►
Description
Day 4 of the 2020 CESM Tutorial featuring Climate variability and change Questions and Answers.
The CESM Tutorial will consist of:
Lectures on simulating the climate system, practical sessions on running CESM, modifying components, and analyzing data.
For more information:
http://www.cesm.ucar.edu/events/tutorials/2020
Community Earth System Model (CESM) is a fully-coupled, community, global climate model that provides state-of-the-art computer simulations of the Earth's past, present, and future climate states.
A
A
B
A
A
You
just
gotta,
make
it
full
screen
and
then,
when
you
do
your
shared
screen
on
zoom
just
select
that
window,
so
you
can
do
other
stuff,
okay
and
then
so
right
now
it
says,
for
me,
sharing
is
pause,
but
since
it's
just
a
picture
it
doesn't
matter
if
it's
real
time
or
not
yeah.
That
should
be
fine.
A
So
are
you
gonna?
Leave
it
open
on
your
end?
Or
do
you
just
want
me
to
do
the
sharing
first
thing,
so
you
can
yeah
if
you
want
to
go
ahead
and
start
sharing,
it'll
kick
me
off
and
it'll
start
your
share.
Okay,
I'm
going
to
have
to
hop
off
soon,
and
then
I
mean
I'm
going
to
keep
an
eye
out
on
the
youtube
stream
and
make
sure
that's
right,
and
then
you
can
still
contact
me
over.
The
google
chat-
because
I
mean
I'll-
be
on
that
too,
and
so,
if
anything,
pops.
B
C
A
A
A
C
C
A
B
All
right
well
I'll,
be
back
on
around
like
9
30
is,
but
let
me
know
in
the
google
chat,
if
you
need
me
or
anything.
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
I'll
respond
to
your
message,
I
just
couldn't
figure
out
how
to
keep
sharing
the
welcome
slide
and
still
do
stuff
behind
the
scenes.
D
Yeah
and-
and
I
will
take
over
the
share
screen
like
in
in
a
second.
D
Oh,
you
mean,
like
you,
have
it's
so
if
you
type
escape
it
will
stop
the
full
screen
option
on
your
on
your
monitor,
if
you,
if
you,
if
you,
if
you
click
on
the
zoom
and
then
type
escape,
and
then
it
should
reduce
your
window,
so
you
can
take
control
of
your
desktop
again.
A
B
Hey
gunter
and
elizabeth:
are
we
going
to
put
the
questions
up
on
the
screen?
You
can
do
that.
You
want
me
to
do
that.
Okay,.
B
D
And
I
have
just
locked
your
your
poll
christine.
B
A
D
D
B
D
All
right:
okay,
it's
8
52!
So
let's
just
get
started
with
the
announcement
good
morning:
everybody,
let's
we're
entering
the
first
day
of
our
tutorial
one
one
more
day
to
go,
and
I'm
just
going
to
start
with
a
few
announcements.
D
There's
not
many
this
morning,
so
we
we
were
talking
yesterday
with
some
of
the
moderators
and
the
helpers
and
we've
we
felt
like
this
virtual
environment
is,
is
tricky
because
it
really
feels
like
especially
during
the
q,
a
sessions
that
a
bunch
of
scientists
are
talking
about
each
other
and
many
of
the
moderators
and
panelists
and
helpers
might
never
have
seen
your
face
yet
so,
if
you
are
not
opposed
to
it,
please
turn
on
your
camera.
D
So
they
could
see
you
even
if
you're
not
saying
anything,
it's
nice
to
see
your
audience
and
you
don't
have
to.
We
would
highly
encourage
you
to
do
so
for
the
reminder
of
the
of
the
tutorial.
D
The
q
a
sessions
have
been
posted
online,
at
least
what
we
had
so
far,
and
you
will
find
them
on
the
main
page
of
the
tutorial
page
and
I'm
going
to
show
it
to
you
here.
D
So
you
know
there's
a
coursework
page
where
you
have
all
the
coursework,
but
on
the
front
page,
you
have
the
q
a
panel
sessions
here
from
monday,
tuesday
and
wednesday,
and
they
have
not
been
trimmed
yet.
So
it
means
that
whenever
our
web
specially
started
recording
until
when
he
stopped
it
captured
everything
so
you're
going
to
have
other
bonus
introductions
and
some
other
matter
before
and
after
the
q
a
session
so
fast
forward
to
wherever
you
would
like
to
be.
D
Let's
see
oops
this
morning
at
some
point,
you're
going
to
receive
a
short
survey
and
this
survey
will
contain
three
questions.
It's
going
to
be
really
quick.
The
first
one
is
going
to
be
about
whether
you
care,
whether
you
allow
us
to
share
email
address
with
the
other
participants
of
this
tutorial.
Typically,
when
you're
in
person,
you
sign
a
waiver,
because
we
take
pictures
of
the
main
seminar,
room
and
other,
and
so
you
sign
a
waiver
that
it's
okay
for
us
to
take
a
picture
and
to
share
some
of
your
contact
information.
D
D
Please
let
us
know
it's
going
to
be
yes
or
no
question,
and
then
tomorrow's
office
hour
is
going
to
be
a
little
bit
particular
different
compared
to
the
other
office
hours
we've
been
experiencing,
because
it's
going
to
be
specific
to
a
typical
component,
so
you're
going
to
have
a
choice
between
working
on
land,
ice
budget,
chemistry,
sea
ice
and
ocean
land,
chemistry,
atmospheric
chemistry,
and
in
order
to
try
to
organize
this
the
best
we
would
like
to
try
to
get
a
distribution
on
where
your
first
main
interest
lie
in.
Of
course,
this
is
non-binding.
D
You
can
always
switch.
You
know,
but
we'd
like
to
know
how
to
best
prepare
for
this.
So
if
you
could
yeah
just
do
this,
this
quick
survey
by
today
by
the
end
of
the
morning
very
good,
it
will
take
30
seconds
and
last
but
not
least
so,
we've
seen
on
the
slack
link
that
you
guys
are
organizing
this
discussion
session,
which
is
great,
and
there
have
been
this
question
whether
scientists
should
be
involved
in
there,
and
we
want
to
leave
this
up
to
you
guys.
D
We
don't
want
to
have
a
bunch
of
scientists,
you
know
taking
over
what
you're
trying
to
accomplish.
You
have
discussing
them
on
yourself
discussing
in
your
ideas,
etc.
But
that
said,
if
you
would
like
to
invite
scientists
to
discussion,
please
feel
free
to
email
them
directly.
If
you
don't
know
the
email
ask
me,
and
I
can
forward
it
to
them
or
ask
elizabeth
and
she
can
forward
it
forward
your
invitation
to
them,
and
we
prefer
it
like.
D
We,
we
kind
of
want
this
to
be
your
space,
rather
than
like
an
anchor
space
like
where
scientists
like
just
you
know,
could
monopolize
the
conversation.
D
A
So
this
is
elizabeth.
I
was
just
going
to
chime
in
on
the
lunch
session
so
from
what
I've
gathered
so
far,
there's
interest
in
an
exoplanet
discussion,
paleo
and
ocean,
so
you
can
go
ahead
and
put
in
the
chat
or
private
chat
to
me
if,
if
I've
misunderstood
this
or
if
there's
some
other
topics
that
you
would
like
to
have
during
during
that
midday.
A
A
E
D
So
on
the
chat
suggesting
a
breakout
today
about.
D
About
making
sense
more
inclusive,
so
you
can
whether
you.
D
So
it
would
be
interesting
to
have
an
inca
staff
member
there
to
know
what
inca
does
and
that's
that's
fine.
So
you
can
go
on
slack
and
suggest
this
idea
to
see
who's
interested
and
if
there's
interest
in
this
discussion,
then
you
can
reach
out
to
some
scientists.
If
you
want-
and
we
can
we
can,
we
can
definitely
try
to
make
this
happen.
A
A
B
D
All
right,
it's
9
00
am
so
let's
get
started
with
a
q.
A
panel
session,
I'm
going
to
introduce
your
moderator
today
after
this
session
is
christine
shields
that
you
have
seen
in
the
office
hours
already
christine
is
a
scientist
in
the
climate
change
research
section
with
expertise
in
simulating
earth's
past
present
and
future
climate
with
csm.
D
B
Thanks
gunter,
so
I
guess
the
first
thing
I'll
do
is
introduce
the
section
which
is
the
climate
variability
and
change
q
a
panel,
and
this
is
a
little
different
than
what
we've
been
doing
previously
in
the
in
our
you
know
our
past
days,
where
it
was
sort
of
focused
on
a
model
component
today.
B
This
is
more
of
sort
of
like
an
all-inclusive
analysis
and
applications
of
the
model
type
of
a
panel,
so
climate
variability,
the
climate,
variability
and
change
is
sort
of
self-explanatory,
and
we've
tried
to
compile
a
team
of
our
scientists
that
are
experts
in
a
variety
of
areas
that
we
thought
people
might
be
interested
in,
and
so
I'm
just
going
to
sort
of
go
through
and
briefly
introduce.
All
of
our
all
of
our
panelists
first
is
clara.
B
Our
second
panelist
is
john
fusulo.
He
wants
to
be
thought
of
as
an
aspiring
generalist
who
studies
climate
variability
and
change,
using
both
observations
and
models,
modelers
think
of
him
as
an
observationalist,
while
observationalists
think
of
him
as
a
modeler.
So
that's
john
third
panelist
is
angie
pendergrass.
B
She
is
an
expert
in
the
hydrological
cycle
and
all
things
precipitation
she
studies
these
features
and
particularly
in
how
they
may
change
in
a
warming
world.
Our
next
expert
is,
she
is
an
expert
in
climate
dynamics
and
variability
in
the
earth
system.
B
Her
emphasis
is
in
extremes
and
in
particular
tropical
cyclones
and
then
finally,
we
have
yaga
victor.
She
is
our
expert
who
studies
the
earth
system
predictability.
So
for
all
those
decadal
predictability,
questions
her
background
is
atmospheric
dynamics,
in
particular
the
middle
atmosphere.
She
studies
things
like
the
qbo
geoengineering
scenarios
and
she
uses
wacom,
which
is
the
whole
atmosphere
model
which
is
sort
of
a
subset
of
csm
for
the
for
the
whole
atmosphere.
So
I
will
stop
there
with
the
introductions
and
now
I
will
share
my
screen.
B
So
we
can
see
the
questions.
Okay,
so
here's
your
screen
desktop
share.
I
hope
you
guys
can
I'll
try
to
make
these
a
little
bit
bigger.
So
we
can
see
everything,
hopefully
that's
good,
so
I
so
what
we'll
do
here
is
I'll
just
I
sort
of
rank
these,
as
you
know,
whoever
the
top
number
of
vote
getters,
but
I
think
we
have
enough
time
to
go
through
all
of
them.
I
think
gunter
said
we
have
about
19
questions.
B
If
we
finish
these
before
the
time
is
up,
you
guys
can
think
about
questions
we'll
have
an
open.
We
can
have
an
open
section
and
we
can.
You
can
ask
questions
of
our
panel
experts,
knowing
what
their
expertise
is.
So
we'll
just
I'll
just
read
the
question
and
then
I
will
let
the
panelists
jump
in
and
if
no
one
responds
I
might
assign
a
panelist
to
answer
the
question.
B
F
So
so,
ultimately,
this
is
a
question
of
attribution
and
to
address
attribution.
You
need
a
model,
you
need
a
model
of
some
kind
and
obviously
cesm
is
one
way
that
you
would
do
that
and
you
can
imagine
model
simulations
that
can
be
designed
to
test
the
source
of
predictability
from
a
forecast.
F
You
really
want
to
sort
through
the
available
simulations
to
see
if
that's
consistent
with
them
and
that's
not
hard
to
do
and
will
save
you
a
lot
of
work
down
the
road,
but
ultimately,
for
example,
if
you
think
that
you
know,
for
example,
microphysical
interactions
are
part
of
your
predictability.
You
can
go
into
the
model,
turn
them
on
and
off
and
then
see.
If
the
predictability
changes,
so
that's
ultimately
where
you'd
want
to
go,
but
not
immediately.
B
So
I
I
just
also
should
say
if
the
person
who
asked
the
question
wants,
the
fault
wants
to
follow
up
with
a
another
question.
You
feel
free
to
do
so
so
either
either
any
student
or
if
another
panelist
wants
to
jump
in
on,
say
something
we
can.
C
Continue,
I
guess
I
thought
maybe
yaga
would
speak
to
this,
but
just
more
generally,
you
know
we
envision
that
on
the
longer
time
scales,
the
ocean
becomes
very
important
in
terms
of
the
source
of
memory
for
the
atmosphere
and
the
climate
system,
but
even
within
the
atmosphere.
There
are
certainly
predictable
aspects,
and
I
know
that
yaga
richter
has
examined
at
length.
The
role
of
the
stratosphere
in
you
know
controlling
aspects
of
tropical
and
extra
tropical
atmospheric
circulation
changes
in
the
troposphere
and
at
the
surface.
So
there
are
many
sources
within
you
know.
C
C
The
overturning
circulation
in
the
ocean
is
another
example,
so
where
there
are
many
processes-
and
I
agree
with
john-
a
model
is
useful
to
then
tease
those
out
with
you
know,
dedicated
experiments
to
test
to
test
hypotheses.
E
So
I'll
add
a
little
bit
to
it,
so,
in
particular
on
the
predictability,
so
the
earth
system
prediction
either
sub-seasonal
or
decadal.
There
is
not
been
a
whole
lot
of
these
experiments
ran
largely
because
they
take
so
many
computing
resources,
so,
for
example,
in
a
subseasonal
scale.
We
believe
that
some
predictably
comes
from
the
land,
some
from
the
atmosphere
and
some
from
the
ocean,
but
to
really
isolate
it.
You
would
need
to
take
out
the
realism
of
some
of
those
components
either
replace
them
with
climatology
or
something
similar,
and
those
are
things
we're
planning.
E
So
we're
planning
that
as
part
of
the
earth
system
prediction
working
group,
but
we
have
not
done
many
of
them
yet.
So
that's
where
the
models
are
really
useful.
You
can
take
out
pieces,
put
them
back
together
and
see
where
you
get
more
predictability
and
over
which
regions
you
lose
it.
For
example,
if
you
take
out
your
sea
surface
temperatures
from
an
enso
vent
large
likely,
you
would
lose
quite
large
regions
of
where
your
temperature
and
precipitations
are
predictable.
C
And
I
might
just
add
that
you
know
models.
We
emphasize
that
here
at
ncar
and
of
course,
they're
very
powerful
and
they
do
truly
provide
test
beds.
But
don't
forget
you
can
analyze
the
observations
and
look
for
you
know,
lead
lag
relationships
and
you
know
it
requires
a
lot
of
thought.
But
ultimately
we
need
to
see
if
there's
any.
You
know,
telltale
signs
in
the
observations
that
there
may
be
some
predictability.
So
I
think
we
can't
lose
sight
of
looking
at
the
observations.
B
That's
a
very
good
point
clara.
Thank
you
always
need
to
look
at
observations
for
sanity
checks.
Okay,
so
all
right,
so
I
will
move
on
to
the
next
question.
I
think
that
one
one
so
we
think
we
answered
that
one
all
right.
The
next
question,
the
resolution
of
lar
of
the
large
ensemble
project
output
is
pretty
coarse.
Could
you
introduce
the
possible
ways
to
down
scale
to
25,
kilometers
and
comment
on
the
pros
and
cons
of
this
method?
C
Well,
maybe
I
can
say
a
few
words
since
people
have
already
done
that,
and
it's
a
very
good
question
and
of
course
you
know,
the
large
ensemble
gives
you
certain
aspects,
but
it
it
is
course
resolution.
C
So
there
have
been
a
handful
of
efforts
at
this
point
that
I'm
aware
of
to
downscale
using
a
dynamical
model
where
you
embed,
you
know
a
higher
resolution
model
over
a
certain
region
and
then
force
or
drive
that
regional
model
with
the
individual
members
of
the
large.
This
the
coarser
resolution,
large
ensemble
project,
so
daniel
swain-
has
done
this
for
actually
for
the
topic
of
atmospheric
rivers
and
their
impact
on
on
california,
precipitation.
C
There's
a
group
in
canada
that
have
actually
downscaled
the
large
ensemble,
I
believe,
over
the
european
region.
Actually,
I
think
they
had
a
collaboration
there
and
there
are
probably
other
efforts,
and
I
think
it's
it's
sort
of
a
great
next
step
or
add-on
for
these
large
ensemble
views
to
do
this
kind
of
downscaling.
C
Yes-
and
we
actually
did
this
for
the
conus,
the
conterminous
us
or
the
western
part
for
selected
members
of
the
csm1
large
ensemble-
we
didn't
actually
get
very
far
in
analyzing
that,
but
if
any
of
the
students
out
there
are
interested
I'd,
be
you
can
follow
up
with
email
with
me
and
I'd
be
happy
to
share
with
you
what
those
runs
are
about.
Thanks
john.
A
This
is
angie.
I
wanted
to
give
my
perspective
the
hydrologic
cycle
perspective
on
this
question,
because
I
think
it's
an
interesting
one.
So
a
lot
of
the
kind
of
downscaling
that's
been
done
has
been
with
wharf
or
models
like
wharf
that
are
kind
of
regional,
climate
mode
or
regional,
often
weather
models,
and
so
one
of
the
things
that's
really
different
about
that
is
that
in
a
climate
model,
we're
we're
putting
a
lot
of
emphasis
on
kind
of
what's
happening
with
the
global
energy
budget
and
the
global
circulation.
A
And
so,
when
you
run
these
limited
area
models,
just
one
of
the
main
ways
that
people
are
downscaling
you're
kind
of
losing
some
of
those
physics
actually-
and
so
I
I
think
that's
one
of
the
potential
cons
of
going
down
to
these
higher
resolutions
is
you
you
can
get
information
that
looks
more
realistic
in
that
the
resolution
looks
higher,
but
it
might
not
have
all
of
the
physics
that
you
would
want
to
have
with
the
global
climate
model.
A
F
And
ultimately,
there's
no
ability
to
feed
back
onto
the
larger
scales
and
so
you're
really
limited
in
terms
of
what
hypotheses
you
can
explore.
A
I
think
that
going
forward
in
the
future,
one
thing
that's
started
to
become
more
popular.
Are
these
variable
resolution
models
and
some
folks
at
ncar
have
also
been
working
on
those,
and
then
you
can
kind
of
refine
the
grid
over
one
region
that
you
care
about.
So
it's
still
computationally
tractable
to
run,
but
those
would
be
kind
of
different
model
simulations
than
the
large
ensemble.
B
Okay,
marta
is
raising
her
hand,
go
ahead
and
un
mute
yourself
and
ask
your
question.
Marta.
A
Yes,
so
I
was
wondering
what
do
you
think
about
a
statistical
downscaling,
because
when
we
think
about
downscaling,
we
think
about
dynamical
downscaling,
but
there's
also
some
researches
that
do
statistics
on
skating.
So
I
was
wondering
if
you
have
any
thoughts
about
it
on
the
advantage
of
disadvantages.
Of
that
I
mean,
I
guess
I'll,
say
something.
First,
I
definitely
think
it's
it's
interesting
and
it's
a
complimentary
approach.
A
I
mean,
then
you
are
losing
more
of
the
physical
constraints,
but
you
can
put
you
know
if
you
choose
assumptions
that
make
sense
for
the
question
that
you're
trying
to
ask
and
for
what
you're
trying
to
look
at
you
can
do
lots
of
new
things
and
you
know
there's
a
lot
of
research.
That's
happening
on
all
kinds
of
statistical
methods,
so
I
think
that
that's
something
that
we're
going
to
see
advance
in
the
future
for
sure.
B
Great
okay,
I
don't
see
anyone
else
raising
their
hands.
Actually,
so
I
mean,
if
I
missed
something
gunter
or
elizabeth,
please
shout
out,
because
I
because
of
my
screen
share,
I
don't
have
the
little
raised
hand
thing
immediately
available
to
me
so,
okay,
well
we'll
move
on
to
the
third
question.
B
C
Well,
that's
a
nice
question.
I'm
sure
john
has
comments.
I
think
they're
they
they.
It
would
be
very
nice
to
use
them
hand
in
hand.
I
think
an
interesting
and
new
area
now
that
is
opened
up
by
these
large
ensembles,
is
to
address
whether
the
internal
variability,
the
characteristics
of
it,
the
patterns,
the
time
scales,
the
amplitudes,
whether
that
does
depend
on
the
external
forcing
and
you
know
are-
is
the
character
of
internal
variability.
C
Does
that
differ
in
a
lat?
You
know
an
lgm
climate,
or
you
know
earlier
time,
climate
versus
a
future
climate
and
especially
on
you
know,
regional
scales.
So
I
think
they,
it's
very
nice,
to
bring
these
two
projects
to
bear
on
that
question,
and
then
we
shouldn't
forget
that
internal
variability
right.
That's
these
long
control
runs
where
there
is
no
change
in
the
external.
Forcing
that
those
are
you
know,
give
us
the
best
sampling
if
you
will
of
internal
variability
without
any
external
influence.
F
F
And
so
you
know
everyone
has
their
wish
list
in
retrospect
of
how
we
had
done
things
differently,
and
I
guess
on
my
list
would
be
that
we
ran
them
with
the
exact
same
version
of
the
model
and,
of
course,
that
gets
very
expensive
as
you
go
to
1
000
year,
simulations
members
with
the
lme,
but
it's
a
real
limit
as
to
how
much
you
can
compare
the
lme
and
the
le
due
to
the
different
resolution
of
the
atmosphere
component.
C
That's
a
really
good
point,
john.
I'm
very
glad
you
brought
that
up.
Yeah,
I
think
enso
in
particular,
is
quite
different
in
those
two
with
those
two
different
resolutions
of
the
model.
F
And
that
comes
up
a
lot
if
you're
dealing
with
multi-model
ensembles
as
well,
that
it's
there's
a
there's,
a
large
range
of
enso
spectra
across
climate
models
and
and
with
the
even
within
climate
models
based
on
the
resolution
at
which
they're
run.
So
beware,.
B
Those
are
all
very
good
points.
Do
any
of
the
students
have
any
follow-up
questions.
B
B
Hand-
okay,
I
don't
think
there
is
so
we'll
move
on
to
our
next
question
for
earth
system
prediction.
Does
the
academic
community
currently
have
a
preferred
initialization
approach?
Generally?
What
factors
do
we
need
to
concern
ourselves
with
when
we
choose
the
initialization
approach,
so
this
this
question
actually
is
there's
a
few
that
are
in
this
theme
in
the
in
the
in
our
panel
of
questions
here
so
so
we
can
spend
a
little
time
on
this.
If
you
want.
E
I
can
perhaps
start
so
the
academic
community-
I
imagine
you,
people
at
universities
and
stuff,
are
usually
the
ones
who
are
not
running
or
system
prediction
simulations.
There
are
a
few
groups
like
ben
kirkman
in
miami.
Does
it,
but
largely
it's
the
bigger
operational
centers
that
have
the
bandwidth
to
do
it
and
there's
a
lot
of
factors
and
the
factors
are
different
on
different
time
skills.
E
So
and
then
the
ocean
was
initialized
with
a
spun
up
ocean
that
was
separately
forced
with
the
jr
jra
55
reanalysis,
and
it's
run
through
several
cycles,
and
then
it's
used
to
start
the
model,
and
there
was,
I
think,
there's
another
question
for
the
downtown
about
full
field
and
anomaly,
initialization
and
so
for
dple
full
field
was
chosen
and
there
wasn't
a
lot
of
time
to
study
anomaly
initialization.
It's
one
of
the
things
that
we
have
on
the
agenda
for
the
next
few
years
to
examine
this
in
detail.
E
The
groups
were
looking
at
it,
but
because
you
will
have
drift
and
model
shack,
no
matter
what
there's
is
really
an
active
area
of
research.
What
is
the
best
way
to
initialize
the
model?
E
Just
for
consistency,
but
the
atmosphere
comes,
let's
see
in
csm1,
we
used
error
interim
for
csm2
we're
using
cfs
initial
conditions,
so
you're
taking
an
atmospheric
condition,
that's
from
a
different
analysis,
different
system.
So
there
will
be
some
inconsistency
when
you
put
it
into
our
model,
but
right
now
at
incar
we
don't
have
a
real-life
data
simulation
that
we
could
have
done
well
and
the
cost
is
really
prohibitive,
not
for
running
forecasts.
E
E
B
Any
other
students
want
to
follow
up.
I
think,
a
yoga
sort
of
covered
the
the
next
three.
The
next
couple
questions
too
about
the
spin
up
and
the
small
perturbations
in
the
initial
field.
So
if
anyone
has
any
more
questions
or
if
there's
confusion
on
on
any
of
this
stuff,
please
please
speak
up.
C
B
C
Chime
in
on
the
how
to
understand
the
small
perturbation
of
the
initial
field
can
represent
internal
climate
variability,
because
we're
asked
that
a
lot
with
regard
to
you
know
decadal
prediction
or
the
large
ensemble
how
to
understand
this.
It's
simply
that
you
know
the
the
atmosphere.
The
fluid
motions
in
the
atmosphere
are,
you
know,
chaotic,
non-linear
and
then
very,
of
course,
unstable
to
small
perturbations,
so
it
just
leads
to
this
so-called
butterfly
effect.
That
then
puts
you
on
a
different
trajectory
of
unpredictable
sequences
of
internal
variability.
C
So
you're,
just
you
know
your
your
your
then,
with
these
small
perturbations.
It's
just
a
way
to
allow
the
climate
system
to
sit
within
its
natural
sort
of
attractor,
I
think,
is
the
language
of
of
its
ver,
its
own
variability.
C
F
So
there
is
some
new
there.
There
is
some
nuance
to
this
question
too,
and
that
is
how
do
you
fully
sample
the
realm
of
internal
variability
in
the
ocean
and
does
your
approach
to
initializing
the
large
ensemble
adequately?
Do
that,
and
so
in
fact,
with
a
large
ensemble,
we
have,
I
think,
on
order
of
10
members
where
we
have
a
different
ocean
state
that
we're
starting
from
and
that
can
play
a
role
and
that
can
take
some
time
to
really.
F
You
know
the
atmosphere
very
quickly
fills
the
realm
of
internal
variability
within
you
know
six
months,
certainly,
and
so
you
get
a
very
good
representation
of
the
the
bounds
but
the
ocean
takes
longer
and
the
dependence
on
the
ocean
state
for
some
questions
is
very
important.
So
that's
a
nuance
to
keep
in
mind.
B
Thanks
for
bringing
it
up,
john,
I
was
gonna
actually
follow
up
and
ask
you
guys
to
sort
of
talk
about
the
differences
between
these
big
changes
and
these
little
prohibition
changes.
But
before
I
think,
quinn
had
a
a
chin
had
a
question
chin.
Do
you
want
to
go
ahead
and
ask
your
question.
F
Yeah,
I
just
can
you
extend
the
spin
up
question
a
little
bit
because
I
I
sorry
I
didn't
follow
quite
well.
E
I
can
explain
it
or
clara:
do
you
want
to
take
a
step?
Go
ahead,
yaga,
okay,
so
for
the
ocean.
We
start.
We
have
observation,
gra
55
observations,
so
an
offline
ocean
is
ran
with
those
fluxes
surface,
flexes
four
spiders
for
my
observations
and
it
cycled
through
five
times.
I
believe
till
you
reach
a
state
that
we
think
is
balanced.
So
then
we're
using
the
last
cycle.
E
So
it's
like
an
ocean
forced
by
observations
that
is
used,
not
observations
directly
that
are
put
into
the
model
and
that's
one
of
the
protocols
that
I
think
for
simup6.
It's
part
of
omip
ocean
modeling,
intercomparison
project,
that's
what
a
lot
of
the
groups
are
doing.
I'm
not
sure
how
exactly
the
details
of
why
it's
five
cycles,
not
eight.
I
don't
know
john
or
clara,
you
know,
but
I
think
that's
just
like
a
a
number.
E
That's
been
arrived
as
an
optimal
number
that,
after
some
time
you
don't
get
any
more
benefit
from
doing
it
longer
and
the
same
for
the
land
you're,
forcing
the
land
model
offline
by
atmospheric
data,
especially
it
deflexes
at
the
surface,
and
you
spin
it
up
till
you
get
to
a
land
model
state
that
is
pretty
comparable
to
what
is
in
present
day.
And
then
you
just
continue
that
forward.
As
you
run,
your
forecasts.
F
So
yaga,
I
actually
have
a
question
about
that,
and,
and
one
could
argue
that
in
nature,
the
ocean
is
actually
not
an
equilibrium
at
any
one.
Point
that
you
have
these
varying
time
scales
of
response
within
the
ocean
from
the
upper
ocean
to
deeper
levels
and
that
it's
continually
adjusting
to
transient
forcing.
And
is
this
something
that's
just
thought
of
as
something
that
we
can't
really
address
with
the
caleb
prediction,
because
you're
obviously
trying
to
get
an
equilibrium
state
to
initialize.
With.
E
E
So
this
is
why
the
anomaly
initialization,
I
think
it's
it
could
be
just
as
good.
So
then,
just
taking
the
anomalies
and
putting
them
on
top
of
your
model
is
another
way
that
people
think
is
a
good
one
and
it's
all
to
try
to
minimize
the
model
shock
and
drift,
because
the
model
will
drift
as
soon
as
you.
You
know,
initialize
it
with
something
that
it
doesn't
want
to
be
at,
but
yeah
it's
an
active
area
of
research
and
because
you
usually
need
a
lot
of
cases
and
you
need
a
lot
of
years.
B
Okay,
if
there
are
no
other
questions
or
follow-ups,
I
think
we'll
move
on.
I
think
there's
probably
some
more
questions
about
this
later
and
later
on,
so
we'll
just
keep
continue
to
plow
through
so
the
next
question
does
csm2
have
a
tipping
point:
does
the
model
have
a
runaway
feedback?
F
So
we
actually
had
some
very
interesting
experiences
in
the
development
of
cesm2
related
to
this
question
and
actually
the
whole
experience
has
really
changed.
My
appreciation
for
whether
or
not
models
can
be
used
to
identify
tipping
points
or
runaway
feedbacks
at
all,
because
I
think
if
you
were
to
encounter
one
while
you're
developing
a
model,
you'll
generally
try
to
get
rid
of
it,
and
so
it's
not
particularly
a
useful
way
to
to
explore
these
things.
F
But
that
said,
I
mean
there
are
some
well-known
tipping
points
in
the
climate
system
and
we
see
them
in
cesm2,
for
example,
the
weakening
of
amok
and
kind
of
sudden
drastic
weakening
of
the
ocean
circulation.
Then
the
atlantic
does
occur
and
it's
a
difficult
thing
to
evaluate
with
observations.
Obviously,
but
you
can
compare
it
across
models
to
get
a
sense
as
to
how
it
falls
within
the
realm
of
models.
F
There
are
other
aspects
of
tipping
points
that
the
model
is
not
able
to
reproduce,
and
so
we
know,
for
example.
A
key
tipping
point
is
the
melting
of
the
antarctic
ice
sheet,
and
the
model
doesn't
have
the
physics
to
deal
with
that.
Yet
we're
working
on
it,
but
eventually
we'll
get
to
a
point
where
we
can
start
to
look
at
that.
F
But
just
as
a
bit
of
a
background
when
we
were
developing
cesm2,
we
had
this
problem
with
the
labrador
c
freezing
over
in
a
control
state
in
a
1850
control
state,
and
this
was
very
problematic
because
we
don't
see
it
in
the
observations
at
the
time
and
it
would
kick
up
and
create
some
very
strange
behavior
in
the
model,
even
in
the
past
100
150
years,
and
so
it's
something
we
worked
very
hard
to
get
to
first
understand
where
it
was
coming
from
and
and
then
to
get
rid
of
it,
because
it
was
very
problematic
for
our
simulations
of
the
historical
era.
F
And
so
that's
part
of
the
basis
for
me
saying
that
if
there
are
tipping
points
in
nature-
and
none
of
these
are
runaway
by
the
way
the
plank
feedback
operates.
At
t
to
the
fourth
and
is
very
powerful
at
constraining
the
system
to
avoid
a
runaway
in
the
climate
system,
but
there
are
tripping
points,
but
I
don't
think
you'll
get
them
from
a
model,
because
I
think
during
the
development
process
you
explore
kind
of
this.
This
parameter
space
that's
pretty
wide
and
you
try
to
avoid
any
of
this
behavior
during
that
exploration.
F
A
Can
I
say
something
about
this:
this
is
angie,
so
I
a
couple
of
the.
I
think
that
tipping
points
can
encompass
a
lot
of
things
and
some
of
us
might
have
different
ideas
actually
of
what
tipping
points
mean.
But
I
think
two
that
come
to
mind
for
me
that
are
perhaps
relevant
to
the
physics
that
are
actually
included
in
cesm
are
kind
of
the
disappearance
of
the
summer
sea
ice.
A
So
you
can
think
of
that
as
a
type
of
tipping
point,
once
there's
no
sea
ice
in
the
summer
at
the
minimum,
that's
pretty
different
and
that's
something
that
should
be
represented
in
cesm.
I
mean
to
zero
with
order,
at
least
I
would
totally
believe.
What's
going
on
with
that,
and
then
another
thing
that
I
always
come
back
to
is
the
runaway
greenhouse.
C
A
Right
and
so
that's
when
you
have
so
much
when
it
gets
so
hot
that
water
keeps
evaporating
and
you
get
radiative
water,
vapor
feedbacks
that
change
pretty
fundamentally
and
that's
something
that
you
wouldn't
expect
to
be
able
to
reach
into
esm
as
it's
formulated,
because
the
radiative
parameterizations
are
kind
of
calibrated
around
a
reasonably
linear,
present
kind
of
state.
And
so
you
can't
actually
before
you
get
to
that
point.
You
would
need
to
change
some
things
about
the
radiative
parameterizations.
In
order
to
accomplish
that.
A
So
that's
a
real
physical
feedback,
but
we've
kind
of
tried
to
make
the
model
realistic
at
the
expense
of
being
able
to
put
it
into
these
really
extreme
states
and
have
it
still
run.
F
To
my
knowledge,
one
one
runaway
feedback
that
has
been
reproduced
in
climate
models
is
the
snowball
earth
so
going
the
other
way
where
you
cool
off,
to
a
point
where
you
increase
the
albedo
and
absorb
less
solar
radiation
and
continue
to
cool,
so
that
has
been
reproduced
in
climate
models,
but
angie's
right
that
a
lot
of
the
parameterizations
actually
in
climate
models.
Generally,
you
know
some
of
them
involve
a
lookup
table
and
that
lookup
table
is
certainly
out
of
range
when
the
by
the
time
you
get
to
a
runaway
greenhouse
effect.
A
Definitely
the
snowball
earth
simulations
that
I
know
about
were
done
in
pretty
idealized
models
too.
I
don't
know
of
anyone,
who's
done
them
with
ces.
F
B
Yeah:
okay!
Well,
thanks
guys
do
any
students
have
any
follow-up.
B
Questions:
okay,
if
not
we'll
move
on
to
our
next
question.
Okay,
another
lens
question
in
lens:
the
effects
of
the
specific
forcing
can
be
deduced
by
subtracting
single
ensemble
mean
from
the
full,
forcing
ensemble
main
this
assumes
the
effects
of
each
forcing
are
linear.
I
wonder
how
good
this
assumption
is,
especially
on
understanding
regional
climate
change.
C
So
maybe
I'll
take
a
stab
at
that.
That's
a
great
question
and
we're
not
assuming
the
effects
are
linear.
C
We
are
fully
aware
that
when,
if,
if
you
use
this
approach
that
in
fact,
you're
not
just
isolating
the
effect
of
the
forcing
that
was
withheld
from
the
simulation,
but
your
you're,
including
any
any
additional
contributions
from
non-linear
interactions,
so,
for
example,
if
we
withhold
the
anthropogenic
aerosols
and
run
an
ensemble
and
then
subtract
that
from
the
full,
all
forcing
ensemble
we're
actually
going
to
deduce
the
effect
of
the
aerosols
and
it's
any
possible
non-linear
interactions
with
the
other,
forcing
agents.
C
So
we
have
not
been
able
to
test
how
lydia,
linear
the
the
linearity.
You
know
assumption
here
or
what's
or
to
the.
To
what
degree
are
these
forcing
agents
actually
linear?
And
I
think
that
would
be
a
very
nice
thing
to
do.
I've
had
it
on
my
to-do
list
to
do
a
run
where
we
just
prescribed
the
anthropogenic
aerosols
and
greenhouse
gases,
and
then
we
can
use
the
existing
single,
forcing
runs
to
really
test
that.
C
C
A
Can
I
jump
in
and
mention
that
clara
has
done
more
than
she's
giving
herself
credit
for.
So
I
think
if
you
twist
this
question,
just
a
tiny
bit.
One
thing
that
perhaps
you've
already
heard
about
is
these:
these
sets
of
simulations
or
we've
kind
of
called
them
all,
but
one
forcing
simulations
we
take
away,
just
the
greenhouse,
gassing
or
just
the
aerosol,
forcing
and
look
at
those
and
compare
them.
A
You
can
compare
those
against
the
ces
of
one
large,
ensemble
and
clara
led
a
paper
that
is
actually
an
early
online
release
right
now
and
she
came
up
with
an
equation
to
add
back
up
the
responses
from
those
single
forcing
simulations
and
then
you
she
actually
in
the
paper,
looks
at
how
well
those
add
up
for
temperature
and
precipitation
in
map
form,
and
so
there
you
can
kind
of
see
where
it
deviates
from
linearity.
A
C
Yes,
thanks
angie
and
there
there
has
been
if
you're
interested
in
whoever
asked
this
question
you
can
read,
take
a
look
at
the
paper
that
we
have
online
early
online
release
in
journal
of
climate
and
in
that
paper
we
have
citations
to
some
very
interesting
work
by
colleagues
at
universities
who
have
examined
this
question
in
even
more
in
a
more
focused
way,
and
I
would
say
from
my
understanding
of
their
work
that
there
are
substantial
interactions
in
the
arctic
region.
C
I
think
it
has
to
do
with
just
the
power
of
sea
ice
albedo
feedback
and
between
the
effects
of
aerosol
cooling
induced
by
aerosols
and
warming
induced
by
greenhouse
gases.
So
there
there
there
have
been
demonstrated
demonstrable
non-linear
inter
interactions,
but
as
angie
said
for
the
rest
of
the
globe,
this
it's
we,
we
don't
find
any
any
significant
evidence
of
non-linearities.
A
I
want
to
add
an
exception
to
that
another
paper
that
I
use
those
simulations
to
look
at
looked
at
a
non-linearity,
that's
actually
in
response
to
temperature
for
extreme
precipitation,
and
so
that
gets
pretty
complicated,
pretty
quickly
to
try
to
interpret
yeah.
So
it
depends
to
some
extent
what
it
is
you're
interested
in.
F
And
I
think,
maybe
to
add
to
that
we
have
a
cesm2
single,
forcing
large
ensemble,
where
we
are
making
more
of
an
attempt
to
allow
the
user
to
get
closure
of
the
individual
terms.
So
we're
we
have
all
of
the
forcing
agents
involved
in
some
of
the
sensitivity
experiments.
So
we
we
can
test
foreclosure.
C
Yeah,
those
are
the
ones
that
are
underway,
so
they're,
not
re
at
all.
You
know
ready
for
for
any
scrutiny.
B
E
This
final
discussion,
I'm
wondering
about
a
linear
forcing
that
has
non-linear
responses
and
then,
if
you
were
to
remove
that
linear,
forcing
how
would
that
affect
the
non-linear
responses.
Further
on
in
the
run.
A
So
what
this
reminds
me
of
is
a
paper
that
natalie
schaller
led
a
few
years
ago.
This
is
focused
specifically
on
precipitation
responses.
I
think-
and
she
saw
some
kind
of
some
hysteresis
in
the
system.
So
if
you
force
the
system
one
way,
then
it
doesn't
quite
get
back
to
where
it
had
started.
So
that's
not
a
very
complete
or
good
answer
to
your
question,
but
there's
some
work
around
looking
at
that.
I
think.
F
So
with
sea
level,
one
aspect
related
to
your
question:
is
that
there's
a
transient
adjustment
time
scale
as
well,
and
so
it's
not
just
the
relationship
between
sea
level
at
any
one
point
in
time
in
the
forcing,
but
rather
this
gradual
adjustment
that
has
a
very
low
frequency
time
scale
where
you
wouldn't
expect
a
one-to-one
relationship
between
the
forcing
and
the
field
in
question,
so
either
sea
level
or
local
ocean
heat
content.
F
In
either
comparing
to
observations
or
comparing
across
models
is
what
that
intrinsic
time.
Scale
of
adjustment
is
and
something
that
you
see
a
lot.
So
so
this
idea
that
there's
any
single
pattern,
for
example,
one
thing
you
look
at
with
climate
climate
change
is
a
pattern
scaling
that
maybe
you
could
take
some
forcing
amount
and
relate
it
to
the
degree
of
change
with
a
given
pattern
and
some
fields
that
are
non-linear
in
the
response.
Don't
lend
themselves
very
well
to
doing
that.
B
Any
other
questions
or
mary
beth
have
a
follow-up
nope.
Thank
you,
okay,
okay!
Well,
thanks!
That
was
great
all
right,
so
we'll
move
on
to
the
next
question.
This
is
what
is
the
status
of
the
season?
2
large
ensemble?
What
is
the
current
thinking
on
how
many
ensemble
members
are
needed
to
realistically
represent
natural
variability.
C
I'm
happy
to
have
this
question
and
happy
to
be
able
to
report
to
everybody,
so
the
csm2
large
ensemble
is
more
than
halfway
complete
when
it's
completed
by
the
end
of
2020,
it
will
have
100
members
from
1850
to
2100.
C
C
Yeah
I
just
yeah.
I
knew
there
was
a
second
thing
yeah,
so
the
current
thinking,
so
it's
that
is
also
a
very
active
field
of
research,
and
there
is
no
single
number
that
for
for
the
answer.
It
very
much
depends
on
your
application.
C
There's
a
very
nice
paper
by
sebastian
milinski
and
the
title
is:
how
large
does
a
large
ensemble
need
to
be?
It's
a
very
self-explanatory
title
and
in
that
he
shows
that
you
know.
Of
course,
it
all
depends
on
your
research
interests,
the
time
scale
of
interest,
the
regional
scale
of
interest
and
also
on
the
sort
of
the
what
moment
of
the
of
the
distribution
you're
interested
in
meaning.
If
you're
interested
in
in
mean
fields
say
averages
over
30
years,
you
may
need
fewer
ensemble
members.
C
If
you're
interested
in
year-to-year
variability,
you
will
need
more
if
you're
interested
in
extreme
events,
the
tales
of
the
distributions,
especially
at
smaller
scales,
you're,
going
to
need
even
more
ensemble
members
to
be
able
to
characterize
the
internal
variability
and
also,
if
there's
a
forced
response
in
the
internal
variability.
So
I
would
point
you
to
his
paper.
C
Also,
we've
written
the
u.s
clive
are
working
group
on
large
ensembles.
We
have
a
recent
publication
in
nature,
climate
change,
that
sort
of
tries
to
talk
about
state-of-the-art
thinking
on
large
ensembles
and
trade-offs
for
the
community.
So
you
might
look
at
that.
That's
led
by
myself
and
it's
it's
on
my
website.
B
B
A
A
Actually
you
need
to
figure
out
what
region
and
what
time
scale
and
what
kind
of
variables
you're
looking
at,
for
example,
for
tropical
cyclones
here,
where
we
never
trust
the
one
degree
or
half
of
a
degree
resolution
climate
models.
We
only
use
quarter
of
a
degree
because
that
half
of
a
degree
that
quarter
of
a
degree
difference
really
makes
your
tropical
cyclone
projection
like
very
different,
because
in
quarter
of
a
degree
we
can
actually
get
very
good
climatology,
including
number
and
intensity
on
our
regional
scales,
especially
in
northwest
pacific
north
atlantic.
A
But
for
a
quarter
of
a
degree
for
a
half
of
a
degree,
your
global
tc
number
would
be
cut
in
half.
So
really
you
the
really
quite
the
real
question
is
what
kind
of
predictions
you're
looking
at
and
and
what
time
scales
are
you
looking
at
yeah
and
that
resolution
depends
on
that.
B
And
any
other
follow-ups
from
the
panelists
or
additional
questions
from
the
students.
B
So
am
I
right
that
this
session
is
over
at
9
50.
B
Right,
let's
go
for
another
one,
let's
see.
I
think
this
next
question
that
yaga
answered
about
the
init
full
field:
initial
initialization
for
the
decatur
dictator
predictability.
So
we'll
we'll
skip
that
one,
because
I
think
was
already
answered,
and
then
here
is
the
next
weather
forecast
tells
us
locally
day-to-day
weather
changes.
Climate
projections
tell
us
climate
trends
and
climatology
of
a
future
period.
B
Okay,
it
is
a
sort
of
a
mix.
A
Babble
about
this,
and
then
maybe
yaga
can
disagree
with
me.
Okay.
So
what
should
we
expect
from
decadal
prediction
in
terms
of
the
spatial
and
time
scale
of
information?
A
A
So
we
know
that
on
a
couple
year
time
scale
we've
been
doing
enso
predictions
which
can
give
you
some
predictability
like
nine
months
ahead
or
so
in
a
little
bit
of
predictability,
and
it
you
know
exactly
how
much
depends
on
where
you
are
and
what
it
is
that
you're
trying
to
look
at
and
then
when
you
go
beyond
that,
perhaps
you
can
have
some
more
I
mean,
and
so
there
are
a
couple
of
ways
that
people
kind
of
quantify
the
predictability,
and
so
I
I
think
you
know
yaga
and
others
have
done
a
lot
of
work
going
in
and
just
trying
to
ask
that
question.
A
I
think
that's
where
the
state
of
the
art
right
now
is
is
how
much
predictability
can
you
get,
and
so
it's
just
such
a
challenge
computationally.
To
just
ask
that
question.
Maybe.
C
C
You
know
we
are
depending
on
that
that
there's
some
deterministic
predictability
and
that's
that's
what
that
is
about
climate
projections.
The
predictability
comes
from
the
fact
that
we
are
prescribing
a
change
in
the
external
forcing,
for
example,
greenhouse
gases.
So
that's
the
forced
response
to
that.
C
To
that
prescribed,
forcing
decadal
prediction
is
the
hybrid
between
the
two
and
it
has
both.
So
it
has.
You
know
you're,
relying
on
that.
There
is
some
under
memory
in
the
system
of
the
initial
condition,
but
you're
also
doing
these
predictions
over
a
long
enough
time
scale
that
the
external,
forcing
the
greenhouse
gases,
for
example,
are
changing
and
that's
also
leading
to
a
a
you
know,
a
predictable
signal.
C
So
decadal
prediction
is
sitting
at
that
nexus
and
when
we
analyze
the
decadal
prediction,
large
ensemble,
for
example,
we
the
beauty
of
having
that
project
done
with
the
same
version
of
the
model
as
the
free
running
csm1
large
ensemble.
The
beauty
of
that
is
that
the
free-running
cesm1
large
ensemble
gives
you
the
response
to
the
change.
In
the
external
forcing
that
you
then
have
to
remove
from
the
dple
decatl
prediction
runs
to
see
what
what
it
was
in
the
in
within
the
climate
system,
from
those
initial
conditions
that
led
to
to
added
predictability.
C
So
you
have
both
types
of
predictability
in
these
decadal
prediction
runs
and
we're
fortunate.
We
can
separate
those
out
with
these
two.
You
know
massive
projects
here
at
ncar,
this
csm1
large
ensemble
and
the
decadal
prediction
large
ensemble.
B
But
so
I
want
to
thank
all
the
panelists
and
all
the
students
for
asking
the
questions.
If
we
didn't
get
your
questions
and
you
would
like
it
answered,
please
send
me
an
email
and
I
will
make
sure
it
goes
to
the
right
person
and
we
can
get
those
questions
answered
for
you.
So
with
that
I
think
we'll
close
gunter.
Do
you
have
anything
else,
you'd
like
to
add.
D
A
Oh
no,
I
was
just
going
to
say
for
everyone:
that's
going
to
participate
in
meet
a
scientist.
You
can
just
take
a
quick,
a
quick
break.