation.
E
So
today,
I
want
to
focus
on
this
box
here
today.
I
want
to
focus
on
what
are
the
effects
of
thermodynamic,
dynamic,
condition
on
cirrus
cloud
micro,
physical
properties,
what
are
the
aerosol
indirect
effects
and
how
to
model
simulate
these
different
properties
of
cirrus
cloud
and
the
factors
affecting
them.
E
So
here
are
the
seven
flight
campaigns
from
NSF
and
we
have
a
composite
dataset
that
went
from
87
degrees
north
to
75
degrees,
south.
In
total
we
have
1,000
flight
hours
and
the
analysis
are
restricted
to
temperature
below
and
equal
minus
40
degrees
Celsius,
and
we
have
a
horizontal
resolution
about
230
meter.
E
That's
for
the
aircraft
data
for
the
instrumentation
we
use
week
so
hydrometer
for
measuring
water
vapor.
It
is
not
very
easy
to
measure
water
vapor
accurately.
So
in
the
past
we
have
used
two
different
system
to
calibrate
the
instrument,
there's
an
audio
path
and
there's
an
environmental
chamber.
So
the
goal
is
to
calibrate
this
open,
passed.
Hydrometer,
which
is
sitting
on
top
of
the
aircraft,
were
all
combination
of
pressure,
temperature
and
water
vapor,
and
combining
that
with
the
temperature
uncertainty.
E
The
relative
humidity
uncertainty
is
about
eight
to
seven
for
the
temperature
regular
interested
the
cloud
measurements
are
all
based
on
the
fast
to
DC
probe
consistently
for
the
seventh
flight
campaigns.
In
addition,
we
have
the
aerosol
probe
the
uses
that
measures
from
16
nanometer
to
about
1000
nanometer.
E
Now
here
we
have
on
the
bottom,
which
is
all
those
instruments
sitting
around
different
parts
of
the
aircraft
for
the
CSM
to
chem
6
model
stimulation.
We
run
two
different
types:
one
is
a
naja
simulation
nosh
towards
the
horizontal
wind
and
temperature
in
ameri
to
data.
The
other
one
is
a
freerunning
simulation.
We
use
the
mg
2
micro
physics
for
both
types
of
simulations
around
about
1
degree,
horizontal
resolution,
and
we
also
try
to
choose
the
similar
okay.
We
also
try
to
choose
a
similar
type
of
variables
between
observation
and
simulation.
E
For
example,
we
choose
the
same
size
range
for
the
ice
micro,
physical
properties.
We
try
to
define
in
cloud
in
the
same
way
as
the
minimum
condition,
seeing
in
the
observation
same
for
relative
humidity,
vertical
velocity
use,
similar
way
to
calculate
it
and
the
aerosols
we
group
them
into
two
different
categories:
larger
ones
and
a
smaller
ones.
E
Now
the
first
thing
we
take
a
look
is
the
original
variability,
so
we
compare
among
topics
mid
latitudes
and
polar
regions.
We
have
the
black
line
for
northern
hemisphere,
solid
black
line
line
for
southern
hemisphere
and
all
the
blacks
are
observation
and
all
the
red
are
model
simulation.
First
thing
we
see
is
that
there
are
higher
ice
water
content
and
n
I,
which
is
ice,
crystal
number
concentration
in
the
northern
hemisphere.
Male
attitude,
then
southern,
haven't
you
me
letter
to
you
as
this
stands
out,
and
also
we
see
higher
anion
topics
and
polar
regions.
E
Eeeek
m6
model
consistently
show
smaller
ice,
water
content
and
diameter
mean
diameter,
and
that's
probably
under
estimating
ice
crystal
sedimentation
and
growth.
There
is
smaller
and
I
in
the
northern
hemisphere
net
latitude
and
that
is
probably
under
estimating
the
aerosol
indirect
effects
for
the
mid-latitude
region.
E
So
most
of
these
results
are
actually
submitted
to
ACP
and
it
should
be
online
and
accessible
in
a
few
weeks
for
the
effects
of
relative
humidity.
We
want
to
see
how
do
thermodynamic
conditions
change,
cirrus
cloud
properties,
so
we
have
ice
for
accountant,
an
I&D
I
plotted
on
y-axis
and
the
x-axis
has
relative
humidity.
E
Now,
the
NI
and
and
ice
water
content
actually
Peaks
around
110
percent
and
the
DI
Peaks
about
130
percent
for
the
simulation
they
have
two
peaks,
which
is
interesting
that
80
percent
and
the
100
percent-
and
this
is
probably
due
to
the
prime
transition
of
I
scale
fraction
in
the
model
that
uses
two
parameters:
the
RH
minimum
and
RH
maximum,
which
locate
at
one
at
80
percent
and
110
percent
respectively.
So
this
may
be
the
prime
transition
now
exactly
following
the
physical
processes.
E
Another
comparison
is
the
variance
of
vertical
velocity
with
the
W
sub
variable
in
the
model.
So
we
have
on
the
top
observation
the
notch
and
free
bound
on
the
bottom
post.
The
observational
model
show
increasing
trend
of
ice,
water
content
and
I
was
the
increase
in
Sigma
value,
but
the
simulation
actually
show
lower
maximum
Sigma
W
as
it
cut
off
right
here.
So
this
is
possibly
due
to
missing
gravity
waves
from
topography,
France
connection
and
etc.
E
You
now,
the
third
effect
we
want
to
look
at
is
aerosol
indirect
effects.
In
order
to
isolate
the
aerosol
indirect
effect,
we
first
have
to
remove
other
effects,
such
as
temperature
effect
and
the
way
we
do
it
is
we
calculate
a
delta
value.
That
is
every
one
second
of
observation
or
every
point
of
simulation
data.
We
subtract
it
from.
E
We
subtract
the
temperature
pin
average
from
these
seconds,
so
we
have
a
delta
value
of
ice,
water
content
or
ni
or
di
on
the
y
axis,
and
we
have
a
delta
value
of
na
100
on
the
x
axis.
So
the
observation
showed
increasing
ice
water
content
and
I
and
di,
especially
at
the
tail,
which
is
when
you
have
much
higher
air,
so
number
concentration
than
the
average
aerosol
member
concentration,
and
we
see
this
non
monotonic
distribution
with
respect
to
both
larger
and
smaller
aerosol.
So
this
part
of
the
work
is
reason
to
publishing
journals.
E
If
you
want
to
read
more
about
the
air,
so
indirect
effect
and
we
further
qualify
the
errors
of
indirect
effect
by
applying
the
linear
regression,
to
these
doubt,
a
value
of
ice
micro,
physical
properties,
with
respect
to
the
Delta
value
of
the
aerosol
number
concentration.
So
here
the
slope
B
value
is
where
we
calculate
the
larger
the
slope
value.
It
means
the
higher
as
icing
direct
effect.
There
is
so
the
na
500,
the
art,
larger,
aerosol
larger
than
500
nanometer.
E
They
are
more
effective
at
higher
temperature,
like
here
highlighted
here,
and
we
don't
show
you
in
this
slide,
but
we
also
tested
birth
velocity.
The
larger
Arizonans
are
also
effective
at
lower
per
to,
but
also
the
indicating
possible
heterogeneous
nucleation
and
a
smaller
aerosol
group
are
more
effective
at
lower
temperature
and
higher
vertical
velocity,
possibly
homogeneous
new
patient.
When
we
put
the
slope
value
B
here
from
our
study
next
to
the
previous
model,
calculated
slope
value
from
a
series
of
model
simulations,
we
have
a
slightly
higher
slope
zurk
in
the
seventh
like
ham
hands.
E
But
please
keep
in
mind
these
three
models
are
only
considering
homogenious
to
patients,
so
there
could
be
underestimation
our
patter
genius,
the
occasion
that
contribute
to
the
difference
in
the
slope
on
the
left
hand,
side
in
this
slide,
we
actually
have
all
the
observations
again
with
respect
to
larger
or
smaller
Arizona.
On
the
right
hand,
side
we
have
the
CEM
sake
simulation
at
one
degree
again
the
simulation
are
able
to
capture
a
slight
increase
of
ice
water
content
and
night,
but
there
is
no
increase
at
all
in
the
diameter.
E
So
this
is
the
side
we
put
everything
together.
We
have
observation,
compare
where's
the
match
the
stimulation
on
the
left,
two
columns.
We
have
the
large
aerosol
in
grey
effect
and
on
the
right
to
calling
have
the
small
aerosol
in
your
effect
each
row
we
have
different
eyes:
micro,
physical
properties,
so
the
chem
six
actually
in
general,
shows
smaller
ice,
water
content
and
smaller,
actually
smaller
di.
E
Compare
with
the
observation
and
for
the
indirect
effect
air,
so
indirect
effect
observation
show
under
an
IDI,
but
these
are
not
very
strong
or
it
could
be
the
opposite
in
the
chem
six
model.
So
eventually
for
the
cloud
fraction,
chem
six
show
increase
in
cloud
fraction,
but
it's
only
ten
times
higher
IR.
So
number
concentration.
That's
when
they
show
it
so
to
summary,
to
sum
up,
we
have
the
original
variation
in
two
hemispheres.
E
We
definitely
see
higher
number
concentration
of
ice
in
the
northern
hemisphere,
mate
latitude,
but
the
motto
show
smaller
ice,
water
content
and
smaller
size
of
aerosol
everywhere.
Oh
sorry,
smaller
size
of
ice
everywhere
come
here
with
the
observation
and
there's
also
smaller
member
concentration.
You
know
net
latitudes
in
the
northern
hemisphere,
for
the
effects
of
relative
humidity
and
vertical
velocity.
We
see
some
differences
of
how
they
correlate
with
the
ice
micro,
physical
properties
and
the
aerosol
indirect
effect
interesting
fairly
serious.
E
There
is
a
non
monotonic
with
showing
the
observation
and
there's
a
sudden
increase
of
these
properties,
especially
a
very
high
value
of
aerosol
number
concentration,
and
we
also
need
to
isolate
the
thermodynamic
and
dynamically
paths
when
we
quantify
these
aerosol
indirect
effects.
The
model
in
general
show
weaker
aerosol
in
direct
effect
and
there's
no
sudden
increase.
E
E
D
E
A
great
comment,
so
for
all
of
the
fast
to
DC
probe,
we
actually
cut
off
at
sixty
two
point:
five
micron.
That
is
a
very
good
point
and
we're
actually
surprised
that,
even
at
these
very
large
sizes,
we
still
see
aerosol
indirect
effect,
pretty
clear
day
for
ice
water
content.
Ni
n
di
we
were
able
to
analyze
different
sizes
in
the
model
and
the
difference
between
the
model
and
absolution
do
not
seem
to
get
smaller.
E
C
D
C
C
F
C
F
We
good
there
yeah,
okay,
good,
sorry,
I'm,
two
screens
here
so
yeah
so
good
morning.
I
just
wanted
to
go
through
some
initial
analysis
that
we've
been
doing
at
NCAR
across
le
s
and
across
single
column,
version
of
cam,
cam
6,
a
little
tiny
bit
of
cam
5,
but
essentially
I'll
describe
some
of
this
activity
and
then
I'll
describe
a
case
study
where
look
we've
constructed
and
looking
at
and
then
some
cam
and
then
a
little
bit
of
Elias
comparison
at
the
end.
F
So
this
is
kind
of
work
came
out
of
a
activity
at
NCAR
across
the
labs.
So
the
intention
was
to
kind
of
consolidate
and
make
best
use
of
the
expertise
and
modeling
expertise,
numbers
of
ational
expertise
for
the
one
purposes.
Multi
scale
parameter
ization
development.
This
involves
a
number
of
le
s
models,
so
we
get
to
really
pro
some
very
different
models.
F
It's
briefly
on
that.
So
the
case
study,
we're
involved
with
right
now
is
part
of
it
was
part
of
the
Southeast
atmosphere
study.
So
this
is
the
kind
umbrella
of
studies,
one
of
which
is
people
familiar
with
there's
no
math
study
and
that
took
place
in
the
summer
of
2013.
So
this
is
a
kind
of
simple
surface
boundary
flat
forested
boundary
and
because
it's
forested
there
was
also
as
well
as
a
meteorology,
boundary
layer,
growth,
etc.
There
was
a
focus
on
the
photochemistry
volatile
organic
compounds,
the
specific
case.
It's
an
ideal
composite
day
of
surface.
F
It's
forcing
you
see
there,
the
latent
sensibly.
He
looks
pretty
some
simple
analytical
forms
over
the
daytime
in
terms
of
increasing
at
a
peak
in
the
early
afternoon,
pretty
simple,
again:
semi
analytical
initial
conditions,
it's
very
dry,
it's
quite
dry,
and
so
there
are
no
clouds
UB
and
so,
but
there's
also
no
large-scale,
attractive
forcings
and
at
this
point
no
chemistry.
So
it's
a
very
simple
kind
of
very
raw
comparison
than
we
can
do
with
le
yes.
F
So
here
we
are
jumping
straight
in
these
are
boundary
layer
depths
in
the
essentially
in
the
control
model
configuration.
So
this
is
30-minute
time
steps
30-minute
coupling
radiation
etc,
and
we
can
see
in
that
top
left
diagram
there
that
we,
you
know,
we
don't
know
for
sure,
of
course,
but
it'll
turn
out
to
be
a
pretty
shallow
boundary
layer
simulation
with
some
noise
I
mean
essentially
there's
24
odd
time
steps
around
this
period.
So
it's
a
pretty
raw
simulation,
so
we're
kind
of
jumping
more
towards
a
faithful.
F
You
know
more
realistic
comparison
with
the
le
s,
so
we've
increased
the
time
step
by
2
minutes
to
2,
admits
and
then
increased
all
the
the
radiation
and
the
surface
coupling
to
the
same
time,
frequency,
and
so
you
know
quite
nicely
the
blue
line
in
the
bottom
left.
We
get
a
much
more
gentle
smooth
evolution
of
the
boundary
layer
depth,
though
much
more
deeper
than
the
default
configuration
which
is
important.
But
essentially
you
know,
camp
6,
scam.
Sex
with
Club
is
much
more
soon
evolution
compared
to
the
you.
F
Do
you
double
your
representation
in
camp
5
in
terms
of
vertical
boundary
layer
evolution,
the
growth,
so
the
vertical
mixing
is
much
more,
is
much
deeper
and
unsurprisingly,
in
camp
six
with
this
high
frequency
temperature
contours
in
the
early
verticals,
it's
a
very
well
mixed
dry
boundary
layer.
What
we
expected
deeper
than
30
minutes
and
then
the
bottom
is
a
kind
of
vertical
diffusion
tendencies
from
club.
F
F
We
can
see
that
the
evolution
of
the
agile
adept
is
pretty
similar,
and
so
you
could
say
well,
32
levels
may
be
enough,
but
if
you
start
looking
at
in
this
case
a
specific
humidity
in
there,
which
we
really
expect
a
sharp
boundary
in
terms
of
Angela
talk
or
drying
just
about
the
top
there.
You
really
are
thinking
that
you
know
they
exist
in
resolution.
32
levels
is
not
enough
and
we
need
probably
at
least
64,
and
then
of
course
you
know,
256
is
very
smooth
and
it
continues
to
sharp
in
that
boundary.
F
Layer
top
and
the
mix
I
mean
important
to
point
out
in
the
depth
of
the
boundary
layer
is
not
perfect
in
terms
of
humidity
is
not
being
mixed
throughout
the
whole
depth.
There
is
obviously
gradient
vertical
greens
and
still
exist
and
in
terms
of
something
associated
the
turbulence.
So
here
we're
looking
at
the
top
level
as
the
variance
and
the
bottom
level
is
the
skewness
of
the
variances
kind
of
internal
to
the
boundary
layer
depth.
You
know,
increase
variant,
increased
determinant.
Can
energy
generally
means
increased
surface
fluxes
vertical
fluxes?
F
A
1
different
cheek
notice
is
that,
as
the
resolution
increases,
the
the
skewness
so
in
some
sense
the
tail
of
the
distribution
of
the
TK
descends,
and
so
that
is
essentially
I.
Think
telling
you
that
this
I
think
this
tail
having
a
stronger
tail
higher
that
boundary
layer
top
is
really
as
expanding
the
reducing
the
stability
of
the
boundary
layer
top,
and
that
is
much
better
in
256,
as
this
Maxima
descends
so
Elias
for
the
last
few
minutes.
F
So
here
I'm,
comparing
with
one
Elias,
it
was
part
of
the
group
the
end
car
Elias.
This
is
the
most
established
one
for
multi-decade
decades
and
Sullivan
kind
of
began.
This
so
on.
The
left
is
cam.
Six,
this
to
meter
256.
You
know
our
supreme
version
has
you
know
a
boundary
layer
that
goes
up
to
two-and-a-half
K.
The
encou
Elias
is
much
shallower
and
actually
in
much
better
agreement
with
the
observations
we
have.
Some
sealeo
meter
estimates
there
below
around
1500
maximizing,
but
there
are
some
caveats
to
this.
F
From
the
perspective
of
cam
is
running
over
a
top
of
the
idea
of
a
land
interactive
land
surface.
Even
though
you're
prescribing
leighton
sensible
heat
flux,
there
is
a
potential
for
an
influence
from
the
spin
up
in
the
surface
temperature,
which
of
course,
would
contribute
to
the
long-wave,
cooling
or
long
wave
heating
aspect.
This
is
just
an
example.
These
initial
cam
six
English's
show
how
the
surface
temperature
grows
and
so
we're
we're
kind
of
trying
to
look
at
having
a
more
equilibrated
idealized
land
surface.
F
F
I
think
from
that
boundary
layer
that
we're
using
the
calculation
death
because
that
Bachelor
top
stability
is
much
lower,
I
think
it's
giving
a
Bachelor
depth
that
is
much
deeper
than
it
is
because
you
look
at
the
top
right
le
s
and
you,
if
you
think
of
duty
to
buddy
z,
it
seems
to
be
actually
deeper
than
the
chem
six,
that's
kind
of
encouraging
specific
humidity.
Again,
you
know
the
difference.
F
Is
you
know
the
evolutions
similar,
but
the
Banja
top
decrease
in
Q
as
much
week,
much
more
stronger
in
the
elias
and
the
mixing
given
by
the
you
know.
More
vertical
contours
of
the
q
is
much
stronger
in
the
elias
or
some
differences,
and
that's
kind
of
you
can
see
that
if
we
look
at
the
vertical
fluxes
of
moisture-
and
we
can
really
see
that
in
the
elias,
these
are
much
more
maximized
closer
to
the
Badger
top
and
there
is
a
very
strong
you
know
drop-off
in
these
fluxes.
F
So
there's
a
you
know
the
flux,
convergence
there
is
very
strong.
Finally,
oh
and
in
terms
of
the
turbulence
in
the
boundary
layer
with
the
this
kind
of
symmetric
turbulence,
the
variance
in
the
vertical
velocity-
it's
not
too
bad-
that
top
figure
a
little
bit
stronger
and
then
car
Ilya
Elias.
But
the
big
difference
is
the
the
asymmetric,
the
scheme's
component,
but
the
vertical
Asti.
F
So
there
is
a
lot
more
kind
of
asymmetric
turbulent
activity
was
seen
in
Elias,
and
this
is
my
final
plot
of
comparison
with
Elias
I
haven't
quite
got
some
right,
color
bars,
because
these
are
the
recent
additions.
But
essentially
these
are
the
three
le
s
is
compared
to
camp
six
in
terms
of
potential
temperature
and
growth
of
the
boundary
layer.
During
this
time-
and
you
know
quite
nicely,
I
think
it
compares
well
with
Fast
Eddie
and
anchor
Elias,
which
are
very
different.
F
Elias
Worf
Elias
is
seems
to
be
doing
something
different
and
there
may
be
some
problems
with
that.
So
give
you
a
quick
tour
through
initial
efforts
to
kind
of
unify
our
workflow
in
terms
of
developing
parameterizations
and
compared
to
observations
and
models.
So
stronk
m6
has
strong
sensitivity
to
a
number
of
things,
timestep
forcing
and
the
coupling
frequency
in
terms
of
PBL.
In
particular,
there
is
a
converter
behavior
with
increased
vertical
resolution.
F
F
And
the
interactive
land
coupling
is
obviously
a
challenge
in
terms
of
the
two
in
terms
of
the
elias
comparisons.
Clearly,
the
asymmetric
turbulent
part,
the
skewness
of
the
vertical
loss
team,
is
playing
a
much
greater
Rawal
in
the
elias
and
there
really
is
a
much
stronger
stability
at
the
top
of
the
boundary
layer
and
moving
forward.
Yeah
like
I,
said
the
collaborated
land
condition
would
be
much
more
useful
chemistry
and
then
again,
more
direct
validation
of
these
higher-order
flux,
components
so
sure
we
associated
with
club
with
the
explicit
versions
in
yes,
okay,
thanks
great.
D
F
D
F
That
is
a
concern
and
I've
not
think
thought
too
deeply
about.
You
know
the
ramifications,
if
you
know
do
doing
you
know
more
frequent.
You
know
there
were
three
things:
the
radiation,
the
time
step
and
the
surface
coupling.
So
the
question
is
I
didn't
get
chance
to
do
this,
but
that's
my
next
kind
of
job
is
to
decompose
these
three
things
and
determine
why
the
boundary
like
that
is
quite
a
bit
shallow
in
their
default
version
of
the
model.
F
F
Quite
sure,
I
guess
in
terms
of
just
describing
somehow
a
surface
temperature
I,
think
it's
kind
of
implying
as
a
rather
than
interactive
land.
Yeah
I
mean
that
yeah.
That's
I
mean
that
eliminates
a
lot
of
sense,
because
there
are
some
differences
that
we
think
we're
driving
even
with
the
forcings
there's
some
suspicion
that
the
models
are
not
quite
seeing
them
this
in
the
same
way,
which
is
the
whole
point
I
think
of
this
exercise.
D
F
Is
not
because
we
I
mean
yeah
I
would
hope
it
would
go
in,
but
I
think
it's
nice
needs
some
fine-tuning.
Still
in
terms
of
you
know
the
start
times
and
the
end
times.
We've
moved
them
a
little
bit
because
it's
obvious
analytical,
and
so
but
yeah
absolutely
I
mean
it
makes
a
lot
of
sense.
I
mean
even
just
the
capability,
just
generally
to
Andrew,
of,
like
I,
want
to
prescribe
this
analytical
function
as
a
surface
flux.
So
this
is
you
know
of
this
profile.
Initial
conditions
I
think
that
capability
of
these
CPUs
yeah.
D
G
B
C
G
You
so
much
videos,
thank
you
for
the
opportunity.
I'll
try
to
be
as
brief
as
possible,
essentially
I'm,
going
to
try
to
introduce
and
summarize
a
new
CBD
project
focused
on
boundary
layer
and
convection
parameterizations,
and
essentially
trying
to
implement
and
evaluate
a
new
multi
blue
area.
Diffusivity,
mass
flux,
approaches,
sort
of
a
unified
boundary
layer,
convection
fan
tradition,
including
different
action,
so
I'm
Glebe
di,
but
the
copy
is,
are
as
important,
if
not
more
than
VI.
G
For
sure
who
do
you
back
my
server,
then
Kelly
O'donnell
GFDL
ROM
foo
at
UCLA,
Jorge
Mateo
at
University
of
Connecticut
and
microvilli
at
UCLA
as
well,
and
we
have
a
variety
of
team
members,
including
Martin
Karofsky
Chi
su
chuan,
Qian
Maria.
She
needs
an
register.
This
is
pretty
much
the
team
I
think
it
will
be
increased
by
one
or
two
people
in
particular
than
car,
but
this
is
the
theme
that
we've
pretty
much
put
together
in
the
last
few
months
and
been
working
together
for
a
while.
G
So
the
project
is
funded
by
NSF
I
would
say
it's
about
80%
funding
from
NSF
and
then
NOAA
funds,
the
GFDL
component.
So
we
can
go
to
the
next
slide,
where
I'll
try
to
simmer
I
started
goals
in
methodology
of
the
PMF
CPT.
So
the
overall
goal
is
to
reduce
some
of
these
essential
biases
that
we
have
that
are
related
to
video
clouds
in
the
connection,
both
in
the
end
car
and
the
GFDL
models.
G
The
methodology
will
be
to
implement
and
evaluate
a
new
PBL
and
convection
unified,
Multipla
medium,
a
parameterization
and
describe
it
very
briefly.
Of
course,
the
DMF,
and
we
can
give
you
all
the
details
that
you
want
later
DMF
is
we
now
call
it
a
fully
unified
turbines
and
convection
parameterization,
essentially
because
we
can
represent
the
transition
from
let's
say:
stable
boundary,
layer
to
dry,
convection,
shallow
compaction
and
even
deep
convection
with
one
single
mixing
branch
ization
now
I'll
show
you
just
one
result
about
that.
G
We
are
focusing
on
the
PBL
and
the
transition
to
deep
convection,
both
from
a
sort
of
spatial
transition
over
the
ocean,
as
well
as
a
temporal
transition
over
land.
So
in
spatial
transition
over
the
ocean
we
go
from
stratocumulus
to
cumulus
to
deep
convection.
If
I
say
we
go
following
the
trades
from
let's
say,
California
into
Hawaii,
Z
and
bio
no
cycle
over
land,
we
can
show
a
results
but
to
simulate
the
transition
from
the
morning
that
stable
boundary
layer,
dry,
convection,
shallow
and
deep
connection.
G
So
we
can
go
to
the
next
slide
where
this
is
the
shortest
slide.
I've
ever
shown
on
EDM
F,
there's
a
bunch
of
mathematical
formalism
and
physical
assumptions
and
theory
behind
it,
but
in
a
very
in
a
nutshell,
as
people
say,
EDM
F
is
an
approach
that
basically
represents
the
sub
grade
scale
vertical
fluxes
of
properties
as
a
combination,
Eddy
diffusivity,
that
represents
the
smaller
scale,
mixing
and
mass
flux,
mixing,
which
is
a
way
of
modeling
plumes,
convective,
plumes
or
thermals
that
are
sort
of
across
the
boundary
layer.
G
G
G
G
D
G
A
G
G
G
So
we
parameterize
the
PDF
of
surface
layer
thermodynamics,
which
is
something
that
we
know
a
little
bit
about
in
terms
of
we,
the
community
know
a
little
bit
about
the
surface
layer
and
then
we
often
use
a
Monte
Carlo
sampling
of
the
PDF
to
produce
multiple
plumes.
So
some
of
these
plumes
will
be
more
energetic
and
sort
of
healthy
from
the
beginning.
But
on
top
of
that
we
also
use
a
stochastic,
lateral
entrainment.
G
You
know
more
buoyancy
from
more
bowing
from
the
beginning,
and
they
will
go
further
than
others
and
that's
what
sort
of
regulates
how
far
the
plumes
go
and
that's
why
we
can
have
different
types
of
plumes
coexisting
in
the
same
model,
green
box,
which
is
a
great
asset
relation.
So
if
you
go
to
the
next
slide,
I'll
show
you
just
one
result:
do
you
have
a
variety
of
papers
on
this?
G
So
the
result
I'm
showing
in
the
next
slide
is
about
the
LBA
case,
which
is
a
diurnal
cycle
of
convection
over
land.
This
is
a
case
that
has
been
you
know,
there's
papers
by
boychik,
hrabosky
and
others
on
in
the
comparisons
that
we
use
the
typical
typical
case-
and
here
we
have
just
two
figures,
so
one
tells
you
cloudberries
in
cloud
tops.
G
This,
of
course,
when
you
go
and
try
to
do
a
transition
from
dry
convection
to
shallow
to
deep
convection.
Of
course,
you
need
to
include
two
microphysics,
which
we
do
in
a
very
simple
way:
solver,
Kessler
type
out
of
conversion
scheme
we
have
down
drafts
and
we
have
a
simple
branch
ization
of
coke
food
effects
as
well.
All
of
these
things
are
still
under
study,
but
what
we
can
show
is
that
and
just
focus
on
on
the
plot
on
the
left.
G
You
can
see
that
we
simulate-
and
you
know,
there's
plenty
of
other
plots
on
papers
about
this-
that
we
can
give
you
the
references
to,
but
we
can
show
that
the
EVM
Efrain
tricky
parameterization
is
able
to
go
from
dry
convection
to
shallow
convection
and
then
to
deep
convection
in
a
way
that
is
very
similar
to
the
early
s,
and
in
that
sense
it
shows
that
you
can
use
a
single
mixing
scheme
that
is
able
to
represent
all
these
different
types
of
convection.
We
don't
have
to
switch
from
one
branch
ization
together.
G
So
let
me
just
in
a
couple
of
minutes
give
you
highlights
so
go
to
the
next
slide,
highlights
of
what
we've
been
doing
or
before
that
I
want
to
just
emphasize
the
fact
that
there's
a
fundamental
tool
behind
a
lot
of
what
we
do,
which
is
larger,
a
simulation
model.
We
just
had
a
talk
that
mentioned
that
as
well
in
this
case,
is
just
to
tell
you
that
the
Arias
model
we're
gonna
use
is
the
university
of
connectivity,
arts
model
developed
essentially
by
george
mateo,
with
some
help
by
daniel
chunk
and
a
couple
of
others.
G
It's
a
model
that
is
based
on
the
original
UCLA
model,
but
then
George
worked
a
lot
while
at
JPL,
and
now
he
is
at
the
University
of
Connecticut
and
one
of
the
most
impressive
things
that
you
can
always
do.
It
Elias
is
to
show
figures
like
the
one.
This
experiment
from
George,
in
which
you
have
the
Ricoh
shuttle
convection
case.
It's
a
precipitating
case.
That's
why
you
have
mister
scale
structures
as
wide
as
they
are.
G
This
is
an
enormous
domain
for
the
typical
Elias
experiments,
it's
about
160
by
160
kilometers,
and
if
you
look
at
the
plot
on
the
left,
for
example,
which
is
a
liquid
water
path,
if
I
told
you
that
this
is
a
picture
from
Excel
light
or
an
airplane,
you
would
I
could
convince
you
because
this
you
know
this
is
testimony
of
how
realistically,
as
can
be,
and
in
particular
this
one
that
give
you
all
this
wonderful
structure
that
we
now
exist
from
measurements
that
we
made
so
just
last.
Two
slides
are
very
brief.
G
Highlights
of
things
that
we've
been
doing.
The
first
one
highlights
the
fact
that
GFDL,
we
have
implemented
a
initially
a
diagnostic
mass
flux,
not
parameterization,
but
just
diagnostic
mascots,
to
see
if
we're
getting
the
right
mass
flux
and
right
now,
we
we
have
a
fully
interactive
EDM
F
that
is
currently
being
tested.
So
it's
been
implemented
in
being
tested
and
results
are
very
promising.
We're
implementing
this
into
a
Mallory,
a
mother
version
of
GFDL,
which
is
kind
of
an
interesting
thing
to
do
as
well.
G
In
the
sense,
that's
Debbie
ate
a
little
bit
from
what
Tefal
does,
but,
but
we
feel
that
you
know,
together
with
the
LGPL
colleagues,
this
may
be
the
best
way
to
move
forward
and
the
plot
on
the
left.
We
can
spend
10
seconds
on
it.
It's
the
evolution
of
and
reached
just
mentioned
a
similar
case,
but
ok,
it's
not
too
different
from
this
one.
G
So
what
we're
working
right
now
is
to
have
the
interactive
version,
correcting
these
problems
and
going
to
the
southernmost
convection
as
well
at
NCAR
and
Michael.
We
D
will
spend
the
next
15
minutes
describing
this
in
more
detail.
We've
tried
a
different
approach
we
decided
to.
Instead,
in
the
beginning,
I
mean
this
is
not
necessarily
the
end
of
the
story,
but
in
the
beginning
we
decided
to
think
so.
If
you
go
to
the
next
slide,
we
tried
to
think
about
not
necessarily
replacing
Club.
G
So
the
assumption
is
that
the
MF
bloom
we'll
represent
some
additional
skewness
and
in
that
sense,
from
a
physics
perspective,
it's
not
unclose
about
to
assume
that
we
can
combine
them,
and
so
that's
what
we
did
was
just
it's
still
a
very
preliminary
work,
but
the
results
are
very
encouraging.
So
what
we
did
in
practice
was
to
couple
the
EMF
plumes
to
the
five
diagonals
over
of
Club,
and
the
results
seem
to
be
promising
encouraging
and
interesting,
and
this
is
just
one
example
of
arm
shallow
convection.
G
If
you
have
questions
but
essentially
EMF
combines
Iggy
nmf
to
represent
in
a
unified
way
through
some
convection
address
is
a
bunch
of
important
science
questions
that
I
didn't
have
time
to
talk
about,
but,
for
example,
it
does
explain
in
a
way
in
a
way
that
is
easy
to
understand
and
physically
plausible
the
Contra
gradient
flux
in
a
way
that
hasn't
been
explained
before,
and
so
we
have
nice
preliminary
results
that
I
just
described
and
I'll.
Let
Michael
get
into
the
details
of
the
work
with
an
current
club.
So
thank
you
for
the
opportunity.
Great.
G
G
What
yeah,
I'll
I
have
to
read
the
questions
done
I
think
there's
a
lot
when
it's
prom,
Julio
I,
think
so
let
me
take
a
look
I'm
happy
to
look
at
them
and
reply
them
by
on
the
chat
box
or
connect
people.
So
no
so
the
question
of
who
Luiz
are
the
plumes
right
now
prognostic
they
are
not
and
as
of
right
now,
we
also
are
working
on
that
aspect
as
well.
The
prognostic
plumes
are
something
that
people
have
talked
about
since
I
started
working
at
this
MW
have
25
years
ago.
G
The
prognostic
plumes
it's
one
of
those
things
that
people
say
it's
easier
said
than
done,
because
they
really
prognostic
rooms
really
imply
almost
a
new
level
of
parameterization
in
the
sense
that
you
need
to
continue
keeping
these
properties
of
these
plumes
in
memory
across
the
the
integration
that
is
extremely
difficult
to
do,
particularly
if
you
have
multiple
plumes,
as
you
can
imagine,
and
if
you
want
the
plumes
to
move
from
green
box
to
agree
box,
then
you
know
it's
almost
forget
about
it
at
this
stage.
So
it's
a
great
question
for
you.
G
H
Thank
you
Peter,
so
yeah
I
guess
I
have
about
a
half
hour
of
introduction
behind
me.
So
thanks
to
Richard
driver
for
setting
you
up
so
well.
So
this
is
work
that
I've
been
doing
with
about
half
of
the
CBT
team,
major
contributions
from
marching,
Karofsky
and
chi-su
shell
at
UCLA,
and
then
all
the
LS
results
I'll
be
comparing
with
our
from
George
Mateo
and
at
UConn.
So
it's
day
by
day,
my
think
drug
did
a
great
job
of
setting
this
up
here.
H
H
Good
variability
here,
then,
the
IDI
component
should
really
explicitly
account
for
that
variability
in
the
environment.
So,
for
example,
perhaps
by
integrating
over
a
zoom
PDF-
and
this
is
what
club
does
compared
with
what
we've
done.
In
the
past,
we
had
done
kind
of
an
exploratory
EDM
F
implementation
in
cam
5,
using
the
whole
slob
of
LPP
of
PBL
scheme,
which,
which
has
really
no
explicit
handling
of
variability.
So
we
felt
like
club
was
a
much
more
statistically
consistent
approach,
and
this
is
particularly
well
illustrated
by
that
figures.
H
You
I
showed
of
you
a
skewed
tail
where
the
mass
flux
plumes
are
sampling,
this
extreme
or
can
be
set
up,
resembles
extreme
tail
of
the
sub-grid
scale,
distribution
of
the
thermodynamic
variables,
although
it
is
not
the
double
Gaussian
by
default.
So
that's
that's
something
for
us
to
explore
now
in
terms
of
the
camp
stakes.
Implementation
as
well
mentioned.
H
Really.
We've
just
made
one
small
tweak
to
clubs,
prognostic
solver
for
the
mean
fields
and
turbulent
fluxes,
and
that
is
demonstrated
by
this
term
in
the
right
here
where
we
are
taking
the
plume
diagnosed,
turbulent
fluxes
of
the
thermodynamic
variables
and
adding
in
that
flux,
divergence
explicitly
to
the
right
side
of
the
solver
equation.
H
We
are
also
diagnosed
in
clue
macro
physics,
which
is
really
just
to
say,
liquid
water
mean
like
water,
mixing
ratio
and
cloud
fraction,
but
these
are
not
yet
intelligently
coupled
to
other
processes
like
radiation
or
aerosol,
and
our
next
big
step
will
be
adding
in
the
blue
micro
physics.
So
we
can
start
looking
at
some
precipitating
cases,
so
stay
tuned
for
that.
H
Okay,
so
jumping
them
right
into
some
single
call
model
results.
I
will
be
comparing
scam,
MLAs
and
just
over
kind
of
that.
The
GZ
SS
comparison
time
span
of
the
four
to
six
hours
into
the
simulation
and
I
have
turned
off
the
deep
convection
scheme,
as
well
as
large-scale
precip,
because
we
were
seeing
quite
a
bit
of
precipitation
with
with
the
precip
schema
so
Andrew.
That's
why
I
was
asking
about
whether
I
use
we're
moving
away
from
kharitonov
Kogan,
since
it's
a
little
bit
too
vigorous.
H
The
radiation
scheme
is
left
on
and
I
Pro
Ola
asterisk
Ihram
I
myself
to
say
that
Bois
max
is
we're
just
looking
at
a
period
where
there's
no
shortwave,
so
there's
some
amount
of
long
way.
The
cooling
that's
probably
happening
here,
but
it
doesn't
really
seem
to
be
having
much
of
an
effect
on
the
simulations
and
I
think
the
rest
of
this
covered.
So,
first
of
all,
I
guess
rich
again
did
a
really
nice
illustration
of
this.
H
Firstly,
let's
just
say:
if
we're
gonna
compare
with
Elias,
we
need
to
have
really
high
resolution,
so
I
didn't
go
all
the
way
up
to
256
levels.
All
of
these
all
of
these
profiles,
I'm
showing
here,
are
just
scan
with
clubs
on
no
mass
flux,
yet
surprise,
surprise:
when
we
go
to
higher
resolution
from
32
to
60
to
120
levels,
we
get
better
agreement
with
the
mean
thermodynamic
fields,
we
get
increased,
moist
tendencies
that
occur
higher
up
and
so
going
forward.
I'll
just
be
showing
results
from
the
120
level.
H
Scam
set
up
that
I've
done
here
all
right.
So,
let's
add
in
the
mass
flux,
plumes
and
let's
see
what
we've
got
so
I'm
going
cross
here.
We've
got
profiles
of
liquid
water
potential
temperature,
total
water,
mixing
ratio,
cloud
fraction
cloud
liquid
water
on
the
top
row
in
the
bottom
row,
moist
tendencies
to
moisture
to
temperature.
Sorry
this
should
be
switched
around
about
the
turbulent
flux
of
state
of
L
and
the
turbulent
flux
of
total
water.
So
it
turns
out
the
meantime
a
dynamic
fields.
We
don't
see
a
huge
impact
from
adding
the
plumes.
H
We
just
see
a
kind
of
a
slight
increase
in
depth
and
a
little
bit
closer
to
the
le
s
same
as
cloud
fraction.
We
don't
see
a
huge
difference
in
the
peak,
but
we
do
see
increased
cloud
fraction
above
that
that
maximum
the
plumes
do
add
quite
a
bit
of
liquid
water,
though,
which
is
what
we
would
expect.
H
The
club,
+
MF
total
into
contributions
from
club
and
from
mass
flux,
and
so
we
can
see
here,
is
that
mass
flux
has
a
pretty
sizable
contribution
to
both
the
theta,
L
and
Qt
fluxes
in
that
cumulus
layer,
which
we'd
expect
and
we're
actually
kind
of
a
little
bit,
maybe
a
little
too
vigorous.
But
we
haven't
done
any
tuning,
so
this
actually
gets
it.
Hannah's
question
is
of
how
we
tune
these
MF
parameters.
H
We
do
a
lot
of
that
in
our
own
SCM
at
JPL
and
a
lot
of
parameters
since
2d
studies
and
because
EDM
F
kind
of
takes
a
little
bit
different
shape
in
terms
of
implementation
across
the
models
that
we
work
with.
This
is
something
that
we'll
need
to
think
about,
specifically
for
for
cam
going
forward,
so
I
think.
That's
all
I
wanted
to
say
about
this
now
now.
H
Here's
one
of
the
big
questions
for
us
is
if
we
take
club-
and
we
just
we
want
to
see
if
mass
flux
can
kind
of
compensate
for
a
loss
of
mixing
from
club.
What
happens
if
we
turn
off
one
of
clubs
gaussians?
So
that's
what
we
tried
next
so
now
I'm
adding
another
two
lines
to
this
figure
or
now.
This
yellow
line
is
the
cloud
layer
unified
by
a
unit
normal
or
so
just
a
single
Gaussian
Club
and
the
purple
curve.
Is
this
single
Gaussian
Club
with
mass
flux?
H
H
Without
that
extra
mixing
from
the
second
Gaussian,
so
I'm
gonna
flip
over
to
a
different
case,
so
where
the
bumex
case
is
kind
of
quasi
steady
state,
we're
really
interested
in
transient
cases.
So
looking
at
diurnal
cycle.
So
for
that
I'm
going
to
bring
back
the
arms
southern
Great
Plains
starts
like
lava
cumulus
case
from
Brown
at
all,
so
now
I'm
sitting
precipitating
cumulus
case
and
it's
no
longer
a
staring
scampia
so
actually
to
adapt
this
case
from
how
it
looked
in
III
sm.
So
it's
a
it's
a
similar
configuration
to
Bo
max.
H
So
now,
looking
at
temperature
and
moisture
evolution,
4
times,
3,
6,
9
and
12
hours
through
the
simulation
again,
we
can
see
club
and
club
does
a
decent
job
of
kind
of
getting
the
rough
state
of
things.
But
mass
flux
has
some
value
in
terms
of
starting
the
mixing
earlier,
which
we
can
see
here
3
hours
and
there's
this
extra
little
deepening
from
the
mass
flux
which
we
really
can
see
in
the
total
water
and
then
as
we
get
further
along
in
the
simulation.
H
So
we
can
see
that
that
mass
flux
is
giving
us
a
little
bit
of
extra
mixing.
What
does
it
look
like
now?
If
we
just
go
straight
to
the
cloud
fields?
So
here
we've
got
curtains
of
cloud
fraction.
The
le
s
is
on
the
bottom
and
the
two
different
scam
configurations
or
appearing
these
top
two
rows.
Overall.
H
The
nature
of
the
la
esperanza
know
where
we're
under
predicting
cloud
fraction
by
a
by
a
little
bit
in
both
scam
configurations
in
terms
of
cloud
liquid
water.
We
see
a
very
similar
story
now
that
these
times,
where
we've
got
the
really
vigorous,
stochastic
plume,
you
can
see
that
cloud
water
extending
deeper
into
the
layer
which
is
embed
agreement
with
dles
now.
Finally,
I
just
want
to
wrap
up
with
a
quick
look
at
the
turbulent
fluxes
here
and.
I
H
Actually
something
really
counterintuitive
going
on
where,
as
we
look
at
these
three
later
times
from
1730
UTC
on
mass
flux,
which
is
this
dotted
line,
appears
to
have
a
somewhat
small
contribution
to
the
total
fluxes.
Even
though
the
total
mixing
is
much
greater
in
the
clogged
+
MF
configuration
so
what's
going
on
here
is
in
these
snapshots,
we're
seeing
apparently
very
little
mass
flux
contribution,
even
though
the
total
fluxes
are
greater.
H
H
Cooler
colors
mean
almost
none
of
it,
and
so
we
can
see
early
on
in
the
simulation
is
that
mass
flux
does
a
lot
of
the
heavy
lifting
to
start
making
this
this
deeper
mixing
happen,
so
once
that's
been
accomplished
and
Club
can
take
over
and
I
think
this
really
gets
at
what
rich
was
showing,
which
is
that
club
can't
do
these
really
extreme
skewed
distributions?
It's
it's!
It's
almost
like
it's
got
some
inertia
in
it
that
needs
to
be
overcome
and
mass
flux
is
able
to
jumpstart
this
process.
H
H
To
club
via
the
prognostic
solver
for
mean
fields
of
turbulent
fluxes
compared
to
club
alone,
the
mass
flux
plumes
increased,
turbulent
fluxes
overall
and
improve
agreement
of
macro
physics
with
the
LS.
We
also
see
some
evidence
that
the
mass
flux
plumes
can
compensate
for
when
club
has
reduced
mixing,
either
by
removing
a
single
gaussian
or
at
the
beginning
of
this.
This
arm
diurnal
our
case
and
in
terms
of
next
steps.
There
are
a
lot
of
them.
The
most
important
thing
will
be
certainly
kind
of
plumb.
H
Some
of
these
mass
flux
diagnosed
quantities
beyond
turbulent
foxes
out
to
other
camp
schemes,
adding
warm
Pacific
recep
to
the
plumes.
We
can
start
again
some
precipitating
cases
and
then,
of
course
the
big
one
is
to
start
doing
some
3d
testing.
So
there's
a
lot
left
to
be
done
here,
but
but
we're
really
excited
about
the
promise
of
a
club
plus
and
enough
scheme
going
forward.
So
with
that
I
will
stop
in
thanks.
So
much
all.
D
J
D
J
Can
you
hear
me
yes?
Well?
Can
you
see
my
slight
yep?
We
can
see
it.
Okay.
Thank
you.
That's
great,
so
I'm
Trin
today,
our
talk
turbulent
dissipation
is
not
pressure.
Dumping
moving
beyond
the
matter
single
master.
Last
skill:
this
is
the
outline
of
my
talk.
So
first
we
will
explain
why
we
need
different
time
scale
for
pressure
turn
and
dissipation
term.
Second,
we
will
introduce
our
new
much
time,
scale
parameter
relation
after
that
we
will
show
some
global
results
and
sensitive
studies.
J
J
Here
is
two
example
wise
crime,
5.5
and
another
is
nice
reason
we
want.
So
we
can
say
that
the
cumulus
is
too
bright,
but
the
strut
qumar's
is
too
Dean
the
small
scale
turbulence.
Mobility
in
moisture
and
heat
content
can
be
dumped
by
the
two
different
processes
for
the
variance
it
is
dissipation
for
the
flux.
It
is
pressure
dumping.
J
These
two
physical
mechanism
stands
in
stark
contrast
to
each
other,
especially
in
the
null
cloudy
layer.
This
patient
has
a
smaller
scale
that
involves
the
turbulence
cascade,
but
for
the
pressure
is
you
react
on
a
larger
scale?
It
doesn't
depend
on
the
turbulence
cascade,
so
we
use
a
large
Eddy
simulation
of
foam
x2
as
a
example.
So
the
left
panel
shows
the
flux
and
the
right
shows
of
variance
we
can
see
between
the
stable
layer.
The
variance
is
nonzero,
because
the
inversion
itself
all
relates
up
and
down.
J
J
So
it
can
be
better
understand
by
you,
the
prognostic
equations.
Here
we
show
two
equations.
First
one
is
it's
a
variance.
We
only
show
the
larger
term
here
is
the
turbulence
production,
dissipation
and
the
turbulence
of
the
reaction
term.
But
for
the
flux
we
have
a
production
term
or
is
it
turn
pressure
term
and
the
advection
term?
J
So
if
we
set
assume
that
the
graduates
already
in
the
steady
States,
so
the
tendency
should
be
zero
and
as
a
generation
first
should
be
balanced
by
the
dissipation
term
and
the
pressure
term
consider
the
street
fighting
inversion
layer
result
any
turbulent
Cascades
for
non-breaking
great
waves.
We
say
the
flux
is
very
close
to
zero,
but
the
variance
is
larger
than
zero.
So
for
the
time
scale
call
for
the
variance
it
should
be
as
larger
as
possible,
but
for
the
variance
of
flux
it
shouldn't
be
as
small
as
small
as
possible.
J
In
order
to
solve
this
problem,
we
developed
a
new
mod
time.
Scale
scheme
is
showing
here
for
the
folder
layer,
a
neutral
and
unstable.
The
point
of
vasila
frequency
you're
ready
less
than
zero,
so
the
unstable
car
has
three
components.
The
first
one
is
a
background.
It
means
the
layer,
our
share
free
and
you
lock
it
above
the
lower
surface
and
then
the
surface
term.
It
usually
peaks
at
the
bottom
levels
which
reach
close
to
the
surface,
and
the
third
turn
is
the
share
damping
term.
J
However,
for
the
stable
layer,
the
point
of
based
on
a
frequency,
you
really
larger
than
zero
and
our
pressure
term
need
a
stronger
thumping,
and
so
we
need
another
extra
contribution.
That
is
a
function.
Sorry,
that
is
a
function
of
the
provision
of
frequency.
We
also
note
that
the
stable
stratification
surprised
the
scatter
flux
more
in
the
null
cloudy
layer
that
it
does
need
to
a
cumulus
layer.
So
we
introduce
another
term
that
only
applies
in
the
clear
layer
with
a
cloud
fraction.
J
The
new
tall
scheme
has
a
several
advantages.
First,
it
simplifies
the
calculation
of
time
scale
and
a
mixing
Lance
scale.
It
decreased
the
run
time
of
club
car
by
about
15
percentage
and
a
second,
it
reduced
the
number
of
tunable
parameters.
We
introduced
five
tunable
parameters,
but
we
get
rid
off
certain.
J
J
So
as
it
results,
the
turbulence
prefer
over
Hawaii
looks
more
humorous
like
and
agree
where,
with
a
large
Eddy
simulation
and
the
observation,
so
at
a
stable
layer,
the
flux
is
near
zero,
but
the
variance
is
still
large.
So
the
new
scheme
can't
dump
the
flux
more
mean
where
it
doesn't:
dump
the
variance
and
the
bright
and
the
coomer's.
J
So
we
put
the
new
tall
scheme
into
Israel
as
we
won
with
car
sales
team,
and
we
did
a
bunch
of
simulation
by
configure
model
in
two
degree
and
one
degrees.
The
result
are
compared
with
default
Israel
assembly,
one
which
user
the
M
scheme
as
a
deep
convection
scheme.
We
only
show
one
degree
result
here,
so
this
slide
shows
three
years
mean,
shall
we
enforcing
in
both
Israel
and
we
won
and
the
club
sales
tone.
J
So
both
configuration
has
a
very
decent
patent
and
the
magnitude
when
compared
with
the
observations,
but
the
history
as
we
won,
underestimated
the
cloud
near
the
coast.
Are
these
overestimated,
the
astronomy
imposing
for
the
criminals?
Regions
sells
tall,
move
the
static
luminous
close
to
the
coast
at
the
same
time
it
deems
the
coolness,
and
it
also
improves
the
distinction
between
straight
humerus
and
the
cumulus.
J
Weepy
called
two
cases
to
understand
why
the
new
scheme
improves
the
coaster
stratocumulus.
So
the
one
case
is
this:
for
co
and
another
is
daikon,
so
the
yarn
interim
are
showing
up
like
and
israel
and
we
won
our
show
improve
our
new
scheme.
Results
are
showing
green,
so
generally
friction
called
the
liquid
award
for
contents,
a
much
more
than
the
eastern
assembly.
One
year's
calls
it
is
because
our
inversion
layer
is
much
sharper
and
mixing
is
stronger.
J
J
We
also
pick
out
to
shun
convection
case.
One
is
the
Rico
another
one
is
Hawaii,
so
look
has
our
crud
fraction
prefers
and
the
called
the
liquid
profiles.
So
you
know
in
your
new
scheme,
the
cloth
fraction
and
cloud
liquid
water
makes
him
near
the
cloud
base,
just
like
the
larger
Eddy
simulation
and
observations,
but
the
folders
thank
you,
but
for
the
III
SM
we
won
the
crowd
a
fraction
and
called
the
liquid
water
table
magazine
as
the
cloud
top
is
more
like
straight
cumulus,
also
the
boundary
layer,
the
realty
wrote
it.
J
J
Our
new
Tao
scheme
also
works
well
in
the
shallow
convection
ratings
in
the
deep
convection
region.
Here
is
an
example,
or
were
the
worst
and
Pacific
weren't
paul,
but
we
only
use
the
club
cells
as
a
thief
scheme.
So
we
can
see
here
the
claude
fraction,
liquid
water
contents,
potential
temperature
and
relative
humidity
as
a
good
agreement
agreement
with
the
m
scheme
and
the
uranium
at
last.
We
want
to
show
some
sensitivity
studies
so
what
happen?
If
we
only
use
one
single
call
for
everything.
So
yes
means
this
study.
J
J
So
here
is
our
conclusion:
the
turbulence
time
scare
of
term
of
dissipation
and
the
pressure
term
should
it
be
different.
Introducing
the
new
multi
timescale,
prime
tradition,
in
higher
order
closer
scheme
is
helpful
and,
as
this
new
scheme
all
asked
to
brighten
the
coast
stratocumulus
and
improve
the
distinction
of
straight
humerus
and
cumin
ours,
that's
it.
Thank
you.
D
Great
thanks,
so
we
have
time
for
a
question
or
two,
let's
see
so
I
think
Adam
Harrington
asked
whether
you
said
that
for
the
club
sales
Tao
Platz,
you
turned
off
the
ZM
scheme.
Is
that
true?
Yes,
yes
and
so
I
had
a
follow-on
question
to
that
which
is
so.
How
do
we
know
that
the
results
you
get
versus
the
III
SM
model
are
due
to
the
turning
off
the
CM
scheming,
using
SILS
versus
using
your
new
Tao
scheme?
Oh.
J
D
So
if
people
didn't
want
to
use
SILS,
do
you
think
that
Tao
is
still
the
way
to
go
for
new
club
simulations?
Yes,
a
chem-7,
okay,
great
thanks,
so
I
think,
let's
see
we're
about
three
minutes
late
for
going
to
our
break.
I
think
that
our
break
is
supposed
to
end
at
9:15
or
sorry
9:15,
my
time,
10:15
Mountain
time.
Let's
continue
that,
and
so
we
will
see
you
in
about
12
minutes
or
so
so
thanks.
Everyone.
D
C
K
K
C
A
D
A
C
D
K
L
My
name
is
Gianni
and
I'm.
A
grad
student
at
Stony,
Brook,
University
and
I've
been
missing
at
NCAR
City
for
a
while
and
the
work
that
I'm
presently
is
made
possible
by
the
tremendous
support
of
my
advisors.
Professor
Reda
move,
as
well
as
the
great
amount
of
support
from
the
ocean
over
resection
and
help
from
some
folks
in
the
world,
learn
what
in
community
so
a
big
shout
out
to
them
as
a
cohan
and
Frank
briefly
mentioned
yesterday.
L
L
There
is
also
a
long
training
of
our
works
using
idealized
through
presentations,
including
ocean
covered
aqua,
as
well
as
a
hierarchy
of
a
signified
land
geometry
representations,
which
can
tell
us
a
lot
about
the
fundamental
roles
of
ocean
circulation
and
how
that
affects
the
climate.
Now
also,
this
model
components
are
incredibly
useful
for
investigating
both
scientific
questions
and
for
model
development
and,
if
evaluation.
But
the
question
is,
from
those
uncoupled
and
extremely
simplified
components,
leading
up
to
the
forty
couple
of
class
kind
of
model
configuration
with
the
complexity
of
the
real
or
system.
L
We
need
something
in
between
that
that
helps
us
understand
a
bit
more
of
the
of
idealized
interactions,
but
but
in
the
cup
of
sense,
and
that's
where
our
model
configuration
tries
to
address
this
gap
and
that's
using
for
the
Cabos
image
class
of
model
within
C
lesson.
But
sending
the
map
is
idealized
configurations
and
to
get
into
that
just
to
show
a
profile
picture
of.
L
What's
those
simplified
cram
planets
could
look
like
the
top
chakra,
which
is
showing,
what's
basically
an
aqua
planet,
but
with
a
forty,
dynamic
couple
ocean
and
just
a
fur
for
the
very
first
look.
There
are
things
that
it
already
looks
very
different
to
what
was
usually
assumed
with
the
atmosphere.
Only
a
ke
kind
of
a
colonist
and
the
bottom
row
is
showing
the
rich
conformation.
Where
this
there
is
a
single
strip
of
protocol
countenance
that
serves
as
a
as
an
idealized
the
ocean
boundary,
which
can
be
seeing
as
a
representation
of
a
gigantic
Pacific.
L
And
we
get
a
little
bit
of
the
action
in
the
Pacific
in
terms
of
the
contrast
of
the
western
and
eastern
size
and
the
role
of
boundary
currents
and
how
that
shapes,
the
climate
and
on
the
rightmost
column.
This
is
like
the
punchline
of
the
current
talk,
one
of
the
applications
that
this
idealized
models
can
be
use
for,
includes
median
or
heat
transport
right
and
for
details
of
how
the
models
are
set
up
for
them.
L
For
the
atmospheric
component,
where
we
are
going
retro
and
using
camphor,
and
that's
because
that's
for
a
balance
of
a
cost,
then
complexity.
Since
these
simulations
require
hundreds
or
even
thousands
of
years
for
the
ocean
model.
We
are
using
the
greatest
and
latest
CSM
component
of
mom
six,
which
would
replace
the
current
ocean
model
in
later
versions
and
the
sea
ice
and
land
components
are
basically
using
was
the
default
of
a
currently
available
in
GSM,
which
will
retain
the
seasonal
cycle
with
idealized
orbital
parameters
and
and
the
climate
and
pretty
much
reaches
quasi-equilibrium.
L
Looking
at
the
mass,
we
mean
sea
surface
temperature
of
year
for
hundreds,
just
just
to
show
some
of
the
actions
that
are
happening
on
this
kindness
and
for
the
aquacase.
We
see
that
there
is
a
global
belt
of
equatorial
upwelling,
that's
making
the
equator
equator
a
little
colder
than
the
waters
right
on
both
sides,
whereas
in
the
rich
case,
because
of
the
presence
of
this
wasn't
boundary,
we
see
the
shifting
with
the
seasonal
cycle.
L
We
we
get
something
that
resembles
pacific
where
we
get
a
a
western,
western
wormhole
and
western
boundary
currents
carrying
the
warm
water
poles
and
on
the
eastern
side
of
the
basin.
We
see
this
coke
town,
which
is
magnified
as
it
slowly
and
and
the
eastern
boundary
current,
such
as
those
on
the
coast
of
north
and
south
america,
and
has
little
role
in
that
as
well
now
to
focus
on
the
hundred
year
climatology
for
the
main
state
of
this
planet.
L
The
first
thing
that
we
we
realized
is
that
those
both
of
these
planets
are
pretty
warm,
and
so
they
do
not
have
any
presence
of
ice
on
them,
and
so
in
the
in
the
in
terms
of
albedo,
then
the
row
of
cows
really
stands
out.
Both
of
these
planets
are
pretty
warm
when
we
sum
up
our
case
being
a
roughly
two
degrees
warmer
than
the
rich
case
and
in
pattern
of
Xia
SS.
L
He
we
see
the
equatorial
code
melt
aqua
and
for
the
rich
case,
because
of
the
western
boundary
we
had,
we
have
the
formation
of
a
Western
wormhole
which,
which
reduces
the
extent
of
of
the
global
equatorial
poorly
I.
Want
you
just
the
eastern
side
of
the
of
the
ocean
basin
and
that's
all
those
asymmetry.
L
These
interactions
of
in
the
atmosphere
we're
in
the
upper
case
code,
equatorial
belt
of
upwelling
is
is
mostly,
is
mostly
dry
and
with
subsidence.
Where
for
reach,
we
see
something
that
looks
a
little
bit
more
like
the
Walker
circulation,
with
with
western
one
core
being
being
moist
and
with
compassion
and
and
these
kinds
of
actions
with
the
water
river
in
the
clouds
affects
the
overall
radiative
budget
and
where
the
aquacase
receives
more.
If
it
were
heating
that
Ridge
and
for
the
ocean
surface,
we
also.
L
We
also
see
that
because
of
the
because
of
the
row
of
shortwave
heating
and
lack
of
latent
heat
loss
and
the
coat
belt
of
aqua
actually
becomes
the
reach
a
region
of
warning
compared
to
reach
in
terms
of
receiving
more
net
heating.
On
the
surface
and
very
much
that
we
have,
we
have
a
equatorial
coat
belt
for
upon
and
that
influence
and
that
surface
temperature
hadn't
influences
the
circulation
at
the
meridional
circulation
in
the
atmosphere
which,
as
you
can
see
in
the
in
the
Eulerian
overturning
for
the
aquacase.
Because
of
that
code.
L
Without
we
get
to
Little
rivers,
Hadley
cells
within
ten
degrees,
north
and
south,
which
is
consistent
with
the
subsidence
and
the
equator
and
for
village
because
of
because
of
the
dollar
asymmetry
and
the
Western
world.
That
surface
reading
lists
them
much
reduced
and
correspondent
is
a
yeah.
The
rivers
happily
are
also
much
reduced
and
on
all
of
these
things
now
go
get
him
back
to
our
hotline
at
the
beginning.
All
of
these
things
has
an
influence
and
Meridian
of
heat
transport.
L
My
bosses
are
clear
and
the
ocean
putting
us
back
for
both
Appa
and
the
rich
cases.
One
thing
that
that
I
think
is
pretty
cool
is
that
outside
of
the
topics
both
of
this
plan
is
actually
doesn't
look
too
far
off
from.
What's
the
realistic
earth
season
look
like
and
the
coolest
parts
are
within
the
deep
tropics
where
the
interesting
actions
are
concerning
the
rivers
Hadley
cells.
L
Ten
minutes
great
thanks
and
for
the
rich
case,
because
of
both
see
ya,
energetic
demons
under
the
influence
of
cloud
and
water,
written
feedback
and
then
and
dynamically.
Those
strengths
of
those
rivers
and
reserves
are
much
reduced.
We
see
that,
even
though
there
is
still
a
little
bit
of
an
acquittal
word,
I
must
repeat
transport
nearly
Twitter,
but
that
is
much
reduced
compared
to
aqua,
and
that
brings
us
to
a
climate
or
in
the
sense
of
a
yellow
heat
transport.
L
That's
much
closer
to
the
realistic
her
system
than
Apple
is
so
that's
just
one
case
of
demonstrating
the
utility
of
this
couple
and
simplifying
models
in
in
helping
sharing
sunlight
under
realistic
person
in
in
other
complex
configurations.
So
to
summarize,
I've
just
briefly
shown
a
one
case.
Where
is
a
young
and
it
is
the
couple.
L
Think
that's
that
and
among
those
is
there
is
some
software
engineering
our
force
on
the
way
to
make
that
process
to
make
that
customer
customization
process
really
simple
ways
in
CSM
and
you're:
welcome
to
contact
her
Scott
Bakula
and
have
a
Simpson
regard
and-
and
in
this,
as
far
as
the
couple
model
goes,
there
are
a
range
of
applications.
It
could
mean
especially
coupled
atmospheric
ocean
dynamics
and
different
disco
in
her
interactions,
which
would
be
permitted
by
changing
the
resolution
of
the
of
the
model
components.
So
I'll
stop
here
and
taken
questions.
It's.
L
K
L
I
believe
that,
well
it's
not
so
much
of
an
emic,
because
what
we
have
is
a
gigantic
Pacific,
and
so
something
that's
we
do
have
overturning,
but
it
doesn't
go
as
a
shape,
looks
very
different
from
the
a
mark,
and
that
has
to
do
with
some
with
the
interaction
between
the
ocean
basins.
There
are
some
other
idealized
study
which
I
study,
if
you,
if
you
put
English
in
another
age
and
you
get
something
like
sorry
marking
the
smaller
ocean
facing,
and
yes
indeed
and
the
scale
of
hundreds
of
years
deep
ocean.
M
K
M
All
right,
sorry,
Fleur
troubles,
and
thank
you
all
for
being
here.
I
hope
everybody
everybody's
here
mean
hearing
me
well,
and
so
my
talk
today
is
to
capo
scam
to
the
damp
gravity,
wave
method
for
simulating
collectively
capo
Kelvin
waves,
and
this
is
the
project
I've
been
working
on
with
my
PhD
adviser,
Brian,
Mays
and
people
had
in
Karluk
reach
and
Andrew
and
I,
especially
like
to
thank
Kevin
Brian
John
Force,
helping
you
set
up
the
some
skin
or
even
King
our
planet
configurations.
M
So
our
ultimate
scientific
goal
is
to
understand
the
interactions
among
stuff,
great
physical
Prem
relations
and
between
them
and
we
soft
dynamics
by
utilizing
Clement
model
here
a
case
that's
been
advocated
by
held
in
2005,
and
this
is
the
how
I
envisioned
the
climate
model
here.
Eric
is
of
Ken
so
from
the
branch
from
the
lower
left
to
right
is
the
hierarchy
of
configurations
for
evaluating
technical
course.
This
is
abducted
from
that.
M
If
you
get
one
of
three
and
ya
Polaski
in
2000
health
and
in
our
project,
we
would
like
to
focus
on
the
range
of
the
upper
left
to
right,
which
is
the
hierarchy
of
configurations
for
evaluating
stuff.
Great
physical
Prem
relations
and
I,
specifically
like
to
emphasize
there's
a
huge
gap
between
the
one-dimensional
and
single
column
can
and
arguably
the
most
idealized
really
can
configurations
which
long
can,
as
an
relative,
combat,
really
bring
over
global
domain.
M
So
in
skin
we
have
to
does
not
use
the
technical
core
and
we
have
to
would
require
some
localized
forcing
they'll
be
obtained
from
either
few
experiments
order,
model
outputs
and
in
this
project
we
would
try
to
introduce
new
configurations
in
which
skin
is
coupled
to
the
permit
parental
rights,
large-scale
dynamics,
and,
specifically
in
this
talk,
the
method.
I
use.
It's
the
demco
active
waste
method
that
could,
when
couple
to
scan
that,
would
stimulate
collectively
a
couple
waves.
M
But
at
this
game
each
W
configurations
is
formulated
as
the
deviations
from
around
the
reference
state
most
clearly
our
RC
state.
So
that
letter,
let
us
to
build
a
scam,
RC
configuration
first
so
in
the
skin.
Rc
configurations
which
tell
a
large
scale
and
fashion
turns
to
zero,
basically
make
us
EMR,
so
you
represent
to
purchase
them
and
average
over
the
broad
tropics
or
a
billion
column
during
and
we're
using.
There
are
some
if
protocols
with
some
modifications,
so
I'll
briefly
review
how
our
formulations
of
a
skin
digital
coupling.
M
We
can
combine
all
this
above
equations
into
a
single
equation:
anomalous,
virtual
temperature
and
an
almost
vertical
pressure
velocity,
and
then
we
assume
the
two
dimensional
and
one
of
chromatic
wave
form
solutions
all
these
prime
variables
plug
in
back
to
that
combined
equations,
and
if
we
pick
up
a
specific
level
locations
and
multiply
it
up
these
combined
creations
by
the
expose.
This
explanation
turns
take
a
real
component.
M
M
Throughout
a
column
we
will
test
Kenna
and
chemist
extender
physics
package,
and
in
this
talk,
I
will
only
show
the
results
using
their
default
vertical
resolution.
We
did
when
a
finally
stopped
vertical
grid,
which
is
60
level,
but
I,
don't
think.
I
would
have
a
time
to
show
them
here
and
we
use
a
600
second
model
times
depth
and
finally,
with
prescribe
of
idealized
radiative
cooling
profile,
as
shown
here,
we
also
have
to
introduce
a
parameterization
of
large-scale
horizontal
advection
of
moisture.
M
This
is
different
from
the
default
scan
or
a
new,
rodent
and
record,
in
which
particular
version
of
moisture
is
computing,
using
a
familiar
quarantine
transport
scheme
and
we
still
a
region
lead
at
175
how
Pascal,
as
the
ref,
our
upper
boundary
condition.
Basically
compos
came
to
a
democratic
wave
method
at
every
time,
step
and
finally,
for
this
to
digit
up
with
free
parameters,
we
test
the
five
thousand
seven
thousand
five
hundred
and
ten
thousand
kilometers
wavelength
and
for
the
damping
time,
our
strands.
M
We
tested
two
four
six
and
eight
days
the
meantime
skill
so
to
run
a
new
scan
digital
simulations.
We
will
initialize
assimilations
used
in
mean
profile
temperature
moisture
from
previous
our
stage
simulations,
and
we
first
run
this
new
skin
digital
simulations
for
two
hundred
days
without
teacher
type
of
coupling
but
which
were
reproduced
in
the
RC
state.
Then
we
take
a
last
100
day.
Temperature
moisture
mean
as
the
reference
day
for
activating
the
HW
coupling
and
run
a
couple
systems
for
another
two
hundred
days.
M
So
this
is
the
time
series
of
energetics
shown
here
and
left
case.
We're
looking
at
here
is
a
stem
Phi
with
ten
thousand
kilometer
wavelength
and
the
sixth
day
10
min
timescale,
and
we
will
see
that
the
wave
gradually
grows
wasting
the
first
60
days
after
the
activation
of
a
digital
coupling
and
there's
a
bias
component
of
excellence.
M
That
means
that
means
superposed,
angular,
wave
disturbance,
this
and
waste
waste
speed
is
approximately
fourteen
point
five
meters
per
second,
and
we
see
that
the
tilted
vertical
structures
of
a
vertical
motion
and
summer
dynamics
did
resemble
that
of
the
west
were
propagating
counterkill
Calvin
waves.
So
this
is
the
mirror
image
that
apparel
as
a
commodity
causal
carbon
ways
from
Clara's
thatö
in
two
thousand
nine.
M
So
that's
the
case
we
see
he
is
just
one
case
if
we
spend
our
simulations
across
different,
wavelengths
and
different
time
in
time
skill,
so
each
subset
of
panel
has
is
composed
of
x,
theory
of
precipitation
and
and
time
pressure
cross-section
of
vertical
velocity
pressure
velocity.
So
we
see
that
cop
assistance
with
longer
wavelengths
tend
to
have
that
characterized
longer
Pierce
oscillations
as
a
feature
of
number
stiffness
and
wave
instability
generally
in
hands
as
the
friends
of
damping
decreases
and
we,
the
bias
of
components.
M
Essence
I
mentioned
earlier-
exist
in
almost
all
s
rungs,
so
that
was
scan
v
and
if
we
look
at
the
stem
space
again,
this
is
just
a
case
of
a
ten
thousand
kilometer
in
a
six
day
time
in
time
scale
as
a
for
a
closer
look,
and
we
feel,
like
features
generally
resemble
that
of
like
invited
upper
case,
but
we
notice
a
stronger
bias
of
assessment
accents
as
here
in
this
plot.
We
need
is
a
larger
scale
in
magnitude
and
and
there's
also
features
the
faster
face.
M
Speed
face
be
nice,
approximately
seventeen
point
four
meters
per
second
again,
we
spend
across
the
different
wave
lands
in
the
Jim
Beam
time
scale.
We
see
that
generally
the
waves,
the
instability
stronger
than
the
skin
five
digit
up
you
last
bets
of
simulations-
and
this
is
probably
due
to
the
already
strong
and
trim
intrinsic
variability
of
skin-
sticks
Asli
e
RS.
We
can
see
here
to
compare
to
the
previous
skin
fight
RC
case
here
and
also
the
couple
systems
with
shorter
wavelengths.
M
M
So
that
what
curious
wide
led
to
an
RC
difference
in
between
skin
fine
skin
sticks
and
there's
a
key
parameter,
that's
called
the
enum
themes,
which
is
the
number
of
negative
buoyancy,
regions
that
are
allowed
for
determining
the
commission,
top
and
cable
might
fall
and
decommission
scheme.
This
is
this
parameter
is
set
to
5
in
C
and
Phi
and
1
in
game
sticks.
M
So,
finally,
we
would
like
to
see
if
that
scammed,
each
W
configurations
could
serve
as
a
tool
for
predicting
the
3d
chem
simulation.
So
here,
I
show
just
yodok.
Our
planet
can
using
the
Euler
internal
core
and
the
t85
horizontal
resolutions
with
602nd
model
x
depth,
and
here
I
only
showed
us
the
magic
component
of
liquidity,
lattice
diagram.
This
is
a
precipitation
from
30
north
to
30,000
degrees.
M
K
N
M
K
O
Is
it
fullscreen?
Yes
great?
So
thank
you,
Christian
and
and
WG
committee
for
allowing
me
to
speak
today.
My
name
is
Alan
Rhodes
and
I'm,
a
research
scientist
at
Lawrence,
Berkeley,
National
Lab
in
the
earth
and
environmental
sciences
area
and
I'm
working
on
both
the
hyper
facets
and
Cascade
project
and
today
I'm
going
to
introduce
some
work.
That's
been
ongoing
for
the
last
year
and
a
half
with
my
co-authors
and
I.
O
We're
all
highlighted
here
to
discuss
some
verbal
resolution,
CSM
work
where
I
won't
be
able
to
go
over
this
in
detail
and
the
current
presentation.
But
we
looked
at
how
the
variable
resolution
refinement
domain
over
say.
The
North
Pacific
might
influence
atmospheric
River,
behavior
and
downstream
impacts
on
the
western
US
hydro
climate.
But
today
I
decided
to
focus
on
our
follow-up
study,
which
is
currently
in
review
and
geophysical
research.
Letters
and
this
looks
at
an
RCP
8.5
future
a
10
century
and
assesses
how
landfalling
a
Arceus
characteristics
in
the
western
US
might
change.
O
O
At
Scripps,
which
tries
to
akin
to
the
hurricane
scales
associated
an
atmospheric
river
or
devise
flavors
of
atmospheric
rivers
from
category
1
through
category
5,
category,
1
being
primarily
beneficial
to
water,
resource
managers
and
category
5
being
particularly
impactful,
and
so
a
ARS
account
in
this
study,
where
associated
with
84%
of
all
flood
insurance
losses
in
the
western
US
states
and
over
51
billion
dollars
in
the
last
40
years.
And
you
can
see
that
on
the
right
year,
there
was
11
cat
5
events
in
these.
O
The
median
of
these
events
was
260
million
dollars
in
flood
insurance
loss
and
could
reach
up
to
a
billion,
whereas
the
cat,
once,
although
there
are
some
that
reach
beyond
100
million
dollars
in
flood
insurance
loss,
primarily
they're,
very
beneficial,
and
so
there's
a
natural
question
of
how
atmospheric
River
character,
and
particularly
two
categories,
might
change
in
the
future.
So
I
would
point
everyone
to
Ashlee
Payne's
new.
O
The
west
us,
so
they
don't
necessarily
represent,
are
at
a
sufficient
resolution
to
represent
topographic
influences,
such
as
orographic
uplifts
on
these
air
events,
and
so
a
lot
of
my
work
over
the
last
five
or
so
years
has
been
in
using
the
variable
resolution.
Capabilities
in
the
community
or
system
model
and
again
just
want
to
highlight
that
I
don't
have
time
today
to
go
over
there
hanyan
paper,
but
it
was
recently
accepted
in
a
jgr.
O
A
special
issue
on
atmospheric
rivers
and
the
QR
code
is
posted
here,
but
this
is
an
example
of
the
grid
that
was
used
for
the
N
century
simulation.
But
in
this
companion
paper
we
also
evaluated
how
the
refinement
domain
size
right
here
over
a
whole
entire
North
Pacific
versus
a
2/3
of
the
Pacific
one-third
of
the
Pacific
might
influence
a
our
behavior
and
integrated
vapor
transport
and
then
the
downstream
impacts
on
snowpack,
surface
temperature
and
precipitation.
O
But
our
other
studies
which
are
highlighted
below
have
also
assessed
more
on
the
horizontal
resolution
of
ER
CSM
microphysics
in
land,
surface
model
dependence
and
so
to
give
some
details
on
our
experiment.
We
ran
all
of
our
simulations
at
nurse
on
Cory
has
well
using
an
a
map
style
simulation
where
we
coupled
that
mr.
and
land
surface,
and
then
we
prescribed
ocean
sea,
surface
temperatures
and
sea
ice
at
monthly
steps,
I'm,
highlighting
that
all
the
non
default
settings
that
we
use
for
our
simulations.
O
So
we
use
km
5.4
spectral
element
dynamical
core,
which
enables
variable
resolution
capabilities.
We
also
had
a
four
times
more
frequent
physics,
time
step
and
tau
was
set
at
two
times
the
frequency,
and
then
we
leveraged
mg,
two
which,
at
the
time
when
we
did
these
simulations,
was
the
the
newest
version
of
the
microfiche
scheme.
We're
excited
to
see
mg
3
and
the
use
of
grapple,
inhale
and
and
I'd
like
to
eventually
evaluate
that.
O
But
importantly,
we
use
CLM
5
satellite
phenology,
because
there's
primarily
a
lot
of
advances
in
the
snow
hydrology
that
we
thought
might
be
important
to
incorporate
in
these
simulations.
And
then
we
used
the
CMAT
5
emission
scenarios
because
we
ran
these
about
a
year
and
a
half
ago
and
the
other
scenarios
were
not
quite
yet
available
and
we
used
similar
to
a
previous
study.
In
2018.
We
used
a
similar
framework
where
we
used
full
bias,
corrected
fully
coupled
CSM
one
simulation
for
SSTs
and
sea
ice,
and
so
just
to
show
you.
O
We
ran
our
historical
simulation
from
1984
to
2015
and
our
future
from
2069
to
2100.
Under
our
CP
8.5,
we
took
the
first
year
as
spin-up,
and
then
we
outputted
monthly,
daily
and
importantly,
six
hourly,
which
we're
going
to
look
at
four
atmospheric
rivers,
and
then
we
also
compared
to
ER
a
five
and
a
few
other
reanalysis
products
in
the
in
the
companion
paper
that
I
mentioned
and
briefly
in
this
follow-up
paper
and
URL.
But
I
just
wanted
to
first
discuss
our
framework
or
methods
that
we
use.
So
we
leveraged
tempest
extremes.
O
You
can
take
all
of
these
ARS
that
occurred
across
a
given
season
or
annual
cycle
and
accumulate
them
so
that
you
can
evaluate
the
climatology
and
there's
a
set
of
metrics
such
as
atmospheric
River,
lifetime,
landfall
location,
duration,
etc.
That
is
also
accumulated
in
this,
and
so
for
this
GRL
article.
O
O
It's
and
there's
actually
a
quite
a
few
unprecedented
ARS
that
start
coming
off
the
Ralpha
tall
scale
quite
shocking,
and
now,
if
you
look
at
the
seasonality
of
they
are
switch,
the
water
years
are
shown
on
the
right
here
from
October
through
September,
and
these
are
30-year
totals
in
terms
of
counts.
You
can
see
in
the
historical
period
we
had
quite
a
few
cat
fives
about
15.
O
O
And
now
so
we,
although
AR
counts
in
their
their
seasonality,
are
important
to
look
at
particularly
for
seeing
if
vr
CSM
is
getting
the
right
timing,
the
number
of
events
etc
compared
to
reanalysis,
but
really
what
it
comes
down
to
is
that
were
interested
in
their
storm.
Total
precip
and
particularly
splitting.
Those
across
categories
which
shown
here
is
the
total
thirty-year
precipitation
across
those
three
coastal
western
US
states,
as
I
mentioned
before,
split
between
non
AR,
which
is
the
white
bars
and
then
each
category
cat
one
through
five.
O
And
so
what
we
can
see
is
that
when
you
compare
1985,
2015
30-year
totals
to
2070
the
2100
that
there
is
quite
a
lot
more
total
precipitation
across
most
months.
Although
there
is
a
little
bit
of
a
sharpening
and
intensification
in
the
kind
of
mid
winter
months
and
more
of
that
precipitation
by
n
century
is
associated
with
AR
events
and
particularly
cat
4
and
cat
5
AR
events-
and
this
is
not
offset
by
a
diminishment
in
the
cat,
1
and
katsue
events,
as
was
indicated
previously
and
I.
O
Don't
have
a
ton
of
time
to
go
over
this.
But
what
was
interesting
is
that
the
historical
simulation
and
the
N
sensory
stimulation,
if
you
were
to
compare
the
maximum
storm
totals
between
between
events
in
those
two
periods,
they're
actually
comparable
to
one
another.
So
what
we
are
finding
is
that
all
those
there's
an
increase
in
total
precipitation
totals
in
the
future.
This
is
primarily
due
to
more
land
following
ARS
that
are
particularly
cat,
4
and
cat
5
and
occur
back-to-back.
O
Cat
v
is
actually
possible
across
most
of
the
water
year
months
and
there's
a
significant
shift
in
our
category
from
mostly
or
primarily
beneficial
to
mostly
a
primarily
hazardous
according
to
the
Ralph,
a
tall
scale
and
a
ten
century
there's
a
net
increase
of
twenty
two
percent.
In
the
thirty
year,
precipitation
totals,
or
six
hundred
six
thousand
four
hundred
millimetres,
predominantly
from
a
our
precipitation,
which
is
interesting
and
no
single
end
century
cat5
air
event,
precipitation
total
was
historically
unprecedented.
O
However,
due
to
the
increased
number
of
cat
four
and
five
landfalling
ARS,
particularly
the
back-to-back
ones,
the
monthly
totals
are,
and
so
with
that.
I
would
just
like
to
again
think
the
am
GG
committee
for
allowing
me
to
speak
today
acknowledge
my
funding
through
the
hyper
facets
and
cascade
projects
and
I'd
be
happy
to
take
on
any
questions.
O
K
O
O
Maybe
some
of
the
physical
reasons
might
be
that
the
ARS
are
there's
more
moisture,
so
there
can
be
more
long-lived
and
and
therefore
more
can
make
landfall
in
the
western
US
and
then
there's
also
some
discussion
that
we
have
and
kind
of
the
discussion
of
the
paper
discussion
section
of
the
paper.
Sorry
that
discusses
the
idea
of
tropical
widening,
which
allows
the
extra
cyclones,
potentially
in
train
more
moisture
as
well,
but
yeah
I
think
those
would
be
kind
of
two
things:
okay,.
K
O
A
great
question
you
know
we
really
wanted
to,
because
snow
is
kind
of
my
wheelhouse
and
I'm
excited
about
always
analyzing
that,
but
unfortunately,
the
simulations
that
we
ran.
The
snow
representation
wasn't
as
as
good
as
we
had
hoped,
and
so
we
just
imagined
that
the
the
snowfall
and
splitting
you
know
precipitation
into
rain
and
snow
their
third
yeah
I
guess
we
could
look
at
it,
but
we
didn't
have
faith
that
it
was
represented.
Well,
thank.
P
P
K
P
Weird
yeah:
what
do
you
see
now?
Yeah
motivation,
okay,
I'll,
just
keep
it
like
this,
then
so
yeah.
So
tropical
cyclone
landfalls
are
expensive
in
2017
alone,
hurricanes,
Harvey,
Irma,
Maria
costs
about
265
billion
dollars
and
the
TC
hazards,
such
as
flood
storm
surge
and
high
winds
that
caused
this
damage
and
also
are
dangerous
to
humans
are
impacted
by
climate
change.
So
we
want
to
understand
how
these
hazards
will
change
in
the
future.
However,
tropical
cyclones
still
are
not
fully
understood.
P
P
So
the
initial
work
looking
at
tropical
cyclones
in
variable
resolution,
cam
was
done
by
Collins
or
Zaki
and
co-authors.
So,
just
to
sum
quickly
summarize
some
of
the
work
they
did.
They
found
that
there
were
no
numerical
issues
when
tropical
cyclones
pass
through
the
transition
regions
in
the
variable
resolution
grids
and
that
the
tropical
cyclones
in
the
cam
and
cam
variable
resolution,
or
nearly
identical
to
those
found
in
global
high
resolution
simulations
and
additionally,
they
found
in
multi-decade
low
simulations
of
North
Atlantic
tropical
cyclones
that
the
TC
counts
and
spatial
patterns
were
similar
to
observations.
P
So
this
work
on
presenting
it's
kind
of
just
an
extension
of
the
work
that
has
been
done
before
specifically
I
want
to
look
at
how
far
out
into
the
Atlantic.
The
high
resolution
domain
must
extend
to
represent
TC
climatology.
With
specific
interests,
see
see
precipitation
over
the
eastern
United
States
and
then
also,
how
do
you
North
Atlantic
tropical
cyclones,
change
under
climate
change,
forcing
so
just
to
quickly
talk
about
the
model
simulations?
P
We
looked
at
three
variable
resolution
grids
shown
below
so
the
western
atlantic
grade
or
the
watt
grid,
had
the
smallest
high-resolution
domain,
with
28
kilometers
spacing
just
over
the
western
Atlantic
and
111
kilometers
spacing
elsewhere.
The
refere
reference
grid
had
the
high
resolution
over
the
entire
North
Atlantic
and
the
extended
or
ext
grid
had
the
is
the
same
as
the
rough
grid,
except
there's
an
additional
high
resolution
domain
over
northern
Africa,
which
may
help
in
representing
African
easterly
waves.
P
So
we
performed
thirty-year,
Amite
style
simulations
using
cam
5,
and
we
had
three
and
Samba
members
for
each
configuration.
Oh
and
we
also
ran
a
global
one
degree
conventional
grid
to
compare
the
variable
resolution
grids
to,
and
so
a
novel
part
of
this
work
was
the
TC
tracking
and
precipitation
extraction
methodology
that
we
did
with
the
updated
tempest
extreme
software,
and
so
the
way
that
this
works
is.
We
first
identify
potential
tropical
cyclones
based
on
sea
level.
Pressure
minima
and
warm
core
characteristics
in
the
model
output.
P
Then
these
potential
tropical
cyclone
points
are
stitched
together
into
tropical
cyclone
paths
and
then
at
each
at
each
model.
Time
step
shown
by
these
dots.
Here
we
compute
a
radial
wind
profile,
which
an
example
is
shown
here
in
step
3
and
from
this
radio
wind
profile.
We
find
the
point
outside
the
eye
wall,
where
the
wind
speed
is
equal
to
8
meters
per
second.
We
call
that
re,
and
that
is
what
we
use
to
define
the
outer
storm
size
and
so
then
to
extract
the
precipitation
associated
with
the
TC.
P
Take
this
re
draw
a
circle
around
the
center
of
the
tropical
cyclone
and
then
all
precipitation
within
that
are
eight
circle
is
recorded
as
the
tropical
cycle
ated
precipitation.
So
this
is
cool
because
the
storm
size
actually
dynamically
changes
with
the
TC,
so
T
C's
are
known
to
change
sizes
throughout
their
lifetime.
So
this
kind
of
tracks
this
and
extracts
the
precipitation
based
on
the
size
at
that
time.
So
this
tracking
and
precipitation
extraction
was
done
on
the
model.
P
Simulations
and
I
just
talked
about
and
then
also
ER
a
five
reanalysis
data,
so
just
to
show
some
results
here
we
have
the
TC
track
density,
and
so
the
track
density
is
defined
here
as
the
hours
per
year
that
each
grid
point
was
within
the
TC
sr8.
So
we
used
the
r8
that
I
just
talked
about
to
kind
of
define
the
the
track
density,
and
so
here
are
the
three
variable
resolution
grids.
The
glob
is
the
global
1
degree,
your
f5s,
the
reanalysis
and
then
ib
tracks
our
observations
and
so
right
off
the
bat.
P
We
see
that
all
of
the
models
and
the
ERI
5
underestimate
the
track
density
over
the
eastern
United
States.
This
is
you
know.
This
suggests
that
not
enough
tropical
cyclones
are
making
landfall
in
the
models,
but
we
do
see
that
compared
to
the
global
1
degree
grid.
The
variable
resolution
version
certainly
do
better
than
the
global
1
degree
and
in
even
in
some
cases
they
even
do
better
than
the
ER,
a
5
reanalysis
for
the
track
density.
P
Looking
at
a
distribution
of
the
outer
sizes.
So
this
is,
these
are
just
normalized
histograms
of
our
eight
and
so
I
won't
spend
too
much
time
on
this,
but
just
to
point
out
that
the
3
VR
distribution,
so
the
the
blue,
red
and
green,
are
similar
to
the
ER
a5,
which
is
the
black
distribution.
So
we
see
that
they
kind
of
they
are
maximized
around
the
same
r8
and
they
kind
of
are
consistent,
and
so
because
the
variable
resolution,
TC
sizes,
are
generally
consistent
with
those
in
a
RI
v.
P
So
here
we're
just
looking
at
some
of
the
extreme
precipitation
results,
so
the
metric
that
we
use
is
rx
5-day,
which
is
the
annual
maximum
five-day
precipitation,
total
and
so
here
on
the
left.
This
is
what
we
consider
observations.
So
what
we
did
here
is
we
combined
the
ER,
a
5
tropical
cyclone
tracks
with
an
observed
precipitation
product
from
the.
P
So
this
just
to
say
that
the
this
observations
are
actually
on
a
bit
underestimated
compared
to
reality,
since
a
ra.5
doesn't
produce
as
many
tropical
cyclones,
but
the
patterns
are
generally
consistent
in
minute.
Thank
you.
So
again
we
see
here
that
the
variable
resolution
versions
and
the
model
definitely
do
better
than
the
global
1
degree,
because
they
simulate
more
landfalling
tropical
cyclones
and
then
comparing
them
to
observations.
P
P
So,
moving
on
to
the
future
climate
simulations,
we
determined
by
looking
at
a
lot
of
metrics
that
the
ref
grid
was
the
best
compromise
between
computational
cost
and
representation
of
TC
climatology.
So
we
used
the
ref
to
run
our
CP
4.5
and
8.5
simulations
for
2070
to
2100.
Again
we
did
three
ensemble
members
per
simulation
and
the
results
were
generally
consistent
with
the
consensus
on
future
TC
climate
projections.
P
P
So
this
is
just
showing
that,
on
an
annual
average,
we
see
a
more
rainfall
per
hour
of
storm
impact
in
the
RCP
runs
and
to
explore
that
a
little
bit
for
we
just
looked
at
the
probabilities
of
different
rainfall
intensity
rates.
So
here
you
can
just
focus
on
the
dashed
lines,
because
the
dash
lines
are
TC
precipitation.
Only
so
looking
in
this
purple
box
here
at
the
most
extreme
precipitation
rates,
we
see
that
the
two
are
CP
simulations.
P
The
green
and
the
red
have
a
higher
probability
of
more
extreme
precipitation
rates
than
the
Amite
pref
simulation,
which
at
least
partially
explains.
Why
are
seeing
more
precipitation
per
hour
of
TC
impact
in
those
simulations
so
to
quickly
go
through
my
conclusions.
The
variable
resolution
versions
of
kam
simulate
north-atlantic
TC
climatology.
Definitely
better
than
the
one
degree
global
cam
and
are
also
comparable
to
observations.
However,
they're.
Definitely
too
few
T
C's
making
landfall
in
the
u.s.
P
in
the
models
compared
to
observations
which
results
in
a
low
bias
in
TC
precipitation
in
the
RCP
simulations
less
TCS
make
landfall
in
the
eastern
United
States,
so
the
annual
average
TC
precipitation
decreases.
However,
we
see
that
the
TC
precipitation
per
hour
of
TC
impact
does
increase
and
so
in
the
future,
I
hope
to
use
idealized
cam
configurations
to
explore
these
TC
changes
a
little
bit
more
in
a
controlled
setting,
and
so,
if
there's
time,
I
can
take
any
questions
on.
P
K
E
C
R
C
Q
Your
speaker,
anyways
I'm,
gonna,
move
on
to
the
next
slot
I
and
so
I
carried
out
the
historical
simulations
with
a
bunch
of
grids
in
available
in
CSM
to
point
to
find
a
fine
2
degree,
fine
Devon,
one
degree
SEC
slam,
one
degree
SEC
slam,
one
degree
with
coarser
resolution
physics
and
then,
of
course,
the
Arctic
and
the
Arctic
crisp
grids
random
from
1979
to
1998.
With
you
know,
time
varying
SST
is
ghd's
or
their
souls.
This
is
the
F
hiss
concepts.
Q
All
these
will
run
out
out
of
the
box
of
CSM
to
point
to
CLM
in
satellite
mode
sysm.
The
ice
sheet
model
is
in
no
evolved
mode,
and
what
that
means
is
that
the
land
model
will
evaluate
surface
mass
balance
at
different
elevation
classes,
hypothetical
vertical
levels
and
use
that
field
to
downscale
the
surface
mass
balance
to
the
high
resolution,
four
kilometer
sysm
grid,
even
though
it
is
not
even
used
to
drive
the
ice
sheet
model,
it
gives
you
a
high
resolution
of
representation
of
the
surface
mass
balance
for
the
variable
resolution
grids.
Q
I
wanted
to
spin
up
the
snow
depths
and
the
surface
mass
balance
from
scratch,
and
so
here's
the
way
you
do
it
so
I
ran
in
the
Arctic
grid
for
20
years
with
constant
1979
boundary
conditions.
But
at
the
beginning
of
the
simulation
you
zero
out
the
snow
and
when
you're
running
for
20
years,
you
output,
high-frequency,
coupler
output
that
can
use
to
force
CLM
offline
by
cycling
over
that
20-year
forcing,
and
it
takes
about
five
hundred
years
to
bring
the
snowpack
into
equal
equilibrium.
And
that's
illustrated
here.
Q
This
is
the
global,
mean
snow
depth
versus
time,
and
this
is
how
long-
and
this
is
what
the
value
looks
like
in
this
time.
Scale
about
500
years
is
governed
by
the
fact
that
it
takes
it's
the
amount
of
time
it
takes
for
grid
cells
with
perennial
snow
to
reach
the
10-meter
snow
depth
cap,
at
which
point
it
can
start
contributing
positively
to
the
surface
mass
balance
and
then
another
relevant
thing
that
is
more
relevant
to
the
other
resolutions.
But
prior
to
see
us
I'm
2.2.
Q
The
clm
model
would
use
a
two
degree,
initial
condition,
file
and
map
it
to
your
target
grid
at
initialization
and
if
you
use
this
is
a
global
quarter.
Degree
simulation
that
I'm,
showing
here
for
the
first
month
of
a
simulation,
and
you
can
see
these
snow
depths.
You
can
see
the
aliasing
from
the
two
degree,
initial
conditions
in
the
quarter,
degrees
solution
and
so
to
get
rid
of
this
ugly.
Q
Q
And
I'm
not
gonna
have
enough
time
to
go
over
everything.
I've
only
had
about
a
month
to
look
at
these
simulations
and
so
I'll
provide
you
guys.
Some
highlights
just
quickly
the
topography.
Obviously
your
resolving
things
in
the
Arctic
Chris
in
the
Arctic
red
that
aren't
resolved
in
the
lower
resolutions,
and
then
it's
always
nice
to
just
have
a
look
at
what
the
landmaster
the
ice
mass
looks
like
in
the
GCM
and
clearly
you
can
see
a
better
representation
at
higher
resolution,
and
so
a
quick
highlight
on
some
interesting
meteorological
features.
Q
This
is
a
stream
function
of
the
lowest
model
level
in
finite
volume.
One
degree
here
and
then
the
Arctic
risk
right
over
here
and
what
I
would
like
to
point
out
is
just
these
Reds
that
show
up
around
the
steep
margins
that
indicate
that
there's
you're
starting
to
resolve
these
katabatic
drainage
wedge
winds
because
they
can
be
accelerated
by
these
now
resolve
steep
margins,
and
so
there's
some
science
to
be
done
with
that
and
working
on
a
visualization
with
sizzles.
Well
to
really
highlight
that.
Q
So
to
prep
this
some
of
the
results
I
just
want
to
go
over
in
the
spirit
of
model
hierarchies,
some
solution,
some
resolution
sensitivity
in
three
plots
that
I
is
from
my
PhD
thesis
and
so
I
did
convergence
experiments
with
aqua
planets
from
any
twenty.
That's
about
one
and
a
half
degree
to
any
one.
Q
That
is
scaling
with
the
vertical
velocities.
And
so
you
know,
larger
vertical
velocity
is
at
higher
resolution.
You'd
expect
more
condensation
and
that
Illustrated
here,
which
is
a
difference
plot
from
any
28
any
40,
and
this
column
is
the
club
tendency
and
I'm
conditionally
sampling.
Here.
All
grid
columns
that
are
ascending
and
you
can
see
an
increase
in
conversational
heating,
with
increased
resolution
in
the
ITCZ
and
the
Jets
and
the
descending
regions.
Q
You
don't
see
any
condensation
heating,
which
is
again,
of
course,
due
to
the
scaling
here,
and
so
this
increased
condensation
on
heating
X
to
warm
the
free
troposphere,
the
mean
state.
So
you
can
see
that
signal
both
the
ascending
and
the
descending
brick
columns
and
that
tends
to
stabilize
the
atmosphere
shut
off
your
deep
convection
scheme.
So
if
we
look
at
just
these
lower
resolution
runs,
for
example,
this
is
the
going
from
2
degree
to
1
degree.
Q
So
now
just
thinking
about
okay
well
going
from
any
32
arctic
and
we're
just
going
to
look
at
jja.
What
does
this
look
like?
And
so
this
is
the
plot
going
from
a
30
arctic
and
you
have
condensation
on
heating
not
only
increasing
in
the
refined
region,
which
is
north
of
where
my
mouse
is,
but
you
also
have
an
increased
conversational
heating
in
the
course
region,
and
this
is
an
unfortunate.
Q
Q
That's
going
out
here
that
you
see
over
here
and
the
difference
and
then
just
doing
the
Arctic,
the
any
30
with
the
higher
physics
time
stuff,
you're,
isolating
the
impact
of
horizontal
resolution
only
and
so
you're
only
seeing
that
greater
condensation
on
heating
in
the
refinement
zone
and
you
get
the
warming,
that's
more
local
to
the
refinement
zone.
Ten.
K
Q
This
is
just
illustrating
your
stationary
waves,
where
you
have
this
anticyclone
upstream
of
the
Rockies,
and
then
this
low-pressure
in
the
Lee
of
the
Rockies,
and
wait
this
wave
activity
flux
from
plum.
This
is
overlaying
on
there,
and
so,
if
I
do
a
difference
plot
this
Leeside
low
over
here
becomes
much
more
red
and
so
you're
reducing
it's
becoming
more
anti
cyclonic
and
and
the
wave
activity
fluxes
are
becoming
weaker,
and
this
results
in
anomalous.
Southerly
winds
over
the
east
Canada
and
anomalous
North
Elise
over
western
Greenland.
Q
That
is
consistent
with
the
warming
you
see,
and
so
our
hypothesis
is
that
this
is
due
to
somehow
resolving
the
Greenland
ice
sheet.
Somehow
facilitates
greater
blocking
events
in
this
region.
Here
that
is
contributing
to
breaking
down
the
stationary
low.
You
see
here
and
we
I
had
computed
some
blocking
I
used
the
blocking
detection
algorithm.
One
of
them
that
is
available
in
tempest
extremes
is
e500
anomaly,
method
and
I
computed.
It
ARCIC
Gris
against
all
the
other
five
lower
resolution
grids
and
the
reason
I
did.
Q
This
is
because
it's
kind
of
an
ensemble
itself
I
guess
it
is
an
ensemble,
and
so
you
need
greater
confidence
of
the
signal
that
you're
seeing
is
due
to
increased
resolution
and
in
all
cases
you
get
increase
in
blocking
frequency.
Great,
where
that
warming
is
occurring,
that's
sort
of
west
of
Greenland
in
all
cases,
and
so
that's
consistent
with
the
hypothesis
that
the
greater
resolution
over
Greenland
is
somehow
facilitating
greater
blocking
events
and
then
just
one
little
sneak
peak
on
the
Greenland
surface.
Q
Mass
balance
I'll
be
presenting
this
tomorrow
at
the
la
land
ice
working
group,
but
these
are
observations.
The
Blues
are
Institutes,
it
scores
and
states.
The
whites
are
radar
from
ice
bridge
and
so
the
radar
from
ice
bridge,
which
is
emotionally
in
the
accumulation
zone,
I,
basically
I
find
the
nearest
CLM
point
there
and
they
do
a
difference
and
then
I
do
a
PDF
of
all
the
differences
and
then
I
flipped
the
PDF
vertically,
and
draw
a
mirror
image
of
the
PDF
to
close
out
the
polygons.
Q
This
is
a
raw
type,
violin
type
plot
and
with
the
different
resolutions
here,
a
nice
bridge.
You
generally,
when
you
go
to
f19
fo9,
you
tend
to
shrink
the
distribution
and
get
a
better
agreement
going
from
PG
2
to
PG
3,
it's
more
or
less
the
same
and
then
going
to
Arctic.
You
do
improvement
and
this
does
the
the
levert
achill
I'm.
Sorry,
the
horizontal
line
in
the
middle
is
the
median.
So
it's
lining
up
pretty
well
with
zero
point
and
then
the
Arctic
Chris
does
as
good
or
better
than
the
Arctic.
Q
In
terms
of
the
differences
with
observations,
the
institution,
the
accumulation
zones
are
a
similar
story
where
you're
thinning
out
the
distribution
getting
closer
to
zero.
With
resolution
p2
p2
and
then
Arctic
NARR
Chris
in
the
ablation
zone,
you
always
will
have
larger.
Biases
melting
rates
are
typically
larger
than
accumulation
rates,
but
nonetheless
you
do
see
an
improvement
going
from
finite
volume,
2
degree
to
1
degree,
PG,
3,
2,
PG,
I,
don't
know
that
could
be
more
or
less
similar,
but
you
do
see
a
clear
improvement
going
to
Arctic
and
then
in
the
Arctic
Chris
grit.
Q
Q
C
Q
N
You
see
my
slides
okay,
thanks
so
I'm
going
to
talk
about
this
investigation.
That's
been
underway
to
look
at
our
options
for
the
vertical
resolution
for
the
next
workhorse
version
of
CES
M.
So
this
has
really
been
a
team
effort
involving
everyone
here
and
I'll,
give
acknowledgement
to
individual
figures
as
we
go
along,
but
I'm
also
not
able
to
show
everything
that
everyone
has
done
towards
this.
N
So
the
task
here
that
was
kind
of
laid
out
by
the
SSC
a
couple
of
months
ago
was
that
we
should
investigate
grid
configurations
for
the
next
workhorse
version
of
CES
M.
So
the
idea
is
this
would
be
the
model
that,
for
example,
be
run
for
cement
7.
It
would
be
released
to
the
community
in
CSM
3
and
should
be
the
grid
that
we'd
expect
to
use
for
the
next
five
to
ten
years.
N
So
this
would
be
a
model
that
doesn't
extend
as
high
as
Wacom,
but
it
extends
higher
than
cam
and
it
has
a
grid
structure
that
has
improvements
in
vertical
resolution
and
the
free
troposphere
and
stratosphere
and
in
the
boundary
layer
to
try
and
capture
certain
features
of
interest.
So
there's
still
be
other
configurations
like
Wacom
would
still
be
built
on
top
of
this,
but
the
idea
is
that
this
would
be.
The
primary
model
would
use
for
exercises
like
scene
it.
N
So
we
were
asked
to
recommend
a
specific
top
height
a
number
of
levels
and
also
come
up
with
a
lower
top
configuration
that
could
be
used
for
a
cheaper
simulation
and
cheating
purposes,
and
we
definitely
think
the
last
of
these
is
important,
but
we
haven't
done
that
yet
so
here
we're
really
talking
about
the
top
height
and
the
number
of
levels.
So
I'm
going
to
show
figures
like
this
a
lot.
N
This
is
showing
the
vertical
grid
spacing
on
the
x-axis
and
then
the
height
of
the
levels
on
the
y-axis,
and
here
you
have
Wacom
the
Wacom
grid
and
the
cam
grid
for
a
reference,
and
so
some
of
the
motivations
are
that
one
is
that
we
currently
run
both
Wacom
and
cam
for
seen
it.
So
camera
is
kind
of
the
workhorse,
but
we
do
run
Wacom
when
we
run
it
first
to
get
with
couple
chemistry
to
get
forcing
fields
that
are
used
to
them
for
scat.
N
So
ideally,
we'd
like
to
skip
that
two-step
process
and
have
one
model
that
we
use
for
that
kind
of
exercise.
You
might
ask
well
why
don't
we
just
use
Wacom,
but
Wacom
is
kind
of
also
in
need
of
some
upgrades.
So
it's
your
toward
middle
atmosphere,
applications,
but
the
resolution
here
in
the
troposphere
and
the
lower
stratosphere
still
isn't
really
enough
to
adequately
represent
the
QBO,
which
is
something
we'd
like
to
have
represented
in
the
model
and
so
I'm
just
demonstrating
that
here.
N
This
is
showing
zonal,
mean
zonal
winds
from
five
sites
to
five
north
and
your
a5
reanalysis,
and
then
in
the
current
Wacom
configuration,
and
you
can
see
that
the
westerly
phase
of
the
QBO
it
doesn't
extend
as
far
down
into
the
lower
stratosphere
in
Wacom
as
it
does
in
year.
Five.
And
so,
if
you
want
to
capture
QBO
impacts
on
the
troposphere,
you
kind
of
want
the
commute
to
get
down.
As
far
as.
N
N
So
if
you
want
to
run
initialize
predictions
and
you
want
to
have
a
representation
of
the
stratosphere,
you
you're
more
likely
to
want
a
model
talk,
that's
around
80
kilometers,
where
you
can
have
some
representation
of
the
stratosphere,
but
you
can
still
initialize
with
reanalysis
products,
and
then
there
are
many
reasons
to
increase
resolution
in
the
boundary
layer
and
rich
talked
about
some
of
those
reasons
this
morning
and
I
listed
many
here.
I
won't
won't
go
through
each
of
these
in
turn,
but
you
kinda
get
the
idea.
N
You'd
expect
that
you'd
improve
fine
scale
features
in
the
clouds
and
input
improve
your
temperature,
humidity
profiles
and
your
near
surface
winds
and
surface
fluxes
and
so
on.
So
the
kind
of
desired
features
of
the
model
are
that
we
like
to
lower
the
lowest
model
level
time
to
buy
10
meters,
increased
resolution
in
the
boundary
layer,
probably
about
a
doubling
of
the
resolution,
improve
the
free,
troposphere
and
stratosphere
resolution
enough
to
represent
the
QB
oh
well,
and
then
have
a
model
top
that's
kind
of
somewhere
in
the
middle
between
Wacom
and
cams.
N
So
partly
because
we'd
like
to
reduce
computational
expense,
we're
going
to
be
adding
in
a
lot
below.
So
we
need
to
kind
of
limit
it
somehow
and
then
also
for
the
initialization
issue
that
I
mentioned
before
so
so
far.
We
really
focused
on
the
free
troposphere
and
the
stratosphere
resolution,
and
that's
what
I'll
talk
about
here,
but
we
kind
of
do
have
an
estimate
of
the
plans
for
the
boundary
layer
that
it
would
amount
to
about
an
extra
10
levels
on
whatever
we
decide.
N
So
first,
we
investigated
what
the
influence
of
the
vertical
grid
spacing.
That
I'll
call
DZ
here
in
the
free
troposphere
and
the
lower
stratosphere
was
so,
if
you
recall,
I
mentioned
there's
this
research
version
of
Wacom,
which
has
110
levels,
and
it
has
a
qbo
that
people
are
satisfied
with,
and
so
this
has
a
grid
spacing
of
500
meters
in
the
troposphere
and
lower
stratosphere.
So
we
took
that
as
their
starting
point
and
we
asked
what
happens
if
you
degrade
it.
N
So
we
systematically
reduce
the
resolution
in
increments
of
a
hundred
meters
dying
to
1000,
leaving
the
model
talk
the
same
so
that
it's
kind
of
a
controlled
test
on
the
resolution
in
the
troposphere
and
the
lower
stratosphere.
These
were
all
run
and
a
matte
mode
would
specify
chemistry
and
they're
20
years
long.
N
N
So
these
are
for
the
cases
500,
600,
700
and
900.
The
black
line
is
showing
the
winds
in
the
in
the
tropics,
and
so
at
these
time
periods
you
have
a
westerly
phase,
that's
maximizing
around
25
kilometers,
and
so
that's
descending
into
the
lower
stratosphere
at
this
time,
and
then
the
blue
and
the
red
lines
are
showing
the
tendency
of
the
zonal
way
and
due
to
the
resolve
and
the
gravity
wave
drag.
N
And
so
what
you
see
is
in
the
500-meter
case
and
the
blue
line
here
is
giving
you
a
westerly
tendency,
so
there's
a
role
for
the
resolve
waves
in
driving
the
descent
of
that
westerly
phase
down
into
the
lower
stratosphere.
The
same
is
true,
but
to
a
lesser
extent,
in
the
600
meter
case,
but
once
you're
at
700
and
Beyond,
you
no
longer
see
a
westerly
tendency.
Do
to
those
resolve
waves,
so
the
qbo,
the
descending
westerly
phase
of
the
qbo
transitions
to
being
only
driven
by
the
gravity
wave,
drag.
N
Having
said
that,
when
you
just
look
at
the
wave
activity
itself,
50
hectopascals
here,
these
are
wave
number
frequency
spectra
of
the
symmetric
mode.
So
the
Kelvin
waves
in
this
blob
here,
which
are
largely
responsible
for
that
descent
of
the
westerly
phase.
They
don't
really
show
a
dependence
on
on
the
vertical
resolution.
So
there's
more
to
understand
here,
it's
going
to
be
something
bit
more
subtle
in
terms
of
the
wave
mean
flow
interactions
associated
with
the
key
vo
that
are
changing.
N
But
when
you
look
at
the
modes
that
are
anti
symmetric
about
the
equator,
there
is
a
clear
dependence
on
the
vertical
resolution
and
the
mixture
of
speed
gravity
waves
show
reduced
power
with
lower
resolution.
Unfortunately,
these
are
not
very
well
constrained.
Observational
E,
as
you
can
see
here,
by
comparing
mariri
analysis
with
ear
at
five,
but
the
difference
here
is
consistent
with
the
resolution
of
the
underlying
model.
Merits?
N
N
So
this
is
the
standard
deviation
of
MgO
filtered
olr,
so
you
have
NGO
activity
over
the
maritime
continent
and
then
each
of
the
configurations
there's
a
bias
in
the
sense
of
reduced
activity,
but
it
kind
of
gets
systematically
worse
at
lower
resolution,
although
there's
not
much
in
it
between
the
500
and
the
600
meter
cases.
So
then
we
look
to
what
happens
if
you
chop
off
the
model
top,
and
so
we
took
a
bunch
of
these
cases,
just
lowered
them
to
80
kilometers.
N
N
So
it
does
seem
like
there's
some
benefits
to
the
500-meter
case,
particularly
when
it
comes
to
the
QBO.
So
if
we
wanted
that,
how
many
levels
would
that
be
well,
if
we
just
took
the
the
straight
up
version
that
we
started
with,
that
would
be
92
levels
and
then
we'd
expect
to
add
on
10,
which
would
make
a
hundred
and
two
level
model,
which
is
a
bit
more
than
we
were
hoping
for.
N
So
we
look
down
and
what
happens
if
you
taper
off
the
resolution,
we
tried
three
different
tests:
tapering
a
twenty
five.
Twenty
and
fifteen
kilometers,
the
main
thing
that
was
sensitive,
there
is
a
QBO
and
so
with,
as
you
taper
lowered
and
the
amplitude
of
the
QBO
gets
progressively
less
and
less,
and
so
really
only
the
25
meter.
Kilometer
case
was
an
option,
so
we're
left
with
that
as
a
possibility
that
would
have
ninety-one
levels,
and
so
it's
may
be
acceptable
and
maybe
a
little
higher
than
what
we
wanted.
N
But
we
do
think
we
should
still
consider
these
cases
as
well,
which
would
reduce
the
number
of
levels,
but
we
would
be
making
compromises
with
the
QBO
and
so
we're
gonna
start
testing
these
in
the
SE
dye
core
and
see
if
the
same
conclusions
hold.
This
was
all
done
in
Fe.
I
probably
should
have
said
that,
okay,
so
to
conclude,
we've
done
lots
of
tests.
I
think
we
can
place
the
model
top
at
80
kilometers,
without
substantial
degradation
and
in
the
troposphere
and
lower
stratosphere.
N
On
top
of
that,
and
we're
currently
testing
that
now
the
500-meter
case
does
seem
to
be
optimal
for
the
qbo,
which
would
leave
us
with
91
91
level
model,
but
we'll
continue
testing
both
the
500
and
the
600
meter
case
before
drawing
firm
conclusions
and
we're
starting
to
work
on
the
boundary
layer
option
as
well.
So
with
that
I'll
take
any
questions
or
any
feedback
thanks
thanks.
K
K
A
A
The
scheme
behaves
the
same
at
all
at
all
resolutions,
except
that
the
drags
are
well
yeah.
Chris
Chris,
just
Chris
just
made
a
good
point
about
tendency
limiters.
There
are,
there
are
tendency
limiters,
which
are
going
to
be
more
active
as
resolution
increases,
because
the
tendencies
may
be
larger,
but
other
than
that
there
is
nothing
explicit.
K
K
Thanks
I
think
we
should
now
move
into
the
discussion
period.
Also.
They
remember
that
we
scheduled
a
discussion
out
at
12:15
and
if,
if
necessary,
we
can
leave
the
room
open
for
longer
if
people
still
want
to
join
the
discussion.
I
think
Julio
wants
to
start
with
a
few
slides
to
engage
everybody
and
to
ask
few
questions,
and
then
we
encourage
us
encourage
you
to.
Let
us
know
whether
you
want
to
talk
and
be
a
chat,
and
we
are
happy
to
on
YouTube,
okay.
A
A
Obviously,
questions
I'm
going
to
go
through
these
slides
first
and
then
we'll
then
we'll
sort
of
circle
back
to
the
top.
First
of
all,
yeah
I
think
obvious
possible
questions.
Are
you
know,
of
course,
if
there
were
questions
for
the
presenters
inputs
on
the
vertical
grid
and
I'm,
particularly
interested
in
any
input,
people
have
on
how
we
go
about
with
high
resolution
in
CSM
and
as
I
will
briefly
touch
on
diagnostics,
asking
Cecile
or
Bryan
maderos
to
chime
in
and
I
want
to
show
I
want
to
alert
people
to
a
SEMA
workshop
and
questionnaire.
A
So
to
remind
you
about
vertical
resume
Norah's
on
ttle
resolution
we
have,
we
have
run
some
horizontal
quarter
degree
runs
with
CSM
physics,
they
show
some.
You
know
depressing
similarities
to
other
runs.
We've
we've
done
at
this
resolution
with
intensified
ITC's
itzz
biases
also
not
shown
our
problematic
wind
profiles
in
tropical
cyclones.
A
So
the
question
is:
do
we
do
we
invest
a
lot
of
effort
in
trying
to
improve
things
in
you
know
CSM
2.1,
essentially
with
its
old,
horizontal
vertical
resolution,
or
do
we
you
know,
wait
until
our
vertical
resolution
effort
is
a
little
more
mature
now
I
will
I
will
ask
Ryan
or
Cecile
to
chime
in
quickly
on
the
Diagnostics
effort?
Are
you
there?
Oh,
yes,.
K
Okay,
then
I
just
say
this
is
the
initial
plan
that
we
present
at
the
atmospheric
working
group
workshop,
and
this
is
to
replace
the
atmospheric
working
group
packages
and
CL
is
going
away.
Then
we
just
put
this
for
reference.
Then
can
you
go
to
the
next
slide?
We
had
an
international
initial
implementation
deadline.
That
was
very
short.
K
It
was
in
April,
and
this
is
they
were
a
lot
of
effort
that
were
done
by
Jessie
and
a
branded
eros
and
now
this
is
basically
what
they
have
done
so
far
and
we
have
a
tool
that
is
working,
it's
still
limited,
but
we
can
take
the
output
of
the
model
and
create
plots
and
again
this
is
for
reference.
The
slide
we
have
here,
I've
done
I
will
not
go
over
it.
Then
let
go
to
the
next
slide
and
we
are
requesting
community
input
on
the
initial
implementation
that
has
been
done.
K
C
A
Right,
thank
you.
Thank
you
very
much
and
yeah.
Please
send
feedback
to
Cecile
or
Brian
I.
Guess
the
seals,
email,
outputs,
Brian's
email
on
there
too,
so
continuing
on
quickly
with
we
would
like
Andrew,
gentlemen,
would
like
some
input
on
on
cam
and
SEMA
the
system
for
integrated
modeling
in
the
atmosphere.
There
are
two
two
surveys
listed
here.
We
ask.
You
know
we'd
like
to
ask
anybody
who'd
like
to
give
their
input
to
please.
Please
do
so,
and
there
is
a
SEMA
community
workshop
occurring
soon.
A
R
So
Julio
this
is
Kevin.
Reid
can
I
ask
a
quick
question,
clarify
something
that
you
were
saying
so
one
of
the
first
things
you
mentioned
was
about
high
horizontal
resolution
and
if
we
should
just
wait
on
that
until
we
have
high
vertical
resolution,
is
that
correct?
Is
that?
What
is
that?
Am
I
interpreting
your
question
correctly?
That's.
A
Essentially,
that's
essentially
the
question.
Yes,
it's
it's
a
lot.
You
know,
as
you
know,
to
tune
up
tune
up.
A
high
resolution
in
the
horizontal
model
will
take
a
lot
of
a
lot
of
person.
You
know
person
hours
and
the
question
is
whether
we
do
that
you
know
with
the
previous
version
of
the
model
or
whether
we
delay
it
until
we
we
have
a
different
vertical
grid,
etc.
Yeah,
so.
R
I
think
so
that
begs
a
question
of
whether
you
know
how
feasible
these
runs
are
gonna,
be
so
I
guess
if
I
was
to
just
give
my
two
cents,
my
two
cents
would
be.
If
we're
gonna
have
a
high-resolution
tuned
version
to
use
for
science,
then
it's
best
that
it
be
as
cheap
as
possible,
meaning
you
know,
the
additional
vertical
levels
is
just
gonna,
make
it
more
expensive
unless.
A
C
On
the
same
topic,
Julio
I
probably
tend
to
agree
with
it
Kevin
just
in
terms
of
how
to
best
use
our
resources,
but
I
also
wanted
to
point
out
as
we're
seeing
that
all
too
common
precipitation
bias
pattern
showing
up.
As
the
resolution
increases,
it
may
be
worth
thinking
really
seriously
about
the
deep
convection
we
do
have
the
to
climate
process
teams,
and
perhaps
they
can
be
asked
in
particular
to
look
seriously
at
the
whole
question
of
scale,
awareness
and
scale
dependence.
C
Q
Just
want
a
second
that
that's
that's
what
I
was
thinking
what
Leah
had
just
said.
It
may
not
be
that
long
before
we
do
have
something:
that's
a
better
deep
connection
scheme,
obviously
I'm
starting
the
job
in
a
month,
and-
and
so,
if
it's
going
to
take
time
to
get
a
new
vertical
resolution,
it
might
even
be
worth
waiting
until
we
get
a
new.
C
Think
we
would
need
to
seriously
consider
as
well
is
you
know,
Julio
have
you
pointed
out
and
as
we've
seen
as
with
kind
of
our
CPT
simulations,
some
issues
with
things
like
the
tropical
cyclone
wind
profiles
and
my
suspicion
is
the
use
of
a
released
version
of
quarter
degree
would
really
target
a
lot
of
those
extreme
phenomena
like
tropical
cyclones
or
like
low-level
jets.
That
were
we're
interested
in
looking
at.
A
C
A
R
Of
course
the
model
is
different,
but
but
they
might
be
further
along
and
tuning
high-res
versions,
and
we
might
be
able
to
learn
something
from
aspects
that
that
they've
done
and
maybe
the
traditional
way
in
which
we
tune
the
model
can
be
tweaked
for
the
high
res
versions
to
try
to
do
it
more
effectively,
at
least
to
get
something.
That's
appropriate
enough
for
science
and
acknowledging
that
it
probably
never
be
as
perfectly
tuned
as
the
conventional
coupled
version.
C
D
C
A
A
Yes,
we
were
discussing,
we
are
discussing
the
question
of
how
much
how
many
resources
to
devote
at
the
moment
to
developing
a
quarter
degree
configuration
with
CSM
to
dot
x.
You
know,
particularly
now
that
you
know
it's
it's
fairly,
certain
we'll
be
going
to
a
different
vertical
resolution.
There
are
also
efforts
in
you
know
there,
the
EDM
F
and
the
club
CPT's
that
we'll
be
making.
You
know,
probably
substantial
changes
to
the
model
physics.
What
what
is
you
know
what
it?
What
is,
what
is
the?
I
I
mean
it
is
a
it's
a
complicated
question
and
there's
probably
a
complicated
answer.
First
of
all,
I
just
wanted
to
make
sure
that
whatever
the
decisions
are
going
to
be
made
by
the
CSM
SSC,
so
I'm
bringing
this
sort
of
Pat
on
on
the
table
and
I
think
the
next
CSM
SSC
meeting
will
be
in
August
sometime
towards
the
end
of
August,
I,
think
and
will
be
discussed
at
length.
I
However,
just
to
provide
a
background,
a
brief
background
is,
you
know
the
present
CSM
high
resolution
model
version
is
literally
based
on
CSM
1.3
version,
and
it
goes
back
all
the
way
to
2010
whatever
early
2010.
So
since
then,
we
have
not
had
any
essentially
a
new
version
of
the
CSM
high
resolution
model
and
it
has
little
ten
years
and
people
are
using
that
and
in
particularly
for
example,
then
the
land
component
is
quite
an
old
version
of
the
model
and
that
folks
have
moved
on.
I
mean
they're.
I
Now
the
newer
versions
of
CLM,
and
if
we
do
not,
if
you
do
not
do
any
high
resolution
versions
with
the
new
CSM
to
model,
then
it'll
be
probably
another
four
or
five
years
before
we
essentially
get
a
new
version
of
the
model
at
high
resolution.
So
in
a
sense
the
community
is
already
using
a
10
year
old
high-res
model.
I
Pretty
much
and
do
we
want
to
keep
the
older
version
and
I
guess,
there's
no
simple
answer,
but
I
wanted
to
visit
this
and
because,
whatever
we
decide
at
this
point,
it
will
have
impacts
for
the
next
5
plus
years
and
I
know
the
issues
with
the
resource
Human
Resources
in
particular.
Does
it
provide
an
answer?
I
A
I
A
I
They're
helping,
essentially
requests
quite
a
few
requests
for
a
higher
conversion
of
the
CSM
to
and
that
at
least
by
quarter
degree,
a
question
land
again
coupled
to
tenth
of
a
degree,
perhaps
motion
model
and
because
people
are
I
mean.
As
I
said,
the
atmospheric
model
is
quite
different
than
what
we
had
in
CSF.
0.3
land
model
is
quite
different,
so
people
wanted
to
see
I
mean
they
want
to
run
errands
simulations
if
possible
and.
A
I
G
C
I
When
I
say
high
resolution,
as
you
know,
I
has
project.
For
example,
we
ended
up
running
low
resolution
coolants
of
pretty
much
everything
that
we
had
done,
I'm
suspecting
that
we
would
need
to
have
also
a
good
low
resolution.
Coolants
of
these
simulations
so
go
hand-in-hand,
perhaps
with
an
AC
type
or
I,
don't
know
what
the
dynamical
fragment
be,
but
of
that
sort
essentially,.
A
Yeah
that
raises
the
further
question
that
when
we
do
when
we
as
we
change
vertical
resolution,
we
will
we
will.
We
will
be
having
to
revisit
some
of
our
low
resolution.
Low,
horizontal
resolution
calculations,
you
know
of
Atrus,
you
know
I
think
we're
going
to
be
wanting
to
do
long.
B,
1850
controls,
possibly
look
at
climate
sensitivity
with
the
new
vertical
resolution
grid,
yeah.
I
I
mean
just
I
mean
Colin
and
a
few
others
have
any
questions
in
terms
of
sort
of
vertical
resolution.
Selection
I
mean
this
vertical.
The
resolution
effort
along
with
how
high
the
cap
top
should
be
so
that
you
don't
have
to
deal
with
the
back
inversion
and
a
cam
version.
It's
a
relatively
new
effort.
I
I
In
terms
of
the
dynamical
course
selection,
SSE
really
doesn't
I
mean
you
don't
have
a
concrete
deadline
yet,
but
my
I
mean
this
is
going
to
begin
further
to
discuss
that
the
SSC
I
mean
we
are
trying
to
do
these
things
sort
of
in
a
proactive
way,
a
dynamical
course
selection
and
what's
needed
for
that
vertical
resolution,
etc.
And
my
thinking
is
that
probably
early
makes
here
late
winter
or
the
spring
next
year
will
be
essentially
focusing
on
really
a
choice
for
our
dynamical
course.
I
F
Mean
this
is
rich,
I
mean
when
it
comes
to
a
question
of
priorities.
I
really
don't
see
how
we
cannot
go
with
vertical
resolution.
First,
you
know
the
high
resolution
always
has
always
been
taking
a
like
a
known
low
resolution
model
and
then
looking
at
that
in
high
resolution
there,
the
thought
of
taking
the
model
we
have
now
and
looking
at
it
in
high
resolution
and
and
and
then
trying
to
understand
that
with
higher
vertical
resolution
is,
is
just
it
just
seems
counterproductive
to
me.
I
F
See
we
really
need
to
hit
it
home
with
the
low
resolution
model,
and
then
we
translate
that
to
the
high
resolution
with
the
high
vertical
resolution
understand
hyper
vertical
resolution
at
low
horizontal
resolution.
We
don't
want
to
be
diving
into
high
horizontal
resolution,
then
figuring
out
in
a
very
slow
way.
What
what's
wrong
in
the
verdict
or
because
there's
more
level
right.
I
So,
in
that
case,
a
path
could
be
rather
than
hymning
a
high-res
version
of
whatever
CSM
one
or
two
then
could
be
have
a
high-resolution
model
version
with
CSM
to
point
X,
where
X
can
be
five
or
six
or
whatever.
If
you
here's
down
the
road
mm-hmm
and
then
the
contour
argue
or
another
argument,
then,
is
well.
We
can
wait
a
few
more
years
after
that,
so
that
we
can
have
the
ocean
model
and
sea
ice
model
and
other
models
updated,
and
that
can
be
a
five
to
six
year,
road
which
may
be
okay.
I
A
So
I
guess,
given
that
there's
still
some
uncertainty
in
the
timeline
for
choosing
the
new
vertical
resolution
or
even
the
dynamical
core,
although
I
don't
know
realistically
I
believe
it
will
probably
be
spectral
element
but
I.
That's
not
that's,
not
a
given
I
think,
maybe
in
the
short
term,
our
effort
should
be
devoted
to
looking
at
the
vertical
resolution
and
that
we
should
not
be
spending.
You
know
we
should
give
this
our
best
shot
in
the
short
term
and
then
maybe
revisit
the
question
of
what
version
of
the
model
to
develop
a
quarter.
Degree.
I
You
just
yeah
I
mean
don't
get
me
wrong,
I'm,
not
necessarily
saying
that
you
should
be
doing
it
I'm,
just
essentially
trying
to
visit
this
high
rents
configuration
and
how
much
what
we
should
be
doing
or
what
we
shouldn't
be
doing
so
I'm.
Just
looking
for
input
at
this
point
yeah,
that's
the
idea.
Yeah.
A
C
I'm
not
sure
if
this
is
relevant
or
not.
This
is
Andrea.
We've
played
around
with
the
horizontal
and
vertical
resolution
combinations
a
lot
at
Gao.
What
we're
finding
certainly
is
that
the
impact
of
the
vertical
resolution
does
not
depend
on
horizontal
resolution.
Our
one
degree
told
us
what
the
doubling
the
vertical
resolution
did,
and
that
answer
was
the
same
at
half
degree.
So
I'm
not
sure,
that's
not
published
anywhere.
That's
just
sort
of
our
own,
take
on
it,
but
I'm,
not
sure
if
it's
relevant
to
this
decision
or
not.
A
C
I
A
K
C
I
You
wanted
to
do
was
high
resolution
prediction
simulation
on
mainly
sub
seasonal
to
slightly
longer
timescales
and,
as
you
know,
with
the
older
code
versions,
with
CLM
and
other
stuff,
there's
not
much
interest
in
the
results
for
that,
especially
from
the
land
group
and
they're
asked
I
mean
if
you
want
brother
community
sort
of
buying
for
these
high
resolution
predictions
they
need
to
be
updated
to
the
new
remand
model.
Perhaps-
and
that
was
another
thing
that
considered
as
I
said,
it's
not
a
simple
discussion.
A
I
C
C
A
So
it's
already
1222
and
I
I.
Don't
expect
that
we'll
come
up
with
a
resolution
today.
But
how
do
people
have
suggestions
for
continuing
this
discussion
either
as
as
a
Google
Doc
or
some
some
some
way
that
we
can
get?
We
can
continue
to
get
a
mwg
community
input
on
on
on
this
issue
before
the
SSC
meeting
in
August.
That.
I
A
H
R
I
A
A
I
A
All
right,
so,
if
if
there
is
no
other,
no
other,
no
other
comment,
I'll
just
go
ahead,
and
and
thank
everyone
for
for
joining
I
know
these.
These
long
do
meetings
are
are,
are
long
and
again.
Thank
you
all
very
much
for
participating
and
we'll
be.
You
know
what
we'll
be
getting
back
to
the
community
with
questions
about
this
horizontal
resolution
issue
and
scheduling
a
zoom
meeting
soon.
So
thank
you
very
much.