►
Description
The 27th Annual CESM Workshop will be a virtual event. Specifically, the Workshop will begin with a full-day schedule on 13 June 2022 with presentations on the state of the CESM; by the award recipients; and two presentations from our invited speakers in the morning, followed by order 15-minute highlight and progress presentations from each of the CESM Working Groups (WG) in the afternoon.
To learn more:
https://www.cesm.ucar.edu/events/workshops/2022/
B
C
You
feeling
actually
better,
but
I'm
in
the
netherlands.
It
was
some
visiting
and
it
was
really
really
bad
timing.
So
that's
why
my
voice
sounds
terrible.
D
E
D
E
B
B
C
Oh,
I
I
in
in
utrecht,
where
I'm
at
right
now
they
the
place,
I'm
staying,
doesn't
have
air
conditioning
it
didn't
even
occur
to
me
tomorrow
is
going
to
be
the
first
really
hot
day.
C
C
E
B
So
something
cool
is
happening
right
now.
I
don't
know
if
you
guys
are
aware,
but
like
they
put
on
the
news
every
morning,
but
there's
one
of
the
lowest
tide
in
like
several
decades.
That's
happening
right
now
with
a
minus
four
coefficient
because
like
because
the
moon
is
at
the
north,
one
of
the
closest
part
of
its
rotation,
so
people
are
excited.
B
C
You
can
move
renee
and
I
up
if
you
want
that's
completely
fine.
B
Well,
yeah
renee's
here.
F
D
We
are
in
the
same
building,
yeah
technically
on
the
same
floor.
Only
in
different
rooms.
D
No,
no,
I
know
in
the
netherlands,
so
now
it's
not
a
problem,
but
I
think
the
reason
why
tree
is
saying
is
because
we
have
a
group
barbecue
outside
in
the
garden.
B
G
So
good
morning
everyone
who's
coming
in,
let's
wait,
maybe
two
more
minutes
or
so
and
then
get
started
with
a
brief
introduction
of
the
co-chairs
and
liaisons.
So
you
know
who
to
talk
to
if
you
have
any
issues
or
questions
about
the
land
ice
or
polar
climate
working
groups,
and
then
we
go
we'll
get
started
with
our
talks.
G
Okay,
we
have
40
participants
already,
which
is
fantastic,
so
I
want
to
get
started
the
recording
in
progress
there
we
go
so
this
is
the
official
start,
welcome
everyone,
everyone
joining
remotely
from
boulder
or
around
boulder
or
all
around
the
world.
We
have
several
people
from
europe
joining,
for
example,
from
other
parts
of
the
world,
so
that's
really
exciting.
G
First
of
all,
my
name
is
john
leonard,
I'm
assistant
professor
at
the
university
of
colorado
boulder
and
also
the
co-chair
of
the
llennis
working
group,
along
with
bill
lipscombe,
and
I'm
going
to
share
the
first
part
of
the
session
here
this
morning.
A
few
things
to
note.
So
there
are
two
kind
of
three
parts
of
our
of
our
morning
session.
Here
we
have
a
first
part
of
six
talks.
Then
we
have
a
break.
G
Then
we
have
three
more
talks
and
then,
after
those
three
talks,
we
have
a
third
and
a
very
exciting
part
of
our
morning
session,
which
will
consist
of
breakout
rooms.
Now
we,
as
co-chairs
and
liaisons,
have
come
up
with
a
list
of
breakout
room
subjects,
topics
to
discuss
and
one
of
us
is
going
to
share
a
link
to
a
google
doc
right
now
there
we
go.
Thank
you
marika.
So
if
you
look
in
the
chat
of
zoom,
there
is
a
link
to
a
google
doc
and
that
will
have
a
list
of
breakout
room
subjects.
G
Now
we
have,
as
you
can
see,
a
list
of
exciting
topics,
but
if
you
think
you're
missing,
one
of
you
know
your
favorite
topics.
You
can
see
that
there
are
two
wild
cards
topics
in
the
on
the
bottom,
so
feel
free
to
edit
or
add
any
of
your
favorite
topic.
So
we
can
add
more
topics,
as
we
run
through
the
first
part,
two
parts
of
our
session.
So
let's
get
started
here
by
first
introducing
our
leadership
of
the
land
eyes
and
polar
climate
working
group.
G
Please
go
ahead.
B
Thank
you,
jan
well.
First,
for
in
the
way,
by
way
of
introductions,
I
wanted
to
introduce
gundry
leggy,
the
science
liaison
for
the
land
ice
working
group
and
kate
thayer
calder,
the
software
engineering
liaison.
B
Also,
I
wanted
very
much
to
thank
john
lennox
for
his
service,
as
co-chair
for,
I
believe,
the
last
six
years,
jan,
it's
been
a
great
pleasure
working
with
you
and
thank
you
very
much
for
your
service
and
also
I
wanted
to
welcome
mirin
viscano,
who
has
accepted
thanks
marin,
who
has
accepted
a
two-year
term
as
the
upcoming
co-chair
of
the
land
ice
working
group.
So
thanks
again
to
yon
and
and
thanks
to
marin
as
well
for
stepping
up
rika.
E
Sure
hi,
so
I'm
erica
holland,
I'm
one
of
the
co-chairs
for
the
polar
climate
working
group
and
my
other
co-chair
is
hansi
singh
from
university
of
victoria.
There
she
is,
and
our
liaison,
who
fills
both
the
science
liaison
and
software
engineering
liaison
roles,
is
dave,
bailey,
so
who's
here
at
encar,
so
reach
out
to
any
of
us.
You
know
if
you
have
questions
about
the
group,
or
you
know,
topics
that
you
want
to
make
sure
we
focus
on
things
like
that.
E
I
also
did
want
to
mention
that
the
discussion
session
today,
which
you
know
you
you,
have
the
link
in
the
chat
for
the
topics
and
so
in
addition
to
adding
topics,
if
you
want
to
know
which
group
you
are
planning
to
go
to
and
if
there's
some
groups
that
have
no
one
interested,
we
might
scrap
them
in
the
interest
of
a
different
group,
but
also
to
note
that
these
breakout
discussions
are
in
part
to
inform
our
computing
proposal,
which
we'll
have
to
write
at
the
end
of
this
summer
and
submit
in
the
fall.
E
And
so,
if
you
have
particularly
exciting
experiments
in
mind
that
you'd
like
to
run
using
polar
climate
or
land
ice
working
group,
resource
resources,
definitely
this
discussion
that
we'll
have
at
the
end
of
the
session
you
can
bring
those
up
because
we're
always
looking
for
exciting
experiments
for
our
computing
resources.
So
this
will
in
part
feed
into
what
we
write
in
that
proposal
and
what
resources
we
end
up
with.
E
So
I
just
wanted
to
note
that
I
don't
know
if
there's
anything
else,
that
my
co-chair
liaison
wanted
to
say
or
anything
else.
The
landice
working
group
wanted
to
make
sure
we
covered.
Oh,
oh
one,
other
thing
I
just
I
don't
have
our
code
of
conduct
up,
but
I
did
just
want
to
mention
that
you
know
we
do
have
a
code
of
conduct,
so
everybody
please
be
respectful
share
the
air
and
you
know,
give
constructive
feedback
to
your
colleagues.
E
G
So
much
marika
and
thanks
bill
for
the
kind
words
and
condolences
to
miranda
indeed
so
marika,
I
don't
know
if
it's
me,
but
I
don't
have
editing
access
to
the
google
doc.
If
so
it's
for
me,
I
think
it's
impossible
to
add
my
own
name
or
change
anything.
So
if
you
can
take
a
look.
E
Okay,
I'll
try
to
change
that,
while
the
talks
are
starting.
Sorry
about
that,
I
thought.
E
G
Yeah
perfect,
so
let's
get
started
with
the
the
science,
so
we
have
six
talks
if
I'm
counting
right,
yes
before
the
break,
the
first
of
which
is
will
be
given
by
renee
weingart
about
exploring
high
mountain
asia
with
vr
cesm.
G
So
just
for
the
speakers,
let's
try
to
limit
your
talks
to
around
12
minutes,
so
we
have
around
three
minutes
for
switching
speakers
and
questions.
Obviously,
for
the
audience,
you
can
ask
your
questions
in
the
chat
or
you
can
speak
up
after
the
the
talk,
obviously,
and
for
the
speakers,
you
can
also
answer
any
remaining
questions
in
the
chat
after
your
talk
or,
of
course,
connect
to
your
audience
after
your
talk
in
in
a
virtual
way,
so
renee
without
further
ado.
G
Please,
if
you
can
share
your
screen
and
we
can
get
started
yes,
and
I
will
give
you
a
two.
D
D
D
Just
a
short
introduction
on
the
highmountain
hvr
7
grid:
it's
a
grid
with
original
refinement
up
to
seven
kilometer
over
high
mounted
asia.
We
applied
this
grid
in
the
following
setup,
using
csm2
with
cam6
cln5
and
the
no
evolve
diagnostic
community
ic
model.
Further,
we
applied
an
lfh
and
class
sound
scaling
scheme
with
36
elevation
classes,
which
is
suitable
for
high
mountain
asia.
D
Some
statistics
about
this
grid
we
applied
chem
physics,
time
step,
225
seconds
dynamics,
timestamp
is
18.75
and
that
eventually
results
in
a
model
cost
of
about
90
000
core
hours
per
simulated
year.
And
if
you
do
that
on
about
1600
cpu,
you
will
end
up
with
10
years,
multiple
years
in
an
actual
month,.
D
We
did
eventually
two
model
simulations.
The
first
simulation
was
completed
around
2020
and
this
simulation
mainly
followed
defaults
on
csm2,
with
exception
of
the
maximum,
so
that
we
found
out,
unfortunately
that
after
this
simulation
that
the
surface
mass
balance
was
not
that
good.
So
that's
why
we
did
another
simulation
which
was
completed
at
the
end
of
last
year,
and
in
this
simulation
we
did
a
moderate
spin-up
procedure.
D
We
changed
some
parameters,
for
example
maximum
snow
depth.
In
order
to
have
a
high
re-freezing
capacity,
we
increased
the
ice
albedo.
We
changed
the
rational
report
tensioning
further.
We
did
some
tunings
on
the
cloud
cover
mainly
in
club
scheme
and
sea
ice,
and
we
included
the
morrison
gatman
3
cloud
micro
physics
scheme,
which
is
actually
the
mt2
plus
crowd
crowdball
and
further
we
applied
an
updated
glacier,
coffer
dataset.
D
All
the
outputs
are
evaluated
by
comparing
them
with
yeah
what
we
have
in
highmountain
asia
and
what
I
could
find
and
that
consists
of
created,
outputs
derived
from
the
globally
uniform
one
degree
grids,
rear
analysis
and
satellite
based
products,
as
well
as
rcm
based
outcomes
from
related
to
glacier
model.
D
What
we
learned
is
that
there
are
one
biases
mainly
in
summer
time
and
also
in
particular,
over
the
northern
regions
of
sub-regions
of
high
mountain
asia,
and
these
warm
biases
generally
become
smaller.
With
increasing
resolution
contrast,
however,
in
the
southwestern
region
of
high
mountain
asia,
particularly
there,
we
see
cold
biases
and
those
coal
biases
also
become
larger,
increasing
resolution.
D
The
pattern
correlation
also
kind
of
demonstrates,
although
the
pattern
correlation,
are
very
good
for
for
yeah
every
season,
but
you
see
that
there's
a
slightly
difference
between
the
winter
and
the
summertime
where
summer
temperature
seems
to
be
have
a
better
pattern.
Simulation
result
compared
to
watch
forcing
every
interim,
then,
in
the
winter
time.
D
If
we
look
into
rainfall
and
snowfall
rainfall
is
overestimated,
mainly
in
summertime
and
in
the
monsoon
dominated
regions
where
the
bias
tends
to
be
yeah.
I
think
yeah
we're
yeah
the
slightly
larger
for
the
one
degree
grid
than
for
the
the
high
mountain
asia
result
simulations
and
if
we
look
at
the
pattern
correlations,
the
pattern
correlates
are
generally
better
for
the
high
mountain
asia
simulations
except
for
in
the
winter
periods.
D
D
Then
we
move
on
to
snow
cover
and
snow
depth.
The
snow
cover
eventually
decided
to
compare
it
with
the
the
weekly
snow
cover
extent
of
the
nsidc
and
from
those
comparisons.
We
learned
that
the
snow
cover
is
slightly
overestimated
during
winter
time,
but
in
generally
well,
except
for
the
northeastern
region
of
high
mountain
asia
and
further
it's
underestimated
in
summer
time.
D
We
also
see
some
improvements
compared
to
the
ne30,
so
the
one
degree
grid
and
that
in
in
alton,
the
best
generally
the
best
match
is
there.
If
we
look
at
the
snow
depth,
we
compare
that
one
with
jr55,
which
is
generally
the
the
the
one
of
the
better
data
sets
in
producing
snow
depth
so
compared
to
grf
55.
We
also
see
overestimations
in
the
in
the
south
western
region,
which
also
could
be
linked
to
the
core
bias
and,
to
some
extent,
to
the
snow
model,
modifications.
D
Then
I
would
like
to
show
you
the
glaciological
surface.
Mass
balance
results
in
this
case.
We
compared
it
with
observation,
so
like
the
glaciological,
geodetical
observations
and
rcm
based
outcomes.
In
this
case
it
will.
It
is
warf
based
glacier
model
outcomes,
and
we
see
that
the
the
high
mountain
asia
vr
surface
mass
balance
is
underestimated
compared
to
yeah
those
outcomes.
D
We
also
see
that
there's
a
big
improvement
in
in
the
second
simulation
and
that's
particularly
in
the
southeastern
region,
the
further
the
number
of
grid
cells
very
positive
surface,
mass
balance
simulated
at
the
end
of
round
triples
from
about
100
to
about
300,
and
that's
mainly
also
in
the
southern
part
of
high
mountain
asia.
D
D
So
that
results
in
also
slightly
more
positive
surface
mass
balance.
And
if
we
look
at
the
coupler
output,
as
is
here
that
goes
into
the
the
ice
sheet
model,
then
we
see
that
in
especially
in
the
higher
elevation
zones
that
there's
even
positive
surface
mass
patterns,
so
that
could
also
have
contributed
to
it,
but
still
yeah.
There's
an
still
some
work
to
do.
The
s
p
bias
is
still
quite
negative,
and
one
of
the
likely
reasons
for
that
is
that
there's
a
cloud
call
for
bias.
D
I
did
not
show
that
kind
of
figures
here
in
this
presentation,
but
we
know
that
the
cloud
cover
is
underestimated
over
highmountain
asia,
which
results
eventually
in
more
incoming
solar
radiation
and
a
smaller
downwelling
long
wave
radiation,
so
the
the
at
the
end.
It
eventually
ends
up
in
a
higher
net
radiation
at
surface,
and
that
might
eventually
contribute
to
a
one
temperature
bias
over
central
asia
as
well
potential
solution,
for
that
could
be
the
notch
cam
with
era
interim
or
5..
D
Also
there's
some
missing
physics.
We
did
not
take
into
account
the
effects
of
hill
slope
aspect
and
orientation
could
also
be
relevant,
especially
over
high
mountain
asia.
About
27
percent
of
the
glaciers
are
north
facing
and
their
somehow
topographic
shading
place
has
some
effect
on
the
solar
illumination.
D
Elevation
down
scaling
is
currently
only
applied
over
glacier
iceland
units
and
that
can
be
a
disadvantage,
especially
in
in
grid
cells
with
a
very
small
ice
fraction,
because
as
we
as
I
see
it,
if
you
have
a
very
small
eye
fraction,
the
other
length
unit
will
mean
most
likely
dominate
the
the
signal
or
the
grid
cell
mean
and
the
effects
of
lfh
and
downscaling
on
the
grid
cell
mean
that
goes
into
camp
is
limited,
and
that
is
also
demonstrated
a
bit
in
this
figure
where
I
plotted
the
surface
mass
balance
against
the
glacial
fraction
in
several
smaller
elevation
zones,
and
everywhere
you
look,
you
see
that
with
a
smaller
glacier
fraction,
you
have
a
more
negative
surface
mass
balance
and
yeah.
D
That's
why
I
think
it
could
be
related
to
that
so
probably
to
solve
this
problem.
We
also
need
to
apply
lfh
downscaling
over
the
factor.
Data
length
fraction
as
well
and
further
could
be
an
idea
to
look
into
improving
the
couple
coupled
the
coupling
between
land
and
atmosphere,
subcrite
heterogeneity.
So
like
the
clasp
thing
that
is
now
done,
I
believe
my
american
folder.
D
So
yeah,
this
is
a
bit
the
summary
of
the
talk
we
see,
indeed,
that
there
are
so
still
some
biases,
but
we
also
see
improvements
in
the
temperature
and
precipitation
the
snow
cover
observation
also
shows
some
better
agreement
with
the
one
that
we
grid
snow
depth
is
mainly
overestimated,
where,
as
we
see
that
the
first
simulation
shows
a
slightly
better
agreement
than
the
second
simulation
most
likely
due
to
the
model
snow
model
modification.
D
But
I
have
to
say
that
this
is
a
kind
of
yeah
how
to
say
that
no,
I
cannot
find
the
right
word,
but
the
idea
is
that
jra55
is
also
a
model,
and
it's
also
based
on
observations
that
are
there.
But
observations
are
quite
scarce
in
high
mountain
asia,
so
that
makes
it
also
a
bit
difficult
to
compare
with
anyway.
The
good
news
is
that
we
can
simulate
a
positive
surface
mass
balance,
but
the
challenge
is
that
we
still
need
to
do
a
lot
of
work
to
improve
the
the
surface.
D
Mass
balance,
mainly
yeah,
to
tackle
with
problems
with
like
cloud
cover,
one
temperature
bias
lacking
model
physics
and
the
elevation
downscaling,
and
this
will
be
actually
my
last
slide
for
today.
This
is
a
bit
about
future
or
the
current
work
I'm
doing
at
email.
D
So
that
means
we
will
end
up
with
a
lower
climate
sensitivity
and
the
reason
why
we
need
it
is
that
eventually,
we
also
want
to
look
into
the
future
to
see
what
sst
or
sea
ice
changes
derived
from
storylines,
how
that
will
eventually
affect
the
polar
climate
or
maybe
even
changes
in
the
lower
latitudes.
D
G
Much
renee,
unfortunately,
since
you've
almost
used
your
15
minutes
of
time.
We
need
to
switch
to
the
next
speaker,
but
if
you
have
any
questions
for
rene,
please
put
them
in
a
chat
and
then
renee
can
answer
there
in
the
in
the
chat.
Also,
while
we
were
switching
to
the
next
speaker
miran,
please
share
your
screen.
G
Please
take
a
look
at
the
chat
and
the
new
link
that
marika
provided
where
you
can
sign
up
for
breakout
rooms
and
or
add
new
ideas
for
breakout
room
topics
so
miran
if
you're
ready
to
go
for
it.
Please
do
miranda
will
talk
about
future
deglaciation
of
the
greenland
ice
sheet
using
the
coupled
csm2
system,
2.
go
ahead,
mirin.
H
Okay,
sorry
yeah,
I
was
panicking
with
the
with
the
mute
okay,
you
can
see
my
screen
and
you
can
hear
me
right.
H
Okay,
that's
good
okay!
So
I'm
gonna
discuss
here
the
complete
degradation
of
the
gryllian
asset
with
the
couple
csmcism.
H
H
Okay,
this
is
a
background
slide
with
a
carving
green
lana.
I
said
mass
law,
so
it
is
contributing
to
sea
level
rise
by
0.7
millimeter
per
year,
and
when
we
look
at
the
projections
in
the
last
ipcc
report,
what
you
can
see
is
that
most
of
them
are
based
on
icing
models
that
are
false
with
the
output
of
a
regional
or
global
climate
models,
and
in
this
case
we
are
missing
the
couplings
between
isis
and
climate
at
all
scales.
H
Now
from
the
global
to
the
regional
and
the
ict
scale,
and
on
the
other
hand,
we
also
have
some
coupled
simulations,
and
in
general
they
they
have
simplified
processes,
for
instance
in
the
melt
or
refreshing
calculations.
H
So
so
here
the
other
value
of
using
csm
cism
is
that
we
are
combining
high
detail
with
the
coupling
together,
and
the
question
that
I
am
addressing
today
is
how
fast
and
by
which
mechanisms
will
the
ice
it
melt
under
high
greenhouse
greenhouse
greenhouse
gas
foresee.
So
this
is
going
to
be
a
one
percent
increase
to
four
times
co2
and
we
already
published
the
results
up
to
year
350
and
we
are
preparing
the
results
from
then
until
they
complete
the
glaciation,
and
this
is
what
I'm
gonna
show
you
today.
H
H
Okay.
So
in
the
simulations
we
apply
coupling
one
to
one
between
the
climate
and
I
said
components
up
to
year,
500
and
from
there
we
accelerated
by
five.
So
we
were
using
one
clam
a
year
to
force
five.
I
said
model
years,
so
by
the
end
of
the
simulation,
we
have
a
global
warming
of
more
than
10
kelvin
and
at
the
beginning,
the
arctic
warming
is
more
than
the
warming
over
the
green
and
ice
head.
H
H
Okay
and
the
I
said
is
lost
in
1700
years,
which
is
very
fast,
and
this
is
the
pattern
of
the
glaciation
here.
You
have
the
percentage
at
different
years,
and
this
percentage
is
the
percentage
of
original
area
so
by
year,
700
more
or
less.
We
are
at
the
half
area
than
we
are
at
year,
one
and
you
can
see
that
we
are
losing
the
southern
part
and
we
don't
leave
anything
there
and
then
the
ice.
It
is
moving
towards
this
location.
H
This
is
the
mean
elevation
of
the
bedrock,
so
you
can
see
how
the
ict
is
retreating
at
the
end,
at
an
area
of
relatively
high
bedrock-
and
you
can
see
also,
what
is
the
evolution
of
the?
How
much
is
contributing
isostasy
and
it
is
very
little
so
it
is
100
details
by
end
of
simulation,
and
this
very
small
contribution
is
because
here
everything
goes
very
fast
and
yeah.
H
It
won't
make
much
of
an
effect
okay.
So
this
is
this,
is
the
partition
of
the
mass
balance
here
is
in
in
the
components
of
smb
discharge
and
basal
melt.
So
please
follow
with
me
the
black
line.
So
when
we
are
here
in
here
200
the
contribution
is
two
millimeter
per
year
and
then
we
go
to
a
maximum
of
eight
millimeter
per
year
here
by
by
tier
500.
So
the
question
is
what
is
causing
this
this
very
strong
acceleration
here.
H
So
this
this
a
huge
increase
in
the
contribution
to
syllable
rise,
so
I'm
gonna
switch
only
I'm
gonna
be
focusing
on
only
in
the
surface
processes,
so
we
are
gonna,
have
contributions
from
precipitation
and
the
contributions
from
runoff.
So,
let's
start
with
precipitation
and
the
fraction
of
snowfall
is
decreasing
from
90
percent
at
the
end
to
30
percent
by
the
end
of
the
simulation
and
okay,
these
are
the
maps
of
how
the
precipitation
is
evolving.
H
It
is
increasing
all
over
the
assets
and-
and
here
are
the
contributions
from
global
climate
from
climate
change.
Elevation
change.
Nice
cover
change
to
determine
those.
What
we
do
is
we
compare
with
a
simulation
that
has
fixed
topography,
so
the
precipitation
is
increasing
both
because
of
the
climate
change,
but
also
because
of
the
elevation
change
that
we
have.
H
The
contribution
from
changing
location
of
the
asset
is
negative,
because
the
ice
is
retreating
from
areas
that
have
very,
very
strong
as
no
very
high
as
no
accumulation
today
as
the
the
atlantic,
the
north
atlantic
margin.
Now
the
southeast
of
of
greenland
has
very
very
high
accumulation
rates
today.
H
Okay,
so
all
of
the
increase
of
the
am
precipitation
goes
to
rainfall.
There's
no
fall.
If
is
practically
constant.
Here
we
are
averaging,
there's
no
fall
and
we
are
the.
We
are
scaling
by
area,
so
we
are
dividing
by
the
total
area
of
the
I
set,
because
if
we
put
total
mass,
the
total
mass
is
gonna
be
going
down
because
the
ic
is
getting
smaller
and
then
we
cannot
distinguish
processes.
We
are
dividing
by
area
in
these
numbers
here,
okay
and
okay.
So
all
goes
to
rainfall
only
increase
the
ablation
areas.
H
They
are
expansing
expanding,
very,
very,
very
fast.
So
you
can.
You
can
see
that
by
the
year
thousand,
all
of
the
acid
is
ablation
area,
and
in
this
period
here
in
the
first
in
between
year
120
and
year,
200,
we
have
a
very,
very
fast
expansion
of
the
ablation
area.
H
How
does
this
compare
with
amy
and
simulation?
It's
it's
very
different,
because
here
we
don't
see
that
super
expansion
of
the
of
the
ablation
area
as
a
percentage
of
the
total
area
of
the
eyes
okay.
So
these
are
the
time
series
for
the
melt
and
the
refreezing,
so
the
re-freezing
so
initially
is,
is
taking
45
percent
of
the
melt
as
the
and
the
melt
is
increasing
and
increasing,
as
the
re-freezing
increases
fast
up
to
more
or
less
the
amount
of
annual
snowfall
rates.
H
And
then
the
percentage
goes
down.
H
Okay,
so
rephrasing
works
for
a
little
while
and
then
cannot
keep
up
with
this
okay,
okay,
so
the
contributions
to
the
melt.
So
the
elevation
change
is
contributing
a
large
amount
by
the
end
of
the
simulation.
So
it
is
by
year
800
that
the
contribution
from
climate
and
elevation
they
become
similar
incoming
radiation,
so
the
solar
is
going
down
and
the
long
wave
incoming
is
going
up
and
if
now
we
look
at
all
of
the
fluxes.
H
What
we
are
seeing
is
that
at
the
beginning,
most
of
the
melt
is
coming
from
the
energy
of
the
incoming
of
the
net
a
long
wave.
But
then
the
albedo
feedback
is
kicking,
and
so
the
shore
wave
radiation,
the
solar
radiation,
is
dominating
the
contribution
to
melt
almost
until
the
end
of
the
simulation
and
at
some
point,
the
sensible
and
latent
heat
fluxes
become
quite
similar
to
the
long
wave
contribution.
H
Okay,
so
going
into
the
shortwave
sorry
latent
and
sensible
contributions,
so
they
they.
They
become
very
important
when,
when
most
of
the
surface
reaches
melt
point
and
then
the
atmospheric
temperature
continues
to
increase.
So
then
we
have
a
very
strong
temperature
inversion
close
to
the
surface,
and
this
is
increasing
very
much
the
fluxes,
except
at
the
end
of
the
simulation
when
the
the
winds
go
down
very
much,
probably
because
the
the
topography
gradient
goes
down
very
much
okay.
So
this
is
the
summary
slide.
H
So
what
we
have
been
seeing
is
that
albedo
feedback
contributes
very
much
to
the
melt.
The
elevation
feedback
as
well.
There
is
a
very
small
contribution
of
the
negative
melt,
esostasy
feedback
to
the
whole
process,
because
this
deglazation
process
is
very
very
fast
and
the
glacialized
static
adjustment
goes
slowly
and
yeah.
It
is
very
fast
and.
H
G
G
I'm
wondering
if
the
total
precipitation
is
changing
in
the
sense
that
instead
of
snowfall,
there
is
more
rainfall,
so
the
total
precipitation
changes,
but
then
the
amount
of
snowfall
does
not
increase
very
much.
Can
you
comment
on
that.
H
Yeah
it
is,
it
is
exactly
as
you
are
saying,
the
the
total
so
here,
because
the
the
size
of
the
ict
is
changing
so
much.
We
have
to
divide
the
gigaton
space
here
by
the
area,
okay
and
and
when
we
do
that
yeah.
So
when
we
do
that,
okay,
you
would
have
to
put
these
three
lines
together.
You
would
have
to
add
them,
and
this
will
give
you
the
the
total
change.
H
But,
okay,
let's
let's
focus
on
the
so
the
elevation
change
and
the
global
climate
change
are
giving
its
200
millimeter
of
increase
right
more
or
less
so
all
of
it
is
going
to
the
rainfall
increase,
and-
and
this
is
the
line,
for
instance
for
the
climate
related
snowfall
increase-
and
it
is
flat
it's
around
the
sido
and
for
the
snowfall
is
going
sorry
for
the
elevation
contribution
is
going
down.
So
as
the
elevation
is
going
down,
we
are
sending
more
to
rainfall,
even
so,
if
even
snowfall
is
losing
extra.
H
G
Yeah
great
time
for
one
more
question.
I
J
Yeah
hi
marin,
thanks
for
that
great
talk,
this
is
super
cool
work,
so
I
guess
my
question
is
the
extent
to
which
these
results
are
sensitive
to
the
background
climate
sensitivity
in
the
model,
just
because
this
model
is
so
sensitive
does
like
so
does
that
mean
that
if
a
model
is
more
realistically
sensitive
like
this
process
will
take
like
2500
years
like
I,
I
just
I
don't
understand
exactly
how
like
the
ice
sheet
dynamics
and
physics
interacts
with
the
climate
sensitivity,
and
so
it
would
be
lovely
to
hear
your
thoughts
on
that.
H
Yeah,
so
it
would
be
great
to
to
check
that
with
a
simulation
where
we
can
correct
for
the
high
climate
sensitivity,
okay,
so
jian-
that
I
think
is
in
the
room
in
the
visual
room.
I
can
tell
you
more
about
this,
but
what
I
am
understanding
so
far
is
that
this
very
high
cloud
feedback
is
is
more
intense
for
warming
over
4
kelvin,
which
is
what
we
have
here.
H
H
These
results
not
so
much
in
the
context
of
a
four-time
co2
simulation
when
we,
when
we
compare
to
other
simulations,
but
we
can
compare
them
to
other
simulations
with
10
kelvin
of
warming
to
make
sense
of
the
time
scale
also.
So
there
is
an
acid
only
simulation
that
uses
ssp
5
8.5
that
melts
the
I
said
within
a
thousand
years,
so
it
is
even
a
shorter,
so
it
is
by
andy
assander
and
colleagues
so
they're.
I
think
they
are
working
also
within
kelvin
of
global
warming.
H
So,
and
they
are
even
faster
than
these-
I
think
it
is
it.
It
is
because
there
are
some
overestimations
there
for
the
lapse
rate
and
also
for
the
warming
that
is
applied
over
greenland
but
yeah.
So
it
is.
We
are
not
even
the
fastest
with
this,
and
it
would
be
great
to
to
revise
the
results
with
a
more
adjusted
timing.
Sensitivity.
Yes,.
G
All
right,
thank
you
so
much
we're
eating
into
the
next
speaker's
time.
So,
let's
move
on.
If
you
have
any
more
questions,
please
put
them
in
the
chat
and
then
feel
free
to
answer
there
and
follow
up.
So
the
next
speaker
is
registry,
3
data,
so
3.
If
you're
ready
and
miranda
stopped
sharing.
So
that's
great
there
we
go
smooth
transition,
so
tree
is
going
to
talk
about
again
to
about
the
vrc
esm
2,
but
now
applied
to
the
antarctic
ice
sheet
so
tree.
The
floor
is
yours.
C
Okay,
so
better.
G
C
Excellent
okay,
so
I
think
jan
already
covered
exactly
what
I'm
gonna
be
talking
about.
I'm
gonna
be
evaluating
the
variable
resolution
csm2,
which,
over
what
I've
been
calling
the
ansi
grid
over
the
historical
record,
and
this
is
from
1979
to
2014..
This
is
with
john
leonard's,
adam
harrington,
david
schneider
and
devin
dunmer.
C
I'd
like
to
thank
nasa
and
the
nsf
as
well
for
for
their
funding
and
support,
okay,
so
to
first
introduce
the
grid.
It's
just
right
over
here.
This
is
a
quarter
degree
resolution
over
the
interior
and
one
degree
in
the
larger
domain
and
the
outputs
that
we're
going
to
be
evaluating
here
from
1979
to
2014,
but
there
are
additional
outputs
from
2015
to
2021,
which
has
moisture
tagging
implemented.
C
These
have
forced
sst
and
cs
concentration,
so
this
is
set
up
just
like
the
amit
simulations
and
uses
cam6sse
and
the
cost
right
now
for
for
this
version.
Without
the
moisture
tagging
is
twenty
three
thousand
four
hours
per
year,
so
we've
also
produced
three
hourly
outputs
for
the
atmosphere
to
facilitate
the
detection
of
atmospheric
rivers,
as
well
as
from
the
surface
in
order
to
potentially
force
fern
models.
C
So
one
actually
one
thing:
it's
an
aside:
it
doesn't
reflect
very
well
here.
An
interesting
factor
of
the
s
this
this
grid
is
that
actually,
the
the
resolution
towards
the
interior
is
is
in
fact
higher
in
the
fv
grid.
One
degree
grid
just
simply
because
the
way
it
converges
so
there's
an
interesting
case
of
where
you're
getting
a
lot
of
resolution
over
the
ocean
in
particular,
but
not
necessarily
towards
the
center
of
the
the
ice
sheet.
C
So
I'm
going
to
keep
that
in
mind,
so
just
to
get
a
picture
of
what
resolution
means,
at
least
towards
the
elderly
of
the
domain.
This
is
over
the
antarctic
peninsula,
where
you
have
a
lot
of
complex
topography
and
the
sc
points
are
shown
in
dots,
whereas
the
one
degree
is
shown
in
in
grid
lines.
C
So,
as
you
can
see,
it
captures
quite
a
bit
more
of
topography
and
it,
and
the
map
on
the
right
is
the
difference
in
surface
topography
of
ansi
minus
the
one
degree
resolution
and
you
can
see,
there's
quite
a
lot
of
differences
and
just
for
for
conventions.
Throughout
this
talk,
I'm
going
to
be
showing
a
lot
of
maps,
it's
always
going
to
be
ansi
minus
the
the
reference
and
where
it's
gray,
that
means
the
standard
deviations
that
are
temporal
for
ansi
and
in
the
case,
amid
ensemble
standard
deviations,
do
not
overlap.
C
So
that's
where
it's
significant
so
going
back
to
this.
What
main
takeaway
here
is
that
you
have
slight
lowering
over
over
ice
shelves
and
a
lot
of
resolution
of
topography,
but
a
higher
interior
and
especially
the
peninsula,
so
the
peninsula
is
going
to
come
up
in
a
second
again.
C
So
the
main
question
we're
asking
here
is:
what
are
the
impacts
of
horizontal
resolution,
horizontal
resolution
alone
fixing
everything
else
and
those
are
effects
over
the
ocean,
as
well
as
the
ice
sheet
and
we're
primarily
focusing
here
on
surface
mass
balance
within
the
space
of
this
top.
So
what
we're
doing
here
is
comparison.
Comparing
these
with
a
map
runs
over
the
same
period
using
a
10
number
ensemble,
but
these
are,
with
the
one
degree
model.
So
what
this
means?
C
We
have
the
same,
forcing
the
same
physics
but
just
a
different
resolution,
and
I
don't
know
why
that
got
cut
off,
but
it's
validated
with
era
5,
as
well
as
the
medley
reconstruction
of
which
uses
furnished
course
to
reconstruct
smb
and
finally,
with
rockmo
2.3,
which
is
a
rcm,
a
regional
climate
model
at
a
higher
resolution.
C
So
what
I'm
showing
here
is
the
difference
in
surface
mass
balance.
Again
it's
nc
minus
amp
on
on
the
left
side
and
a
lot
of
the
differences
conform
towards
to
topography.
But
you
know
what
I
want
to
point
out,
especially
is
just
the
the
loss
of
smb
that's
over
here
and
how
much
how
much
of
a
change
that
is,
but
you
know
mostly
towards
the
interior.
You
have
a
amount,
a
gain
towards
the
interior
and
east
antarctica.
You
have
a
really
strong
gain
of
smb.
C
C
When
has
that
diverged
and
so
in
green
is
where
the
the
bias
has
been
reduced
and
and
in
purple
is
where,
where
the
bias
has
been
increased
in
anc
as
compared
to
ameth,
and
so
we
can
see
here.
Is
you
have
some
pretty
stark
patterns,
which
is
you
know
towards
this
part
of
west
antarctica?
You
have
some
improvements
in
symbian,
that's
mostly
just
where
it's
decreased,
whereas
over
east
antarctica
there's
a
little
bit
of
a
disagreement
between
the
reconstruction
versus
rockmo.
So
it's
there's
some
dispute
there.
C
But
if
we
look
at
the
integrated
s
b
and
surface
mount,
what
I'm
showing
here
is
for
the
full
entire
full
ice
sheet
for
the
grounded
ice
sheet
and
for
ice
shelves
is
surface
mass
balance
and
with
east
antarctica,
west
antarctica
and
the
antarctic
peninsula
split
up.
C
They're
kind
of
main
takeaways
here
is
that
this
is
the
standard
deviation
for
the
ensemble
for
ansi
ansi
is
shown
in
blue
and
oranges
is
for
aimip
and
most
of
the
time
the
the
the
standard
deviation
actually
overlaps
in,
except
in
the
case
of
the
antarctic
peninsula,
or
you
know
in
most
regions,
you
know
over
the
ice
shelves.
It
does,
but
not
every
antarctica
and
not
over
the
antarctic
peninsula
and
whoa.
C
Sorry
and
the
other
thing
is
I'm
touching
on
surface
melt
here
just
really
briefly,
but
that's
not
the
major
focus
of
this
top.
You
have
a
lot
more
surface
mount
over
east
antarctic
ice
shelves
in
particular,
and
that's
kind
of
important,
but
the
main
thing
is,
even
though
you
have
a
major
increase
in
surface
melt
overall
in
nancy
versus
emma
the
difference
between
amip
and
the
best
case
of
observations
that
we
have
is
that
it's
it's
already
pretty
enormous,
so
it's
it
makes
a
already
bad
bias
and
aimip
a
little
bit
worse.
C
So
the
main
reason
going
back
to
the
surface
mass
balance.
Most
of
this
increase
in
smb
is
driven
by
winter
precipitation.
So
what
I'm
showing
here
is
the
smb
first
of
all,
and
then
the
difference
in
large
precipitation
the
mean
difference
in
over
winter
and
over
summer,
and
this
is
relative.
So
it's
it's
nc,
minus
a
map
divided
by
ameth
times
100..
C
Well,
we
can
mostly
see
is
two
major
takeaways
here.
One
is
that
most
of
the
precipitation
is
happening
during
winter
and
over
the
east
antarctic,
which
matches
up
with
with
a
lot
of
this
gain,
and
the
other
thing
is,
I
want
to
point
out
the
wet
lc
sector,
so
this
is
some
of
the
effects
of
enhanced
topography
over
the
entirety
peninsula.
So
you
can
see
enhanced
precipitation.
C
Now,
what
I'm
showing
here
is
taking
a
step
back
for
that
you
know-
and
this
is
all
during
winter-
is
what
are
the
sources
of
moisture?
Where
is
this
coming
from
so
we
know
that
there's
a
lot
of
precipitation
in
winter,
over
east
antarctica,
and
so
I'm
showing
here
is
integrated
vapor
transport,
and
this
is
for
meridianal
here
and
zonal
here
on
the
right
and
the
major
takeaways
here
are
that
you
have.
C
You
have
a
lot
of
increased
meridian
transport
over
east
antarctica
generally
in
in
ansi,
but
that
it's
coming
from
sectors
over
here
and
over
here
and
I'll
be
piecing
apart.
The
winds
in
a
second,
but
the
other
thing
is
that
zonally,
you
have
a
lot
more.
What
zonal
transport
that's
occurring
exactly
where
you
have
a
you
know,
a
gain
and
a
loss
in
in
precipitation.
C
So
summer
is
a
little
bit
less
sustained,
but
I
just
thought:
I'd
show
it
anyway,
just
to
to
reinforce
a
couple.
Things
is
that
that
you
know
you
have
a
lot
of
the
same
pattern
and
you
know
with
a
little
bit
with
with
a
lot
more
meridianal
transport
towards
east
antarctica
in
summer.
As
well
but
generally,
the
the
effect
is
somewhat
weaker.
C
Sorry
so
we're
looking
at
so
I'm
taking
a
step
back
and
looking
at
this
changes
in
circulation.
That's
that
drive
that
ready,
not
vapor
transport
and
what
I'm
showing
here.
So
the
vapor
transport
for
for
winter
is
shown
here,
I'm
showing
those
errands
values
for
u
500
and
just
the
the
mean
case
and
then
for
for
v
500.
So
this
is,
you
know,
meridian
I'll
transport
towards
the
south
pole
and
positive
and
ansi
minus
amit.
C
What
I'm
showing
here
is
that
there's
been
overall
reduction
in
zonal
wind
and
this
actually
conforms
more
with
era
5,
but
an
increase
towards
the
the
on
the
continent,
and
so
similarly
you
can
see
this
this.
This
is
a
difference
between
antiminus
radionoly
and
you
can
see
this
increase
in
in
winds
over
here,
which
is
helping
to
drive
a
lot
of
the
vapor
transport.
C
This
isn't
a
place
where,
where
universally
you've
had
a
a
mean
change
in
in
wind
speeds,
so
that's
something
to
piece
apart
in
the
future
is
just
to
figure
out
exactly
what
conditions
are
creating
that
kind
of
a
vapor
transport.
C
So
now
taking
another
step
back
and
thinking
about
the
change
changes
in
temperature,
so
ansi
overall
has
an
increase
in
temperature
at
the
surface
and
I'm
showing
here,
ansi
minus
sorry,
nc
minus
a
mip
on
the
right
again
as
usual,
and
and
as
you
can
see,
it's
generally
an
overall
increase
in
temperature
across
the
board,
except
for
some
places
in
the
interior
and
the
weddell
c
sector
to
the
east
of
of
the
antarctic
peninsula.
C
And
then
this
is
sorry.
I
didn't
label
this.
This
is
this
is
the
in
in
the
near
surface
and
on
the
bottom
is
a
t500,
so
500
hecta
pascals,
and
so
what
we
can
see
here
is
that
the
bias
is
you
know,
increases
the
warm
bias
as
compared
to
era
five.
C
So
taking
another
step
back,
I'm
looking
at
changes
of
temperatures
compared
to
aws
data,
and
there
are
bins.
There's
133
aws
stations
and
they're,
divided
into
bins,
from
500
meters
to
500
and
thousand
meters
to
a
thousand
two
thousand
two
thousand
three
thousand
then
greater
to
three
thousand,
and
what
we
can
see
here
overall
is
that
it
matches
up
mostly
with
the
the
area
five
nicest,
except
for
the
fact
that
this
is.
C
You
know
the
zero
degree
point
and
where
you
have
a
positive
bias,
it's
on
the
right
and
then
a
negative
bias
is
on
the
left
and
in
general,
there's
been
more
warming
overall
in
ansi,
except
for
at
very
low
elevations,
which
is
kind
of
interesting
and
in
some
cases
the
the
bias
is
actually
a
correction.
C
C
Whoa,
okay,
so
I'm
gonna
move
on
to
the
most
interesting
part,
which
is
a
change
in
cloud
cover.
What
you
have
when
you
have
increased
temperatures,
is
enhanced,
moisture
and
enhanced
clouds.
C
So
the
main
takeaway
here
is:
I
want
to
point
you
just
to
the
bottom,
which
is
enhanced
total
cloud
liquid
water
path,
which
is
much
higher
in
ansi
and
the
interest
of
in
and
the
effect
is,
is
higher
down
than
where
le
net
radiation.
C
And
since
I'm
I'm
going
to
lose
time
here
and
I
can't
go
into
the
radiation
bad
budget
anyway.
The
final
thing
I'll
point
out
is
that
okay,
the
main
takeaways
here
are
that
higher
surface
mass
balance
and
ansi
is
driven
primarily
by
enhanced
snowfall
in
winter,
and
that
these
higher
temp
temperatures
can
lead
to
increased,
moisture
cloud
cover,
precipitation
and
and
and
and
produce
enhancement
for
moisture
transport
in
east
antarctica.
C
So
the
remaining
remaining
question
is
really:
where
is
the
the
that
energy
coming
from,
and
is
that
not
energy
coming
from
the
ocean
and
so
we'll
be
examining
the
the
differences
in
the
the
energy
balance,
in
particular
to
sort
of
answer
that
question
and
thanks
and
please
feel
free
to
contact
me
here
and
my
my
twitter
handle
is
here
as
well.
G
G
No
hands
up:
okay,
let's,
let's
move
to
the
next
speaker
and
if
you
have
questions,
please
put
them
in
the
chat
entry,
please
feel
free
to
respond
there.
Next
speaker,
we
have
is
mira
berdau
from
university
of
washington
mira.
Are
you
ready
to
go.
K
Yeah,
so
I'm
gonna
talk
today
about
some
work.
I've
been
doing
with
the
cism
model
to
explore
ice
sheet
model
sensitivity
to
thermal
forcing,
and
this
is
work
I'm
doing
with
contour
leggy
bill
lipscomb
and
nathan
irvin,
so
this
work
was
motivated
by
the
results.
Actually,
let
me
just
minimize
this
thing
yeah.
K
Here
we
go
from
a
2021
cism,
izmit
6
paper
written
by
bill
gunter
and
others
in
which
it
became
clear
that
some
parameters
that
are
chosen
during
the
model
spin-up
actually
ended
up
having
a
significant
effect
on
the
ice
loss
amounts
during
the
forward
runs.
So,
for
example,
in
the
izmip6
runs.
The
ice
sheet
is
spun
up
by
nudging
toward
observed
thickness
by
adjusting
basal,
friction
and
thermal,
forcing
correction
factor.
K
So
at
the
end
of
the
spin
up,
you
get
an
ice
sheet
configuration
that
matches,
observed,
grounding
lines
and
thickness,
and
it's
a
near
steady
state.
So,
overall,
the
advantage
of
such
a
method
is
that
it
works
well
to
keep
grounded
ice
near
observed
thicknesses,
but
a
disadvantage
is
that
the
parameters
used
during
your
spin-up
can
impact
the
sensitivity
of
the
ice
sheet
during
your
forward
runs.
K
So
the
positive
values
here
shown
in
the
oranges
and
reds
indicate
where
the
spin
up
is
thicker
than
observed
and
the
black
lines
indicate
the
boundaries
of
the
floating
ice
at
the
end
of
the
spin-up.
So
the
thing
to
notice
here
is
that
both
of
these
spun
up
states
look
very
similar,
despite
the
fact
that
they
use
different
parameters
during
their
spin-up.
K
So
based
on
the
previous
work,
there
were
two
parameters
that
were
found
to
be
very
important
in
the
sensitivity
of
the
ice
sheet:
first
gamma,
which
is
a
constant
in
the
thermal
forcing
parametrization,
and
it
describes
how
much
melt
occurs
per
ocean
warming.
So
in
other
words,
it
affects
the
sensitivity
of
the
ice
to
the
ocean
warming.
So
in
the
azmet
6
experiments,
for
example,
modelers
were
able
to
sample
a
low,
medium
and
high
value
of
gamma
as
sort
of
a
nominal
exploration
of
the
sensitivity
of
ice
sheet
projections.
K
Bill's
paper
also
showed
that
mass
loss
from
the
ice
sheet
was
strongly
dependent
on
the
degree
of
water
pressure
support
from
the
ocean.
So
the
parameter
p
here
controls
the
amount
of
water
pressure
support
in
marine
based
ice
sheets,
and
it
therefore
impacts
the
sort
of
the
effective
pressure
near
the
grounding
line.
So
in
other
words
it
controls
the
basal
slipperiness,
particularly
in
marine
based
ice.
K
K
So
first
we
ran
a
25-member
ensemble
of
spin-ups,
so
each
run
was
10
to
20
000
years
long
and
we
generated
combinations
of
p
and
gamma,
where
we
sampled
more
heavily
toward
higher
p
values
and
lower
gamma
values.
And
so
the
sampling
show
is
shown
in
the
green,
with
the
p-value
sampling
density
on
the
y-axis
and
the
gamma
on
the
x.
K
So
the
previous
work
had
shown
that
there
was
less
impact
between
p
for
0
to
0.5
and
then
so
we
decided
to
sample
more
heavily
toward
one
and
the
gamma
sampling
density
yeah.
It's
shown
in
the
x
and
it's
weighted
more
heavily
toward
lower
gamma
values
because
that's
considered
more
likely,
and
so
the
blue
dots
that
are
overlaid
on
top
of
the
plot
show
the
actual
p
and
gamma
combinations
that
we
used
in
our
spin
up
ensemble.
K
K
K
So
here
I'm
showing
the
thermal
forcing
anomaly
profile
for
the
decade
average
from
2090
to
2100
as
an
anomaly
with
respect
to
the
1995
to
2005
profile.
They
are
divided
into
regions
following
the
alarm
basins
seen
in
the
bottom
right
map,
and
this
plot
shows
the
maximum
thermal
forcing
profile
with
depth
as
seen
by
sysim.
K
K
K
It
shows
that
the
warmest
model
produces
the
largest
sea
level
with
over
202
sorry
with
over
2
meters
by
500
years
into
the
simulation
and
the
coolest
model
results
in
about
75
millimeters
sea
level
rise
and
on
the
right.
I'm
showing
the
final
sea
level
rise,
spread
for
each
model.
So
remember
the
spread
that
you
see
here
is
purely
coming
from
the
p
and
gamma
value
choices
during
the
spin
up.
So
it
appears
that
with
sufficient,
forcing
such
as
in
this,
the
easy
earth
model.
K
So
those
are
like
the
pale
purple
and
the
brown,
the
choice
of
pink
gamma
can
become
quite
important
for
multi-century
sea
level,
rise
outcomes
and
the
simulations
that
are
forced
with
the
cooler
ocean
temperatures
like
the
bcc
and
cams
those
little
blue
ones.
They
tend
to
show
less
sensitivity
to
p
and
gamma,
simply
because
there's
not
enough
forcing
to
generate
much
sea
level
and
rise
in
the
first
place.
K
So
next,
let's
take
a
look
at
how
the
final
sea
level
rise
varies
with
the
choice
of
p
and
gamma.
So
here
I'm
showing
the
final
sea
level
rise
values
at
500
years
into
the
simulation.
As
a
function
of
gamma
in
the
top
panel
and
as
a
function
of
p
on
the
bottom
panel,
the
linear
fits
are
included
here
and
the
the
r
squared
values
for
those
fits
are
shown
to
the
right
in
the
legends,
so
you'll
just
notice
that
the
fits
are
very
strong
with
gamma
and
less
so
with
p.
K
K
So
we
can
plot
this
a
little
differently
to
get
a
sense
of
the
qualitative
structure
of
the
final
sea
level
rise
as
a
function
of
p
and
gamma.
So
every
panel
here
is
showing
a
final
sea
level
contribution
for
different
model
forcing,
and
you
should
note
that
every
color
bar
here
is
on
a
different
scale,
so
on
each
y
axis
is
gamma
and
an
h
on
each
x-axis
is
p.
K
Interestingly,
we
don't
see
much
contribution
from
the
amundsen
in
any
of
our
cmip6
force
experiments
and
we
figure.
This
is
likely
a
result
of
two
main
reasons.
First,
the
forcings
generated
for
the
end
of
the
century
by
these
six
models
are
fairly
weak.
They
remain
largely
between
zero
and
one
degree
and
second,
as
part
of
the
spin
methodology,
we
assume
the
ice
is
in
equilibrium,
with
the
current
melt
rates
and
so
in
order
to
keep
the
ice
from
retreating
under
current
conditions.
K
We
have
to
apply
thermal,
forcing
correction
factor
that
essentially
cools
the
ocean
in
that
region
and
therefore
our
future
forcings
must
overcome
this
correction
factor
and
generally,
that
leads
to
an
underestimate
of
mass
loss
in
the
amundsen.
K
All
right,
I'm
just
going
to
take
a
quicker,
a
closer
look
at
just
one
example
case,
so
here
I'm
showing
the
eye
thickness
change
for
two
simulations
that
are
both
forced
with
the
same
thermal
forcing
so
the
only
difference
is
the
p
and
gamma
values
we
chose
during
spin
up
so
on
the
left.
We
with
moderate,
p
and
low
gamma
values.
The
ic
contributes
about
one
and
a
quarter
meters
to
sea
level,
whereas
with
a
high
p
and
gamma,
we
can
contribute
over
three
meters.
K
So
we
applied
a
synthetic
thermal,
forcing
in
only
one
one
and
a
half
and
two
degrees
to
those
25
spin-ups,
and
the
figure
shows
the
sea
level
contributions
under
each
experiment.
So
basically,
what
we
find
is
that
the
weights,
for
example,
the
weights
collapse,
is
possible
only
when
there
is
a
partial
to
full
water
pressure
support
and
when
melt
rates
are
sufficiently
high,
and
we
also
find
that
in
the
amazon,
this
region
exhibits
a
threshold
of
instability
in
the
range
of
one
and
a
half
to
two
degrees.
K
So
we
believe
this
temperature
threshold
is
likely
associated
with
seafloor
topography
that
acts
as
an
obstacle
to
ice
shelf
flow.
So
our
runs
show
that
in
the
amundsen
the
grounding
line
tends
to
stabilize
on
a
few
high
seafloor
ridges
and
under
sufficient
thermal,
forcing
the
ice
ungrounds
and
it
enables
retreat
okay
so
to
wrap
up.
Our
study
highlights
the
potential
downstream
effects
of
ice
conditioning
during
model
spin-up
and
since
it's
possible
to
achieve
a
similar
spun
up
state
with
different
sensitivities
to
ocean
warming.
K
We
find
that
the
choice
of
p
and
gamma
alone
can
impact
the
final
sea
level
estimates
by
up
to
two
meters.
The
ross
and
weddle
regions
dominate
the
sea
level.
Contributions
in
the
semi-six-force
forward
simulations
and
mass
loss
in
these
areas
is
largely
controlled
by
gamma,
rather
than
p,
which
implies
a
dominance
of
ocean,
forcing
parameters
on
ice
loss
over
ice
sheet
parameters.
K
And
finally,
we
find
that
the
amundsen
exhibits,
a
mix
of
ocean
ice
and
temperature
thresholds
that
together
determine
the
sensitivity
of
the
sector.
The
cement
six
force
runs
failed
to
produce
widespread
waste
collapse
after
500
years.
However,
with
additional
synthetic
forcing
testing,
we
find
that
the
almonds
and
mass
loss
can
be
triggered
with
thermal
forcing
anomalies
between
one
and
a
half
and
two
degrees,
and
in
these
cases
the
grounding
line
retreats
from
topographic
pinning
points,
and
without
these
stabilizing
points,
the
grounded
ice
in
the
basin
collapses.
G
All
right,
I'm
sure
there
are
questions.
So
if
people
process
this
a
little
bit
more
and
have
a
question,
please
put
them
in
the
chat
and
we
are
please
follow
up
there.
Okay,
let's
we're
nicely
on
time.
So,
let's,
let's
continue
with
the
next
speaker,
which
is
gen
k.
L
So
ann
was
not
available
to
make
this
presentation
this
morning,
but
was
really
excited
for
me
to
share
the
results
with
you.
All.
I'm
also
especially
excited
to
share
these
results,
because
the
original
discussions
about
this
topic
came
at
a
workshop
that
marika
organized
about
bringing
together
observations
and
modeling.
I
think
it
was
in
2018,
but
mike
steele
and
tristan,
and
I
went
to
lunch
and
talked
about
these
topics
and
of
course
this
is
just
really
cool.
L
But
this
is
ann
in
the
yellow
here
recently
presenting
at
a
group
meeting
at
cu,
and
if
you
have
any
questions
about
the
work
in
detail,
obviously
I'll
do
my
best
to
answer
them,
but
also
you
can
always
contact
ann
at
the
email.
That's
listed.
L
L
And
so
why
was
that
important?
Well,
clouds
strongly
influence
radiative
fluxes.
So
the
presence
of
clouds,
of
course,
can
strongly
impact
the
ocean
heat
gain
during
the
melt
season
and
so
therefore
can
influence
the
maximum
sst.
L
And,
of
course,
the
influence
of
clouds
is
somewhat
complicated
because
there's
both
the
shortwave
effects
and
the
long
wave
effects
during
the
melt
season,
the
shortwave
effects
dominate.
So
clouds
would
actually
be.
You
know,
blocking
incoming
sunlight
and
leading
to
potentially
a
lower
maximum
sst.
L
But
it's
important
to
acknowledge
that
the
long
wave
is
also
important
here
and
so
the
exact
seasonal
timing,
and
it's
not
an
easy
thing
to
just
you
know,
pen
and
pencil,
like
it's
really
nice,
to
have
some
observations
or
some
models
to
kind
of
explore
these
topics.
L
L
So
we
set
up
some
cesm2
cam
6
simulations.
We
just
branched.
300-Year
runs
from
the
1
increasing
co2
per
year,
cesm2
run,
so
we
have
a
pre-industrial
run.
That's
just
continuing!
For
100
years
we
have
one
that
we
branched
at
the
present
day
or
pretty
close
co2
concentrations.
Yes,
actually
next
year
we
probably
will
hit
424
parts
per
million.
So
if
you're
thinking
that's
high,
you
should
probably
be
looking
up
what
the
observed
co2
actually
is
right
now,
so
it's
pretty
good
for
present
and
then
a
four
times
co2.
L
So
just
a
couple
of
definitions
which
we
took
from
steele
and
dickinson,
I
think
they
set
this
framework
up
really
nice.
So
they
talk
about
the
the
melt
season.
So
that's
just
when
the
sea
ice
is
melting
and
then
the
heat
season,
and
so
that's
when
the
upper
ocean
is
warming.
So
you
can
imagine
this
heat
onset,
there's
maybe
different
ways
to
define
it,
but
the
way
they
defined.
It
is
the
last
day
where
you
have
sea
ice
concentration,
greater
than
0.15
and
the
sst
max,
so
the
maximum
warming
that
occurs.
L
You
know,
sometime
in
the
late
summer,
early
fall
where
the
the
heat
has
all
gone
into
the
ocean
and
you
reach
the
maximum
sst
and
then
after
that,
of
course,
it's
the
descent
into
arctic
winter.
So
the
ssts
are
decreasing
and
at
some
point
you
know
you
form
sea
ice
again.
So
this
sst
max
and
the
heat
season
are
really
important
because
they're
very
related
right.
L
This
timing
of
how
long
you
heat
the
ocean
is
going
to
affect
how
much
the
sst
is,
but
then
also
we're
interested
in,
in
particular
how
the
clouds
during
this
heat
season,
affect
this
sst
max
so
yeah,
just
to
kind
of
put
our
our
simulations
and
the
framework
of
this
heat
season
and
melt
season
definitions.
L
We
have
melt
onset
at
the
top
here,
the
heat
onset
and
the
second
row,
and
then
the
heat
season
length
at
the
bottom-
and
you
can
see
the
heat
season
length
you
know-
will
depend
on
where
you
are
especially
within
the
arctic.
But
it
of
course,
is
getting
a
lot
longer
as
you
get
more
co2
in
the
atmosphere
and
these
very
idealized
simulations.
L
So
you
know
at
287
parts
per
million-
it's
just
you
know,
weeks
in
the
interior,
maybe
up
to
months
at
lower
latitudes,
but
by
the
time
you
get
to
four
times
co2.
There's
multiple
months
where
you
have
open
ocean
and
the
heat
season
is
quite
long
so
a
bit
about
the
clouds
in
cesm2.
L
And
then
we
just
show
the
calypso
observations,
not
because
it's
a
perfect
comparison
to
make
observations,
but
just
to
kind
of
show
that
you
know
the
seasonal
cycle
and
the
cloud
amounts
look
pretty
reasonable
compared
to
modern
day
satellite
observations.
L
So
these
are
all
values
for
ocean
grid
cells,
polar
to
70
degrees,
north,
and
I
encourage
you
to
check
out
a
special
issue
paper
from
the
cesm2,
led
by
ellen
mcgill
hatton,
for
comparisons
of
the
clouds
in
this
generation
of
cesm,
so
cesm2
and
the
last
generation
cesm1,
and
the
basic
point
from
the
milk
elhan
paper.
Is
that
there's
a
lot
more
liquid
in
the
clouds
in
the
cesm2
and
that's
a
more
reasonable
representation
of
the
clouds?
Then
in
cesm1.
L
L
And
so
we
can
look
at
that
as
just
a
sensitivity
of
the
change
in
the
downwind
long
wave
on
the
top
or
the
downwind
shortwave
on
the
bottom
to
a
change
in
a
cloud
property.
So,
for
example,
here
the
sensitivity
to
a
change
in
the
liquid
water
path
or
and
on
the
right,
a
change
in
the
calypso
cloud
fraction.
L
And
so
you
can
see
you
know.
The
long
wave
is
always
warming.
If
you
increase
the
liquid
water
path,
you're
always
getting
more
downloading
long
wave
radiation.
But
the
the
timing
and
the
time
of
the
year
does
depend
on
how
sensitive
things
are
and
in
fact
you
have
a
much
stronger
influence,
especially
in
the
the
winter
and
the
non-melting
seasons.
L
And
then,
of
course,
the
shortwave
is
very
tied
to
the
solstice
and
that's
just
because
that's
when
you
have
the
most
downwind
shorewave
radiation,
and
so
that's
where
clouds
can
block
the
most
downloading
radiation.
L
And
so
I
want
you
to
keep
in
mind,
especially
this
timing,
where
the
shortwave
is
really
it's
really
locked
to
the
june
solstice,
and
so,
as
we
you
know,
can
imagine
if
we
start
to
warm
the
arctic
and
melt
earlier,
we're
going
to
push
this
melt
time
and
the
heat
season
closer
to
this
shortwave
optimum
for
clouds
to
have
a
maximum
effect,
so
some
results
at
low
co2.
Basically,
an
early
start
to
the
heat
season
leads
to
a
higher
maximum
sst.
So
this
is
basically
the
mechanism
that
was
outlined
in
steel
and
dickinson.
L
It
makes
total
sense.
It's
negatively
correlated
such
that
you
know
if
you
start
earlier,
an
earlier
day
of
the
year,
you're
going
to
end
up
with
a
higher
maximum
sst.
L
If
we
look
at
correlations
with
clouds
both
of
these
kind
of
pre-industrial
and
sort
of
present
day
they're
somewhat
weaker
than
the
correlations
with
the
day
of
the
year.
So
just
the
length
of
the
heat
season
is
really
what's
showing
the
strongest
correlations.
L
In
contrast,
as
we
start
to
see
the
heat
season
align
with
the
june
solstice,
the
clouds
increasingly
explain
the
maximum
sst,
and
so
we
can
start
to
see
these
correlations,
especially
between,
for
example,
the
heat
season
and
the
liquid
water
path
and
the
maximum
sst
really
start
to
increase
quite
a
bit
and
they
actually
become
larger
correlations
than
the
actual
length
of
the
melt
season
or
the
heat
season.
G
L
And
the
reason
for
this
is
just
as
already
kind
of
motivated
the
alignment
of
the
heat
season
onset
with
the
june
solstice
and
the
peak
influence
of
clouds
on
shortwave
fluxes,
and
this
basically
just
gets
optimized
as
co2
increases.
So
here,
for
example,
we
can
see
these
are
just
histograms,
showing
when
the
heat
onset
occurs
and
on
the
top.
These
same
sensitivity
plots,
for
example,
the
downwelling
shortwave
to
the
change
in
liquid
water
path.
So
you
can
see
as
we
move
into
these
reddish
colors
at
the
four
times
co2.
L
Basically,
the
timing
of
when
the
heat
onset
is
starting
is
starting
to
align
with
when
clouds
have
their
maximum
shortwave
cooling
effect.
L
So
just
a
summary
here,
this
is
work
and
prep
that
ann
plans
to
submit
to
grl
very
shortly
that
with
low
co2.
It's
the
cs,
retreat
timing,
not
clouds
that
generally
controls
the
maximum
annual
arctic
sst's
as
the
co2
increases.
L
The
sea
ice
retreats
earlier
closer
to
this
june
solstice
and
the
clouds
are
increasingly
explaining
sst
variability
and
in
particular
it
is
through
the
influence
of
the
clouds
on
the
shortwave
fluxes,
and
this
is
just
because
of
this
alignment
of
the
maximum
sensitivity
to
changes
in
cloud
properties
with
the
summer
solstice
and
so
just
to
kind
of
give
you
a
hint
at
one
of
our
key
three
points
for
the
grl
max
paper.
L
The.
If
we
look
at
sort
of
the
slopes
of
these
correlations,
we
see
that
the
seasonally
free
arctic
ocean
is
three
times
more
sensitive
to
clouds.
L
So
I
will
leave
it
there
and
hopefully
you
will
contact
anne
if
you
have
any
questions,
but
I'm
happy
to
entertain
some
here
as
well,
and
I
want
to
thank
also
the
polar
climate
working
group
resources
that
were
used
for
this
presentation
and
again
to
marika
for
organizing
the
workshop
that
kind
of
started
many
of
these
discussions.
G
Thank
you
so
much
jen.
We
have
time
for
a
couple
of
questions.
I
know
that
paul
kushner
asked
the
question
chad
paul.
Are
you,
okay
with
sure
yeah.
M
Just
that-
and
that
was
just
to
make
sure
I
understood
so
you're
the
four-time
co2
simulation-
seemed
to
be
still
equilibrating.
So
right,
it's
it
seems
like
it's
still,
warming
warming
up
and
I'm
wondering
if,
but
it
seems
like
your
conclusions,
wouldn't
really
be
sensitive
to
that.
I
I'm
I'm
imagining
right
like
that.
Basically
you're,
just
seeing
on
a
warmer
climate
as
the
as
the
season
as
the
ice
free
season
gets
sort
of
closer
and
closer
to
the
solstice
you're
you're,
going
to.
L
We
just
wanted
to
hit
it
with
a
big
hammer
so
that
we
could
align
those
two
times
better
and
yeah.
I
totally
agree
with
you
like.
This
is
not
necessarily
like,
what's
gonna
happen
or
a
forecast
for
when
this
is
going
to
happen,
but
more
just
like
a
this
is
a
physical
alignment
of
two
things
and
they.
L
The
climate
warms-
and
so
I
think
it's
a
really
nice
kind
of
process
based
understanding
of
how
these
two
things
might
change,
and
so,
when
we
might
see
this
in
observations,
are
we
already
seeing
it
observations?
We
just
wanted
to
kind
of
understand
the
parameter
space
before
starting
to
look
at
observations
to
see
for
evidence
of
this.
If
that
makes
sense,.
M
E
Hi
jen
nice
to
talk.
I
was
just
wondering
if
how
much
the
variability
in
the
sst
actually
changes
in
these
different
climates?
And
do
you
see
it's
much
more
variable
in
the
four-time
co2
climate
or
just
the
magnitude
of
the
variability?
Do
you
know
how
much?
L
Yeah,
I
would
have,
I
would
point
to
anne
for
details,
but
I
can
only
imagine
that,
with
a
longer
heat
season,
you're
going
to
be
reaching
larger
maximum
sst's
and
we're
already
seeing
that
in
the
arctic.
These
days,
where
I
think
you
know
certain
years,
I
I
think
it's
maybe
you
know
two
or
three
or
four
or
five
degrees
at
the
end
of
the
melt
season,
and
there
is
some
sensitivity
to
you
know
the
start,
of
course
of
that
melt
season.
In
the
current
observations.
L
Actual
yeah,
you
just
have
more
time
to
heat
it,
and
you
know
I
guess
if
it
was
a
particularly
cloudy
summer
that
would
suggest
that
that
this
would
help
the
clouds
might
help
dampen
how
much
heating
you
get.
But
I
can
only
imagine
if
you're
heating
for
months
versus
weeks
that
you're
gonna
get
warmer.
G
Adds
please
go
ahead.
N
I
had
a
very
similar
quest
to
marika,
but
I
guess
I
guess,
with
a
longer
open
water
season,
you
just
have
longer
time
for
the
high
frequency
variability
to
up
to
give
you
low
frequency
so
yeah.
I
think
I
think,
but
I
guess
my
the
the
other
thing
I
thought
about
was:
would
your
cloud
variability
itself
change
too,
and
would
that
increase
and
I'm
not?
I
don't
know
that
I
have
an
immediate
gut
reaction
for
that.
L
Well,
I
do
because
we've
looked
a
lot
at
how
summer
clouds
are
sensitive
or
not
sensitive
to
the
underlying
changes
in
the
surface
ocean
and
in
general
in
the
summer,
there's
very
little
cloud
response
to
what's
going
on
at
the
surface.
So
I
would
expect
this
is
much
more
of
just
a
top
down
clouds
influencing
the
sst
as
opposed
to
a
response.
L
Now,
a
response
to
those
ssts
being
much
warmer
and
that
heat
coming
out
of
the
ocean
in
the
fall
is
definitely
warranted.
And
you
know
we've
seen
this
through
analysis
of
satellite
observations
in
the
present
day
record
and
then
also
looking
at
future
cesm
simulations
and
I'm
happy
to
put
those
papers
in
the
chat
they
were
both
led
by
a
former
phd
student.
Ariel
morrison
thanks.
G
Thank
you.
Everyone.
Thank
you,
jen.
Let's
move
to
our
last
speaker
of
the
first
session,
but
stick
with
clouds
and
sea
ice
king
wadding
are
going,
is
going
to
talk
about
summertime
atmosphere,
cis
connections
and
the
role
of
clouds
in
csm
and
cmm6
models.
Please
go
ahead.
Thank
you.
O
Yeah,
I
hope
everyone
can
see
my
slide
right
now.
Yes,
okay,
thank
you
so
yeah.
This
is
the
work
led
by
my
student,
ray
law
and
other
collaborator
from
like
jfdl
and
usb
and
a
piano
and
also
some
institute
from
china.
So
the
main
goal
of
this
study
actually
is
to
look
at
something
like
the.
O
We
call
the
lc
air
atmospheric
sea
ice
connection
in
the
polar
region
in
summer
time
in
arctic
in
the
summer,
in
the
summertime-
and
I
won't
talk
about
like
the
rc
coupling
anomaly,
we
focus
on
like
the
tropical
region
like
related
enzo
or
iod,
but
I
think
that
this
is
still
very
new
phenomenon
in
arctic
region,
because
we
don't
really
know
how
these
two
components
talk
to
each
other
which
influence
another
more.
O
So
to
start
this,
we
we
looked
at
the
observation
first,
so
this
is
a
uri
data
to
the
right
you
can
see
just
to
to
show
this
result,
basically
we're
just
using
it
mostly.
I
say
a
c
is
the
area
index
for
each
month,
like
arctic
cs
index,
for
each
month
and
correlated
with
this
zonal
mean
perpetual
height
and
in
the
arctic
region
from
60
north
to
90
knots
so
no
mean
component.
O
So
by
doing
this,
we
try
to
see
doing
a
different
like
a
leader
like
we
want
to
see
which
variable
leave
another
variable
more,
so
you
can
see
very
clearly
if,
when
the
zoom
z
lead
like
the
other
months
like
the
sea
ice,
is
you
see
very
significant
and
negative
correlation?
O
The
negative
correlation
means
like
the
warmer
warm
air
or
higher
potential
high
lead
states
decrease
in
the
following
months
and
when
the
ci
somehow
seems
is,
can
occur
with
the
japanese
high
or
lead
the
potential
high,
as
means
the
cs
may
have
a
player
as
a
frozen
for
following
z
change.
O
We
don't
see
clear
as
I
like
a
correlation,
so
that
means
from
this
is
a
test
we
can
simply
understand,
probably
during
the
summer
time
that
atoms
atoms
here
influence
the
cs
more
easily,
rather
than
other
way
around.
So
because
this
is
so
simple
to
calculate.
We
just
are
using
the
september
cis
this
month,
according
with
the
jjizu
mean
as
like
a
benchmark
or
or
like
the
pattern
to
evaluate
to
the
how
significant
sips,
how
cm6
model
and
the
csm
capture
the
similar
feature
and
dynamically.
O
We
think
that
probably
this
is
related
to
something
we
call
the.
This
is
the
something
like
the
atmospheric
impact
on
sea
ice
through
this
adiabatic
warming,
because
we
know
in
arctic
when
there
is
a
high
pressure.
Normally
is
a
biotopic
structure,
and
this
one
can
induce
is
the
adiabatic
warming
and
details
the
surface.
Friction
and
warm
air
can
in
the
troposphere,
can
immediately
increase
a
longer
radiation
to
melt
this.
So
somehow
this
is
the
idea,
the
thing
that
is
a
correlation
pattern
and
yeah.
O
As
I
said,
this
is
so
easy
to
use
to
calculate
the
pattern
just
to
see
september
cis
per
arctic
index
correlated
with
the
jj,
so
no
mean
we
try
to
use
this
evaluator
or
other
or
model
available
model
in
cinema,
5
and
76,
and
also
a
large
example
archive.
We
find
that
if
you
do
the
same
thing,
no
matter
it's
the
historical
one
or
like
the
pre-industrial
con
industrial
simulation.
We
find
the
correlation.
It's
a
coupling
strategies
or
connection
strength
is
pretty
weak
compared
with
observation.
O
The
observation
is
a
negative
0.6,
because
you
used
to
determine
the
data
you
can.
We
do
the
same
thing
for
all
the
model
results
same
length,
40
years
or
100
years,
or
whatever
less
we
use.
We
always
find
a
much
weaker
connection
between
the
two
variables.
A
comparison
observation
76
seem
to
be
worse
than
cm5
and
the
csm
large
example,
as
that's
pretty
good,
compared
with
others
for
model.
O
In
this
large
example-
and
we
also
looked
at
csm2
very
latest
version
of
the
csm2
different
version,
it's
a
little
bit
different
and
this
seemed
to
seem
to
like
this
negative
center
shift
to
the
lower
latitude.
So
since
that
seems
like
it's
important,
if
this
is
real,
I
think
it's
real,
because
we
find
it's
a
really
robust
connection
in
observation,
and
is
this
a
real
and
indicate
some
physical
meaning?
O
So
we
try
to
understand
why
and
because
we
know
this.
The
long-range
radiation
is
a
key
component.
So
we
just
simply
do
this.
We
call
the
partial
correlation
of
this.
This
analysis
and
very
easy
to
understand.
O
So
just
select
the
same
plot,
but
we
just
do
the
arctic
average,
just
like
the
average
from
74
to
19
hours
and
we're
using
using
the
number
to
show
there's
a
connection,
so
you
can
see
with
reverse
sign
right,
so
we
can
see
a
z,
a
z,
j,
z
and
a
september
sea
ice
is
pretty
strong
correlated
at
0.6
and
the
mirrors
show
same
thing
and
the
model
is
weaker,
but
still
can
capture
some
strength.
O
And
then
we
try
to
remove
to
simply
do
this:
statistical,
partial
correlation
removal,
impact
of
q
net
and
each
component
of
q
net.
We
we
want
to
see
that
if
it's
a
correlation
job
really
strongly.
So
that
means
that
term
is
important
and
you
can
see
if
you
remove
some
impact
from
like
little
heat.
Sensible
heat
sends
a
throttle
wave
and
it
doesn't
change
too
much.
It's
a
stress,
but
once
we
remove
this
clear
sky,
they
are
long
wave
radiation
and
the
total
is
just
substantially
dropped
to
something
like
this.
O
And
if
you
remove
two
net,
is
it's
no
connection
almost
so
this
means
that
this
connection
is
the
same
q
net
up
within
the
q
net.
It's
more
likely
like
a
dlr,
especially
clear
sky.
Dr
is
important,
and
so
now
we
just
try
to
our
model.
Do
the
same
thing
even
model.
The
total
stress
is
weaker,
but
remove
the
dr
effect
from
model
relationship
model.
O
Also,
this
connection
model
is
also
dropped
fast
to
almost
like
a
neutral,
and
so
now
we
think
that
we
should
focus
on
the
how
the
circulation
linked
to
this
br,
dr,
is
like
jj
adr
on
the
surface
and
we
find
even
er5
and
the
mirror
show
a
little
bit
different
structure.
This
is
simply
just
a
jd,
dr.
It's
periodic
just
a
70
to
90
north
average
quality
with
the
zoology
z.
O
You
find
something
like
the
high
pressure
here
and
a
warm
air
emit
the
long
wave
variation,
but
the
mirror
shows
a
little
bit
different
structure.
So
now
we
do
that.
We're
using
the
significance
model,
surface
3333
model
to
redo
this,
to
recalculate
this
plot,
and
we
find
using
some
method
to
do
this.
A
clustering
and
we
find
basically
order
performance
can
be
represented
by
these
six
nodes
or
we
call
six
groups
and
the
sum
is
really
weak.
O
0.3
something's,
really
strong,
and
if
you
look
at
the
detail
of
each
group,
you
can
find
in
node
six.
That's
what
we
call
the
strong
group
and
the
csm
welcome
is
over
there
and
another
one.
You
can
see
very
interesting,
some
model
different
name
but
they're,
using
chem
as
their
atmospheric
component.
O
So
but
the
other
three
model
it
seems
like
the
r2
model,
using
like
a
matte,
a
hot
and
center
model.
I
don't
know
why
that,
but
I
think
maybe
this
indicates
this.
This
group
might
have
some
same
cyclical
component.
Another
interesting
thing
is
that
aerosol
is
all
prognostic
aerosol
but
more
wise.
This
group
is
like
the
prescribed,
aerosol
and
a
very
different
various
atom,
fair
component,
and
the
csm2
is
over
here
csm2
another
version
over
here
over
four.
But
why
can't
seem
to
be
to
capture
this
one?
O
And
so
now
we
focus
on
at
the
individual
component
of
the
r
and
we
separately
calculate
the
correlation
between
the
sky,
clear
sky,
dr
and
cloud.
Dr
with
each
variable,
so
this
is
a
pretty
busy
plot,
but
basically
just
to
show
the
correlation
of
this
one
jja
in
order
antarctica
correlated
with
the
sonoma
z,
zono
mean
temperature
like
the
vertical
structure
and
just
with
the
humidity
and
omega
and
separately.
O
So
the
first
row
is
the
total
dr
and
this
one
era,
the
first
economy,
area,
mara
or
the
model
36
model,
the
good
group
and
the
weak
group.
The
goal
is
to
see
how
the
each
atom
very
variable
connect
to
the
different
components
of
the
r.
So
we
find
it's
pretty
clear.
The
clear
sky
or
model
real
analysis
is
pretty
consistent
and
the
main
troublemaker
for
this.
O
For
this,
the
connection
is
it's
a
cloud
of
dr
and
even
his
object.
So
in
the
strong
group
he
selected
increase
here
but
and
the
weak
group
is
the
decrease,
and
so
this
is
a
the
weak
group
really
similar
to
something
in
the
mirror
and
the
strong
group
really
similar
to
the
era.
Two
minutes.
Okay.
So
so
this
is
idea,
and
we
also
looked
at
the
cloud
structure.
O
O
The
better
cloud
cloud
measurement
is
the
most
similar
to
to
the
marathon
data,
and,
if
you
look
at
the
z
at
the
upper
level,
z
same
place
and
the
correlated
with
the
cloud
fl
cloud
zone
cloud
cover
structure,
it
seems
like
the
high
pressure
flavor
out
of
phase
change
of
the
cloud
and
the
marrow
should
increase,
and
this
really
seems
to
the
strong
group
in
era
and
the
weak
groups
seem
like
the
matter
too,
and
we
think
that
this
red
humidity
is
the
key
thing,
because
we
just
zoom
in
here
and
look
at
the
domain
average
correlation
between
the
z
and
each
relative
humidity
and
the
cloud
we
find
the
same
thing
just
like
the
there's.
O
A
car.
The
convert
change
is
mainly
determined
by
the
relative
humidity,
and,
to
summarize
this,
I
think
the
main
idea
is
pretty
simple.
Just
when
there's
a
high
pressure,
a
strong
group
there's
a
the
model
responding
to
different
type
of
regime
and
the
the
strong
group
really
similar
to
the
er
e5
and
because
the
the
cloud
is
out
of
phase,
so
this
can
cancel
this
account
of
forcing
so
make
that
is
a
total.
Dr,
a
clear
sky,
dr,
is
more
dominant
in
determining
this
in
determining
there's
a
total
dr.
O
So
that's
that's
the
reason
why
circulation
can
easily
talk
to
cx,
but
here
is
a
uniform
decrease.
So
this
is
a
decrease.
There
are
cloud
forcing
and
cancel
somehow
here
make
the
the
total
dr
is
not
as
strong
as
a
strong
group.
So
that's
probably
that
in
favor
is
the
talking
between
the
the
circulation
and
the
cx.
So
there's
a
lot
of
plot-
and
I
want
to
show
here-
is
that
so
we
speculate.
O
Probably
the
the
reason
to
understand
this
is
that
probably
due
to
circulation,
frost
is
not
as
strong
in
this
of
model
historical
simulation
or
pre-industrial
stimulation.
So
we,
just
probably
due
to
this
attack
connection,
is
not
a
well
simulated
in
the
model.
If
you
can
see
here,
you
have
a
really
strong
standing
region
with
jji
z
and
the
model
gen
is
a
little
bit
weaker,
and
that
means
the
wind
is
not
strong
as
observed.
So
now
we
just
put
opposite
wings.
O
We
put
the
yari5
wing
in
the
model
and
we
don't
change
other
things,
and
if
you
do,
we
do
the
same
same
with
simple
correlation.
We
find
that
this
is
the
coupling
that
is
pretty
close
to
observation
csm2.
We
can
find
also
enhanced.
This
means.
The
the
physical
part
is
the
is
okay,
but
just
probably
one
reason
is
that
the
wing
is
weaker
in
the
model,
so
I
want
to
stop
here
and
take
any
question.
O
G
Thank
you
so
much
any
questions.
Jen,
yes,
go
for
it.
L
Hi
to
you
always
interesting
to
see
what
you're
analyzing
and
looking
at
I
guess.
I
have
a
couple
of
questions
that
are
questions
about
methodology
like
first
I'm
a
little
bit
concerned
about
making
comparisons
between
cesm
and
clouds
in
re-analysis,
especially
when
you're
making
no
effort
to
make
sort
of
a
definition
aware,
scale-aware
comparison
between
them.
I
mean
the
clause
and
re-analysis
are
not
all
that
different
than
the
clouds
in
a
climate
model,
they're
heavily
parameterized
and
the
radiator
fluxes
are
not
even
constrained
in
the
re-analysis.
L
So
I
guess
I'm
a
little
bit
skeptical
about
some
of
those
comparisons.
I
mean.
Maybe
that's
just
a
comment
and
I'll
stop
there.
Then
I
do
have
a
question
so
in
these
last
comparisons
that
you
were
making
with
the
wind
nudged
runs
and
the
large
ensembles.
Are
you
looking
at
the
full
large
ensemble
spread
when
you're,
making
those
comparisons?
You're
not
comparing
to
the
ensemble
mean.
Are
you
I'm
a
little
bit.
O
Dude
yeah,
that's
good
question.
We
really
appreciate
that
we
do
each
member
and
add
them
together.
So
there's
some
spread,
but
the?
U
overall
is
a
relatively
weak,
like
the
average
elite,
just
like
0.3
in
observations
around
0.6
and
some
member
can
reach
to
0.4,
but
some
numbers
go
to
0.2.
So
add
them
together
is
0.3.
So
we
do
that
each
member
first
then
average
them.
L
K
O
F
O
N
Yeah
again,
it
was
quite
related
to
to
chance
by
very
thought-provoking
talk,
and
but
I
did
actually
you
know
thinking
about
this,
comparing
like
40
years
in
observations
with
one
of
your
earlier
slides.
Actually,
you
know
forgetting
about
clouds
and
just
looking
at
the
relationship
between
your
potential
height
and
sea
ice,
and
I
think
you
saw
like
different
csm2
versions.
N
I
don't
know
how
many
years
you
were
using,
but
how
robust
basically
are
40
years
right
and
I
think,
maybe
with
a
long
control
model,
you
can
take
a
running
40
year
windows
and
I
I
know-
we've
talked
about
different
some
of
this
teleconnection
work
right,
but
dave
bonin,
and
I
have
that
gerald
paper
right
there.
N
Some
of
the
models
can
capture
the
observed.
Teleconnection.
Like
the
I
mean,
this
is
not
the
pacific
arctic,
one
for
over
40
year
periods,
but
never
with
the
mean
state.
So
I
I
I
would
invite
kind
of
that
perspective
too.
With
this
work
and
maybe
like
split
the
observations,
the
first
20
years,
second
20
years
and
yeah-
try
and
get
a
sense
of
how
robust
this
is.
That,
I
think,
would
be
incredibly
good.
Yeah,
thank
you
all
right.
I'd
love
to
talk
more
about
it.
G
Everyone,
yeah
and
paul.
I
think
your
question
was
very
similar
to
jensen
and
and
ads,
but
feel
free
to
continue
in
the
chat.
Yes
thumbs
up
all
right.
So
it's
time
for
a
break
thanks
to
all
the
speakers
and
to
all
of
you
for
your
interest
and
sticking
to
the
time
and
very
interesting
science
that
we've
had
so
far,
we
will
reconvene
using
the
same
link,
so
this
link
will
stay
active
in
around
25
minutes
from
now
so
at
10,
40
mountain
time
and
then
we'll
see
each
other
again.
G
Maybe
someone
can
repost
the
link
to
the
breakout
sessions.
So
if
you
haven't
put
down
your
name
in
one
of
the
breakout
sessions
and
you're
interested
in
joining
one
of
them,
please
do
as
soon
as
we
as
you
can.
So
we
have
a
view
on
the
interest
for
the
specific
sessions
all
right
thanks
so
much
and
we'll
see
each
other
again
in
25
minutes.
J
One
thing
I
just
wanted
to
say
about
the
breakout
session:
sorry
is
that
todd
will
not
be
assigning
you.
You
will
have
to
go
into
your
breakout
session,
so
this
is
just
you
know
like
a
statement
of
maybe
you're
interested
in
those,
but
you
know,
even
if
you
don't
get
to
write
your
name
under
one
of
those
sessions,
that's
totally
fine.
Those
are
we're
going
to
basically
put
notes
in
there
regarding
perhaps
some
research
priority
areas
and
model
experiment
proposals
so
yeah.
I
just
wanted
to
clarify
that.
Thank
you.
I
I
I
E
Yeah
I'll
also
just
say
we
we
are
planning
to
have
the
zoom,
the
zoom
link
open
after
the
meeting
adjourns.
If
people
want
to
just
stick
around
and
eat
lunch
together
and
chat,
so
we
kind
of
have
an
informal
just.
You
know
hangout
together
session,
when
the
meeting
adjourns-
and
I
figure
people
will
probably
leave
for
10
or
15
minutes,
get
lunch,
etc
and
hopefully
rejoin
so.
I
A
Oh
yeah,
I'm
not
sure
a
number
of
us
are
going
down
to
the
southern
sun
around
four,
so
I'm
going
to
pass
on
the
hike
myself.
I
G
I
A
Right
so
yeah,
I
I
think
this
meeting
was
big
enough.
That
encar
was
concerned
about
us
having
an
in-person
element,
and
so
so
gokan
had
to
declare
it
completely
virtual.
So.
I
M
A
Okay,
alice
pointed
out
that
it's
partly
because
we
were
not
in
the
green
phase,
also
in
time
for
us
to
be
able
to
declare.
I
think
there
was
some
uncertainty
too,
about
rooms
available
anyway,
yeah,
whatever.
I
A
A
Yeah
the
well,
as
I
said,
we're
doing
the
happy
hour.
Some
of
us
are
going
to
a
happy
hour
at
4
at
the
southern
zone,.
I
A
Today,
yep
and
and
then
there's
the
the
madison
ams
polar
meeting,
that's
on
track
to
be
in
person
so.
I
I
A
Well,
it's
nice
because
it
does
allow
european
speakers
to
join
us,
and
things
like
that.
So,
but
I
mean
there
is
the
time
difference
of
course,
but
it's
still
it's,
it
still
does
add
for
a
flexibility
in
a
wider
audience
right.
A
I
A
So
has
everyone
got
a
chance
to
sign
up
for
one
of
the
discussion
breakouts,
I
see
a
few
names
in
there.
B
F
E
A
E
I
expect
he
is
okay,.
A
E
B
E
B
E
Did
you
have
graduation
duties?
Did
you
have
like
in
graduate
students
or
I
went
to
graduation
for
the
first
time
ever.
B
And
it
was
the
catch-up
graduation
for
the
last
two
years
when
it
was
canceled,
it
was
really
fun
actually
because
it
was
much
smaller.
You
know
the
main
one
was
50
000
people.
E
The
ketchup
one
was
about
four
thousand
yeah,
my
nephew
graduated
in
the
main
one.
So
did
you
have
to
be
a
flag
bearer?
You
know
he
for
arts
and
sciences.
I
think
he
carried
in
a
a
banner
because
he's
yeah
but
and
my
my
daughter
anna's
is
there
now
as
a
freshman.
So
are
you
dub,
yeah
yeah?
B
E
He'll
see
her
biology
is
her
plan,
probably
a
minor
in
chemistry.
So
let's
see
she
just
finished
her
freshman
year.
She
loved
it.
So
it's
great.
She
just
got
home
two
days
ago
because
she
drove
back
with
a
friend.
B
E
A
M
J
Lots
of
things
to
run
y'all,
yeah,
okay.
I
know
it's
very
interesting.
Actually,
we've
never
done
this
experiment
of
like
if
you
had
scientists
running
everything
like
what
would
it
be
like
you
know,
versus
like
politicians
like.
Would
that
change
anything
yeah
anyway?
Okay,
so,
thank
you
all
for
being
here,
and
it
is
now
time
for
us
to
resume
some
talks
and
then,
after
that,
I'm
super
excited
about
all
of
these
working
groups
or
little
breakout
groups
that
we're
gonna
do
so
to
begin.
J
We
will
have
lorenzo
zampieri.
Did
I
say
that
right?
I
hope
I
said
that
right.
Okay,
awesome,
tell
us
about
clear
sky
biases
over
the
arctic.
F
F
It's
not
really
100
related
to
csm,
but
we
all
use
re-analysis
to
verify
our
models
or
also
as
baseline
for
other
our
analysis
and
most
of
us
use
them
as
boundary
condition
for
running
single
model
components,
and
I
will
try
to
explain
how
we
can
improve
the
atmospheric
analysis
over
the
arctic
skies
really
close
to
the
surface
so
yeah.
F
This
is
what
I
will
try
to
explain
in
the
next
one
minutes,
so
the
shortcomings
of
the
current
generations
of
the
analysis,
in
particular,
of
the
temperature
field,
the
causes
and
the
consequences
of
these
issues
and
what
we
can
do
for
improving
things
yeah.
Just
as
a
side
note,
this
work
is
available
as
a
preprint,
so
yeah.
If
you
want
more
detail,
please
have
a
look
there
or
just
drop
me
an
email,
I'm
happy
to
discuss
further
with
you.
F
F
In
particular,
the
arctic
cis
tended
to
be
too
thin
and
they
found
a
workaround
to
solve
this
problem,
which
was
to
apply
some
offsets
to
the
either
temperature
field
or
the
radiation,
and
this
has
been
done
consistently
in
the
past
years
for
core
forcing
gerry5.
F
Otherwise,
we
don't
really
get
a
good
sea
ice
simulations.
If
we
don't
do
that
and
if
we
jump
of
15
years,
this
paper
was
published
by
trucking
miller,
where
they
identified
the
issue
with
atmospheric
analysis
and
they
link
these
biases
to
the
missing
representation
of
the
snow
or
diatic
cis
in
the
atmospheric
models
used
to
produce
the
analysis.
F
So
what
happens
is
in
practice
is
following,
so
here
we
see
the
cs.
Representation
in
a
typical
cmap
model
can
be
csm.
For
example,
we
have
the
thickness
discretization
over
each
thickness
classes.
We
have
snow,
which
has
a
very
low
conductivity
thermal
conductivity
and
so
in
general,
when
we
couple
this
to
the
atmospheric
model,
the
model
has
a
good
idea
about
the
heat
fluxes
and
I
would
say
they:
they
are
quite
decently,
constrained.
F
On
the
other
side,
if
we
look
at
an
atmospheric
model
used
for
dairy
analysis,
the
representation
of
the
cis
is
very,
very
simple.
So,
first
of
all,
we
have
a
binary
cs
cover.
We
don't
really
have
a
concept
of
cs
concentration.
F
We
have
constant
thickness
everywhere
in
space
and
time,
no
matter
the
season
and
the
the
location,
and
also
we
don't
have
the
snow
layer
on
top
of
the
sea
ice.
So
we
miss
this
insulating
layer
and
therefore
the
surface
tends
to
radiate
and
we
have
temperature
biases
and
yeah.
Here
is
just
an
example.
In
r5
we
have
an
assumption
of
one
enough
inter
thickness
for
the
ice
in
jerry
55..
It's
two
meter
thickness
and
yeah.
The
mechanism
behind
this
bias
is
the
following.
F
So
on
one
side
we
have
the
analysis,
cis,
which
is
constant
thickness.
If
the
real
word
cis
is
thicker
than
that,
then
the
reanalysis
will
be
to
warm.
So
we
have
a
positive
surface
temperature
bias
on
the
other
side.
If
the
analysis
cis
is
thicker
than
the
real
world
size,
we
will
have
a
negative
temperature
bias
and
when
I'm
I'm
considering
thickness
here,
I'm
also
including
the
effect
of
the
snow
for
simplicity.
So
it's
the
ice
thickness
equivalent.
F
F
We
have
era
5
j55,
and
here
we
have
the
observations
from
basically
a
thermal
sensor
on
board
of
the
satellite,
and
you
can
see
that
we
have
positive
temperature
biases
up
to
six
degrees
or
even
larger
in
some
locations,
and
we
also
see
that
this
bias
is
not
really
constant
everywhere.
It's
very
heterogeneous
and
of
course,
if
we
apply
then
this
forcing
to
as
mandatory
condition
to
our
model
simulations,
we
will
have
issues
and
the
eyes
will
be
definitely
too
thin.
F
Yes,
just
a
bit
of
a
technical
consideration
is
that
the
cloud
state
is
relevant
for
this
study
and
for
two
main
reasons
so
on
on
one
side,
clouds
basically
mask
the
sea
ice,
so
the
satellite
observations
cannot
really
see
the
surface.
When
there
is
cloud
they
measure
the
temperature
of
the
top
of
the
cloud
and
therefore
we
cannot
really
trust
observations
in
cloudy
conditions
and
also
when
we
have
clear
sky
conditions.
F
So
when
we
don't
have
clouds,
the
bias
tends
to
be
particularly
enhanced,
because
in
these
conditions,
the
surface
can
radiate
much
more
effectively
to
space,
in
particular
during
winter
and
yeah.
So
we
need
somehow
to
identify
the
cloud
conditions
from
re-analysis
and
we
do
this
in
two
ways.
So
there
you
can
see
for
the
same
day
the
total
cloud
cover
in
earth
again
era,
5
and
05,
and
the
long
wave
state-
and
you
see
that,
even
though
they
refer
to
the
same
atmospheric
state,
they
are
very
different.
F
Now,
how
do
we
formulate
an
effective
bias
correction
strategy?
So
what
we
want
to
do
basically
is
to
have
a
correction
which
is
a
step
forward
compared
to
the
previous
attempts.
So
we
don't
want
something
climatological,
that's
applied
everywhere,
but
we
want
a
correction.
That's
state
dependent
and
the
first
step
is
to
quantify
the
temperature
bias
at
the
day,
the
time
scale
when
we
have
observations-
and
we
can
do
that-
because
we
have
quite
good
observations
nowadays
of
the
surface
temperature
from
remote
sensing
products.
F
The
second
step
is
to
identify
some
predictors
that
we
can
associate
to
the
bias.
Of
course,
we
could
develop
a
correction
that
relies
on
the
own
observations,
but
the
problem
is
that
observations
don't
go
back
far
enough
in
the
analysis
record
and
they
are
not
available
everywhere
every
time.
So
we
need
to
have
some
more
reliable
products
that
are
based
on
the
analysis.
On
atmospheric
analysis
and
ski
spray
analysis
that
inform
us
on
the
nature
of
the
bias-
and
here
we
use
the
surface
temperature
itself,
the
downward
long
wave
radiation.
F
F
F
So
here
you
can
see
basically
a
general
view
of
the
bias
and
the
correction
predicted
by
the
neural
network
as
function
of
the
four
predictors.
So
we
have
the
temperature,
the
downward
long
wave
radiation.
And
here
in
this
plot
we
have
cs
thickness
and
there's
no
thickness,
and
we
can
see
that
yeah.
The
structure
of
the
bias
somehow
makes
sense.
F
F
So
we
have
this
two
type
of
regimes
and
you
can
see
that
the
neural
network
is
able
to
capture.
Well
this
these
bias
and
yeah
just
for
reference.
This
data
were
not
was
not
used
in
the
training
of
the
network,
so
it's
a
sort
of
an
independent
comparison,
but
now,
let's
have
a
look
and
how
the
situation
looks
like
when
we
look
at
maps.
F
So
when
we
apply
our
correction
at
the
reanalysis
time
scale,
so
three
hours
for
jerry
55
one
hour
for
r5,
and
then
we
take
the
average
over
for
almost
40
years.
F
We
see
that
the
correction
that
we
get
as
a
structure,
so
it's
particularly
it,
tends
to
be
stronger
over
thicker
eyes,
and
this
makes
sense,
and
also
we
see
that
in
particular
in
winter
we
have
this
double
like.
We
have
this
negative
correction
in
central
arctic,
where
the
ice
is
thicker
than
the
analysis
assumption,
but
also
we
have
this
positive
correction
where
the
ice
is
thinner
than
there
is
assumption,
and
this
makes
makes
sense
and
it's
quite
convincing.
F
We
also
observe
that
there
is
a
seasonality
related
with
this
correction
and
this
seasonality
emerges
from
the
from
the
the
model
itself,
so
we
didn't
really
impose
it,
and
this
is
also
quite
quite
nice
because
it
means
that
the
neuron
might
work
somehow,
even
though
it's
not
really
physically
informed,
but
it
can
yeah,
it
can
produce
a
physically
meaningful
correction.
F
So
let's
have
a
bit
more
a
detailed
look
into
this
seasonality
of
the
temperature
correction.
So
here
what
I'm
showing
is
the
average
correction
north
of
70
north.
F
So
it's
the
difference
between
the
original
and
the
corrective
analysis
field,
and
we
can
see
basically
the
ammo
cycle
splitted
into
four
decades,
and
we
can
see
that
so,
first
of
all
in
summer
we
don't
really
have
a
correction
on
the
temperature
field,
and
this
is
because
we
basically
have
temperatures
that
are
close
to
melting,
so
yeah.
We
we
don't
really
have
bias
in
the
analysis.
F
We
see
that
we
have
a
general
maximum
in
when
the
temperature
are
the
coldest
during
winter,
and
also
we
see
that
we
have
a
decrease
in
the
strength
of
this
correction
with
time,
and
this
makes
sense,
because
we
we
can
think
that
the
current
cis
conditions
are
more
compatible
with
the
analysis
assumption
compared
to
the
sea
ice
conditions
that
were
observed
in
the
80s
and
in
the
90s,
and
we
see
that
yeah.
F
This
structure
is
consistent,
I
would
say,
between
era,
5
and
05,
although
g65,
we
don't
really
observe
this
strong
and
this
correction
trend,
let's
call
it
in
in
winter,
but
more
rather
in
the
in
the
fall
during
the
freezing
season.
F
Now
we
we
also
checked
if
this
correction
has
an
impact
on
the
multi-decadal
warming
trend
of
the
arctic,
and
it
has
so.
I
would
say
it
has
a
small
impact
overall.
Compared
to
this
strong
signal
that
we
see
in
polar
regions
and
in
general,
we
can
see
that
the
corrected
field
tends
to
have
a
slightly
larger
warming
trend,
but
yeah
I
mean,
I
think
it's
I'm
not
sure.
If
this
is
really
a
significant
result,
but
yeah,
we
will
check
this
more
in
detail
so
yeah.
F
This
brings
me
to
a
quick
summary,
so
the
correction
is,
first
of
all,
state
dependent
the
way
we
developed
it
and-
and
this
is
coherent
with
means
that
it's
coherent
with
the
observed
size
conditions
and
with
the
local
weather
and
it
also
favor,
clear
sky
events
in
agreement
with
the
observed
nature
of
the
bias.
F
Also,
the
predictors
that
we
chose
are
physically
compatible
with
the
mechanism
that
explains
the
bias,
so
they
are
somehow
related
to
this
misrepresented,
snow,
thickness
and
ice
thickness,
and
also
we
are
able
to
capture
this
positive
and
negative
corrections
depending
on
the
ice
state,
which
I
think
this
didn't
necessarily
happen
with
previous
correctional
tents
and
also
a
self-emerging
property
of
the
correction
is
that
we
have
a
seasonality
but
also
an
internal
trend
in
the
analysis,
which
was
definitely
not
the
case
in
previous
correction
strategies
and
yeah.
F
I
leave
you
with
this
animations,
where
we
show
how
the
correction
looks
like
in
a
one
hour
time
scale.
So
when
it's
applied
in
on
the
error
file
field
and
yeah,
you
can
see,
of
course,
that
it's
restrained
to
clear
sky
conditions
and
you
can
see
the
effect,
for
example,
of
clouds
appearing
and
disappearing
and
yeah.
We're
not
sure
if
this
is
a
limitation
or
not
of
this
method.
For
sure
the
correction
that
we
developed
is
consistent
with
the
analysis
field
yeah.
F
It
might
be
not
hundred
percent
compatible
with
observation,
but
that's
something
we
will
investigate.
Also,
when
we
apply
this
re-analysis
to
run
as
boundary
condition
for
cis
models
and
that's
the
next
step.
So
thank
you
and
I'm
happy
to
take
questions.
J
Thanks
lorenzo
dave,
go
ahead
and
ask
your
question
real,
quick
and
then
we'll
leave
the
rest
of
the
questions
for
lorenzo
on
the
chat.
So
so
please
ask
him
on
the
chat
just
because
we're
running
just
slightly
over
so
sure
no
problem.
J
Oh
okay,
I
think
dave's
gonna
ask
you
on
the
chat.
Okay,
so
I'm
gonna.
B
M
See
how
how's
that
working?
Can
you
see
my
lead
slide?
That's
great
love!
You,
okay,
great
yeah,
so
I'm
presenting
today
on
a
project
that
is
led
by
a
phd
student
in
my
group,
alexander
odette,
and
also
a
new
student
luke,
fraser,
leach
and
alexander's.
M
Has
I'm
mainly
going
to
be
talking
about
his
project,
and
this
is
about
looking
at
sea
ice
perturbation
experiments
and
it's
it's
outside
a
lot
of
what's
been
talked
about
today
so
far,
but
I'd
really
like
to
hear
your
feedback.
So
that's
why
I
rather
selfishly
I'm
asked
to
come
talk
to
you
so
just
by
way
of
introduction.
M
So
this
is
a
a
talk
where
I'm
thinking
about
how
to
idealize
views
of
polar
climate
variability
in
the
in
the
climate
model
hierarchy,
and
I
think
so
this
is
kind
of
a
maybe
I
think
of
it
as
a
iconic
figure.
I
don't
know
I've
seen
it
a
lot,
I'm
showing
sorry
my
my
cursor
has
been
lost
there.
We
go
okay,
nope,
that's
not
helping
there.
We
go
is
that
we
can.
We
see
my
christian.
G
M
Okay,
I'll
just
go
over
here,
so
basically
we
have
this
whole
big,
a
big
model
hierarchy
right
where
we
can
try
to
model
the
earth
system
with
various
degrees
of
complexity.
I
don't
want
to
belabor
this,
but
oh,
this
is
terrible.
M
Oh
there
we
go
okay,
but
what
I'm
trying
to
do
here
is
talk
about
a
particular
set
of
experiments
that
play
with
the
boundary
conditions
and
climate
models
where
we
go
from
a
comprehensive
earth
system
model
like
csm,
we
can
remove
the
dynamical
ocean,
go
to
a
slab
ocean
model
and
we
can
go
to
a
no
ocean
model,
an
agcm
and
and
replace
the
the
ocean,
boundary
conditions
and
sea
ice
boundary
conditions
with
prescribed
fields
all
right.
M
So
the
project
that
I'm
I'm
gonna
be
talking
about
is
the
polar
amplification
model
into
comparison
project
pa
map
all
right-
and
this
is
a
set
of
it's
a
model
into
comparison.
M
It's
a
set
of
constrained
sea
ice
experiments
that
are
used
to
elucidate
the
causes
and
consequences
of
polar
amplification,
and
there
are
pmf
experiments
that
in
this
sort
of
first
group
here
that
looked
at
agcm
simulations
and
tried
to
understand
the
circulation
response
to
sea
ice
loss
for
understanding
polar
mid-latitude
connections,
but
there's
also
a
set
of
coupled
simulations
and
those
are
called
pmf
group
six.
Now
the
payment
prescription
is
it's
inspired
by
step
by
pictures
like
these.
M
This
is
taken
from
the
earlier
work
from
clary
duster's
group
and
look
just
comparing
when
you
take
a
a
model
like
ccsm4
and
perturb
the
sea
ice
in
it,
either
in
an
agcm
setting
or
in
a
coupled
ocean
atmosphere.
Gcm
setting
you
get
a
very
different,
polar
response
right,
you
get,
you
get
a
lot
more
warming,
deeper
warming
and
in
the
slab
ocean.
So
there's
some
also
dramatic,
inter
differences
too
between
the
agcm
and
the
slab
ocean.
M
Basically,
as
you
introduce
more
coupling,
you
get
more
thermodynamic
coupling
into
the
the
deeper
free
troposphere
and
even
coupling
from
the
northern
hemisphere
to
the
southern
hemisphere.
So
pmip
is,
I
think,
inspired
in
part
by
model
experiments
like
this,
and
so
what
pianip
does
so?
The
the
group,
one
experiments,
are
kind
of
these
agcm
experiments
and
you
impose
a
ci,
a
sea
ice
concentration
change.
That
looks
like
the
map
here
in
the
upper
right
and
also
a
ci
thickness
change
and
these
perturbations.
M
This
is
kind
of
important
they're,
drawn
from
the
kind
of
a
an
aggregate
of
cement.
Five
sea
ice
changes
under
two
degrees
of
global
warming
and
there's
actually
our
antarctic
and
antarctic
experiments
that
are
part
of
this
prescription,
and
these
are
all
documented
in
the
paper
by
smith
and
others
in
2019,
and
so
we
what
we,
what
we
did
and
what
several
other
groups
have
done
is
start
to.
M
You
know,
use
these
perturbations
to
to
force
models,
so
this
is
a
an
example
of
the
an
agcm
experiment:
okay,
the
surface
temperature
response
to
that
sea
ice
loss
with
with
radiative
forcing
held
fixed
and
here's
an
experiment
with
ghost
nudging.
So
this
is
basically
adding
heat
heat
to
the
bottom
of
the
sea
ice
and
looking
at
the
responses
in
in
temperature
and
also
in
the
zoning
as
well.
M
Okay-
and
what
you
might
notice
is
that
in
the
agcm
simulation
using
this
particular
forcing
and
model
so
the
model
is
the
specified
chemistry
version
of
wacom
4..
So
it's
an
older
generation
model,
but
it
it.
It
was
good
for
this
purpose.
M
I
think,
and
what
we
noticed
was
that
in
the
coupled
simulations
for
some
reason
we're
getting
unexpectedly
weak
warming
compared
to
what
we
saw
when
we
imposed
sea
ice
concentration
and
thickness
changes
in
the
agcm
and
also
the
the
warming
here
is,
you
know
not
quite
as
deep
and
it's
not
quite
as
strong
and
the
arctic
amplification.
We
get
isn't
because
we're
quite
as
strong,
so
we
were
trying
to
understand
this
and
the
issue
of
method
dependence
and
so
the
the
research
question
here.
M
It
relates
to
this
protocol
and
I'm
trying
to
make
a
very
simple
point.
So
I
hope
I
don't
make
it
too
complicated
in
the
talk,
but
basically
the
pmic
group
six,
this
coupled
protocol.
It
doesn't
really
prescribe
a
method
for
driving
sea
ice
laws,
because
different
modeling
groups
had
different
ways
of
approaching.
Coupled
modeling
of
you
know,
prescribed
sea
ice.
M
My
own
observation
was
that
the
pmf
group,
through
its
workshops
there
weren't
very
many
sea
ice
modeling
experts
in
that
there
were
a
lot
of
people
interested
in
the
atmospheric
science
and
people
who
worked
in
sea
ice
prediction
for
sure.
But
in
terms
of
like
really
thinking
about
the
protocols,
I
didn't
hear
that
those
those
voices
very
often
so
I
think
this
this
is
something
that
could
be
improved
on
and
what
we.
M
What
we
were
thinking
about
was
the
the
method,
dependence
of
the
responses
all
right,
and
certainly,
if
there's
differences
in
the
arctic
responses,
then
we
expect
differences
in
the
extra
tropical
and
tropical
responses
as
well,
and
there
have
been
studies
by
by
james
screen
by
lentil
sun
that
have
looked
at
the
method.
Dependence
in
these
coupled
model
experiments
so.
H
M
We
were
trying
to
do
is
develop
a
standard
method
for
that
could
be
easily
implemented
and
by
various
groups
to
control
sea
ice
a
bit
more
carefully
and
what
kind
of
results
we
might
get
from
that
and
we
have
these
pmf
specifications
that
to
to
go
by,
and
the
final
question
is
actually
a
bit
more
of
a
follow-on
research
question,
and
that
is
whether
we're
doing
the
right
thing
by
running
these
perturbation
experiments
at
all.
So
I'll.
Try
to
get
to
that
here.
M
So,
first
of
all
I'll
just
describe
this
method,
and
so
alexander
is
calling
this
the
hybrid
nudging
method,
and
basically
we
have
this
target,
which
I've
shown
you
before,
and
the
standard
go
split,
ghost
flux,
nudging
method:
this
is
what
was
used
in
seven
and
several
papers.
Alanthou
sun
had
a
bit
of
a
review
of
this
in
a
in
a
short
paper
of
different
methods.
Kelly
mccusker
used
this
in
her
work.
M
The
idea
is
you,
you,
you,
it's
a
basal
heat
flux
that
you
add
to
the
to
the
bottom
of
the
of
the
sea
ice
model,
and
you
can
use
that
to
thin
and
thickness
again
ice
in
all
categories,
and
a
little
tweak
we've
introduced
here
is
to
apply
some
nudging
to
the
concentration
but
to
focus
first
on
the
thinnest
sea
ice
category
available
in
the
grid
cell.
So
that's
what
that's
the
that's,
where
we'll
do?
M
Our
concentration,
nudging
and
alexander
adjusts
for
for
salt
conservation
just
to
make
sure
that
we're
conserving
water
and
and
we're,
but
we're
still
adding
some
we're,
still
nudging
towards
a
target
thickness,
and
so
we
have
we're
using
both
these
targets
at
once,
whereas
the
ghost
flux
nudging
is
focused
on
cs
volume
or
or
like
the
sea
ice
thickness
for
all
categories,
we're
trying
to
sort
of
separate
out
the
thickness
and
the
concentration
nudging.
So,
as
I
said,
it
is
a
simple
idea:
there's
nothing!
M
It's
not
rocket
science,
but
it's
it's
just
checking
if
this.
If
this
helps
or
not-
and
so
it's
a
hybrid
because
it's
a
combination
of
a
sea
ice
specification
which
kind
of
gives
a
jumpy
sort
of
nudging
from
time
step
to
time,
step
and
that's
been
used
in
some
some
experiments
from
other
modeling
groups
and
using
the
thickness
nudging,
which
tends
to
have
nice
smooth,
smooth
behavior
over
time.
Okay,
so-
and
so
alexander
did
some
tuning
of
this.
M
So
we
can
look
at
the
rms
error
as
a
function
of
the
sea
ice
concentration,
time
scale
and
sea
ice
thickness,
nudging
time
scale.
And
might
you
know
what
you'll
find
is
if
you,
if
you
go
to
very
short
nudging
time
scales,
you'll
get
you
you'll,
get
the
the
the
target
field
right?
Okay,
so,
but
somewhere
in
the
middle,
between
sort
of
you
know
just
a
few
days
and
and
many
days
for
sea
ice
thickness,
and
we
we
actually
looked
at
very
short
time
scales
for
the
ci's
concentration.
M
You
can
kind
of
rank
with
the
responses
and
your
ability
to
get
to
have
a
relatively
less
bias
and
and
better
a
better,
lower
rms
error
and
basically
alexander
hit
on
this
combination
of
the
parameters
for
this
particular
sc,
wacom
4
set
that
he
hit
on.
So
this
is
a
one
day:
sea
ice,
thickness,
sea
ice
concentration
on
nudging
and
a
five
day
sea
ice
thickness,
but
results
may
vary
if
you're
doing
this
with
your
own
with
other
models.
M
So
that's
that's
to
be
determined,
but
you
know-
and
these
were
all
10-year
simulations-
so
it's
a
it's
nice
to
have
a
a
relatively
coarse
resolution
model
to
work
with
to
do
some
of
this
work,
and
I
just
should
mention
that
he
also
did
this
with
scythe
standalone
and
got
pretty
consistent
results
and
and
the
antarctic
as
well.
So
the
the
bottom
line
is
is
shown
here
in
terms
of
like
the
overall
impact.
M
So
the
key
thing
about
all
these
methods
is
that
they
they
really
do
get
the
the
volume
right
so
the
effective
thickness.
So
the
average
thickness
over
the
arctic
and
the
target
is
an
x.
Is
here
the
solid
is,
is
hybrid
and
the
dotted
is
the
ghost
method,
the
the
ghost
method.
M
So
that's
the
older
method
and
the
gray
is
control
and
the
the
the
warm
climate
or
the
sea
ice
perturbation
climate
is
in
purple,
okay,
and
what
you
find
is
that
the
the
what
we
found
is
that
the
hybrid
method
is
better
able
to
capture
the
time,
the
the
seasonal
cycle
of
the
thickness
and
its
response
for,
and
that's
important
and
also
for
the
the
total
sea
ice
area
all
right.
M
So
the
key
thing
is
that
the
the
ghost
nudging
method
overestimates
the
summertime
sea
ice
minimum
in
the
in
the
control
simulation,
and
it
underestimates
the
wintertime
melt
under
two
degrees
of
anthropogenic
global
warming,
all
right
and
so
we're
trying
to
and
when
we
we
do
this
nudging
to
the
thinnest
sea
ice.
That
helps
us
get
kind
of
a
a
a
capture
that
target
with
a
relatively
light
nudging
and
capture
the
full
pnf
target.
M
And
it's
and
it's
all
connected
in
a
sensible
way
to
the
surface
energy
budget
and
there's
some
figures
about
that
in
the
in
the
publication
or
in
the
paper,
that's
in
review.
So
what
does
that
do
to
the
response?
Okay,
so
here's
the
agcm
response.
I
showed
you
before
here's
the
response
from
the
nudging
experiments.
I
also
showed
you
this
before
and
here's
the
results
from
the
hybrid
nudging.
M
So
when
we
focus
on
that
on
the
event,
I'm
nudging,
the
thinnest
the
ice
and,
of
course,
we're
nudging
marginalized
more
favorably
with
this
two
degrees
of
warming,
kind
of
perturbation.
M
We
get
additional
warming
through
the
arctic
that
for-
and
this
is
not
an
observational
target-
we're
just
kind
of
trying
to
capture
what
we
typically
see
in
in
these
coupled
simulations
and
actually
quite
a
deep
warming
in
in
in
the
zonal
mean
all
right,
and
we
find
some
interesting
results
too
for
antarctic,
which
I
have
in
the
extra
slide.
M
But
I
don't
have
time
to
show
it
all
right
and
there
was
one
mystery
in
this
and
which
is
why
we
weren't
seeing
a
strong
response
from
the
ghost
nudging
that
matched
previous
publications
when
we
were
basically
just
using.
M
You
know
the
previous
methods
like
lantau's
methods,
and
the
fact
is
that
the
the
two
degrees
of
warming
sea
ice
perturbation-
that's
actually
pretty
relatively
weak,
forcing
compared
to
end
of
century
forcing
for
it
for
these
simulations,
and
so
we
actually
are
able
to
recover
a
deeper
warming
when
we
use
a
stronger
sea
ice
forcing.
M
But
if
we're
trying
to
capture
the
pa
mitt
protocol,
which
is
you
know-
which
a
lot
of
groups
have
worked
on
already
and
is
already
approved
as
a
project
we
we
have,
we
found
this
sea
ice
concentration.
Nudging
method
worked
better.
So
the
quick
question
I
want
to
sort
of
address
before
concluding
is
just
to
ask.
Well:
are
we
on
the
right
track
with
these
sea
ice
perturbation
experiments
at
all
and
the
and
the
paper
that
I'm
worried
about
and
what
we've
been
talking
about
in
the
group?
M
Quite
a
bit
is
a
paper.
That's
in
review
by
mark
england
and
others
and
mark
has
been
talking
about
this
paper
in
pmp
workshops
and
other
forums
and
the
and
what
they're,
what
they're
saying
this
is
what
work
with
these
and
eisenmann
and
wagoner
as
well,
and
what
they're
saying-
and
this
is
a
result
that
they
showed
within
the
context
of
an
energy
balance
model
and
we've
reproduced
their
results.
M
It's
pretty
pretty
simple
to
do
so,
and
so
this,
if
this
is
the
result
in
temperature
at
the
surface,
for
you
know
typical
global
warming
experiment
under
radiative,
forcing
just
an
idealization
an
ebm,
what
they.
What
they
say
is
that
the
actual
sea
ice
effect
can
be
constrained
by
just
looking
at
the
top
of
the
atmosphere,
surface,
energy,
top
of
atmosphere,
energy
balance
and,
and
if
you
prescribe
the
albedo,
the
planetary
albedo
change
in
in
that
setting.
M
M
So
what
luke
has
done
here
is:
do
a
bunch
of
different
nudging,
experiments,
albedo
experiments
and
so
on,
and
their
point
is
that,
basically,
this
is
a
seasonal
cycle
of
the
response,
and
the
idea
is
that
if
you
try
to
target
a
certain
amount
of
sea
ice
loss
in
a
simulation,
you're
going
to
have
too
much
albedo
change
in
these
perturbation
experiments
and
that
will
lead
to
too
much
polar
warming.
M
Relative
and
overemphasizes
the
role
of
sea
ice
as
a
trigger
of
polar
amplification,
and
it
might
mask
other
effects-
and
that's
basically
what's
shown
here
in
this-
is
in
the
seasonal
cycle.
Basically,
the
specified
albedo
simulation,
which
is
the
one
where
you
you
figure,
the
top
of
atmosphere,
albedo
change
under
under
global
warming.
That
actually
doesn't
give
you
enough
forcing
to
retreat
the
sea
ice
to
to
the
value,
that's
seen
under,
say,
doubled
co2
and
that.
But
if
you
use
a
nudging,
you
you
get
that
target.
M
M
So
that's
kind
of
a
a
bit
of
a
cloud
on
the
horizon
and
I
think
it's
worth
thinking
about
so
just
to
wrap
up
we're,
proposing
a
simple,
hybrid
sea
ice
nudging
method
to
better
explore
coupled
ocean
atmosphere,
responses
to
sea
ice
loss
and
so
we're
trying
to
target
both
the
concentration
and
thickness
at
the
same
time
in
the
pmf
prescription
and
he's
alexander's
been
looking
at
this
for
looking
at
different
models
and
and
looking
at
forward
heat
transport
as
a
kind
of
scientific
focus,
and
we
see
a
very
strong
dependence
of
of
the
mo
of
the
results
in
these
coupled
simulations
on
the
sea
ice
constraint.
M
So
we
have
to
be
careful
on
on
how
we
do
these
experiments,
which
are
very
expensive,
and
there
are
concerns
that
these
experiments
drive
serious
polar
amplification
and
those
concerns,
I
don't
think,
should
be.
They
shouldn't
be
ignored.
There
are
cause
for
concern
thanks.
J
Thanks
so
much
paul,
so
I
think
let's,
let's
have
one
question
on
the
floor
and
then
maybe
the
remainder
of
the
questions
in
the
chat.
Does
anyone
want
to
ask
a
question
to
paul.
A
Sure
I
mean
one
of
the
issues
I've
talked
with
one
towel.
A
lot
about
this
is
is
just
how
do
you
apply
this
nudging
over
the
subgrid
scale
by
thickness
distribution?
You
know:
do
you
just
apply
it
evenly
into
all
the
categories,
or
do
you
just
resort
back
to
a
single
category.
M
No,
so
the
the
well,
we
do
different
things
so
for
the
concentration
we
we
we
nudge
concentration,
starting
with
the
thinnest
sea
ice
category
right.
So
we
sort
of
use
up
or
increase
the
the
thinnest
sea
ice
category,
so
we're
selective
there.
But
for
the
heat
flux
we
just
try
to
copy
lantau's
method
or
which
is
not
his
originally.
M
I
mean
it's,
but
it's
the
one
he's
been
using
and
we
used,
you
know
we're
drawing
from
his
code
and
then
we
so
we
just
apply
that
heat
flux
proportional
to
the
amount
of
thickness
in
each
cut
to
the
amount
of
sea
ice
in
each
thickness
category.
But
it's
the
same
yeah.
J
Thanks
so
much
paul,
it
sounds
like
paul
wants
a
lot
of
feedback,
so
please
continue
to
give
paul
feedback
within
the
chat,
and
we
will
move
on
to
our
next
speaker
and
our
next
speaker
is
patrick
ugranow.
J
B
K
P
Great
all
right
so
good
morning,
my
name
is
patrick
ebernow.
I
am
an
undergraduate
student
at
the
university
of
colorado,
boulder
and
today
I'll
be
talking
to
you
about
a
little
bit
of
my
college,
undergraduate
research
that
I've
got
gotten
to
participate
in
and
we're
focusing
on
the
north,
open
water
polyneia
and
how
it's
going
to
respond
to
a
warmer
climate.
P
So,
with
a
little
bit
of
background
information,
what
is
plenty
on
why
it's
important
so
our
north
open
water
pollenia
is
located
in
northern
batham
bay,
just
west
of
greenland
and
on
average
it's
approximately
80
000
kilometers
squared
at
its
greatest
extent,
and
by
definition
a
pelinia
is
an
area
of
open
water,
surrounded
by
sea
ice
and
our
north
water
plenty
supports
a
vibrant
marine
ecosystem
here
in
the
arctic
and
in
this
image
it's
just
a
kind
of
a
reference
to
see
where
our
polenia
is
and
what
it
is.
P
Additionally,
local
communities
do
rely
on
this
ecosystem
as
a
form
of
sustenance
as
well.
So
understanding
how
global
temperatures
arise
in
global
temperatures
will
affect
our
polenia
is
is
vital
to
the
future
of
this
ecosystem.
P
Okay,
so
we
looked
at
two
data
sets
to
actually
capture
our
polenia
and
we
utilize
csm1
and
nsidc
data
in
order
to-
and
we
plot
that
here
in
our
seasonal
plot
of
area-
and
we
can
see
both
capture
it
fairly
well
and
similar
in
similar
fashions
too.
They
both
model
the
pelini
opening
in
the
same
time
peaking
and
dissipating
at
the
same
time
as
well,
and
we
can
see
that
here
as
well.
P
They
both
model
it
in
slightly
different
fashions,
but
both
captured,
very
well
and
and
how
we
actually
calculate
the
area
of
the
polynia.
Is
we
defined
a
pelina
region
which
is
this
red
box
here,
and
we
summed
grid
cells
that
were
below
70
sea
ice
concentration
into
the
total
area
within
this
region,
and
the
black
contour
line
is
actually
our
70
concentration
line.
P
So
this
dark
blue
is
what's
being
added
into
the
total
area
in
both
of
these
spatial
plots,
and
you
can
see
why
the
area
is
varying
a
little
bit
between
the
cesm
and
nsidc,
as
our
observational
data
actually
models
it
forming
further
south.
While
our
model
sees
it
forming
further
east
instead
of
south.
P
So,
additionally,
in
order
to
calculate
the
area
we
utilized
a
flagged
region
in
eastern
bathroom
bay
as
well
and
this
flagged
region.
Is
this
orange
dashed
box
right
here
and
we
utilize
this
flag,
a
flag
region
to
identify
winter
pelinia
is
no
longer
surrounded
by
sea
ice,
but
rather
when
it
connects
to
those
waters
in
southern
baffin
bay.
P
So
there's
multiple
thresholds
that
we
can
utilize
for
our
flag
and
here's
just
an
example
of
a
few
of
them
that
we
looked
at,
and
an
interesting
note
is
that
we
find
the
threshold
we
use
for
our
flagged
region
doesn't
affect
the
opening
of
our
polenia
as
each
seasonal
plot
captures
it
opening
in
april,
approximately,
though
the
flag
does
highly
affect
the
peak
of
our
polenia
and
its
dissipation,
as
can
be
seen
throughout
these
six
subplots.
P
So
you
know
how
we
calculate
area
so
we'll
get
into
how
the
plenum
is
actually
changing,
as
temperatures
increase
and
we
looked
at
three
warming
levels
to
actually
determine
what
is
changing,
and
that
includes
the
one
and
a
half
two
and
four
degree
celsius
scenarios
and
as
these
temperatures
increase,
we
found
that
our
polenia
opens
up
earlier.
So
it's
forming
earlier
and
dissipates
sooner
and
when
we
say
dissipate,
we
just
mean
that
it's
no
longer
surrounded
by
ice
and
it's
connected
to
waters
in
southern
bathurn
bay.
P
So
here,
looking
at
our
seasonal
plot,
we'll
focus
on
its
earlier
dissipation.
First
historically,
we
see
that
our
polenia
is
present
in
july
at
approximately
30
000
kilometers
give
or
take,
though,
as
we
hit
the
one
and
a
half
degree
warming
scenario,
we
see
that
it
decreases
drastic
drastically
to
nearly
5
000
kilometers
squared
and
if
we
already
hit
the
two
or
four
degree
warming
scenario.
P
The
polenia
is
no
longer
present
in
july
and
is
really
just
open
water,
and
we
can
see
that
here
in
our
our
spatial
plots.
Our
flagged
region
still
has
large
ice
concentrations
within
it,
keeping
it
the
plinia
isolated
and
additionally,
we
still
see
somewhat
of
that
same
concentration
here,
but
as
we
get
to
the
two
and
four
degree,
and
especially
in
the
four
degree,
we
can
see
that
sea
ice
concentrations
are
nearly
they're
not
even
present.
P
In
the
four
degree,
warming
scenario,
which
is
showing
how
our
polenia
is,
is
completely
gone,
and
it's
just
open
water
by
july.
P
Okay,
so
we
focus
on
dissipation,
it's
earlier
dissipation
and
we'll
focus
on
now.
It's
it's
earlier
formation
as
well-
and
this
is
the
same
seasonal
plot
from
the
previous
slide,
but
we
just
zoomed
in
to
really
capture
what
was
going
on
and
when
we
look
at
april.
Historically,
we
see
that
the
plenus
just
starting
to
form
it's
approximately
a
thousand
kilometers
squared,
and
we
can
see
that
here
along
the
eastern
portion
of
our
plenum.
It
starts
to
form
a
little
bit
right
here,
maybe
a
little
bit
on
our
western
edge.
P
We
see
that
the
plinia
is
nearly
three
times
larger
in
size
in
april,
at
about
three
thousand
kilometers
squared,
and
we
can
see
the
signature
a
bit
stronger
now
as
it's
starting
to
open
up
a
little
more
along
this
eastern
portion
and
maybe
a
couple
spots
on
our
western
side
as
well,
and
really,
we
really
see
a
big
change
by
the
four
degree
warming
scenario,
where
it's
nearly
five
times
bigger
than
historical
data
and
it's
very
strong
signature.
P
So
we
know
how
sea
ice
concentrations
are
going
to
be
changing
and
how
temperature
global
temperature
increases
is
going
to
affect
our
plenty
of,
but
we
also
want
to
look
at
ocean
conditions
as
well,
and
specifically,
we've
looked
at
a
mixed
layer
depth
and
we
utilize
the
same
three
warming
scenarios:
the
one
and
a
half
two
and
four
degrees
scenario
for
mixed
layer
depth
as
well
and
in
this
seasonal
plot.
We
can
see
that
the
mixed
layer
depth
increases
drastically
in
the
early
months.
The
pallenia
is
open.
P
P
So
as
soon
as
our
polenia
connects
to
open
waters,
which
tends
to
happen
in
july,
as
temperatures
increase,
we
see
that
mixed
layer
depth
decreases
drastically
to
nearly
10
meters
in
depth
in
july,
and
I
should
note
that
this
is
a
seasonal
plot
of
just
one
grid
cell
within
our
polenia,
and
it's
just
a
rough
estimate
of
where
we
are
on
this
image.
P
76,
north
and
285
east,
but
you'll
see
that
better
in
the
next
slide,
so
we
have
the
same
seasonal
plot
of
that
one
grid
cell
and
that
dot
in
each
of
our
spatial
plots
represents
that
one
grid
cell
and
very
quickly.
We
noticed
that
the
mixed
layer
is
deepening
and
is
great.
This
deepening
is
greatest
on
the
western
side
of
our
paulinha.
P
As
you
can
see
here
this.
This
dark
blue
signature
is
about
approximately
30
40
meters
in
depth.
But
if
we
are
to
hit
that
one
and
a
half
degree
warming
scenario,
we
see
a
drastic
deepening
and
deepening
is
largest.
We
noticed
when
we
are.
We
limit
warming
to
this
one
and
a
half
degrees
scenario,
though
under
greater
warming,
and
especially
in
the
four
degree
warming
scenario,
we
note
that
mixed
layer
depth
becomes
much
more
uniform
throughout
our
pallenia
and
in
baffin
bay
as
whole
as
a
whole.
P
P
So,
in
summary,
we'll
just
cover
the
quick
key
points:
the
north
water
plena.
P
Although
again,
we
didn't
know
that,
regardless
of
what
threshold
we
use
for
the
eastern
flag,
it
did
not.
It
does
not
affect
the
opening
of
our
pulling
the
timing
of
the
opening
about
bulimia.
P
Additionally,
we
looked
at
mixed
layer
depth
and
we
noted
that
it's
projected
to
increase
drastically
early
in
the
year
and
we
saw
that
in
up
to
a
50
meter
increase
is
expected
along
the
western
side
of
our
pallenia.
If
temperatures
are
if
warming
is
limited
to
one
and
a
half
degrees
centigrade.
P
Additionally,
if
we
continue
on
through
those
warming
scenarios
and
hit
a
four
degree
scenario,
we
saw
that
mixed
layer
depth
becomes
more
uniform
throughout
our
polenia
in
baffin
bay
as
well.
P
That's
the
extent
of
what
we've
looked
at
currently,
but
we
are
currently
looking
at
vertical
velocities
and
biological
productivity
in
our
palinear
region.
As
well
and
we're
trying
to
understand
how
these
other
ocean
conditions
are
changing,
as
these
are
the
key
characteristics
that
support
this
vibrant
marine
ecosystem,
so
understanding
how
these
ocean
conditions
are
changing
as
well,
will
give
us
insight
into
how
this
pelenia
and
how
this
ecosystem
will
be
affected
as
global
temperatures
increase.
P
But
that's
it
I'll
try
and
answer
any
questions.
You
only
have.
J
Thanks
so
much
patrick,
so
I
can
see
that
dave
has
a
question
dave.
Do
you
want
to
ask
your
question.
J
Just
just
barely
okay,
okay,
patrick,
go
ahead
and
take
the
questions
on
the
chat
and
thank
you
so
much
for
your
talk
and
now,
let's
see
if
the
breakout
rooms
are
available,
oh
they
are
not
yet
available.
Maybe
let's
see
okay
thanks
todd.
J
Wait
are
there
29
breakout
rooms?
Oh
okay,
there's
people
not
assigned,
I
see
okay,
so
we
will
head
into
the
breakout
rooms
to
discuss
sort
of
future
directions
for
the
polar
climate
working
group
and
where
our
interests
lie
in
terms
of
our
scientific
objectives.
So,
looking
forward
to
talking
to
you
all.
B
K
B
I'm
not
hearing
you,
oh
you're,
on
mute.
Oh
sorry,
can
you
hear
me
now
yep?
Could
you
put
me
in
the
same
room
as
betty
and
bear
just
tell
me
the
name
of
the
room?
It's
a
multi-millionaire
ice
sheet,
modeling.
D
B
K
J
Well,
that
was
an
exciting
discussion.
We
all
got
like
just
shoved
into
here
when
everyone
was
mid
mid-sentence.
So
thank
you
all
for
for
that
exciting
session
today,
and
I
think
that
we
are
just
about
ready
to
wrap
up
unless
there's
any
final
thoughts
that
any
of
my
fellow
co-chairs
or
anyone
else
wants
to
contribute.
And
then
I
think
after
this
we
are
all
planning
to
perhaps
go
find
some
lunch
or
snacks
to
eat
and
we
might
join
up
again
here
just
for
some
catch
up.
E
Yeah
so
I'll
just
say,
thanks
also
to
everyone
and
and
thanks
for
the
great
discussion,
and
at
least
you
know,
the
co-chairs
all
oh
and
other
people
that
were
sort
of
in
the
rooms
overseeing
the
discussions
will
will
coordinating
and
get
kind
of
a
group
input.
That'll
inform
our
computing
proposal
and
other
things
moving
forward,
and
hopefully
some
of
you
stick
around
for
lunch
too.
A
It
does
look
like
they're
keeping
this
zoom
session
open
for
for
the
lunch
networking
so.