►
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
B
A
A
And
we
have
like
some
Logistics
or
housekeeping
to
note
about
the
room.
One
is
speakers:
will
have
these
microphones
on
put
them
in
the
middle
of
your
of
your
body,
so
that,
if
you
turn
your
head,
you
will
still
be
picked
up
when
you
take
it
off
turn
it
off,
so
that
we're
not
getting
like
static
in
the
transition.
Don't
put
your
microphone
in
your
phone
in
your
pocket
with
your
phone,
because
that
will
also
give
static
and
there's
microphones
on
the
ceilings.
Those
two
discs
with
the
red
when
those
are
green.
A
Oh,
we
need
to
log
in
online
so
that
we
can
monitor
the
chat
all
right,
the
the
folks
who
are
online.
You
know,
please
use
the
chat
if
anybody
in
the
room
wants
to
log
in
so
you
can
participate
in
the
chat.
That
would
be.
That
would
be
great
we'd
like
to
make
it
as
fully
interactive
with
the
folks
who
are
here
locally
and
the
folks
who
are
remote
as
possible.
B
E
Sorry
there
you
go
so
to
start
off.
We
all
know
that
plants
have
stomata
on
leaves
that
regulate
gas
exchange
with
atmosphere.
So
we
haven't
opened
stomach
with
open
stomach
or
high
small
conductance.
Then
we
have
a
closed
amount
of
low
small
conductance,
and
so
why
do
we
care
about
small
conductance?
E
Well,
changes
in
small
conductance
Drive
changes
in
water
flux,
which
can
impact
the
water
cycle
and
affect
the
water
availability
and
the
risk
of
extreme
climate
events
like
droughts,
heat
waves
and
floods
and
and
see,
and
it
changes
in
some
more
conductance
can
affect
the
effects.
Surface
properties
like
evaporative
resistance
in
Murray
siligur
at
all,
2019
has
shown
that
changes
in
evaporated
resistance
can
then
further
affect
surface
temperatures
in
CLM.
Small
conductance
is
represented
by
the
Medlin
model
equation
shown
here
and
we're
interested
in
the
melon
slope,
which
is
boxed
and.
E
E
So
this
is
a
platform
with
adult
2015,
showing
the
estimated
melon
slope
based
on
field
observation.
So
on
the
y-axis
here
we
have
the
estimated
Mountain
slope
and
on
the
x-axis
we
have
temperature.
Each
color
represents
a
plant
type
and,
as
we
can
see
here,
there's
a
huge
range
of
metal
slope,
values
across
and
within
plant
types,
and
the
this
large
range
can
lead
to
large
variants
for
plant
related
processes.
E
So
in
CLM
one
melon
slope
values
is
used
to
represent
each
plant
type.
So,
for
example,
this
is
the
estimate
amount
of
values
for
the
broadleaf
evergreen
tropical
trees,
based
on
the
lineadale
2015
data,
and
we
have
a
low
of
1.7
to
a
high
of
9.11.
But
a
default
value
of
4.45
is
the
one
value
used
for
CLM,
and
so
this
really
begs
the
question.
What
would
happen
if
we
used
a
different
melon
slope
value,
so
we're
going
to
focus
on
the
month-up
parameter?
E
And
why
are
you
sufficiency
is
related
to
mount
soaps
proportionally
and
it's
defined
as
photosynthesis
over
transpiration,
and
this
is
a
simple
visualization
showing
the
relationship
between
water
sufficiency
in
the
middle
cell
parameter.
So
with
the
increase
in
melon
slope,
we
would
expect
a
decrease
in
water
sufficiency.
However,
if
we
don't
consider
water
stress,
we
see
a
different
relationship
between
for
the
synthetic
rate
and
the
metal
and
slope
parameter.
So
with
an
increase
of
parameter.
E
So
this
is
a
spatial
plot
of
the
differences
in
photosynthesis
between
the
high
and
default
myelin
slope,
where
a
brown
corresponds
to
a
decrease
in
photosynthesis
and
the
green
corresponds
to
an
increase
in
photosynthesis
and
we
can
see
for
the
high
melon
slope
case.
That
is
only
really
Brown
in
the
plot,
and
so
it
decreases
in
photosynthesis
are
pretty
consistent
globally,
oh
okay,
and
if
we
compare
it
against
the
unstressed
for
the
synthetic
rate
plot
we
can.
E
We
would
have
if
that,
if
this
plot
were
to
hold
true,
we
would
have
expected
to
increase
in
photosynthesis,
which
we
clearly
don't
see.
So
it's
clear
that
the
water
use
efficiency
plot
holds
a
more
dominant,
is
a
more
dominant
player
here
or
holds
true
more.
But
if
we
take
a
look
at
the
low
melon
slip
case.
So
that's
the
spatial
plot
on
the
left.
E
Here
we
see
a
lot
of
color
a
lot
of
green,
a
lot
of
brown,
and
this
really
shows
that
the
response
to
a
low
melon
sub
is
regionally
dependent
or
yeah
the
result
yeah.
And
since
it's
it's
pretty
complic,
it's
a
pretty
we're
gonna
Focus,
regionally
break
down.
Regionally
to
understand
what's
happening
for
the
low
Mountain
sub
case,
and
we
chose
to
focus
on
Amazon
first,
so
that
red
box
is
just
to
constrain
the
data
we
analyze
and
photosynthesis
increases
in
the
Amazon
with
low
metal
and
slope.
Our
water
use
efficiency
plot
still
holds
true.
E
We
see
an
increase
in
water
sufficiency,
with
a
low
metal
and
slope
and
I'm
going
to
walk
through
mechanistically.
Why
we?
That
might
be
the
case.
Let
me
see
here
increase
the
photosynthesis
with
that
increase
in
water
sufficiency,
so
we
see
an
increase
in
both.
Why
you
sufficient?
Why
use
efficiency
in
photosynthesis?
We
saw
decreases
in
transpiration,
which
is
expected
with
an
increase
in
more
use
efficiency.
E
We
also
saw
changes
in
heat,
flux
or
decreasingly
heat
flux
and
increase
and
sensible
heat
flux,
which
is
also
as
expected,
but
we
actually
also
see
a
decrease,
pretty
large
decrease
in
water
stress
and
an
increase
in
soil
liquid
content.
So
we
think
that
the
changes
in
water
due
to
the
aluminum
slope
and
an
increase
in
water
use
efficiency
caused
a
increase
in
photosynthesis.
E
However,
this
was
for
the
uncoupled
case.
If
we
turn
on
the
atmosphere,
so
we
couple
the
atmosphere,
we
actually
see
something
different
happens
to
photosynthesis.
So
on
the
left,
we
have
a
spatial
plot
in
the
of
the
Amazon,
with
a
couple
case,
with
the
difference
in
photosynthesis
for
the
low
and
d
low,
minus
default,
metal
and
slope,
and
on
the
right
we
have
the
same
spatial
plot
we
saw
earlier
for
the
uncoupled
case
and
if
our
worry
is
sufficient
increase
in
water
use,
efficiency
were
to
hold
true
or
a
decrease
in
photosynthesis.
E
It
turns
out
that
the
impacts
on
water
are
similar
in
both
cases.
There's
similar
decreases
in
water
stress
and
increases
and
soil
with
all
your
content,
and
it
actually
also
turns
out
that,
due
to
atmospheric
feedbacks,
we
see
a
much
more
significant
increase
in
temperature
and
we
think
this
is
why
we
see
a
decrease
in
further.
E
So
the
first
is
my
atmosphere,
coupling
that
I
showed
previously.
The
next
is
we're
interested
in
what
happens
when
we
increase
the
CO2
since
that
changes,
the
relationship
between
carbon
water
fluxes
and
we're
also
interested
in
Dynamic
leap
year,
compared
against
a
fixed
Leaf
area,
since
Leaf
area
would
change
the
relationship
between
water's
efficiency
and
the
total
fluxes
of
carbon
water,
and
on
top
of
all
of
these
factors,
we're
also
going
to
continue
to
look
at
the
low
and
high
melon
slope
compared
against
the
default
myelin
subcases.
E
The
two
little
boxes
in
red
are
the
two
configurations
that
I
showed
earlier
that
we've
compared
so
there's
a
lot
of
a
lot
more
runs
to
analyze
and
compare
and
yeah
to
cook
analyze
and
compare
but
I'm
going
to
go
back
to
the
the
two
configurations
I
showed
earlier
in
the
bar
chart
form
now.
So
this
is
a
bar
chart
showing
the
percentage
differences
of
photosynthesis
in
my
Amazon.
E
So
the
y-axis
is
the
percent
difference
in
photosynthesis
and
we
have
the
couple
and
then
coupled
case
that
I
showed
earlier
here,
and
we
see
that
it's
two
completely
opposite
signs
of
response.
So
now
I'm
going
to
keep
the
uncoupled
case
and
compare
it
against
which
was
a
fixed
Leaf
errand
compared
against
the
dynamic
Leaf
area
case.
And
we
see
it
amplification
of
response
for
the
synthetic
response
and
now
I'm
going
to
take
that
same
uncoupled,
fixed
Lai
case,
which
is
actually
one
times
pre-industrial
CO2
and
compare
it
to
doubling
the
CO2.
E
E
B
E
I
Katie
great
talk,
Amy.
This
is
really
cool
to
see
all
the
differences
in
this
bar
chart.
I
guess
I'm
wondering
if
you're
thinking
about
how
the
different
factors
combine
if
they're,
linear
or
not,
are
there
ways
to
kind
of
tease
out
when
you
have
one
thing:
changing
versus
multiple
things,
changing
it's
a.
E
I
J
I
was
just
going
back
to
sanjeev's
question:
do
you
think
it's
the
coupling
or
is
it
the
fact
that
the
climate
is
different?
That
is
the
reason
why
you're
saying
such
a
strong
difference
between
you
couples.
So
if
you
were
to
take
the
output,
let's
say
of
a
coupled
run
for
and
put
that
through
a
couple
that
and
put
it
in
uncoupled,
so
you
just
did.
That
was
very
forcing
do
you
think
you
would
see
the
same
response
is
what
I'm
trying
to
answer
is
the
coupling
or
whether
it's
the
climate
difference?
E
C
B
E
E
K
L
B
B
C
B
E
M
So
in
several
of
yesterday's
talks,
we
saw
that
there's
a
lot
of
uncertainty
in
land,
carbon
sink
estimation,
and
a
lot
of
that
has
to
do
with
our
understanding
of
how
vegetation
will
respond
to
the
carbon
dioxide
fertilization
effect.
So
some
of
that
response
is
likely
governed
by
vegetation
Stoichiometry,
which
is
the
proportion
of
carbon
to
nutrient
elements
in
plant
tissues.
M
So
today,
I'm
going
to
focus
mostly
on
foliar
carbon
to
nitrogen
ratios
and
as
atmospheric
CO2
Rises,
those
carbon
to
nitrogen
ratios
increase
as
plants
assimilate
more
carbon,
but
they
might
need
to
take
up
a
matching
amount
of
nitrogen
phosphorus.
To
maintain
some
of
these
internal
stoichiometric
ratios
as
that
carbon
increase,
carbon
assimilation
increases
and
we're
not
certain
about
the
degree
of
that
flexibility
of
that
Stoichiometry.
But
there
is
some
evidence
in
a
lot
of
Earth's
ecosystems
that
carbon
to
nitrogen
ratios
increase
under
elevated
CO2.
M
So
this
study
is
a
meta-analysis
of
free
air,
carbon
enrichment
study
data
and
they
demonstrated
that
across
plant
tissues,
as
well
as
in
soil
pools.
There
is
an
increase
in
C
to
n,
under
elevated
CO2
relative
to
ambient
CO2
and
today
I'm,
going
to
focus
on
these
Leaf
foliar
values,
there's
about
15
to
20
percent
increase
in
foliar
C
to
n
under
elevated
CO2,
and
this
is
seen
in
these
Freer
carbon
enrichment
studies.
M
It's
also
seen
in
a
number
of
natural
systems
and
is
a
fairly
common
phenomenon
across
ecosystems
as
atmospheric
CO2
Rises,
and
this
then,
is
likely
to
influence
how
productive
these
ecosystems
can
be
in
the
future
because
of
those
internal
stoichiometric
balances.
However,
we
don't
currently
have
a
representation
of
Ceta
and
foliar
C
to
n
in
models
that
changes
in
response
to
elevated
CO2
so
that
it
increases
progressively
over
time,
as
is
seen
in
some
of
these
ecosystems.
M
So
we
generated
this
new
equation
based
off
of
some
of
those
face,
meta-analyzes
as
well
as
lter
syntheses
to
project
the
leaf
carbon
to
nitrogen
ratio
as
a
function
of
atmospheric
CO2,
which
we
implemented
into
CLM
and
ran
that
offline
with
land,
only
dswp3
forcing
and
an
ssp-370
anomaly.
Now
you
might
be
familiar
with
a
flexible
C
to
n
term
that
is
already
in
clm5.
M
However,
the
fun
module
that
controls,
the
nitrogen
cycling
has
fixed
plant,
functional
type,
specific
C
to
n
values
and
that
sort
of
outweighs
the
flexible
parameterization
in
CLM,
and
so
that
means
that
the
Fourier
C
to
n
values,
don't
currently
really
change
over
time
very
much
in
CLM,
which
you
can
see
here
in
this
orange
line
at
the
bottom
of
this
plot.
So
that's
our
fixed
foliage
data
and
scenario
with
the
current
parameterization
of
CLM,
and
this
blue
line.
M
Flex
is
our
new
parameterization
with
that
CO2
driven
foliar
C
to
n,
which
then
goes
up
over
time
out
through
the
end
of
the
century.
So
this
essentially
shows
that
our
new
parameterization
was
doing
what
we
hoped
it
would
do
by
increasing
the
foyer
C
to
n
as
CO2
changes
in
the
model,
and
in
that
fixed
foyer
see
the
end
scenario.
There
are
different
C
to
n
values
specific
to
each
plant,
functional
type,
so
those
range
from
about
20
to
50
across
the
globe.
However,
those
don't
really
change
over
the
course
of
the
simulation.
M
However,
in
looking
at
the
difference
of
the
fixed
and
flex
scenarios
by
the
end
of
the
century,
you
can
see
that
there's
an
increase
of
about
20
in
the
C
to
n
values
across
most
of
the
pfts,
showing
again
that
our
new
parametrization
is
doing
what
we
had
hoped
it
would
do
in
the
model.
Now
we
wanted
to
first
see
how
that
would
then
influence
some
of
the
carbon
cycling
variables
that
were
of
interest
to
us.
First
thinking
about
the
cumulative
land
carbon
uptake,
this
looks
at
the
land
carbon
sink
over
the
historical
period.
M
First,
you
can
see
our
fixed
scenario
here
in
Orange
compared
to
the
flexible
scenario
in
blue.
Both
scenarios
were
well
within
observationally
Drive
data
from
the
global
carbon
project,
which
is
in
black
and
the
gray
area,
is
the
confidence
interval.
They
were
also
within
the
confidence
interval
of
11,
an
11
member
Ensemble
of
cement
6
data,
which
is
in
purple.
So
this
suggests
that
both
are
fixed
and
flexible
scenarios
are
plausible
scenarios
to
use
to
project
the
land
carbon
sink.
M
However,
when
we
look
at
these
out
through
the
end
of
the
century,
they
really
strongly
diverge
with
our
flexible
scenario,
being
about
200
pedigrams
of
carbon,
less
than
the
fixed
C
to
n
scenario,
suggesting
that
how
we're
representing
foliar
carbon
to
nitrogen
in
the
model
is
likely
to
have
really
major
implications
for
projections
of
the
land
carbon
sink.
Now,
this
change
in
carbon
uptake
was
not
evenly
distributed
across
the
globe,
so
this
map
looks
at
the
difference
in
the
fixed
and
the
flex
line
by
the
end
of
the
century
spatially.
M
We
can
see
that
they're
the
greatest
reductions
in
the
land,
carbon
sink
in
tropical
regions,
as
well
as
in
some
Boreal
regions,
and
this
owes
largely
to
a
reduction
in
photosynthetic
capacity
in
the
model.
So
this
is
demonstrated
here
for
NPP
on
the
y-axis,
which
is
reduced
pretty
strongly
by
the
end
of
the
century.
In
that
flexible
scenario
again
shown
in
blue,
we
also
showed
similar
responses
for
gpp
leaf
area
index
and
heterotrophic
respiration,
which
I
won't
show
here.
But
if
you
are
interested
in
this
data,
I
have
them
elsewhere
and
can
share
them
later.
M
This
change
in
photosynthetic
capacity
was
similarly
spatially
distributed
as
changes
to
the
land
carbon
sink,
so
this
again
shows
a
difference
between
the
fixed
and
the
flex
line
and
the
last
plot
across
the
globe,
with
the
greatest
changes
being
in
tropical
regions
again,
and
this
actually
had
some
really
interesting
implications
for
biophysical
cycling.
That
I
hadn't
really
necessarily
thought
a
lot
about
before
running
this
experiment,
but
these
changes
to
MPP
and
vegetation
productivity
specifically
had
really
major
implications
on
the
hydrologic
cycle.
M
Given
the
role
of
tropical
evapotranspiration,
especially
in
driving
large-scale
precipitation
patterns,
we
did
also
see
that
there
were
changes
in
nitrogen
cycling
in
our
flexible
scenarios,
so
this
demonstrates
that
there
was
a
reduction
in
net
nitrogen
mineralization
rates.
Nitrogen
cycling
rates
were
generally
lower
in
the
flux
scenario,
which
again
is
shown
here
in
blue.
This
mostly
reflects
that
change
in
productivity.
That
I
showed
you
earlier,
because
the
reduction
in
productivity
reduces
the
vegetation,
nitrogen
demand
and
that
then
reduces
nitrogen
cycling
rates.
M
However,
this
doesn't
necessarily
allow
us
to
easily
parse
whether
there's
some
below
ground
feedbacks
going
on
that
may
be
further
changing
productivity
So.
Currently,
the
model
doesn't
represent
flexible
soil
Stoichiometry.
So
this
doesn't
necessarily
allow
us
to
see
if
there
is
a
feedback
between
changes
in
foliar,
Stoichiometry
and
below
ground,
nitrogen
cycling
and
soil
Stoichiometry.
That
could
further
lead
back
into
changes
in
productivity.
M
So
that's
one
of
the
next
steps
that
I
hope
to
address
with
this
research,
and
hopefully
we'll
continue
investigating
them
that,
through
potentially
through
using
the
mimics
model
and
see
if
we
can
examine
some
of
those
feedbacks
a
little
further
another
Next
Step.
That
I
would
be
curious
to
hear
your
thoughts
on
for
this
research
is
whether
or
not
this
parametrization
might
be
useful
in
ctsm,
because
it
is
not
currently
representing
this
sort
of
directional
change
in
foliar
Stoichiometry,
but
hopefully
at
some
point
in
the
future
would
be
a
representation
represented
by
the
model.
N
Yeah
hi
thanks,
that's
really
interesting,
so
I
guess
I
was
so
in
as
you
as
you,
as
you
alluded
to
it's,
not
strictly
fixed
acetylene
ratio
in
Salem
Five.
We
have
this.
This
function,
which
changes
the
Theta
and
ratio
as
a
function
of
how
expensive
it
is
for
the
plants
to
get
nitrogen
out
the
soil,
as
as
is
the
kind
of
concept
embedded
in
the
fund
model
right.
N
M
I
think
that
that,
as
I
understand
it,
that
would
kind
of
be
a
goal
so
that
the
C
to
end
does
change
as
a
Plant
Exchange
cost,
but
because
there
is
something
that
is
currently
sort
of
overriding
the
ability
of
the
sea
to
end
to
be
flexible
in
CLM.
It's
not
necessarily
performing
that,
which
is
why
we
introduced
this
sort
of
empirically
drive.
If.
N
I'm
getting
that
right
yeah,
so
it's
not
it's
not
it's
not!
It
doesn't
override
it.
It's
just
it's
parameterized
as
you've.
Illustrated
is
parameterized
such
that
it's
not
very
variable
so
that
it's
yeah.
It's
quite
logic.
It's
quite
a
big
difference
between
being
invariant
and
not
very
variable,
yeah,
so
yeah,
so
I
guess
I
guess
what
I
take
from
your
study
is
that
that
is
that
that's
parameterized
on
the.
O
N
Side
so
I
guess
I
guess
I'm
interested
in
whether
it's
possible
to
parameterize,
not
if
not
the
function
we
have
now,
admittedly,
but
a
function
which
is
tied
to
the
cost
of
nitrogen
from
the
soil,
because
you'd
imagine
that
that
would
be
the
driving
factor
of
this
of
this
flexible
Stoichiometry
right.
You
see
you
can't
just
keep
taking
up
as
much
nitrogen
as
you
want.
It
gets
more
expensive
as
CO2
fertilization
carries
on
and
so
yeah
I
think
that'd
be
interesting
and
we
can
talk
offline
if
you
like,
but.
M
M
P
P
Yeah,
so
we
just
cycle
over
the
last
I
think
it
is
15
years
of
the
gswp3
forcing
and
then
apply
an
anomaly
that
kind
of
matches
the
change
in
in
driver
variables
that
the
cesm2
represented
so
that
way
we're
kind
of
able
to
smoothly
transition
from
the
historical
period
to
the
SSP.
But
you
do
get
that
periodic
oscillation,
because
we
are.
P
C
R
C
B
B
S
Okay,
okay,
thank
you!
So
much
so
hello,
everyone,
my
name
is
Jenning
bomb
and
I'm
a
second
year
PhD
student
working
with
Professor
Pierre
Jenkins
at
Columbia
University,
so
first
of
all,
I
would
like
to
apologize
for
all
of
some
reason.
Simple
scheduling
and
today,
I
will
be
presenting
this
talk
on
behalf
of
Professor
gentine
nesbales
Dr
Julian,
who
is
a
postdoc
in
our
group,
also
working
on
different
classes.
S
So
the
topic
of
this
talk
today
will
be
reconstruction
of
a
long-term
spatially.
Contiguous
solar
induced
fluorescence
data
over
1982
to
2021,
and
the
main
motivation
for
this
project,
in
particular,
is
implication
for
land
service.
Modeling
is
that
many
times
when
it
comes
to
monitoring
long-term
vegetation
Dynamics,
we
do
not
have
very
good
Global
constraints.
We
have
satellite
observations
from,
for
example,
avhr
back
in
the
1980s.
S
So
in
order
to
address
this
issue,
we
believe
that
there
is
a
need
for
a
long-term
solar
induced
fluorescence
product
solar
induced
fluorescence
might
be
helpful
in
this
case,
because
it
is
a
signal
that
is
more
closely
related
to
the
physiological
response
of
vegetation
photosynthesis.
So
maybe
it
is
a
proxy
that
is
closer
to
the
the
biophysical
biophysiological
effects
of
garlic
capture
and
previous
study
have
shown
that
there's
a
great
potential
using
save
to
monitoring
gpp.
S
By
relying
on
some
of
this,
this
statistical
relationships
between
Sith
and
reflectance,
because
we
can
monitor
reflectance
for
us
for
a
longer
period
and
also
Reflections
data
sets
are
spatial
in
contiguous
and
there
have
been
multiple
successful
attempts
of
reconstructionists
globally
confusive.
But
a
question
remains
the
extent
to
which
these
reconstructive
system
can
be
actually
attributed
to
physiological
signals.
And
if
we
were
to
extend
the
these
reconstructive
steps
to,
for
example,
early
1980s,
then
we
need
to
use
earlier
satellite
records
of
reflectors.
S
But
that
again
it's
a
challenging
topic
because
we
don't
know
how
much
confidence
we
have
in
these
earlier
satellites
and
also
when
it
comes
to
avhr.
There
are
only
two
bands,
like
broader
bands
of
Reflections
in
the
rad
and
here
Emperor
regions,
whereas
where
many
of
these
reconstructed
sips,
they
use
the
first
warbands
of
modus.
So
it
is
unclear
whether
we
can
have
a
reliable
reconstruction
of
solar,
angios
viruses
all
the
way
back
to
the
early
1980s.
S
And
but
here
we
decided
to
take
on
this
Challenge
and
we
want
to
see
how
how
far
we
can
go
in
this
direction.
So
in
order
to
make
sure
that
we
have
some
confidence
in
the
ahr
data
set,
we
apply
a
relatively
systematic
approach
to
calibrate
avhr
against
about
this.
So
we
apply
all
the
relevant
quality
black
in
the
ltdr
data
and
also
we
apply
orbital
drip
correction.
S
So,
basically,
we
removed
the
potential
effects
of
swords
and
its
angle,
the
avhr
reflectance
data,
using
a
method
that
was
previously
found
to
be
successful
in
a
paper
by
John
at
all,
at
2018
and
after
orbital
correction
that
avhr
data
set.
We
combine
it
with
modus
and
we
cross
calibrated
with
modus,
and
here
we
use
a
double
correction
procedure
in
order
to
leverage
both
the
temporal
and
spatial
correlational
structures
between
lhr
and
Modis.
S
So,
first
we
apply
a
pixel,
wise
linear
model
and
that's
mainly
based
on
a
local
correlational
structure
from
the
temporal
domain,
and
we
built
that
linear
model
to
First
apply
the
first
layer
of
corruption,
but
we
think
that
there
might
be
other
systematic
biases
between
your
two
data
sets
due
to,
for
example,
differences
in
reflectance
bands,
different
atmospheric
calibration
procedures.
So
the
next
stage
we
apply
a
global
calibration
using
a
machine
learning
model
and
we
use
aerosol
Optical
desks.
Nope.
That's
cloud
cover
elevation
and
energy
indexes
predictors
and
that
is
to
Leverage
The
Globe.
S
S
That
uses
the
first
four
bands
of
the
Modis
and
we
also
examine
the
potential
residuals
orbital
Effects
by
validating
our
data
set
at
this
pseudely
variant
calibration
size.
So
these
are
desert
locations
in
Sahara
or
the
Arabic
Peninsula
doubt.
Theoretically,
there
shouldn't
be
any
vegetation
there,
and
so
any
trends
that
we
observe
should
be
due
to
some
systematic,
biases
or
or
noises,
and
we
found
that
there's
little
correlation
between
these
signal
as
a
picf
sites
and
the
global
anomaly
of
the
Civ.
S
S
We
also
found
that
our
product
has
the
potential
of
capturing
vegetation
responses
to
drug
events.
So
we
look
into
various
significant
Central,
Advanced,
Regional
drought.
Events
we
found
at
our
JSA
is
able
to
capture
this
decrease
in
photosynthesis,
so
our
data
will
be
made
publicly
publicly
available
in
May
this
year
and
there's
a
trial
version
that,
if
you're
interested
you
can
send
us
an
email,
and
we
will
give
you
access
to
a
preliminary
data
release.
S
Don't
note
that
for
the
land
modeling
Community,
our
model
has
the
potential.
Our
data
has
a
potential
of
accessing
long-term
legislation,
productivity
in
response
to
climate,
CO2
and
also
land
use
plant
cover
change,
and
it
can
also
be
used
as
a
potential
Benchmark.
Data
set
for
comparing
or
System
model
results,
as
well
as
assimilating
into
Global
biogeochemical
models.
So,
thank
you
very
much
and
that's
the
end
of
our
presentation.
Any
questions
will
be
welcomed.
S
A
Sort
of
the
validation
Fair
product
it
looked
like
there
were
about.
You
know
two
units
of
dynamic
range
yeah,
the
previous
slide,
that
you
I
think
it's
the
one
previous
two
units
of
dynamic
variability
on
the
Sip,
and
it's
like
the
scatter
about
the
line
was
at
least
like
one
and
a
half
units
of
dynamic
variability.
A
So
you
know
I
know
stiff
suffers
from
imprecision,
but
that
would
suggest
to
me
that
there's
something
problematic
about
the
spatial
temporal
skills
that
you're
you're,
providing
the
data.
At
that
you
know
there
has
to
be
some
recommendation
to
been
or
aggregate
the
data
to
deal
with
that
imprecision.
Can
you
comment
on
that?
A
little
bit.
S
Oh
yes,
so
that's
a
very
good,
very
good
comment,
and
indeed
we
see
that
there's.
So
we
see
that
during
the
modus
period
the
if
you
look
at
the
anomaly,
they
are
generally
consistent
with
each
other.
The
reconstruct
is
it
and
and
that's
not
surprising,
because
they
actually
use
the
same
import
and
output
data
set.
So
they
built
all
of
these
UCL
CO2
Civic
and
all
of
these
use
we've
noticed
as
predictors.
S
So
so
that's
why
the
the
inter-annual
anomaly
are
consistent,
but
when
it
comes
to
the
absolute
change
that
can
that
can
be
different
due
to
the
differences
in
the
algorithms,
and
it
can
also
be
due
to
the
different
ways
like
some
of
the
data
gaps
in
the
data
set.
So
so,
when
I,
when
I
see
people
do
do,
assimilation
of
these
data
sets,
they
generally
look
at
the
normalized,
so
they
gen.
B
I
think
in
the
interest
of
time
we'll
need
to
move
on
so
Peter.
You
could
add
a
question
to
the
chat
if
you
like
and
I
thought
there
was
another
question
for
you
there
in
the
chat.
If
you
want
to
answer
it
next
up
we're
going
to
have
James
King,
who
is
online
James?
Are
we
there
great,
so
I
think
we
need
to
stop
sharing
one
screen
and
then
we
can
start
sharing
the
other.
L
B
C
C
L
To
give
you
a
more
conflict
story
about,
I
was
and
I
will
talk
about
how
I
came
up
with
this
idea
from
very
beginning.
So
this
is
a
water
drainage
pool
in
the
basement
of
my
house
and
there's
only
one
light
bulb
in
this
room,
and
you
can
see
this
light
bulb
makes
a
shadow
in
the
pool
and
in
the
shadow
area
there
was
no
Legend
but
outside
of
the
Shadow.
L
There
grows
a
lot
of
Legend
and
it
made
me
think
of
the
importance
of
light
in
determining
the
growth
of
plant.
So
initially
on,
I
was
thinking.
Can
we
light
up
the
the
whole
dark
side
of
the
earth
to
increase
the
carbon
uptake
on
of
the
whole
biosphere?
So
I
came
up
with
two
plants.
The
first
is
to
launch
visible
light
lasers
to
the
geostationary
orbit.
L
So
the
temperature
increase
depends
on
this
angle.
So
the
larger
this
angle
is
the
larger.
The
air
view
will
be
it's
going
to
be.
Illuminated
will
be
so
you
can
see,
there's
still
a
very
high
temperature
increase
due
to
this
plant
and
the
reason
is
most
of
the
radiation
is
absorbed
by
the
lens
surface
for
hitting.
So
this
was
when
I
shifted
the
idea
to
Forest,
because
theoretically,
most
part
of
the
solar
radiation
that
reaches
the
forest
surface
will
be
used
for
photosynthesis
or
used
for
chemical
energy
storage.
L
So
my
work
is
just
to
explore
the
the
carbon
cycle.
Responses
to
this
plan
lighting
up
chocolate
first
at
night
via
lamb
networks
above
the
forest
canopy,
and
to
qualify
the
influence
on
carbon
sextration,
local
climate
and
eco-environment
and
Technical
and
economic
feasibility
I
conducted
numerical
experiments
by
fully
coupled
csm2,
and
this
work
was
supported
by
encouraging
high
performance
Computing
platform.
L
So
the
nominal
horizontal
resolution
is
one
degree.
Tropical
forests
are
defined
by
broad
leaf
evergreen
tree
heavier
percentage,
larger
than
60
percentage
and
in
csm2.
The
solar
radiation
consists
of
four
components:
direct
visible
light,
diffuse,
visible
light,
direct
near
infrared
and
it
diffuse
near
in
fat.
So
in
the
experiment
on
all
the
other,
three
components
were
specified
to
be
zero.
The
diffuse,
visible
light
was
specified
to
be
200
watt,
primitive
squares
and
the
reason
for
this
choosing
this
specific
number
will
be
introduced
later.
L
So
this
is
the
methodology
framework.
The
whole
experiment
consists
of
a
series
of
sub-experiments.
First
is
on
14
years.
Historical
simulation
and
starting
from
2015
I
did
a
series
of
lighting
experiment.
So
first
is
24
hour
experiment
to
find
the
appropriate
lighting
power.
So
this
is
what
so,
from
this
stage
of
experiment,
I
decided
to
choose
200,
watt,
Community,
Square
and
then
folded
by
a
16-year
lighting
experiment
and
at
2031
I
stopped.
L
The
lighting
experiment
then
kept
running
the
model
for
another
20
years
to
see
how
the
carbon
cycle
and
climate
responded
to
the
experiment
termination
on,
and
this
is
20
years
and
all
this
experiments
were
along
with
control
simulation
and
also
in
two
members
to
estimate.
The
accident
is
all
right.
So
here,
due
to
the
time
limit,
I
omitted
the
24
hour,
experiment
results,
but
if
you're
interested,
we
can
definitely
talk
about
this
24
hour.
L
So
this
is
the
global
tropical
forest,
carbon
flux
and
climate
responses
to
the
16-year
lighting
experiment
and,
as
I
said,
then
I
stopped
the
lighting
experiment
and
keep
running
the
model
for
another
20
years.
Then,
let's
look
at
the
tropical
forest
response
responses
to
this
experiment
termination.
So
when
you
stop
the
lighting
experiment,
the
gpp
and
autotrophic
respiration
dropped
down
quickly,
while
soil
respiration
started
decreasing,
but
also
at
a
much
lower
rate
than
on
gpp
and
plant
respiration.
L
So,
as
a
result,
the
net
carbon
uptake
the
tropical
forest
turned
into
a
net
carbon
Source
at
the
termination
of
the
lighting
experiment
and
local
air
temperature
came
back
to
the
control
level,
as
well
as
the
precipitation
and
soil
moisture.
So
because
tropical
forest
now
started
releasing
carbon
dioxide
into
the
atmosphere.
The
global
CO2
concentration
started
increasing
again,
but
overall,
the
global
average
air
temperature
is
still
lower
than
the
control
simulation
and
then
I
look
at
where
the
the
net
absorbs
carbon
go.
L
So
there
are
three
major
carbon
poles
in
choco,
first
vegetation,
carbon
pole,
coursewood,
adapter
and
litter
carbon
pole
and
then
soil
carbon
pool.
So
you
can
see
that
most
of
the
carbon
most
of
the
net
absorbed
carbon
went
into
the
vegetation
carbon
pool.
But
since
the
lighting
experiment
terminated
the
vegetation
carbon
pool
and
coswood
adapter
in
the
later
carbon
post
started
decreasing,
started
shrinking,
but
the
soil
organic
matter.
L
So
several
discussion
about
this
results.
First
is
the
ecosystem
level.
Food
experiments
are
needed,
so
physiological
responses
of
tropical
trees
to
longer
for
the
periods
at
the
ecosystem
level
Remain,
the
one
of
the
biggest
instrument
is
and
the
tree
growth
might
be
limited
by
nutrient
and
water
supply.
So
this
ecosystem
level
field
experiments
are
needed
to
understand
to
better
understand
the
tropical
forest
ecosystem
responses.
L
The
second
discussion
point
is
csm2.
Overestimated
local
air
temperature
increases
due
to
the
omission
of
chemical
energy
stored
during
photosynthesis.
So
this
is
the
canopy
and
the
canopy
energy
conservation
or
energy
equation
used
to
calculate
temperature.
So
this
is
solar
radiation,
along
with
radiation,
sensible
heat
and
Latin
heat.
So
this
is
energy
equation
used
to
calculate
temperature,
however,
in
csm2
and
so
in
csm2
and
other
modern
Earth
system
models.
The
chemical
energy
that
is
stored
during
photosynthesis
and
released
by
respiration
is
ignored
at
the
net.
L
Chemical
energy
usually
amounts
to
less
than
one
percent
of
absorbed
insulation,
so
this
is
the
normal
condition.
However,
in
our
lighting
experiment,
16-year
light
experiment,
17
of
absorbed
installation
was
fixed
in
the
ecosystem
as
chemical
energy
and
didn't
contribute
to
local
air
temperature
increase,
and
the
model
failed
to
exclude
this
chemical
energy
storage
from
the
energy
equation.
L
So
the
model
overestimated
the
local
temperature
increase,
and
the
implication
of
this
point
is
the
temperature
simulation
results
in
especially
vegetation
temperature
stimulation
results
should
be
treated
carefully
when
Earth
system
models
are
used
to
do
experiments
related
to
solar
radiation
modifications,
then
the
third
poet
discussion
point
is
the
post-action
CO2
outgassing
from
tropical
first.
So
how
do
how
to
understand
this
phenomena
during
the
control
simulation
period?
L
You
see
the
vegetative
primary
productivity
or
MPP
is
on
a
similar
level
of
soil
respiration.
So
you
see
the
screen
line
represents
the
net
ecosystem
productivity.
However,
when
you
give
additional
lighting
to
a
forest
ecosystem,
the
Pro,
the
productivity
of
vegetation
increased
significantly
and
in
a
much
faster
rate
than
soil
respiration,
so
soil
respiration
also
increased.
So
in
a
much
lower
rate
and
as
a
result,
the
net
carbon
uptake
increased.
L
Well,
however,
when
you
stop
the
artificial
water
treatment,
the
the
gpp
decreased
and
soil
respiration
also
starts
decreasing,
but
in
a
much
lower
rate,
but
it
started
the
the
whole
ecosystem
turned
into
a
carbon
source,
and
this
is
pretty
similar
to
just
imagine.
If
you
take
stimulants,
the
stimulants
are
gonna
enhance
your
body
function,
but
when
stimulants
lose
effect,
it's
going
to
destroy
your
body
function,
so
I
just
name
it
on
a
stimulant
effect,
so
giving
forest
ecosystem
additional
lighting.
It's
similar
to
you're,
giving
stimulants
to
human
bodies.
L
Okay,
so
sorry
very
quickly,
so
you
can
also
observe
similar
phenomenon,
the
over
show
scenario
simulations
and
on
the
free
air,
carbon
dioxide
enrichment
experiment.
So
it
implicates
that
CO2
removal
message
focused
on
enhancing
ecosystem
carbon
substration
by
altering
environmental
factors
in
the
short
term
could
induce
this
post-action
suit,
outgassing
very
quickly,
conclusions.
I
think
the
first
is
implications
for
Earth
System
model
users.
The
temperature
simulation
without
should
be
treated
carefully
when
Earth
system
models
are
used
to
do.
L
Experiments
relates
to
solar
radiation
modifications
and
implications
for
nature-based
climate
Solutions,
enhancing
land
ecosystem,
carbon
sequestration
by
changing
environmental
factors
might
be
an
inefficient
approach
and
finally,
for
geoengineering
measures,
it's
important
to
study
the
post,
geoengineering
reactions
well,
I
think
that's
it.
Finally,
I
want
to
say
this.
Article
has
been
published
in
Earth,
System,
Dynamics
and
I'll
be
giving
a
departmental
seminar
tomorrow.
So
just
feel
free
to
reach
out
to
me
we're
going
to
discuss
a
little
more
details
on
so
can
you
shop
for
me
for
a
zoom
link
thanks.
L
B
F
F
Yeah
perfect
I.
Thank
you
very
much.
My
name
is
James
King
I'm,
a
postdoc
at
the
University
of
Sheffield,
here
in
the
UK
and
today
I'm
going
to
talk
to
you
about
some
work
we've
been
doing
with
with
clm5
and
with
the
broader
season
2
model,
looking
at
the
the
atmospheric
and
climate
and
environmental
implications
of
tree
planting
as
a
as
a
carbon
dioxide
removal
strategy.
F
So
first
you
just
acknowledge
my
my
colleagues
James
and
Maria
Maria,
leading
the
project
and
also
Peter
Lawrence,
who
we've
been
working
with
at
ncar
and
also
broadly
the
the
members
of
the
land
model.
Working
group
I
was
fortunate
to
spend
some
time
at
ncar
over
summer
last
year
and
I
had
some
really
productive
discussions
and
thanks
to
everyone,
who's
contributed
to
these
fantastic
modeling
tools,
foreign.
F
Tree
planting
on
a
global
scale
has
really
attracted
a
lot
of
attention
from
governments,
ngos
and
Civil
Society
as
a
strategy
to
to
take
carbon
dioxide
out
of
the
atmosphere.
So
lots
of
countries
have
made
national
commitments,
usually
with
large
and
suspiciously
round
numbers
to
begin
to
plant
two
billion
trees.
F
We're
going
to
plant
three
billion
trees,
there
are
also
International
agreements,
such
as
the
bond
challenge,
countries
agreeing
to
restore
degraded
forests
and
also,
more
recently,
the
the
Glasgow
declaration
as
part
of
last
year's
cop
made
big
commitments
towards
landscape
preservation
and
restoration,
and
it
makes
sense.
You
know
why
invent
direct
air
capture
if
you
can
just
use
a
tree,
but
obviously
trees
have
more
impact
on
the
environment.
F
Beyond
just
absorbing
carbon,
and
some
of
the
motivation
for
this
work
is
the
realization
that
a
lot
of
big
splashy
papers
talking
about
tree
planting
have
focused
mainly
on
how
much
carbon
will
they
absorb
and
there's
been
less
attention.
Although
that
that's
a
picture,
that's
changing
now
on
the
climatic
and
environmental
impacts
of
trees,
which
are
numerous,
obviously
trees,
effect
evapotranspiration
they
affect
Albedo.
So
there's
there's
implications
there
for
surface
energy
balance.
F
Trees
emit
bvocs
such
as
isoprene
monoterpine
sesquiterpenes,
which
have
a
variety
of
complex
interactions
with
atmospheric
chemistry,
potentially
enhancing
the
lifetime
of
methane,
Downstream
bvocs
produce
aerosols,
which
then
interact
with
clouds
and
interact
with
radiation.
We
also
see,
via
nitrogen
cycle
impact
on
ozone,
which
obviously
high
up
in
the
atmosphere,
is
a
good
thing,
but
in
the
troposphere
is
a
bad
thing
and
from
a
land
surface
perspective,
a
shift
from
grassland
or
crop
land
to
trees
could
be
expected
to
increase
water
stress,
decrease
soil
moisture.
F
So
there's
questions
there
for
the
plausibility
of
of
tree
planting
as
a
strategy
as
well
I'm
going
to
talk
a
bit
about
land
stuff.
Today,
I
will
add
the
caveat
that
I'm
not
by
by
training
a
land
surface
expert.
So
one
of
the
reasons
that
I'm
talking
to
the
this
distinguished
group
here
is
to
kind
of
ask
for
feedback
and
and
point
to
see
if
we're
on
the
right
track.
F
So
how
are
we
investigating
Global
tree
planting
commitments
so
we're
using
a
CSM
on
We've,
successfully
ported
the
model
to
Archer
2,
which
is
the
new
British
Flagship
HBC
system?
It's
very
shiny,
it's
very
fast.
It
also
doesn't
work
a
lot
of
the
time,
but
we're
working
on
that
for
this
particular
set
of
experiments.
We're
running
the
model
fully
coupled
so
sort
of
everything
on
we
have
the
active
ocean.
F
We
have
three
experiments,
all
of
which
follow
ssp-12.6
in
terms
of
greenhouse
gas
concentrations
and
sort
of
anthropogenic
emissions.
The
experiments
differ
in
the
land
use.
So
we
have
a
base
case
which
follows
the
ssp-1
2.6,
a
Max
Forest
scenario,
which
was
came
out
of
a
piece
of
work
that
Peter
Lawrence
did,
which
I'll
talk
about
in
a
minute
and
a
control
where
we
hold
lime,
juice
and
land
cover
constant
at
2015
levels.
F
We
run
the
model
over
the
period
of
2015
to
2100,
using
obviously
spun
up
initial
conditions,
and
we
will
be
running
three
Ensemble
members.
So
that's
that's
three
times
three
runs
eventually,
once
they've
found
some
more
storage
for
us
to
stable
that
up,
but
the
Microsoft
scenario
is
developed
by
Peter
is
we
think,
a
quite
significant
advance
on
a
lot
of
the
the
the
studies
that
you
see
looking
at
Global
tree
planting,
which
are
often
it's
a
little
more
straightforward?
F
F
This
is
what
it
looks
like
on
the
globe,
so
that
kind
of
cycle
for
a
few
moments-
and
the
color
bar
here
is
the
percentage
of
natural
vegetation
made
up
of
tree
pfts.
Those
of
you
obviously
know
how
CLM
fiber
portions
the
land,
and
we
can
see
that
essentially
in
this
scenario,
existing
tree
cover
expands
and
primarily
we
see
that
expansion
in
tropical
rainforest,
but
we
also
see
significant
expansion
in
North,
America,
Europe,
South
and
parts
of
east
and
south
Asia
you'll.
F
If
you
look
at
this
scenario
in
context
so
on
this
plot
here,
the
max
Forest
scenario
is
the
dotted
line,
the
solid
lines,
the
various
SSP
scenarios,
and
we
compare
the
max
Forest
scenario.
We
see
it
fits
in
favorably
with
policy
commitments
that
have
been
announced
so
that
it
fits
in
nicely
with
the
bomb
challenge
GRC.
There
is
a
database
of
global
reforestation
and
restoration
commitments
up
to
2050,
and
we
see
that
the
max
Forest
scenario
on
a
global
scale
is
slap
bang.
F
In
the
middle
of
that
we
expand.
Forest
expansion
is
constrained
by
climatic
suitability,
so
you
don't
plant
trees
and
deserts
where
they
won't
grow.
We
take
into
account
the
demand
for
agricultural
land
by
holding
that
constant
at
2015
levels,
with
a
few
minor
adjustments
and
tree
planting
peaks
in
mid-century,
which
is
aligned
with
the
commitments
and
sort
of
asymptotes
towards
2100..
F
If
we
look
at
the
scenario
regionally,
we
can
see
that
the
majority
of
the
tree
planting
the
scenario
takes
place
in
the
tropical
rainforests
in
or
temperate
environments
such
as
Europe,
it's
comparable
to
ssp1,
and
this
is
quite
a
good
thing.
We
think,
because
you
know
the
tropical
rainforests-
you
don't
have
so
much
of
a
water
limitation
problem.
The
Albedo
problem
essentially
is
negated
by
latent
heat
release,
and
you
know
biodiversity
benefits
are
substantial,
so
this
is
a
great
great
bit
of
work
by
by
Peter.
F
That
he's
kindly
allowed
us
to
to
use
and
kind
of
run
with
a
little.
So
what
happens
when
we
run
this
scenario
in
the
model?
For
a
start,
we
see
across
the
globe,
increases
in
in
net
primary
production
or
net
primary
productivity,
sorry
associated
with
a
shift
to
more
tree
cover.
This
is
associated
with
carbon
dioxide
removal,
so
concentrations
of
CO2
are
lower
in
the
max
Forest
case
than
the
base
and
the
no
land
use
cases
as
with
most
of
the
variables.
F
F
F
F
F
F
So
yeah,
that's
where
we
are
at
the
moment.
Cdr
potential
and
cooling
do
certainly
exist.
If
you
have
a
global
scale,
afforestation,
reforestation
and
Forest
restoration
pledge.
That's
in
line
with
policies,
but
there
is
also
increased
water
stress,
which
has
kind
of
implications
for
for
the
viability
of
those
plants.
If
it
wasn't
in
a
scenario
where
we
were
sort
of
imposing
them,
other
implications
are
under
investigation.
Mark.
Any
any
comments
on
on
what
we're
doing
so
far.
Thanks
a
lot.
C
B
I
actually
have
a
question
you
you
said
it
was
fully
coupled,
but
I
wasn't
actually
sure
if
that
meant
that
you
had
prognostic
atmospheric
CO2
I'm
guessing,
probably
not,
but
maybe
you
did
and
either
way
could
you
comment
on
what
the
carbon
feedbacks
might
have
been
because
you're
changing
photosynthesis.
At
the
same
time,.
F
So
yeah
the
the
carbon
in
the
atmosphere
is,
is
prescribed
following
ssps,
so
it's
it's
it's
it's
not
an
emissions
driven
so
effectively
the
the
the
runs
that
we
do
have
no
impact
on
the
CO2
in
the
atmosphere.
You
know
the
land
draws
down
carbon,
but
the
the
carbon
in
the
atmosphere
is
prescribed,
which
obviously
is
limitation
of
this
kind
of
study
and
we're
doing
some
work
with
a
simplified
UK
model
called
fair
to
kind
of
try
and
constrain
that
more
effectively
we
haven't
as
yet
partially
for
this
reason
are
set
up.
F
It
makes
it
quite
tricky
to
work
out
what
would
be
the
cooling
effect
from
changes
in
the
carbon
cycle
from
enhanced
biosphere
carbon
directly.
We
can
make
indirect
calculations
based
on
the
size
of
that
carbon
sink
or
the
size
of
that
the
increase
in
that
carbon
sink
and
look
at
what
the
reductions
in
atmospheric
CO2
would
be
if
we
did
have
prognostic
CO2,
which
is
something
we're
working
on
at
the
moment.
It's
small
but
not
insignificant.
F
F
F
It's
It's
tricky
to
isolate
Cloud
signal
amongst
Cloud
noise,
as
is
often
the
case
with
a
single
set
of
ensembles,
but
the
the
next
set
of
ensembles
are
in
the
works
and
and
starting
to
tee
up
and
run,
and
hopefully
we'll
we'll
have
a
clearer
picture
of
the
cloud
response.
Once
we
have
those
ensembles,
you
know.
N
U
F
We
are
currently
running
because
it's
quite
tricky
to
extract
that
direct
impact
of
water
vapor
amongst
a
a
scenario
which
we're
allowing
the
ocean
to
evolve.
So
that
tends
to
be
what
dry
is
driving.
The
atmospheric
moisture
is
an
SST
effect
rather
than
a
land
surface
effect,
so
we'll
have
clearer
answers
on
that.
I
think.
Once
we've
done
the
ensembles,
the
the
picture,
that's
emerging
from
some
work,
we're
doing
with
a
sort
of
offline
cloud
and
water-based
radiative
forcing
calculations.
F
B
Great
well
thanks
very
much
and
we
are
up
for
our
break
next
for
folks
online.
There
is
an
online
breakout
group
that
Elizabeth
sent
out
you're
welcome
to
go
there
and
chat
for
folks
here
we
will
have
Refreshments
outside
the
room
and
we're
reconvening
10
20.
10
20,
27
minutes.
25
minutes
we'll
see
you
back
in
here.
Okay,.
V
Thanks
Gretchen
yeah,
so
I'm
gonna.
Can
we
talk
to
you.
Tell
him
much
of
the
zero
missions.
Commitment
occurs
before
reaching
net
zero
emissions,
just
maybe
a
confusing
title,
but
hopefully
I'll
explain
it
to
you
and
yeah,
so
I'm
going
to
start
off
the
the
motivation
for
this
comes
from
this
figure,
which
you
know
I,
think
it's
sort
of,
arguably
like
the
most
important
figure
in
our
field
of
sort
of
coupled
carbon
climate,
which
shows
the
linearity
and
the
path,
dependence
or
path.
V
So
what's
going
on
there,
why
is
there
only
limited
evidence
and
and
why
they
sort
of
put
out
there?
Is
this
caveat?
So
I
was
curious
about
that
and
into
the
Led
Led
to
this
stuff?
Okay,
so,
what's
going
on
with
with
why
why
this
tcre
relationship
might
not
hold
under
that
negative
emissions?
This
idea
comes
from
this
paper
led
by
Christian
ziegfelder
on
2016.,
where
they
did
a
one
percent
per
year.
V
Concentration
reversal,
experiment
and
what
they
found
is
under
that
that
there's
this
systematic
asymmetry
between
this
relationship
between
the
the
temperature
anomaly
and
the
cumulative
CO2
emissions
under
a
positive
under
under
negative
emissions
right.
So
after
you
get
to
Net
Zero
and
go
go
back,
there's
this!
This
is
consistent,
positive
bias
and
you
can
see
that
in
in
the
CDC
map,
6
cdrmip
experiments
as
well
that
this
move
six
models.
V
Also,
this
consistent
positive
asymmetry
so
that,
after
going
to
negative
emissions,
there's
still
you
know
the
temperature
is
higher
for
a
given
cumulative
emissions
than
than
in
the
upswing.
So
this
is,
there
might
be
like
that.
You
know
this.
This
consistent
bias
after
cumulative
of
after
Net
Zero,
and
so
that's
one
of
these
idealized
experiments,
whereas,
interestingly,
under
a
non-idealized
experiment.
If
you
look
at
the
SSP
experiments
that
go
past
Net
Zero
to
net
negative
emissions,
sp1-2.6
and
SPS
3.4
overshoot,
those
actually
don't
show
this
consistent,
positive
asymmetry.
V
In
fact,
they
show
something
much
much
more,
that
the
proportionality
actually
does
still
hold,
and
so
there's
kind
of
this
weird
scenario
where
the
the
non-idealized
experiments
are
are
behaving
more.
Ideally
than
the
idealized
experiments,
I
should
say
in
these
non-idealized
experiments.
If
you
look
at
the
cm6
models
that
there
there
is,
any
individual
model
will
have
some
proportionality
from,
but
of
the
the
sort
of
sorry
the
the
proportionality
is
conserved.
V
You
know,
with
some
asymmetry
but
and
that
asymmetry
is,
is
it
for
each
model
is
well
explained
by
this
quantity
called
the
zero
emission
script
or
the
Zac
for
each
model,
so
the
so
the
thing
that
that
after
you
get
to
Net
Zero,
the
relationship
between
global
warming
and
cumulative
is
is
to
find
the
proportionality
within
asymmetry.
That's
this
other
metric,
called
Zach
and
and
Zach
is
this
other
measure
of
how
much
warming
would
occur
if
CO2
emissions
were
to
abruptly
stop.
V
So
it's
measured
by
going
from
a
one
percent
concentration
experiment
and
at
roughly
two
two
times
CO2,
that
you
take
the
model
and
allows
CO2
to
become
transient
or
prognostic
with
no
zero
further
missions.
What
happens
is
CO2
concentrations
decrease,
but
the
temperature
remains
constant
and
so
those
two
and
the
reason.
Why
is
because
you
know
sinks
continue,
but
the
the
climate
sensitivity
tends
to
increase
because
ocean
heat
uptake
decreases
and
the
the
physical
climate
sensitivity
tends
to
increase
in
time.
V
So
so
what
what's
going
on
here?
Our
hypothesis
is
that
that
the
asymmetry
and
the
one
percent
concentration
is
is
really
kind
of
an
artifact
that
it
that
it's
that
that,
if
you
look
at
the
the
accumulative
or
the
emissions
rates
that
you
are
required
to
give
you
that
one
percent
per
year
concentration
reversal,
what
you
see
on
the
on
the
right
hand
side
is
that
it
implies
this
50
pentagram
per
year
jump
in
emissions
from
about
plus
30
to
about
minus
20
pentagrams
of
carbon
per
year
instantaneously.
V
Right,
and
that's
that's
a
lot
to
ask
of
path,
Independence
that
that
there'll
be
path
invariant.
Even
if
you
jump
instantaneously,
50
pentagrams
of
carbon
per
year
right
rough,
like
currently
we're
only
emitting
10
pedograms
of
carbon
per
year.
So
it's
an
enormous
jump
in
in
in
in
in
in
carbon
fluxes
that
we're
asking
the
you
know
this
path
Independence
to
hold
over.
V
You
can't
actually
allow
you
to
to
estimate
the
tcre,
because
the
CCR
is
a
carbon
climate,
only
quantity
without
other
greenhouse
gases.
Tsm2
is
really
ideal
for
this,
because
it
has
a
very
negative
Zach
so
because
it
has
this
very
negative.
Technology
actually
allows
you
to
separate
this
tcre
proportionality
from
the
tcrd
plus
the
exactly
and
the
negative
emissions
phase.
V
Okay.
So
so
what
do
we?
So?
So?
What
are
the
results?
So
if
we
do
this,
this
is
what
the
CO2
flux
is
in.
This
experiment
look
like
so
again,
you
know
the
the
for
a
first
150
years.
We
do
positive
emissions,
the
second
150
years,
you
do
negative
information,
they've
been
rented
out
for
another
50
years
with
zero
emissions.
V
After
that,
the
this,
both
the
land
of
ocean
tanks,
follow
the
emissions
to,
and
they
flip
sign
in
response
to
the
emissions
to
become
they
go,
the
sinks
become
sources
after
emissions
become
negative
emissions,
and
so
you
can
start
to
see
these
these
kind
of
roughly.
You
know
these
Decatur
lags
start
to
to
to
happen.
V
The
the
client
climate
response
follows
kind
of
the
the
expected
gaussian
relationship,
proportionality
to
our
missions,
but
because
our
missions,
our
cumulative
missions,
are
following
this
gaussian
function.
We
expect
our
temperature
to
also
follow
this
sort
of
gaussian
function
and
it
roughly
does.
But,
interestingly,
it
also
leads
the
cumulative
emissions
for
most
of
this
scenario.
V
So
temperatures
the
the
the
red
line-
and
you
can
see
it-
you
know-
sort
of
drops
down
faster
than
the
the
negative
emission
or
the
cumulative
emissions
drop
down
under
negative
emissions
for
most
of
this
scenario,
and
then
it
warms
a
bit
at
the
end.
So
there's
some
interesting
stuff
going
on
there
and
that's
that's
kind
of
on
the
that's
kind
of
unexpected
right.
V
Why
would
temperature
actually
lead
the
cumulative
missions
and,
and
our
again
our
hypothesis
was
that
you
know
the
warming
would
follow
this
TCR
even
partiality
on
the
up
slope
and
then
tcre,
plus
the
Zach,
which
for
csm2
case
is
negative
on
the
downtell
I
mean
it
mostly
holds
right
that
you
can
see
the
lines
kind
of
you
know
these
two
diagonal
lines
here.
These
are
the
ttre
and
the
T3
plus
Zac,
and
so
it's
doing
this.
V
The
interesting
thing
is
that
it
actually
switches
from
this
tcre
line
to
the
tcre
exact
line
before
getting
to
Net
Zero
right.
So
this
is.
This
is
kind
of
the
the
surprise
we
expected
to
follow
the
blacklight
on
the
up
slope
and
the
blue
line
and
the
down
slope,
and
it
does
that,
but
it
trigly,
but
it
switches
actually
earlier
than
we
were
expecting
it
to
and
so
yeah.
So
you
know
what
what's
going
on
there?
V
V
So
it's
three
different
things
as
a
function
of
either
the
tcru
or
the
Zach.
What
we
see
is
that
you
know
kind
of
as
we
expect.
The
peak
warming
is
mainly
governed
by
the
tcre,
although,
as
you
know,
the
teacher,
a
plus
half
of
the
Zach
actually
correlates
with
an
even
higher
r
squared,
then
this
this
T3
only
the
timing
of
the
peak
warming
I.
V
Current
climate
policy
says
that
you
know
once
we
get
to
Net
Zero,
we
expect
temperatures
to
stabilize,
and
this
says
well,
if
our,
if
we
live
in
a
planet
with
a
negative
Zach,
temperatures
can
actually
stabilize
slightly
before
we
get
to
Net
Zero,
which
is
actually
good
news.
The
co
the
covers
for
that
is
also
true.
V
If
you
live
in
a
planet
with
a
large
positive
Zach
warming
could
could,
you
know,
could
might
may
not
happen
too
well
after
we
get
to
Net
Zero,
but
so
this
this
says
that
this
deck
and
this
relative
timing
of
peak
warming
and
pql
Emissions
are
actually
you
know
very
highly
correlated
to
each
other.
This
is
an
important
you
know:
policy
relevant
Point,
interestingly,
in
this
experiment,
where
you
then
pull
out
as
much
CO2
from
the
atmosphere
as
we
put
into
it.
V
This
end,
warming
that
you
know
above
the
pre-industrial
amount,
is
also
largely
governed
by
Zac
as
well,
so
just
to
conclude,
CO2
syncs
fall
emissions.
Under
this
kind
of
scenario,
reverse
side
becomes
sources
after
emissions
reversed
with
a
decade
time
scale
lag
the
lag
between
CO2
flux
and
Emissions
causes
atmospheric
CO2
concentrations
to
actually
lead
emissions.
V
The
tcred
proportionately
holds
that
are
negatives
CO2
emissions
subject
to
an
asymmetry
that
is
well
Quantified
by
attack,
so
I
think
what
this
says
is
that
the
the
the
result
from
the
one
percent
per
year
concentration
reversal.
Experiment
is
largely.
You
know
that
there's
this
positive
asymmetry
of
the
you
know
relative
to
the
tcre
after
negative
emissions
is
largely
an
artifact
of
that
experimental
design
and
that
that
that
huge
insaneous
jump
in
emissions,
and
if
we
avoid
that,
then
then
actually
the
TCR
proportionality
does
hold
quite
well.
V
Lastly-
and
this
is
kind
of
a
key
Point,
much
of
the
Zach
appears
before
reaching
the
net
zero
and
thus
Jack
also
governs
the
timing
of
peak
forming
relative
to
Net
Zero,
and
so
that's
an
important
point.
For
you
know
policy,
like
you,
know,
relevant
things
of
saying
you
know,
because
if
we
you
know
get
to
Net,
Zero
and
climate
doesn't
stabilize
that's,
that's
a
really
important
thing
that
we're
going
to
have
to
be
able
to
explain-
and
you
know
likewise
with
climate
stabilities
Society
before
then.
C
V
A
bunch
of
interesting
questions
too
right,
one
of
which
I
don't
think
we
have
a
good
answer
for
is:
why
does
csm2
have
such
a
Negative
Zach?
You
know
I
think
we
know
why
it
has
a
such
a
large
tcre.
You
know
because
it
has
a
high
climativity
I.
Don't
think
it's
as
obvious,
why
it
has
such
a
Negative
Zach.
Another
point
is
if,
if
the
csm2
zekrons
are
extended
longer,
do
we
start
to
see
these
kind
of
weird
restraints
like
we
do
in
this
experiment
and
in
the
long
term
over
should
experiment?
V
How
does
how
does
land
use
change
into
some
of
this
stuff?
You
know
how
how
what
aspects
of
these
Dynamics
are
sensitive
to
Perimeter
and
structural
things?
You
know
differences
and
you
know,
would
this
be
a
useful
and
practical
cement?
Seven
experiment,
you
know
I
think
I
argue.
We
argue
in
this
paper
that
that
you
know
it's
it's.
G
V
Lastly,
I
just
want
to
end
this
on
how
else
can
we
use
some
emissions
driven
cesm
experiments
to
explore
some
of
these
Cloud
mitigation,
cereals
and
then
actually
extra?
Lastly,
I
also
just
want
to
plug
that.
We've
got
a
postdoc
opening
that
with
a
link
here
for
anybody
interested
in
postdocs
on
vegetation,
Dynamics
and
so
I
will
end
there
thanks.
W
A
F
C
X
X
So
a
quick
overview.
Biological
nitrogen
fixation
is
the
dominant
natural
input
of
new
nitrogen
into
the
terrestrial
biosphere
here
I'm
showing
the
tree
Robina
pseudo
Acacia
black
Locus,
it's
the
dominant
nitrogen
fixing
tree
in
the
U.S
and
nitrogen
fixing
plants.
They
have
root
nodules,
where
atmospheric
nitrogen
is
converted
to
ammonium
in
exchange
for
carbon
and
there's
been
multiple
approaches
to
modeling.
Nitrogen
fixation
in
land
models,
for
example,
CLM
4.5
used
to
do
this
as
a
function
of
NPP.
X
Now
CLM
5.1,
it's
more
of
a
process
based
approach
using
the
fun
model
where
n
fixation
is
a
function
of
endemand
plant
available
carbon
and
temperature
and
temperature
makes
sense.
For
example,
a
global
analysis
of
nitrogen
fixing
tree
abundance
showed
that
maximum
temperature
is
the
best
predictor
of
where
nitrogen
fixing
trees
are
followed
by
temperature
in
the
warmest
and
wettest
quarter
as
the
third
and
fourth
best
predictors
and
notice
that
where
temperatures
are
warmer
or
you
have
more
nitrogen
fixing
trees.
X
So
traditionally,
the
temperature
function
that
has
been
in
use
in
models
comes
from
Holton
at
all
2008.,
and
this
has
been
the
best
guess
up
till
some
of
the
data
that
I'll
show
you.
This
has
been
based
on
Six
studies,
five
of
which
were
actually
from
Free
Living
bacteria
not
symbiotic
nitrogen
fixers,
but
it's
been
extrapolated
to
the
global
scale
to
model
symbiotic
nitrogen
fixation,
so
we
really
wanted
to
get
at
the
temperature
response
of
symbiotic
nitrogen
fixation
and
we
focused
on
tree
fixtures
for
this.
X
X
We
did
it
across
four
species,
so
two
of
which
are
tropical,
two
of
which
are
temperate
and
then
the
two
types
of
symbioses,
so
either
actinarizal
or
rhizobial
symbiosis,
and
we
did
this
across
a
range
of
growing
temperatures
spanning
10
degrees,
Celsius
and
here
I'm.
Showing
each
panel
has
three
different
individuals
and
then
we
fit
function
to
each
column
so
for
each
species
across
growing
temperature,
where
parameters
of
the
function
can
acclimate
to
Growing
temperature.
X
That
range
is
driven
by
acclimation
to
Growing
temperature,
particularly
in
the
tropical
species,
and
we
think
that
our
curves
are
shifted
to
the
right
relative
to
the
Holton
function,
possibly
for
a
couple
of
reasons.
One
is
that
was
based
on
Free,
Living
bacteria
and
not
plant
bacterial
symbioses,
and
also
it
had
a
latitudinal
bias
to
high
latitudes.
X
Additionally,
I'd
like
to
point
out
that
even
temperate
fixtures
grown
at
cold
temperatures
had
higher
Optima
than
the
function
in
use
and
then
way
higher
Optima
occurred
in
the
tropical
fixers
grown
at
high
temperatures.
Those
are
the
two
top
red
curves
so
now
for
the
modeling
part
I'm
leading
a
model
in
our
comparison
project
across
seven
Lan
models.
X
So
this
kind
of
relates
to
this
one
of
the
implications
may
be
testing.
This
projected
decline
in
tropical
nitrogen
fixation
with
climate
warming,
which
has
been
predicted
using
the
Holton
function
here
on
the
right
I'm,
showing
this
from
the
cable
model
for
three
different
levels
of
CO2,
and
you
know
the
question
that
kind
of
I'm
asking
myself
now
is
how
sensitive
is
this
result
to
the
temperature
response
function.
Event,
fixation
that
is
being
used,
so
we're
doing
a
few
different
approaches
with
the
functions
in
use.
X
First,
we
have
the
original
Holton
function,
then,
from
our
nature
plants
paper,
we
have
a
separate
temperate
and
tropical
function,
plants
kind
of
been
together
well
by
temperate
versus
tropical
origin,
not
So
Much
by
symbiotic
type,
and
this
one
does
not
include
acclimation,
and
then
we
have
a
third
scenario
where
we
do
include
acclimation,
but
it
only
occurs
between
the
range
of
temperatures
we
have
data
for
so
in
the
in
CLM.
The
way
this
is
implemented
is
via
the
fun
model
and
there's
a
temperature
dependent
cost
carbon
cost
of
nitrogen
fixation.
X
So
here
I'm
plotting
up
what
that
cost
looks
like
for
the
Holton
function
and
our
temperate
and
tropical
function
that
does
not
include
acclimation.
Essentially
it
flips
the
function
and
sets
the
temperature
response.
Function
is
essentially
flipped
and
the
minimum
is
set
to
six
grams
of
carbon
per
gram
of
nitrogen.
We've
also
started
empirically
measuring
this
cost
and
are
shapes
for
flipping.
It
actually
work
pretty
well.
X
So
just
to
sum
that
up,
we
have
three
approaches
to
representing
nitrogen
fixation
and
some
models.
We
include
a
fourth
one,
for
example,
Orca
day
before
this
project,
I
used
a
function
of
NPP
scaled
by
plant
and
the
P
ratios,
so
they
have
a
fourth
simulation
and
right
now
we're
doing
site
level
simulations
for
a
tropical,
a
temperate
and
a
boreal
site,
and
the
next
step
will
be
Global
level
simulations.
X
We
have
a
historical
time
period
for
the
simulations
from
1850
to
2014
and
then
two
future
scenarios.
One
is
future
warming
with
with
fixed
CO2
and
undeposition,
and
one
is
future
warming
with
transient,
CO2
and
undeposition,
and
we're
using
easy
myth
gfdl,
forcing
the
future
scenarios
using
SSP
585.
X
So
just
to
give
you
a
flavor
of
some
of
the
results
that
we're
starting
to
see.
It's
all
super
preliminary
I'll
start
by
just
showing
what
soil
temperature
in
the
layer
that
and
fixation
responds
to
looks
like
these
are
monthly
means
for
the
temperate
and
the
tropical
site,
and
then
so,
if
we
go
now
to
the
temperate
site
here,
I'll
show
how
n
fixation,
Through
Time
looks
like,
and
this
is
for
the
transient,
CO2
and
end
up
position
scenarios
and
we're
summing
right
now.
Symbiotic
and
Free.
Living
fixation
in
these
plots.
X
X
We
have
the
temperate
function
in
general,
with
climate
change,
fixation
rates,
increase
at
the
temperate
site,
but
there's
variation
in
magnitude
depending
on
what
model
is
in
use
and
then,
if
we
look
at
the
difference
in
fixation
between
our
temperate
function
and
the
Holton
function,
all
this
is
no
acclimation.
By
the
way,
we
see
a
consistently
higher
or
lower
fixation
rate
using
the
temperate
function
and
that's
consistent
through
time,
so
that
doesn't
get
affected
by
climate
change.
X
X
X
X
X
Yeah,
so
we're
gonna
after
these
site
level
simulations
a
subset
of
the
models.
Not
all
of
them
run
globally
will
be
used
for
Global
simulations
to
to
test
those
kind
of
questions,
but
yeah
I'm
curious
right,
like
will
nitrogen
Supply
from
fixation,
increase
or
decrease
with
climate
change
in
the
tropics,
it
may
be
a
different
direction
than
previously
thought.
A
X
Yeah
I'm
not
sure
how
the
NPP
factors
into
that
right
now.
Daniel
Galt
just
recently
rewrote
the
formulation
for
fixation
there
before
it
did
not
have
a
temperature
function.
So
I'll
have
to
check
that
but
yeah.
Thank
you
for
bringing
that
up,
but
yeah.
That
response
was
much
stronger,
particularly
first
classic
from
Canadian
esm.
T
Is
the
temperature
control
on
nitrogen
fixation
related
to
plant
side
or
bacteria
side,
because
you
know
bacteria
can
adapt
incredibly
fast,
so
over
long-term
climate
change
they
might
change
their
ability.
You
know
their
optimal
temperature
trees,
not
so
much.
They
don't
have
as
much
time
to
adapt.
X
Of
course
that's
not
adaptation
within
within
a
species
or
within
bacterial
lineage,
but
then
acclimation
was
very
different,
but
there
has
been
some
work
on
Free
Living
fixation,
temperature
response,
curves
that
came
out
a
few
years
ago
and
they
were
seeing
quite
a
bit
of
adaptation
within
a
certain
lineage
of
bacteria
that
associated
with
lichens,
so
I
think
there's
quite
a
possibility
for
that.
We
just
haven't
really
observed
it
yet.
A
Y
Z
Days,
it's
been
a
while,
since
I've
been
to
a
land
model
working
group
meeting,
so
it's
been
super
fun.
So
I'm
gonna
switch
gears
a
little
bit
and
spend
some
time
talking
about
how
we've
used
Isotopes
to
study,
plant
drought,
stress
responses
in
North,
American
ecosystems,
I
just
want
to
shout
out
to
there's
a
whole
bunch
of
collaborators
on
this
from
the
modeling
side,
as
well
as
for
from
the
data
collection,
side
and
gml.
So
just
a
big
shout
out
to
everybody
and
also
funding
from
the
NOAA
map
program.
Z
Z
This
figure
shows
changes
in
the
model
conductance
as
a
function
of
vapor
pressure
deficit,
where
VPD
is
essentially
a
measure
of
stress
and
as
you're
increasing
vapor
pressure
deficit
or
increasing
stress.
That's
the
model
conductance
drops
off,
but
depending
on
whether
you're
a
forest
or
a
grassland
or
something
between
the
rate
of
drop-off
is
different.
Z
But
at
the
same
time,
whether
you're,
a
forest
or
a
grassland
or
something
in
between
you
can
still
maintain
a
certain
level
of
gross
ecosystem
productivity,
and
this
is
all
down
to
this
plants
to
model
response
which
you've
heard
mentioned
several
times
this
week,
and
that
is
the
focus
of
our
study.
Is
this
the
model,
interaction
between
plants
and
the
environment
and
the
simulations
I'm
going
to
show
you
with
the
model
that
we've
been
running?
Z
Z
The
theoretical
basis
for
this
is
that
plants
prefer
carbon-12
or
the
lighter
carbon
for
photosynthesis.
But
when
they're
stressed
out,
they
don't
have
a
choice,
so
they
take
whatever
they
can,
and
so
that
leads
to
less
discrimination
which
leads
to
relationships
like
this,
where
you
see
a
different
isotopic
discrimination
signature
as
a
function
of
net
carbon
uptake.
Z
And
then
you
can
construct
an
equation
like
this
and
then
calculate
as
a
result
of
all
of
these
process
is
put
together
what
your
final
photosynthetic
fractionation
signal
is
going
to
be,
and
in
sib4,
which
is
the
model
that
I'll
get
into
in
the
next
slide
and
tell
you
a
little
bit
more
about.
We
modeled
this
the
model
conductance
as
a
function
of
ballberry,
and
so
that's
primarily
how
the
the
stress
response
in
the
plants
is
connected
to
this.
The
model
conductance
Factor,
so
this
was
put
into
sib
2.5
we've
developed.
Z
Z
It's
simplified,
as
the
name
suggests,
compared
to
CLM,
and
my
focus
has
been
on
developing
the
land
surface
part
of
this
model
and
particularly
the
carbon
cycle,
part
of
it
so
I
mentioned,
we've
got
fractionation
during
photosynthesis.
We
also
put
in
a
parallel
pool
structure.
So
now
the
CO2
molecule
as
it
passes
through
the
different
carbon
pools.
Z
We
also
have
a
C13
CO2
molecule
that
passes
through
all
the
different
C13
pools
and
then
just
some
technical
details
about
the
model
we
spin
it
up
in
pre-industrial
and
then
run
it
forward
for
172
years
and
we
do
vary
the
boundary
conditions,
as
I
mentioned
in
the
beginning,
both
the
atmospheric
Delta
C13,
based
on
the
a
compiled
non-in
star
record,
as
well
as
the
CO2
mixed
layer
concentration
over
time.
Okay.
So
what
does
this
look
like
so
I'm?
Z
Showing
you
results
here
for
an
evergreen
needle
Leaf
site
in
Wyoming
and
I've
just
pulled
the
summer
months
for
the
years
between
2000
and
2020..
So
each
of
these
points
is
an
average
over
those
that
that
summer
growing
season
period
and
I've
colored
the
Dots
here
as
a
function
of
Di
SCI,
which
is
the
drought
severity
coverage
index.
It's
a
composite
drought,
monitor
that
comes
from
the
UNL
drop
monitor.
So
it's
a
composite
index
that
takes
into
account
the
severity
of
all
the
different
categories
and
bends
it
as
one
number.
Z
But
the
main
take
home
here
is
that
the
darker
colors,
so
dsci
greater
than
300
to
400
are
are
more
severe
drought
conditions
and
then
the
the
yellows
are
are
less
severe
drought
conditions
and
you
can
see
that
separated
as
a
function
of
vapor
pressure
deficit,
and
the
nice
thing
here
is
that
we're
seeing
civ4
kind
of
out
of
the
box
is
doing
what
we
expect
in
terms
of
the
isotopic
fractionation
signature
being
different.
Depending
on
how
severe
the
drought
is.
Z
Z
So
we
can
also
look
at
it
in
anomaly
space.
So
each
of
these
plots
I'm
plotting
the
anomaly
in
the
fractionation
against
the
anomaly
and
Nee,
which
is
the
net
carbon
uptake,
and
then
the
colors
are
all
representing
drought
categories,
where
the
Bluer
colors
are
less
severe.
Red
is
more
severe
drought.
So
we
do
see,
as
we
expect,
that
under
more
severe
drought
conditions,
we're
seeing
both
positive
any
e
anomalies,
which
is
carbon
being
released
as
opposed
to
being
taken
up
and
then
also
positive,
Delta
C13
anomalies.
Z
So
again,
the
model
is
behaving
as
we
expect,
which
is
good,
we're
also
not
seeing
changes
for
C4
plants
also
as
expected,
because
they're
they're
more
hot
adapted,
but
also
they
don't
have
as
many
of
this
the
model
steps
involved
in
C4
versus
the
C3
pathway,
which
I
won't
get
into,
and
then
forests.
This
is
kind
of
interesting
for
for
shrub
lens.
You
kind
of
see
this
like
gradual
departure.
Z
If
we
look
at
the
spatial
distribution
of
these
across
the
U.S,
most
of
the
reason
for
the
color
scale,
difference
is
because
of
the
proportion
of
C3
versus
C4
plants,
so
C3
and
C4
plants
have
have
different
isotopic
signatures.
C4S
are
more
in
the
-4
range
and
then
C3
is
more
minus
18
and
then
the
subtle
coloring
of
the
oranges
and
Breads
and
yellows
are
all
climate
stress
essentially,
and
then
the
higher
granule
variability
in
kind
of
the
central
U.S
is
also
driven
by
this
mixed
grid.
Pixels,
as
well
as
different
climate
impacts.
Z
So
I
showed
you
that
graph
in
the
beginning
of
the
isotopic
Discrimination
as
a
function
of
any
e
and
how
the
Peters
at
all
paper
found
this
nice
relationship
for
net
carbon
uptake,
and
so
we
did
the
same
kind
of
thing.
But
we
tried
to
break
it
down
a
little
bit
more
into
plant
functional
types.
And
so
what
I'm
showing
here
is
that
if
we
do
it
for
Evergreen
needle
a
force
in
the
U.S,
we
are
seeing
this
really
nice
anti-correlative
relationship
that
we
were
hoping
to
see,
but
not
so
much
for
this
E3
grasslands.
Z
And
so
this
speaks
to
obviously
the
plant
physiological
differences
between
these
functional
types.
But
it's
also
a
cautionary
note
that
when
we're
doing
this
kind
of
regional
modeling
across
different
regions
that
include
multiple
plant
functional
types,
we
need
to
be
careful
as
to
how
we
interpret
them.
Okay,
so
I'm
gonna
spend
the
last
few
minutes
talking
about
how
we're
using
our
atmospheric
data
at
gml.
So
we
have
a
really
rich
record
of
flask
data
across
the
US
for
CO2,
Delta,
13c
and
CO2
and
a
whole
bunch
of
other
Trace
gases.
Z
But
what
we
did
was
we're
trying
to
use
that
atmospheric
Delta
C13
signal
from
the
observation
side
and
then
modeling
it
as
well
as
a
function
of
these
tracers,
which
I'll
get
into
a
minute
and
then
just
trying
to
see.
If
that
tells
us
something
about.
You
know
what
that
model
data
mismatch,
tells
us
about
what's
wrong
in
our
models
and
what
we
could
improve.
So
I'm
not
going
to
get
in
the
details
of
all
of
these
different
terms,
but
this
is
essentially
how
you
would
construct
a
C13
budget.
Z
Some
of
these
terms,
like
the
fossil
Delta
C13
signature,
the
net
ocean
to
some
extent
the
ocean
to
sequilibrium
as
well,
they're,
reasonably
well
known.
Quite
a
lot
of
the
variability
comes
from
the
land
terms,
though,
and
so
if
we
look
at
this
kind
of
forward
modeling
scenario
or
you
take
each
of
those
tracers
and
then
you
transport
them
for
it
through
an
atmospheric
transport
model
and
in
our
case
we're
using
tm5.
Z
And
then
you
can
compare
the
time
series
that
you
get
from
the
model
which
is
in
red
to
the
observations
which
are
in
Black.
This
is
the
SGP
site,
which
is
a
site
for
which
we
have
a
a
really
nice
long
record,
and
you
can
see
that
there
is
two
things.
One
is
that
we're
offset
from
the
one
to
one
line,
but
we're
there's
also
a
trend
in
the
model
which
is
not
there
in
the
observations,
and
we
think
that
this
is
due
to
this
land.
Z
Disequilibrium
term
and
I
won't
dwell
on
it
too
much
in
this
talk.
But
we
are
for
a
separate
project,
developing
a
different
modeling
scenario
for
the
Landis
equilibrium,
where
we've
adjusted
that
based
on
observed
carbon
resonance
time.
So,
if
anyone's
interested
in
this
topic,
we
can
chat
about
it
offline
or
afterwards,
but
adjusting
that
carbon
residence
time
and
civ4
because
we
don't
step
four
carbon
pools
are
too
young.
Essentially.
Z
So
if
we
adjust
that
based
on
observations
and
scale
them,
then
we
see
a
slightly
better
match
with
the
trend,
although
we're
still
offset
from
the
one-to-one
line.
So,
okay,
coming
back
to
the
fact
that
we're
okay,
so
we're
gonna,
ignore
the
fact
that
we
don't
get
the
trend.
But
then
we
can
de-trend
it
and
look
at
the
results
in
anomaly
space
and
what
we're
seeing
is
that
we
actually
get
the
phasing
and
timing
of
the
seasonal
Cycles
relatively
well.
Z
But
we're
missing
the
magnitude
so
we're
not
seeing
the
dynamic
range
and
we're
not
capturing
the
highs
and
the
lows,
which
is
why
we're
much
more
angle
out
the
101
line.
We
can
also
look
at
atmospheric
CO2,
and
so
if
we
do
the
same
type
of
analysis,
where
here
I'm
showing
you
the
CO2
absolute
values,
we
do
a
much
better,
better
job
matching
the
absolute
CO2
values
which
is
hopefully
to
be
expected
and
a
good
thing,
and
it's
actually
even
better.
Z
If
we
look
at
some
of
our
background
sites,
this
is
a
continental
site
and
so
there's
some
more
noise
around
the
one-to-one
line.
But
if
you
look
at
a
place
like
monologue
or
South
Pole,
this
is
a
really
tight
relationship
both
for
the
absolute
values
as
well
as
for
the
anomaly
values,
but
as
at
this
SGP
site.
What
we're
seeing
is
that,
even
in
CO2
space,
there
are
some
really
high
CO2
values
that
we're
still
missing
in
our
Ford
modeling
scenario.
Z
So
that
hints
that
there's
something
in
our
model
that
we're
missing
in
terms
of
mortality
or
stress
or
something
that
we
should
be
causing
the
model
values
to
be
higher
than
they
actually
are.
So.
Finally,
we
can
bring
it
back
to
looking
at
a
drought
index,
which
is
what
I
had
mentioned:
I'm,
trying
to
kind
of
not
Benchmark
exactly
but
use
that
as
a
way
to
parse
our
atmospheric
data.
Z
So
this
is
again
the
dsci
drought,
severity
coverage
index,
but
for
the
SGP
site
and
then
I
went
ahead
and
bend
it
in
categories
of
of
a
hundred.
And
what
we're
seeing
here
again
is
that
the
dynamic
range
which
is
so
the
the
size
of
the
red
bars,
is
much
less
than
the
observation
so
kind
of
backing
up.
Z
What
I
showed
you
in
the
previous
slide,
but
that
our
model
is
basically
missing
the
dynamic
range
for
the
anomalies
and
then,
except
for
the
most
severe
drought
category,
we're
kind
of
about
neutral,
both
in
observation
space
as
well
as
model
space,
and
so
the
the
model
does
kind
of
track.
The
trend
and
the
observations
fairly
well,
we're
just
not
capturing
the
dynamic
range
and
I'm
done
so
just
wanted
to
throw
out
there
that
we're
using
atmospheric
CO2
as
a
tracer
and
I'd
love
for
this
to
become
more
of
a
CLM
collaboration.
Z
At
some
point.
In
the
future,
I
think
that'd
be
really
cool
and
that
we
have
a
ways
to
go
in
terms
of
figuring
out.
You
know
we're
experimenting
with
different
stress,
setups
and
I
didn't
get
into
how
we
even
model
stress,
but
we
can
talk
about
that
offline
again,
so
I'm
happy
to
take
questions.
Thank
you.
U
W
Z
I'll
I'll
make
that
plot
and
get
back
to
you
I,
don't
know
we
were
interested
in
actually
seeing
if
we
could
capture
the
subseasonal
variability.
So
that's
why
we
did
it
in
this
space
but
on
an
annual,
mean
basis,
I
think
it
would
look
something
like
that
plot
up
there,
because
those
are
annual
means,
but
this
site.
So
the
slight
complication
with
you
using
NOAA
data
is
that
the
site
is
not
one
PFT
right.
U
AB
Z
Z
Z
Oh
there's
definitely
there's
definitely
a
fractionation
associated
with
primary
production.
I
was
going
to
say
photosynthesis
in
the
ocean,
but
I
don't
think
that's
the
right
term,
but
I
haven't
looked
at
that
I
know
so
part
of
our
budget
is
the
ocean
exchange
and
there's
other
people
working
on
that.
So
I'll
ask
them.
H
You
see
me
okay,
great,
thank
you
for
having
me
here
myself,
Manoj
and
I'm
from
India
I'm,
a
graduate
student
over
there
and
currently
I'm
a
visitor
here
at
ncar,
thanks
to
Danica
and
Professor
diagi,
for
that.
Well,
today,
I'll
be
discussing
about
the
effect
of
aerosols
over
net
primary
productivity
across
different
agroclimatic
regions
over
India.
H
So
for
this,
as
you
all
know,
that
like
carbon,
has
both
positive
and
negative
effects
on
the
ecosystem
as
well
as
in
the
climate.
But
then
there
are
other
factors
that
determine
the
intensity
of
these
effects
and
one
such
effect
that
we'll
be
discussing
about
us
like
the
radiative
transfer
effect.
Well,
they
have
a
possibility
of
transferring
like
the
because
of
the
diffuse
radiation
they
either
promote
the
carbon
Flex
or
like
the
demote
based
upon
the
carbon.
H
Sorry
seawater
Fertilization
in
the
ecosystem,
but
then
they
leave
us
with
a
one
particular
question
that,
like
at
what
degree
these
effects
of
aerosols
on
the
primary
productivity
is
inverse
like
whether
they
tip
this
effect
in
a
positive
manner
or
in
a
negative
manner.
So,
in
order
to
understand
this
question,
so
we
frame
this
study
with
a
simple
remote
sensing
process
based
model
called
Casa,
which
is
Carnegie,
aims,
Stanford
or
press
model,
and
we
used
aerosol
modus
aerosol
products.
H
In
order
to
compare
this
relation
so
to
find
the
radiative
effect,
we
used
SB
Dart
model
and
we
coupled
it
with
in
order
to
find
the
spatial
sensitivity
across
different
regions
of
India.
So
to
understand
this,
we
distinguishly
classified
India
into
15,
equal
climatic
divisions
based
upon
the
climatic
conditions
and
the
vegetational
background.
Well,
the
plot
here
you
are
seeing
is
like
the
aerosol
trend
for
the
period
of
20
years.
H
Where
you
can
see
there
is
a
graph
I
mean,
whereas
a
very
distinct
trend
of
positive
aerosol
accumulation
over
the
Eastern
Front,
which
is
mostly
dominated
by
the
Agro
regions.
That's
usually
like
this
intensity
of
aerosols
is
usually
maintained
by
the
seasonal
cropping
pattern
and
the
biomass
burning
that
has
been
taking
place
over
the
region.
On
the
other
hand,
if
you
took
a
if
you
look
at
the
primary
productivity,
you
can
see
a
strong
positive
trend
over
the
period,
especially
on
the
Central
Central
Zone,
on
the
North
and
the
northern
Zone.
H
But
then,
if
you
just
take
like
the
Western
Front
and
the
Northeast
and
front,
you
have
a
gradual
decrease
in
the
net
primary
productivity,
and
that's
where
you
have
most
of
the
forest
based
ecosystems
present.
So
you
have
Evergreen
broadly
forest
and
Evergreen
deciduous
forest
over
these
two
regions
and
they
account
a
significant
loss
of
about
five
percentage
of
total
loss
of
net
primary
productivity
for
the
whole
period.
H
If
you
look
at
a
temporal
extent
where
you
can
see,
we
compared
the
casa
model
with
the
20
Ensemble
and
with
the
modest
product
where
we
we
could
able
to
see.
There
is
a
Harmony
in
between
these
two
products,
where
you
have
a
gradual
increase
in
the
trend.
But
then,
if
you
look
at
the
decadal
shift,
I
mean
at
the
decaded
level.
You
have
a
steady
phase
at
the
second
decade,
where
the
trend
seems
to
be
little
poised
on
because
of
the
loss
of
productivity.
H
In
the
in
the
forest-based
ecosystem,
but
here
the
gray
shaded
regions
represents
the
aod
period
when
the
Avery
is
like
lesser
than
the
actual
threshold
for
the
over
overall
average
skill.
But
then
the
red
phase
indicates
the
average
aerosols
are
like
higher
than
the
average
threshold,
and
that's
usually
I
mean
it
it.
It's
synchronized
with
the
the
trended
anomalies,
where
the
maximum
productivity
has
been
observed
into
in
the
I
mean
the
previous
decade,
but
then
later
on.
H
But
then,
if
you
look
CLM,
particularly
it
almost
like
underrepresented,
a
CLM
compared
to
the
other
models,
particularly
if
you
look
at
the
monsoon,
where
the
shift
has
been
transferred
towards
Monsoon
rather
than
at
the
post
Monsoon.
If
we
are
comparing
the
other
model
on
a
spatial
extent.
If
we
look
at
the
address
or
variability
across
different
regions,
where
you
can
see,
you
have
higher
seasonal
variability
over
the
regions
that
is
being
plotted
in
Orange,
they
are,
they
usually
represents
the
region
that
is
have
aggro
based
ecosystems
on
the
region.
H
But
then,
if
you
look
at
the
green
phase
over
here,
so
this
is
mostly
the
region
where
you
have
like
higher
higher
Forest
based
ecosystem,
where
you
have
the
poised
aerosol
trend.
On
the
other
hand,
if
you
look
at
the
NPP
Trend,
this
represents
the
whole
like
seasonal
variability
of
NPP.
Again,
you
can
see
a
higher
variability
of
NPP
over
there
in
the
Agro
ecosystems
rather
than
at
the
forest
based
ecosystems.
H
But
then
what
we
did
is
like
we
segregated
years
or
like
the
period
when
you
have
lower
aod,
where
you
can
see
the
green
lines
represents.
That's
the
productivity
when
the
arrows
over
the
particular
region
has
been
lesser.
I
mean
lesser
than
the
actual
thousand,
and
that's
when
you
have
aod
has
been
like
I
mean
the
NPP
is
like
increased
over
these
regions,
especially
on
the
forest-based
ecosystem,
rather
than
in
the
Agro
ecosystems.
H
But
then,
when
you,
when
the
aod
was
like
higher
than
the
actual
threshold
over
the
region,
you
could
see
the
NPP
has
been
decreased
at
irrespective
of
the
ecosystems.
If
you
compare
the
Agro
ecosystems
as
well
as
in
the
forest-based
ecosystem,
NPP
seems
to
be
drastically
reduced
at
different
level.
H
So
that's
when
we
used
like
the
radiative
transfer
model
in
order
to
understand
how
this
shift
has
been.
The
sensitivity
has
been
playing
between
aod
and
NPP,
so
we
figured
it
out
that,
like
NPP
has
been
having
like
inversionally
proportional
to
aod,
and
it
is
like
having
higher
sensitivity
at
different
zones
at
like
different
pins.
H
On
the
other
hand,
if
you
look
at
the
Agro
ecosystems,
where
the
growth
of
NPP
was
restricted,
when
it
increased,
six
I
mean
0.6
and
above,
but
the
maximum
growth
of
NPP
was
absorbed
when
the
aerosol
hit
like
0.5.
So,
on
the
overall
scale
we
understood
like
when
the
aerosol
when
aerosol
of
the
particular
region
it
it
increases
to
42
percentage
of
the
overall
region.
H
It
stops
the
NPP
productivity
at
a
significant
level,
whereas
if
the
aod
is
like
lesser
than
the
12
lesser
than
14
percentage
of
the
oral
region
threshold,
it
promotes
NPP
at
a
significant
rate.
So
this
sensitivity
helps
us
to
understand
like
to
calculate
the
sensitivity
between
different
regions
based
upon
this
one,
the
maximum
negative
effects
has
been
found
on
the
western
plateau
and
the
Eastern
Plateau
front.
H
So
to
understand
this
analysis,
we
specially
found
the
regions
that
has
been
sensitive
to
aod
and
that
has
been
marked
in
a
bread
box
with
a
stipulator
point
when
we
plot
on
a
normal
scale,
where
you
have
a
poised
level
of
NPP
over
the
region.
But
then,
if
you
just
look
at
those
particular
points
where
the
red
red
line
indicates,
that's
when
you
have
higher
NPP
productivity
over
the
post,
monsoon
season
been
able
is
like
lesser.
H
But
then,
if
you
consider,
when
there
is
like
actually
higher
aod
in
the
monsoon
season,
the
NPP
Trend
seems
to
be
reduced
in
this
region.
So
this
region
is
like
the
Eastern
coastal
plain
where
you
have
where
you
have
the
mixed
canopy
layers
it.
It
includes
that,
like
both
Evergreen,
broadly
forests
and
deciduous
forest,
so
we
extrapolated
this
analysis
to
few
other
heterogenetic
background
of
regions.
H
But
this
region
is
like
represents
the
region
that
has
active
agricultural
background,
and
this
region
represents
both
have
multiple
layer,
canopy
system,
and
you
can
see
when
the
average
is
like
having
higher
it.
Almost
like
induces
the
NPP,
rather
than
at
the
when
A4,
is
like
lesser
than
the
optimal
thousand.
But
then,
if
you
look
at
the
other
with
a
multiple
layered
canopy
region,
it
doesn't
show
much
variability
over
this
over
the
period
So.
H
So
to
summarize,
all
the
study
we
found
that
like
India,
is
having
a
gradual
increase
in
the
productivity,
but
then
when,
if
we
note
a
dedicated
scale,
it
seems
to
be
like
the
later
decades
are
not
dip.
But
if
we
look
at
a
sensitive
level
where,
based
upon
the
canopy
architecture,
the
forest-based
ecosystem
has
placed
a
major
role,
then
in
the
Agro
ecosystems
and
in
case
of
some
drawbacks
we
had
as
well.
So
this
is
like
it's
a.
This
is
like
a
mono
scenario
study.
H
But
then,
if
you,
if
you
compare
it
with
like
multiple
scenario
study,
we
could
able
to
identify
more
sensitive
regions
across
India
and
the
other
thing
is
like
we,
since
it's
like,
we
spatially
average
all
the
pixels
so
that
we
might
be
we
might.
We
fail
to
capture
a
phenological
difference
between
like
each
regions
and
with
that
concluding.
Thank
you.
A
So
I
have
a
question
I
feel
like
in
some
of
the
observational
based
studies.
They
actually
show
that
there's
a
bigger
effect
of
diffuse
light
induced
from
aerosol
and
crops
as
opposed
to
to
Forest.
Like
do
you
have
any
sort
of
observational
targets
that
you
can
compare
here
to
see?
If
you
know
the
system
behaves
differently
in
the
subtropics
or
given
the
the
types
of
crops
that
are
being
used.
H
So
that's
one
of
the
major
drawbacks
of
the
study
because,
like
we
don't
have
like
India's,
not
well
sophisticated
with
the
flux
networks
or
any
other
network,
so
we
couldn't
able
to
like
validate
this
study
based
upon
at
the
ground
line,
but
otherwise
that's
why
we
use
mostly.
We
use
the
remote
scenario.
C
A
A
G
So
my
name
is
Keith
Lindsey
and
I'm
going
to
be
talking.
This
is
very
much
a
I'm
going
to
even
update
on
some
fast
spin-up
techniques
for
different
components
of
an
earth
System
model.
This
is
very
much
a
work
in
progress,
so
this
is
not
a
final
product.
That's
ready
to
to
be
used
by
others.
Yet
that's
the
goal
to
be
there.
This
is
work
that
I'm
doing
jointly
with
Sam
Levis
and
will
leader
leader,
Matt,
broadly
speaking.
What
this
this.
G
The
statement
of
the
spin-up
problem
is
to
generate
tracers
or
pool
distributions
that
are,
in
balance
with
respect
to
some
time
varying
forcing
there
are
many
different
applications
for
this.
One
is
to
initialize
transient
experiments,
analyze,
Dynamics
or
properties
of
Spun,
up
tracers
being
spun
up.
If
you
have
non-linear
Dynamics
analyzing
the
dynamic
of
those
non-linear
Dynamics
when
you're,
not
spun
up
by
you,
know
different
analysis
results
than
when
they
are
spun
up.
G
You
want
to
compare
your
tracers
or
pools
to
observations
you
might
want
to
do
optimization
of
parameters
to
reduce
model
bias,
and
that
has
shown
up
in
a
variety
of
talks
that
have
happened
in
the
lmwg
earlier
this
week.
This
requires
the
ability
to
spin
up
repeatedly,
so
you
want
something
that's
fast
and
efficient,
and
that
has
shown
up
in
a
number
of
talks.
G
Doing
a
brute
force
is
prohibitively
expensive,
particularly
for
Ocean
Models,
where
running
for
2000
years
might
take
you
40
wall
clock
days
and
that's
not
something
you
can
do
repeatedly
so
I'm
a
mathematician
by
training,
so
I'm
translating
this
into
a
mathematical
framework
and
then
use
tools
of
mathematics
to
attack
this.
So
the
notation
that
I'm
introducing
is
that
C
of
T
is
the
Tracer
or
pool
state
of
the
model
in
the
ocean.
That's
Tracer
concentrations
where
this
might
be
pools
in
the
land
model.
G
So
that's
the
result
of
integrating
all
tendencies
in
the
model
forward
in
time
from
an
initial
state
to
time
t
the
spin-up
problem
is
to
find
in
a
special
initial
condition:
c-star
such
that
when
you
initialize
the
model
with
c-star
and
run
it
with
period
capital
T
you
get
back
to
where
you
are
so.
This
is
the
end
state
is
the
beginning
state
in
this
big
high
dimensional
space
and
T.
Here
is
the
period
of
forcing
of
Interest.
G
So
this
might
be
10
or
20
years
that
captures
inter-annual
variability
or,
if
you've
got
something
you
might
do
a
shorter
if
you're
just
looking
at
some
sort
of
synthetic
annual,
forcing
but
typically
with
climate
applications,
spinning
up
for
initializing
coupled
models.
This
might
be
10
to
20
years
now
to
apply
the
tools
of
mathematics,
I'm,
going
to
take
this
equation
of.
G
G
We're
going
to
use
Newton's
method
to
find,
because,
with
the
zero
of
that
function,
Newton's
method
is
an
iterative
method
that
you
might
have
learned
about
in
Calculus
class
for
one-dimensional
variables.
This
is
just
like
that,
except
that
it's
in
a
million
dimensional
space
instead
of
one
dimension.
G
What
Newton's
method
does
is
it
generates
a
sequence
of
iterates
that
converge
to
the
solution
for
a
system
of
equations
in
the
way
that
Newton's
method
is
derived?
Is
you
take
your
function
that
you're
trying
to
find
the
zero
of
G
you're
evaluating
it?
You've
got
a
current
iterate
C
sub
K,
and
you
write
the
value
of
the
next.
G
The
function
of
your
next
iterate
G
of
K
plus
1
do
a
Taylor
series
expansion
about
your
previous
existing
iterate,
so
G
of
C,
K,
plus
1,
equals
g
of
C
K,
plus
a
partial
derivative
times
the
difference
here.
This
is
a
Taylor
series
expansion
and
then
we
drop
all
the
non-linear
terms.
So
we're
doing
the
linearization
of
this
big
SP
of
this
big
function.
With
respect
to
your
difference
between
your
iterates,
we
set,
we
drop
all
the
higher
order
terms
and
solve
for
CK,
plus
one.
G
What
what
CK,
plus
one
with
that
linearization
gives
you
a
zero,
and
if
you
rearrange
terms,
then
you
end
up
getting
C.
You
get
this
equation
right
here,
so
the
equation
for
the
this
Newton
increment.
Is
this
what
you're,
adding
to
your
current
iterate
to
get
to
the
next
iterate
and
that
solves
a
linear
system
of
equations?
So
there's
a
derivative
times,
an
increment
is
equal
to
our
right
hand,
side.
G
G
However,
that's
easier
said
than
done.
Computing.
That
increment
is
solving
the
system
of
equations,
but
I
mentioned
this
is
a.
We
are
in
a
million
dimensional
space
or
10
million
dimensional
space
in
this
linear
system
of
equations
tends
to
be
dense.
So
it's
not
feasible
to
actually
compute
that
Matrix
of
partial
derivatives
much
less
store
it.
If
you've
got
a
10
million
by
10
million
Matrix.
That's
you!
We
don't
have
enough
disk
space
even
on
our
supercomputers,
to
store
that.
So
that's
not
a
you!
Don't
solve
this
system
of
linear
equations
directly.
G
You
need
to
find
a
way
to
approximate
the
linear
system
of
equations,
and
so
what
I'm
using
in
this
technique
is
another
iterative
method
to
solve
for
the
Newton
increment.
It's
an
iterative
method
inside
an
iterative
method
which
is
a
using
a
crylov
method
for
which
is
a
particular
class
of
iterative
methods
for
systems
of
linear
equations.
G
That's
well
suited
for
this
scenario:
I'm
not
going
to
go
into
the
details
of
krylov
methods,
but
the
key
feature
is
that
you
end
up
Computing
a
matrix
Vector
multiply
in
each
iteration
of
the
cryolov
method
and
in
order
to
compute
this.
In
this
scenario,
we're
we
don't
actually
have
this
Matrix,
so
we
can't
really
compute
a
matrix
Vector
multiply.
G
So
if
you
start
thinking
globally,
what's
going
on,
we've
got
this
outer
iterative
method.
Newton's
method
inside
that
is
another
iterative
method.
The
cryolav
method.
In
each
step
of
that
inner
loop
is
doing
a
forward
model
run
of
time
t
in
order
for
the
krylov
method
to
actually
converge
efficiently.
G
We
need
to
apply
a
technique
called
preconditioning
to
that
system
of
linear
equations,
where
you
are
approximating
the
inverse
of
that
Matrix,
and
that
is
where
a
lot
of
the
The
Art
of
Doing,
these
types
of
techniques
comes
into
play
is
constructing
a
preconditioner
that
is
a
good
approximation,
but
is
also
feasible
to
compute
and
I'm
not
going
to
go
into
the
details
of
of
that.
But
suffice
it
to
say
that's
where
a
lot
of
the
work
comes
into
play.
G
G
Some
issues
that
end
up
coming
into
the
challenges
and
issues
that
end
up
coming
arising
are
that
that
Newton
increment
can
lead
to
non-physical
values.
So
you
can,
you
might
get
a
negative
Tracer
concentration
or
a
negative
carbon
carbon
pool
in
the
land
model,
and
there
are
some
ways
that
you
can
scale
down
the
increment
to
ensure
that
you
get
physical
values
that
you
don't
shoot
off
your
increments
into
a
non-physical
regime.
G
Another
challenge
is
that
Newton's
method
is
doing
this
linearization
of
your
function
over
the
time
scale
of
big
capital
T.
But
if
your
model
has
strong
non-linearities
on
shorter
time
scales,
the
linearization
they
may
not
play
friendly
with
the
linearization
of
Newton's
method
on
the
time
scales
that
you're
applying
it
to,
and
so
one
of
the
techniques
that
is
that
we're
applying
to
these
is
apply.
Newton's
method
to
I
mean
what
I'm
calling
Shadow
tracers
so
basically
cut
off.
G
Some
of
the
feedbacks
that
lead
to
between
some
pools
in
your
system
cut
off
the
feedbacks
that
lead
to
those
fast
non-linearities
so,
for
example,
in
the
soil,
pools
cut
off
the
feedback
between
the
soil,
decomposition
and
gpp,
and
then
just
speed
up
spin
up
your
soil
pools
with
fixed
gpp,
with
fixed
inputs
into
your
soil
pool
and
then
apply
the
Newton's
method.
To
that,
then
copy
them
back
to
the
real
pool.
Let
your
fast
pools
adjust
and
then
iterate
again.
This
is
another
iteration
on
top
of
all
these
tested
iterations.
G
There's
been
some
talk
about
applying
this
to
CLM
for
many
years.
Actually,
when
I've
talked
about
this
for
in
Ocean
model
working
group,
some
folks
in
the
Land
Group
have
asked
me:
I've
got
some
emails
back
from
2015
where
people
have
asked
me.
Oh,
is
this
possible
to
apply
this
to
CLM
in
being
young
and
naive,
I
thought
well
sure
this
all
the
math
would
work
fine
well.
G
At
least
one
of
those
statements
is
true
about
being
young
and
naive,
and
so
in
more
recent
I
there's
been
talk
about
doing
this
for
many
years,
but
actually
the
work
to
do
it
is
actually
more
transpired
in
the
last
year
or
so
collaborating
with
Will
in
with
Sam
and
some
of
the
fundamental
questions.
So
the
background
is
that
there
is
a
current
approach
for
doing
spin
up
in
the
land
model
of
accelerated
decomposition.
G
That
many
of
you
in
this
room
are
probably
familiar
with
to
spin
up
the
below
ground
pools,
and
one
open
question
is:
if
we
get
Newton
cryolov
working
for
the
land
model,
is
it
faster
than
accelerated
decomposition
and
another?
This
would
be
relevant
for
doing
high-res
simulations
where
ad
becomes
expensive
or
for
things
like
the
PPE
that
we
heard
about
yesterday
or
other
experiments
where
people
are
trying
to
do
multiple
spin-ups.
G
We
ended
up
with
a
very
simple
plan,
very
naive,
planned,
get
NK
working
with
a
single
column
without
mimics
get
it
working
in
the
global
model
then
extend
to
mimics,
as
I
said,
this
is
naive,
easier
said
than
done.
I'm
going
to
skip
some
of
this
description
about
the
iterative
nature
of
how
this
project
has
proceeded,
and
sometimes
the
iterations
and
solving
problems
has
been
easy
to
do
like
just
adding
more
history
variables
to
the
land
model.
G
We
recently
got
this
working
with
a
single
Boreal
column.
Forest
I
mean
this
is
now
the
horizontal
axis
is
Newton
iterations,
and
this
is
a
penalty
function
written
in
log
space
where
we're
getting
exponential
convergence.
So
this
is
spinning
up
that
boreal
forest
column
in
300
model
years,
instead
of
what
would
take
thousands
of
years
to
do.
Brute
Force,
so
we've
tried
to
extend
this
to
more
complicated
situations.
G
Is
we're
now
applying
this?
This
is
an
example
of
applying
it
to
a
global
course
simulation,
and
this
is
my
favorite
column,
column
108
in
a
in
a
course
resolution
model
where
we
get
exponential
convergence,
even
converges,
even
faster
than
that
boreal
forest
over
10,
iterations
converging
very,
very
quickly.
G
So
this
is
a
work
in
progress
of,
and
this
is
now
a
map
of
NE
after
zero
iterations
in
one
10
Newton
iterations
on
a
sort
of
a
logarithmic
scale.
So
you
can
see
in
many
places
the
solver
is
really
getting
towards
convergence,
we're
going
from
the
yellows
into
the
deeper
colors.
But
in
some
points
it's
getting
worse.
So
there's
a
lot
going
on
and
we
need
to
figure
out
what's
going
on
with
the
preconditioner
of
what
how
to
improve
things.
So,
as
I
said,
this
is
a
work
in
progress.
G
What's
going
on
with
the
problematic
runs,
why
does
it
require
so
many
newton
iterations
when
I
showed
that
single
column,
it
was
taking
50
Newton
Newton
iterations
I'm
used
to
seeing
about
three
Newton
iterations
for
convergence
in
the
ocean
model?
So
it's
actually
going
a
lot
slower
than
I
would
hope
it
would
go
and
there's
some
other
outstanding
questions
that
we're
working
on
that
outstanding
questions
like
can.
We
even
extend
this
to
something
like
the
fates
model.
It's
all
close
there.
G
A
B
A
C
C
A
R
R
R
R
The
the
composition
model
of
CLM
is
represented
here
in
this
picture,
that
I
took
from
the
documentation
where
you
can
see
different
carbon
pools
first,
starting
with
the
course
wood
and
then
the
litter
pools,
and
then
the
soil
pools.
And
you
see
that
there
is
a
vertical
yeah
vertical
tools
as
well.
So
each
sole
layer
has
one
pool,
and
then
you
can
have
carbon
transferred
from
the
the
less
decomposed
carbon
pools,
which
are
on
top
to
the
more
decomposed
pulls
on
the
bottom
and
so
in
each
pool.
R
Carbon
has
some
turnover
time,
which
is
in
years
here,
and
so
when
carbon
is
transferred
from
one
pool
to
the
other.
There
is
a
small
flux
which
is
the
respiration,
the
fraction
of
respiration,
and
so
our
first
idea
was
to
add
a
small
term
which
will
be
the
doc
production.
So
every
time
carbon
is
transferred
from
one
pool
to
the
other.
R
R
So
we
have
everything
that's
in
this
produced
at
the
surface.
We
put
this
in
a
surface,
geosc
pool
and
everything
that's
produced
below
the
surface
we
put
in
the
subsurface
doc
pool,
and
then
here
are
some
maps
of
preliminary
results
showing
how
much
carbon
is
produced
in
the
subsurface
and
in
the
at
the
surface.
We
see
that
there
is
much
more
produced
in
all
the
soil
layers
than
just
the
surface
layer
and
it's
also
a
different
distribution
and
yeah
it's
much
smaller
than
the
respiration.
R
That's
because
of
the
parameters
that
are
much
far
as
well.
Here
we
can
see
different
sole
layers
from
so
so
layer,
0
to
soil,
layer,
14
and
obviously
there's
more
DLC
produced
in
the
top
players,
because
it's
proportional
to
the
to
the
decomposition
of
of
carbon,
and
so
it
depends
on
temperature,
water
and
so
on.
R
R
R
So
there
is
very
little
surface
in
DLC
accumulating
in
ctsm.
Almost
everything
is
sent
to
Mozart.
So
Mozart
is
one
of
the
river
models
in
ctsm.
There
has
been
another
presentation
with
Missy
roots,
which
is
more
like
water,
like
basins
with
different
shapes,
polygonals
shapes
I.
Think,
and
here
it's
it's
more
like
gritzel
based
I,
think
so
it's
the
the
Mozart
model
is
divided
in
three
areas,
and
so
the
surface
water
will
go
to
the
hill
slope
part
and
the
hill
slope.
R
And
so
what
I
do?
I
put
the
surface
you
see
with
the
surface
water
into
the
hill
slope
and
the
subsurface
DLC
straight
into
the
subnetwork
rooting
part
with
the
subsurface
water,
and
so
here
I
showed
the
example
of
how
I
will
make
the
DLC
move,
along
with
the
water
I
showed
the
example
for
the
main
Channel,
but
it's
more
or
less
it's
the
same
concept
for
the
hill
slope
and
the
subnetwork
part.
R
So
what
we
do
is
that
we
have
the
concentration
in
each
of
the
parts
of
the
model
and
so
While
We're
looping
over
each
of
those
parts
we
will
have.
The
concentration
is
updated
at
every
time
step
and
so
I
will
compute.
How
much
do
you
see
the
mass
of
DLC
already
in
each
of
those
parts
and
I
will
add
the
mask
coming
in,
remove
the
mask
that's
going
out
and
divide
by
the
total
amount
of
water
in
this
part
of
the
model
and
so
yeah.
These
are
the
inputs.
R
So
it's
the
surface,
DLC
subsurface
DLC
is
the
same
as
I
showed
in
Mozart.
It's
just
different
units
and
yeah
liquid
water
below
it's
just
different
units
as
well,
but
it's
the
same
that
what's
going
out
of
of
ctsm,
and
so
here
we
see
the
concentration
I
change
the
scale
a
little
bit
so
that
we
can
see
something
but
I
check.
All
the
values
are
between
zero
and
0.3
grams
per
liter,
so
3
0.3
kilograms
per
cubic
meter.
R
And
then
here
I
showed
storage
of
the
sea
in
kilograms
compared
to
the
storage
of
liquid
water
in
the
model
and
the
river
discharge
over
land
of
the
Sea
and
the
river
discharged
to
the
ocean
of
DLC,
and
so
obviously,
I've
been
working
only
a
couple
of
months
on
this,
so
I'm
not
an
expert
at
all,
and
this
is
just
Premier.
Premiere
Premier
really
work.
R
But
there
are
many
things
to
do,
and
right
now
I'm
trying
to
verify
that
the
DLC
mass
balance
is
around
zero
when
I
integrate
a
simulation
over
time,
because
I
think
there
is
more
DLC
coming
out
than
that's
coming
in
into
the
model
weirdly
and
that
may
be,
because
of
because
of
the
watering
motor
that
sometimes
is
like
the
water
storage
is
below
zero.
And
so
there
are
some
problems
when
the
water
goes
from
positive
to
negative
values
in
the
different
pools
of
the
model
and
then
when
it
goes
back
to
positive
values.
R
So
I
have
to
double
check
this.
But
then
there
are
many
things
to
do
ahead.
Like
compare
these
two
observations
that
we
have
around
and
in
Norway,
and
then
there
are
many
nice
things
that
we
can
do
in
addition
to
this:
firstly
probably
improve
the
DLC
production
parameters
that
I
showed,
which
are
just
broad
estimations
and
maybe
even
change
the
equation
so
have
an
independent
equation
for
Doc
production.
R
That
is
not
part
of
this
decomposition
equation
and
then
yeah
have
some
differentiation
for
records,
reconcitrant
and
labile
doc
pools,
add
Don
as
a
tracer,
because
right
now,
I
added
this
whole.
The
whole
car
is
an
array
that
is
transported
to
OC,
but
it
will
be
very
easy
to
just
add
the
orn,
because
I
do
it
the
same
way.
Liquid
and
Ice
are
transported
into
the
model.
R
It's
like
just
a
tracer,
so
it
will
be
really
easy
to
add
Don
once
we
have
it
as
an
output
of
ctsm
and
then
adding
Dom
going
back
to
land,
but
then
yeah,
I
I
right
now,
I,
don't
do
anything
with
the
doc.
That's
going
into
the
ocean
and
I
don't
do
anything
with
the
DLC.
That's
accumulating
in
the
the
pools
over
land.
So
that's
some
stuff
I
had
to
do
as
well.
A
C
R
P
Do
you
have
any
idea
if
I
I
really
like
your
simple
approach
like
we
don't
know
what
these
numbers
should
be?
We
just
want
to
generate
flexes,
but
do
you
have
any
idea
if
the
like
the
concentrations
you're
getting
of
Doc
and
solution,
either
on
land
or
in
export
or
at
all
reasonable,
I,
don't
even
know
what
reason
would
be.
R
Yeah,
so
I
I've
done
a
quick
check
for
the
0.3
grams
per
liter,
which
should
be
fine
as
like
a
maximum
value,
and
so
this
was
actually
first
I.
I
did
not
even
have
this
accumulated
Pavilion
ctsm,
but
then
I
would
have
like
a
lot
of
DLC
in
desert
regions
where
there
was
no
water,
and
this
DLC
will
just
accumulate
in
Mozart
and
result
in
very,
very
high
concentrations.
R
So
zero
three
is
theoretically
possible,
at
least,
and
then
those
tiny
parameters
that
I
have
in
the
the
composition
Cascade.
Those
are
so
Dev,
which
was
working
before
me
on
this
project,
had
found
this
tiny
parameters
which
should
be
theoretically
possible
as
well.
It's
just
not
very
Dynamic
when
we
compare
to
jewels
and
arcade,
which
probably
have
like
equation
for
DFC
production,
which
is
independent
from
the
composition
and
where
it
depends
on
temperature
and
and
water
react
independently.
AA
It
was
I,
guess
it's
a
commenter,
we
probably
should
talk
with
you
know.
Developments
of
ctsm
for
csm3
were
kind
of
avoiding
those
negative
water,
States
and
Mozart,
because
the
ocean
model
doesn't
doesn't
like
them
right.
So
we
understandably
so
we
should.
We
should
just
talk
and
there
might
be.
You
may
not
have
that
problem
anymore.
If
you
use
our
most
updated,
parameter,
settings
and
nameless
settings
and
things
like
that
right
so.
E
A
AB
I
was
really
excited
about
Charlie's
simulation
of
native
admissions.
That
has
a
somewhat
more
realistic,
although
still
somewhat
unrealistic
time.
Evolution
of
the
emissions
and
I
love
the
idea
of
having
more
CSM
simulations
that
are
emissions
based
rather
than
prescribed
at
mrex
YouTube
concentrations
and
I
can
see
how
at
least
the
ocean
Community
could
begin
to
use
those
simulations
and
so
I'd
love
to
have
a
conversation
with
this
working
group
about
whether
that's
something
we
should
be
doing,
whether
we
should
try
to
get
resources
to
do
that.
AB
AA
AA
To
do
a
whole
new
set
of
emissions
driven
runs
going
out
in
that
case,
in
addition
to
doing
new
ones,
doing
it
with
low
CO2
emissions
levels,
which
we
haven't
really
done,
which
I
think
were
you
know
the
mitigation
efforts
are
more
impactful,
probably
and
also
going
out
to
2300,
which
we
also
haven't
done.
So
there
is
some
sort
of
funding
associated
with
the
rubisco
project.
To
try
to
you
know,
I
I,
don't
know,
get
get
more
of
the
knowledge
base
out
of
Keith's
brain
and
into
into
a
wider
range
of
of
capabilities.
AA
AB
AA
These
would
be
like
SSP
scenarios,
so
they'd
be
realistic,
yeah
yeah,
okay,
we
haven't
decided
exactly
which
scenario
we're
going
to
run,
but
just
you
know
the
only
I
think
the
only
missions
runs.
We've
ever
done
with
with
CSM
really
has
been
the
ssp-5
like
the
really
high
Keith.
You
can
correct
me
if
I'm
wrong,
but
I
think
we've
only
gone
for
the
really
high
one
I.
G
B
G
B
B
Yeah
I
want
to
say
even
there's
like
running
them
and
then
there's
just
more
people
needing
to
look
at
what's
happening
in
them,
because
I
there's
all
kinds
of
emergent
time
scales
of.
What's
going
on
when
you
look
at
Charlie's
plots
of
like
Global
fluxes.
But
why
are
those
the
time
scales
and
what
are
they
dependent
on
I
feel
like
nobody's
even
scratched
the
surface
of
looking
at
what
that
depends
on
yeah.
AA
Absolutely
whatever
we
do
would
be
a
community
run
to
you
know
for
everybody
to
look
at
so
I.
Think
yeah,
I,
just
I
feel
like
it's
important
that
we
take
the
take
the
effort
to
try
to
be
able
to
do
these
right
because
now
I
think
we're
all
a
little
bit
fearful
of
them
because
they're
hard
to
set
up,
and
so
we
just
need
to
kind
of
get
over.
That.
T
This
might
be
a
really
stupid
point
or
question
I,
just
sort
of
summing
up
what's
been
happening
over
the
last
couple
days,
and
and
just
my
experience
over
the
last
couple
years,
still
learning
a
lot
about
this
model
and
feeling
like
I
finally
can
do
things
with
it.
You
know
that
work
and
but
there's
also
at
the
same
time
we
have
these.
You
know,
there's
lots
of
things
coming
at
the
model,
most
of
which
are
like
great.
It's
like
okay.
T
We
are
able
to
we're
adding
processes
that
are
important
or
we're
finding
processes,
making
it
more
realistic,
but
the
model's
still
getting
more
complicated
and
simultaneously
there's
all
these
other
approaches
that
are
requiring
people
to
run
the
model,
hundreds
or
thousands
of
times
we
have
to
spin
the
model
up
more.
We
have
to
do
it
just
seems
like,
and
people
are
always
you
know
when
you're
talking
about
designing
an
experiment,
people
will
say:
well,
you
really
should
use
an
ensemble.
You
should
just
force
this
with
one
thing:
you
should
force
it
with
an
ensemble.
T
You
should
try
different
parameters
and
it's
just
multiplying.
You
know
there's
a
lot
of.
We
have
the
capacity
and
the
understanding
of
the
model
to
basically
know
you
have
to
do
more
stuff,
and
it
just,
but
it
makes
it
seem
like
things
are
beginning
are
just
snowballing
and
I
know.
That's
something.
That's
probably
happened
through
the
history
of
all
modeling,
but
my
question
is
within
the
community.
How
do
people
feel
that
things
are
getting
simplified?
T
You
know,
there's
you
can
always
find
ways.
You
can
always
find
ways
that
things
are
getting
more
complicated,
but
how
are
we
actually
simplifying
things
and
making
it
less
more
manageable,
especially,
you
know.
This
is
a
tough
thing.
When
people
come
into
the
community,
it
can
be
very
overwhelming
to
think
about
all
the
stuff
you
might
have
to
do.
C
C
D
Yeah
I'd
say
that
two
things
that
I've
found
ex
extremely
hopeful
the
new
up,
C
driver
and
being
able
to
run
a
single
point
just
by
entering
coordinates,
that's
been
pretty
much
a
game
changer
and
the
subset
data
script.
So
I
think
there
have
been
a
lot
of
tools
that
have
been
worked
on
over
the
years
and
like
just
this
morning.
I
just
use
subset
data
and
created
a
bunch
of
surface
data
sites
for
for
sites
that
I
need
to
run,
and
it's
just
so
much
easier
than
it
was
when
I
started.
D
T
Q
Q
You
can
actually
think
of
those
things
as
making
it
more
complex,
but
what
I
was
sort
of
saying
in
my
talk
is
that
you
know
the
physics
and
Physiology
are
more
theoretically
sound
than
what
we
have
so
in
some
ways
it's
simpler,
because
you
can
actually
sort
of
understand
the
equations
that
are
in
the
model.
I
I,
say:
I,
think
the
same
is
true
with
Fades.
Q
You
know
it's
more
ecologically
realistic
and
sort
of
trying
to
operate
at
the
way
in
which
sort
of
particularly
in
forest
forest
Dynamics
actually
works,
whereas
in
a
biogeochemical
model
you
can't
get
that
so
you
just
sort
of
try
to
make
it
work
and
you're
adjusting
things
all
the
time
to
try
to
get
the
right
answer
and
so
I
think
in
some
ways
these
models
that
are
coming
in
are
actually
making
it.
It's
complex
in
more
lines
of
code,
but
it's
simpler
in
terms
of
the
theory
that
goes
into
these
mods.
C
B
B
That's
not
really
accessible
until
you
kind
of
learn
how
everything
works
together,
but,
for
example,
the
multiple
complexity,
models
of
complexity,
modes
and
Fates
like
well,
they
might
be
a
little
bit
hard
to
understand
at
first
will
allow
people
to
more
directly
run
a
version,
that's
relevant
for
the
question
that
they're
asking
or
we
made
a
really
simple
land
model,
so
I
hope
that
those
options
give
people
a
more
direct
path
to
a
answering
the
question.
They
want
to
answer.
Charlie.
V
Yeah,
on
this
point
and
and
to
Keith
one
of
your
bullet
points
in
your
last
slide
about
applying
your
incredible
method
to
Faith
I
mean
you
know.
I
do
think
that
right
now,
because
the
set
of
people
who
are
using
Fates
is
you
know,
it's
still
relatively
small
and,
and
you
know
the
spin
up,
it's
been
a
bit
is
difficult
right
and
as
we
as
we,
you
know,
more
people
use
it.
V
You
know
spirit
with
faith
is,
is
considerably
slower
than
the
in
the
big
leaf
model,
and
that's
going
to
get
worse
when,
when
we
include
nitrogen
cycles
and
things
like
that,
so
I'm
curious,
like
if
you've
thought
at
all
about
how
we
might
you
know,
approach
this
been
a
problem
better.
You
know,
if
you're
you
had
the
ball
boy
in
your
talk
for
Fades,
would
you
go.
G
G
V
G
C
V
A
P
P
I'm
super
happy
with
how
our
first
remote,
hybrid
I'm,
sorry
hybrid,
meeting,
worked
and
I'm
looking
forward
to
be
able
to
offer
this
offer
this
in
the
future
yeah.
So
I
talked
a
little
bit
about
this
later
award
earlier
in
the
week
and
we
had
a
distinguished
panel
of
of
scientists.
Try
to
wrestle
with.
You
know,
out
of
the
30
some
presentations
that
we
had
from
graduate
students
and
postdocs.
P
You
know
who
who
really
made
some
significant
contributions
to
the
to
the
working
group
and
so
we're
really
impressed
by
the
clmu
team,
and
this
is
kind
of
the
first
time
we've
ever
given
the
award
to
a
team,
but
we
thought
that
Joyce
Yang,
Kathy
Lee,
cure
cures
on
and
Bowen
Fang
Bowen
didn't
present,
but
it's
been
really
integral
in
the
development
of
the
transient.
Urban
capabilities
in
clmu
really
did
a
nice
job,
so
our
first
award
goes
to
the
CLM.
P
P
P
T
AA
Y
Oh,
this
feature
shows
the
CO2
emissions
to
the
atmosphere
and
also
the
carbon
sinks
into
the
ocean
and
the
land.
So
from
this
figure
we
can
see
how
much
carbon
is
remained
in
the
atmosphere
and
how
much
carbon
could
be
stored
in
the
ocean
and
in
the
line.
So
we
are
interested
in
the
ocean
carbon
zinc.
This
is
because
the
ocean
takes
about
25
percent
of
anthropogenic
CO2
from
atmosphere
over
the
industrial
area,
and
this
figure
this
figure
shows
the
mechanisms
that
can
affect
the
ocean.
Carbon
things
are
pretty
complex.
Y
It
is
important
for
us
to
distinguish
the
natural
carbon
cycle
and
arthropogenic
carbon
cycle.
So
if
we
look
at
the
atmospheric
carbon
in
the
in
the
corner,
we
can
see
that
in
the
pre-industrial
area
in
steady
state,
the
atmospheric
CO2
concentration
was
either
increasing
or
decreasing.
So
18
study
said,
and
on
top
of
that
we
are
importing
anthropogenic
partitions.
That
start
with
the
natural
carbon
gives
the
total
carbon
of
total
carbon
of
in
the
in
the
atmosphere.
Y
Y
So
again,
the
atmospheric
CO2
is
the
first
driver
of
the
ocean
carbon
sink.
However,
the
ocean
circulation
patterns
could
also
affect
the
ocean
carbon
sink
by
multiple
ways,
and
this
figure
is
showing
the
global
ERC
CO2
flux
under
a
calc,
multiple
warming
scenarios,
and
we
see
that
there
are
great
authorities
of
the
site
of
ocean
carbon
sinks
across
symmetics
climate
models
under
each
climate
scenario,
and
we
also
see
that
under
each
climbing
scenario,
the
atmospheric
CO2
in
each
model
is
the
same.
Y
So
the
thing
is,
the
large
authorities
are
driven
by
ocean
processes
and
in
this
study
we
pick
up
three
climbing
scenarios
from
the
most
small
transforming
scenario,
to
the
basic
social
warming
scenario
and
for
the
models
we
choose,
all
the
available
same
website
models
plus
one
specific
or
System
model
CSM,
and
also
an
offline
model
to
to
do
this
study.
So
the
scientific
question
here
is:
how
do
ocean
circulation
changes
could
affect
the
ocean
convincing
in
simulated
models
and
in
the
CSM
under
climate
change?
Y
So
here
is
my
results,
so
first
figure
showing
here
is
the
changes
of
both
amok
and
Ice.
Mark
allow
a
long
climate
changes
under
different
warming
scenarios
from
SST
1
2.6
to
SFP
58.5,
so
amalc
on
ismos
is
the
upper
and
a
basal
cell
of
overturning
circulation.
You
can
look
into
this
schematic,
so
we
see
from
this
figure
is
the
meridian
overturning
circulation,
slow
Star,
both
in
the
upper
and
in
the
orbital
cell,
no
matter
in
which
warming
scenario,
but
the
strings
of
the
Slowdown
will
highly
depend
on
the
scenario.
Y
So
over
10
years
circulation
will
slow
more
under
high
end
warming
scenario,
but
atrial
flow,
slides
under
modulatory
forming
scenario
and
to
the
euro
2300.
There
are
only
around
5
or
500
jobs
or
fast
charge,
email
class
under
business
as
usual
warming
scenario,
but
in
terms
of
the
ice
Mark,
which
is
basal
cell
overtime
circulation,
we
see
that
it
almost
fully
shuts
down.
Y
So,
in
this
study,
we're
trying
to
separate
the
law
of
Iceberg
slowdown
on
the
biological
pump
and
the
solubility
pump
to
affect
the
ocean
carbon
sink.
So
the
first
one
is
the
first
way
is
the
ice.
Mark
slowdown
will
how
the
ice
marks
slow
down
will
affect
the
accumulation
of
the
regenerative
carbon.
Y
Y
So
this
figure
shows
shows
out
shows
the
changes
of
ice
marks
amok
and
the
iceberg
production
at
100
meter
in
each
individual
cmf6
models.
So
the
total
we
fully
understand.
We
fully
studied
13
cmf6
models
to
include
all
available
models
that
we
can
do
and
we
we
noticed
that
almost
all
same
F6
models
show
larger
slowdown
of
iMac
and
Ice
mark
on
their
high
end.
Warming
scenarios
and
moderate
warming
scenario
see
the
panel
a
and
panel
B
here,
and
we
also
noticed
that
all
team
F6
models
show
decreasing
biological
productivity.
Y
Y
The
next
feature
we
showed
accumulation
of
preformed
carbon
and
regenerated
carbon
in
the
ocean
and
the
upper
panel.
The
upper
the
top
row
here
shows
the
accumulation
of
the
preformed
and
regenerated
di
regenerative
carbon
iot
intermediate
depth
from
100
to
2000
meter,
and
the
bottom
panel
here
shows
the
carbon
accumulation
in
the
deep
ocean,
and
we
see
that
almost
all
symmetics
models
show
increasing
preformed
carbon,
which
is
panel
e
and
G,
and
regenerated
carbon
10
iPhone
H.
But
we
can
also
notice
that
the
preformed
carbon
accumulates
more
in
the
intermediate
depth.
Y
The
next
figure
we
are
trying
to
understand
how
the
ice
marks
slowdown
could
affect
the
accumulation
of
regenerated
di
regenerated
and
preformed
carbon
across
all
the
same
F6
models.
So
the
these
figures
here
shows
the
changes
of
Iceberg
production
over
the
changes
of
ice
marks
versus
the
accumulation
of
regenerated
DIC
in
the
form
water
column
of
the
water.
Y
Y
So
this
part
showing
how
the
Slowdown
of
ice
marks
will
contribute
negatively
to
the
streams
of
the
solubility
pump
and
the
last.
The
last
panel
here
shows
the
Knight
effect.
That
is
how
the
Slowdown
of
overturning
circulation
could
affect
the
total
carbon
outtake
in
the
ocean.
So
we
we
still
see
there
is
some
positive
correlation
that
indicates
the
net
effect
of
slowdown
of
ice.
Y
So
we
see
that
the
over
10
year
circulation,
slow
down,
will
cause.
We
will
lead
to
the
ocean
lead
to
will
sequester
the
nutrients
in
the
deep
ocean.
This
will
further
decrease
the
iceberg
production,
so
it
sees
about
biology.
Part
will
keep
euclation
from
as
climate
worms
compared
to
the
pre-industrial
level.
However,
the
circulation
slowdown
could
also
allow
the
carbon
to
stay
in
the
ocean
for
a
longer
time,
so
that
we
see
the
removal
of
this
carbon
from
the
ocean
back
into
the
atmosphere
will
also
decree
will
also
decrease.
Y
This
will
lead
to
a
nice
positive
carbon
regenerative
carbon
accumulation
in
the
ocean
and
the
next
effect
of
overturned
circulation.
Slow
down
to
the
biological
carbon
pump
is
the
overtime
circulation.
Soda
could
help
the
ocean
to
accumulate
more
regenerative.
Carbon
adapts,
and
so
here
is
a
summary
on
the
Slowdown
of
over
10
year.
Speculation
depends
on
the
scenario
and
it
slows
more
on
high-end
scenarios
and
a
slow
size
under
moderate
scenarios,
and
also
we
see
the
Slowdown
of
ice
Mark
will
decrease
the
iceberg
production
but
in
case
the
Cardinal
secretration
time
in
the
ocean.
Y
So
this
will
automatic
effect,
will
accumulate
more
regenerative.
Carbon
in
the
ocean
or
the
streams
of
the
biological
part
will
go
up,
but
the
Slowdown
of
overtime
circulation
will
finally
reduce
or
reduce
its
accumulation
of
the
preformed
carbon
in
the
ocean,
even
though
the
specific
mechanisms
are
not
very
and
very
well
understand,
and
the
net
effect
is
the
Slowdown
of
overtime
circulation
will
reduce
the
size
of
the
ocean
carbon
sink,
and
this
will
explain
around
one
third
of
the
Sprite
across
all
the
same
FCX
models.
C
Y
C
P
N
B
C
O
Have
using
cesm1bc,
where
we're
trying
to
quantify
the
feedbacks
between
phytoplankton,
Elemental,
Stoichiometry
and
Marine
biogeochemical
cycles,
so
are
this
group
of
people
is,
is
going
to
be
very
familiar
with
the
sort
of
context
for
this
for
This
research.
So
we
know
that
carbon
export
is
linked
to
nutrients
via
stoichiometry.
O
Yeah,
so
we
know
that
carbon
export
is
linked
to
nutrients
via
Stoichiometry
and
the
portion
of
the
biological
pump.
That
I'm
going
to
focus
on
here
is
phytoplankton
photosynthesis,
an
uptake
of
nutrients
and
incorporation
into
phytoplankton
biomass
or
their
Elemental
ratios
of
carbon,
nitrogen
phosphorus
iron
and
silica
as
well.
O
This
has
been
pretty
well
observed
at
this
point
via
Ocean
observations
of
cellular
Stoichiometry,
as
well
as
particular
organic
matter
Stoichiometry,
and
that
variations
in
phytoplankton
nitrogen
phosphorus
are
pretty
directly
tied
to
nitrogen
phosphorus
Supply.
So
what
we
want
to
know
is
what
impacts
does
variable
Stoichiometry
have
on
the
sensitivity
of
biogeochemical
cycles.
O
O
They
have
our
three
phytoplankton
groups,
our
small
phytoplankton,
diatoms
and
diasotrophs,
with
fully
variable
carbon,
nitrogen
phosphorus
iron
and
s-I,
and
we
have
a
variety
of
iron
sources
that
have
basic,
mostly
been
updated
to
resemble
that
of
cesm2.
So
we
have
atmosphere,
sediments
Rivers
as
well
as
hydrothermal
vents
and
bottom
Scavenging,
so
we
have
a
fairly
robust
Marine
ecosystem
as
well
as
iron
cycling.
Y
O
So
basically,
what
we
have
is
when
we
have
a
lot
of
a
nutrient
X,
then
our
phytoplankton
are
setting
their
ratio
at
the
highest
possible
nutrient
to
carbon
and
then
when
they,
when
the
ambient
nutrients
fall
below
a
prescribed,
optimal
value,
the
phytoplankton
begin
decreasing
their
nutrient
to
carbon
ratios
until
they
hit
a
prescribed
minimum,
and
so
this
is
how
all
four
of
these
nutrients
are
now
represented
in
the
model.
We
also
have
some
added
complexity
where,
for
example,
for
the
P
quotas.
O
When
nitrate
is
low,
we
reduce
but
reduce
both
nitrogen
and
phosphorus
uptake
in
order
to
maintain
endopy
uptake
that
is
appropriate
for
ambient
phosphate
levels,
and
then
we
also
have
an
added
sort
of
complexity
for
our
SI,
which
is
when
s
I
is
in
when
there's
a
lot
of
or
when
the
dissolved
SI
concentration
is
lower.
You
have
lower
acid
carbon
and
when
it's
replete
but
iron
is
lower,
you
get
a
further
increase
in
acid
carbon
and
that's
sort
of
based
off
of
some
physiological
studies.
O
So
these
are
the
ratios
that
we
have
currently
implemented
into
the
BEC.
So
each
group
of
phytoplankton
has
their
own
prescribed
range
which
have
been
constrained
from
the
Ghost
Ship
pom
observations.
Oh,
this
is
a
okay.
The
tanyoka
at
all
is
now
published
was
published
last
year,
and
then
we
have
our
end
of
P.
O
O
I've
got
constant
pre-industrial
CO2
run
up
300
years
averaged
over
the
last
20
years.
We
have
a
fixed
model
where
the
C
to
end
the
p
is
fixed
at
96
to
16,
to
1.
iron
to
carbon
7
s.
I
to
n
is
one
and
then
our
variable
Stoichiometry,
that
I
reported
on
the
previous
page
and
all
units
are
typical,
and
then
we
have
our
PM
database.
This
is
just
showing
where
our
observations
are
from
and
I'm
going
to
be.
O
O
So
here
we
have
the
observation
and
phytoplankton
observations
in
Blue
export
in
red
phytoplankton
in
Black,
showing
the
C
to
n
and
C
to
P
versus
dissolved,
inorganic
nitrogen
and
po4
at
the
bottom.
So
in
general
you
know
we
have
our
phytoplankton
have
much
higher
carbon
and
nitrogen
than
our
export
and
there's
sort
of
General
agreement
between
our
PLM
observations
and
our
export
ratios,
because
those
are
the
closest
comparisons
to
make,
and
then
we
also
have
our
c2p
as
well.
O
So
these
aren't
final
results.
Yet
here
we
have
global
maps
of
this
is
the
export
at
100,
meter,
carbon
and
nitrogen,
and
then
overlaid
and
squares
are
observations.
So,
as
you
can
see,
we
get
General
agreement
in
the
patterns
of
our
carbon
and
nitrogen,
especially
if
you
look
at
our
Pacific
Crews
there
that's
p18,
we
see
lower
carbon
and
nutrient
values
in
the
equatorial
upwelling
region
and
then
elevated
values
in
the
South
Pacific
gyre.
O
O
O
Here
is
a
specific,
Cruise
transect.
So
this
is
that
p18
that
Pacific
North
South
South
Cruise
so
going
from
south
to
North,
and
we
have
our
C
to
n
and
a
p
c
to
P.
Our
nitrate,
dissolved
iron
and
phosphate
for
blue
is
our
observations
and
yellow
is
our
model
and
for
the
model
I'm
showing
in
the
dotted
line,
is
the
just
combined
Zone
phytoplankton
and
then
the
yellow
line
is
the
actual
export.
So
you
can
see
that
those
lines
are
actually
quite
similar
here
and
it's
usually
in
a
previous
version.
O
They
weren't
as
similar,
but
in
general
we
see
you
know
we're
getting
that
pattern
of
elevated
values
in
the
gyres
which
we
want
to
see.
We're
reproducing
our
observed
nutrients
quite
well,
so
yeah
and
the
the
low
end
to
P
is
driven
by
the
low
C
to
P,
whereas
our
C
to
n,
actually
I
think
looks
the
best
it
ever
has,
which
is
nice,
and
so
here
we
have
our
actual
sort
of
science
results.
O
We
have
the
comparisons
of
our
phytoplankton
surface,
nutrient
limitations
for
a
fixed,
Stoichiometry
and
variable
Stoichiometry
scenario.
We
can
see
for
our
diatoms.
We
get
a
transition
from
nitrogen
nitrogen
Limited
in
most
of
the
gyres
to
primarily
SI
limited,
as
well
as
expansion
of
our
P
limited
regions,
and
we
also
see
large
increases
in
si
limitation
where
iron
was
previously
limiting.
We
have
the
same
for
our
small
phytoplankton
again
in
our
fixed
Stoichiometry.
O
You
see
that
expansion
of
P
limitation-
this
is
most
likely
because
our
phytoplankton
have
far
greater
P
flexibility
than
and
flexibility.
So
when
you
fix
both
n
and
p,
the
rigidity
of
that
P
quota
dominates
the
sort
of
patterns
of
new
of
limitation
and
then
for
our
nitrogen
fixers.
We
see
sort
of
increases
in
the
severity
of
our
P
limitation
in
the
Atlantic,
for
example,
as
well
as
reductions
in
the
sort
of
regions
that
are
replete.
O
So
here
are
some
impacts
on
sort
of
globally
integrated
fluxes
for
our
variable
and
fixed
scenarios.
So
for
net
primary
productivity,
we
actually
see
a
slight
increase
when
we
use
our
fixed
with
an
increase
of
roughly
three
percent
for
our
POC
export.
We
see
a
decrease
in
10
percent
for
nitrogen
fixation
and
water
column
denitrification.
We
see
much
more
drastic
decreases
of
roughly
25
to
to
54
for
both.
So
when
we're
using
this
sort
of
fixed
Stoichiometry
version
model,
we're
really
underestimating
these
key
nitrogen
fluxes
other
impacts.
O
There
were
significant
changes
to
the
community
composition
when
fixed
Stoichiometry
was
used.
So,
for
example,
in
our
variable
run,
our
dryers
had
up
to
15
over
15
diatom
composition
compared
to
the
small
phytoplankton,
whereas
it
was
the
diatom's
almost
disappeared
from
the
gyrus
entirely
in
the
fixed
model.
We
also
saw
a
global
increase
in
the
n-star
bias,
greater
than
20,
with
positive
n-star
values
in
all
ocean
gyres
in
the
fixed
run
from
zero
to
350
meters
depth.
O
Part
of
this
is
exacerbated
by
the
end
of
P
being
too
low
in
our
export,
but
still
because
we
see
that
change
from
a
you
know
normal
negative
n
star
in
the
global
ocean
to
positive
everywhere.
We
still
expect
to
see
a
bias.
It
just
will
be
probably
slightly
smaller.
I
also
did
some
additional
experiments.
These
are
new.
O
The
fixed
SI
I
think
probably
finished
this
morning,
but
I
didn't
get
it
in
here
in
time,
but
so
I
iterated,
where
we
fixed
only
one
nutrient
at
a
time
to
sort
of
try
to
pull
out.
You
know
which,
which
nutrient
bigfix
has
the
greatest
impacts
for
those
same
majorly
integrated
fields
and
as
you
can
see,
our
fixed
an
interesting
result
is
that
our
fixed
nitrogen
actually
results
in
increases
in
some
of
our
carbon
and
Nitro
nutrient
Fields,
NPP
and
POC
flux.
O
Don't
change
much,
but
our
nitrogen
fixation
water
column
denaturation
do
increase
slightly,
whereas
with
our
fixed
P,
we
see
those
decreases
in
POC
export
and
fixation
and
water
Quantum
denitrification
that
dominate
in
when
you
fix
all
of
the
ratios
and
the
fixing
of
the
iron.
The
impact
of
that
on
the
nitrogen
cycle
is
something
that
is
explored
in
the
paper
that
I
currently
have
in
review.
So
I
didn't
really
go
into
that
here.
O
So
in
General,
Dynamic
Stoichiometry
is
necessary
for
under
understanding
interactions
in
Ocean
biogeochemistry.
Changing
Stoichiometry
is
likely
to
modify
climate
change
impacts
and
fixed
nutrients
can
have
opposing
impacts
on
the
BGC
cycling,
and
hopefully
we
will
have
our
final
version
of
this
soon
and
and
we'll
be
submitting
it
thereafter.
So
thank
you.
AB
C
O
So
that's
a
that's
a
great
question
I.
That
would
be
my
assumption
and
I
would
need
to
actually
look
at
the
total
DIC
inventory.
But
that's
something
that
we're
interested
in
is
because
some
of
these
studies
that
I've
done
comparisons
of
fixed
and
variable
Stoichiometry
see
competing
conclusions
where
some
see
increases
some
see
decreases,
and
so
we're
hoping
that
a
lot
of
those
have
focused
on
either
fixed
C
to
n
or
fixed
c2p,
but
not
fixing
and
variable
both.
So
that's
something
that
we
hope
with
this
research.
AC
Okay,
so
hi
everyone,
my
name
is
Jin
I'm
a
second
year
PhD
student
in
UC,
Irvine
working
with
Professor
kismur.
We
are
collaborating
with
csm2
marble
developers
in
Encore
about
expanding
numerical
system,
our
marble
and
this
project
is
still
ongoing
in
progress.
So
we
just
wanted
to
share
some
preliminary
results
here.
AC
So
first,
let
me
start
with
the
big
picture
of
why
we
need
to
do
this,
so
the
biological
Palm
plays
a
critical
role
in
storing
carbon
from
the
atmosphere,
and
an
efficiency
of
the
biological
pump
is
largely
controlled
by
the
ecosystem.
Community
compensation.
So,
as
we
can
see
in
the
I,
think
the
rest,
my
mouse
so
as
we
can
see
in
the
stratified
oligotrophic
ocean,
that
kind
of
ecosystem
is
dominating
by
the
smaller
size,
phytoplankton
and
so
pentane.
AC
As
a
result,
the
POC
export
is
kind
of
weak,
but
in
the
misotropic
and
eotrophic
ocean,
that
kind
of
ecosystem
is
dominating
by
a
larger
size,
phytoplankton
and
so
plantain.
As
a
result,
you
can
see
the
POC
exports
is
stronger
and
the
expense
of
biological
pump
is
also
stronger.
But
if
we
look
at
the
two
degree
warming
in
the
future,
we
can
see
that
the
global
warming
will
cause
the
global
ocean
to
be
more
stratified
and
the
oligon
traffic
condition
will
also
expand
globally.
AC
As
a
result,
in
the
previously
in
the
in
the
region
previously
is
dominated
by
larger
phytoplankton
and
zoplankton
in
the
future.
It
won't
will
not
be
the
same.
It
will
be.
The
ecosystem
will
shift
to
the
smaller
size,
phytoplendent
and
so
pentane.
So
we
can
see
that
the
such
condition
will
actually
hurt
the
strength
of
the
biological
pump
in
the
future.
So
there's
a
question
mark
here.
AC
How
much
uncertain
do
we
know
about
the
future
changes
in
in
the
American
export,
so
in
the
recent
paper
from
Henson
2022,
and
it
shows
that
the
prediction
in
PLC
export
in
the
current
Sigma
Six
models
do
not
agree
with
each
other,
both
in
terms
of
the
future.
The
direction
of
the
change,
and
also
in
terms
of
the
magnitude
of
the
change
and
is
mostly
the
many
reasons,
can
cause
that,
and
the
main
reason
is
that
there
are
the
Marine
Corps
system.
AC
Representation
in
this
most
of
the
76
model
are
very
simplified.
There's
lack
of
patent
size
diversity
there.
Typically,
it
only
includes
two
to
three
types
of
phytoplankton
and
one
to
two
type
of
zooplankton
and
such
a
Simplicity
will
introduce
our
sentences
of
the
estimation
of
future
changing
future
future
changes
or
POC
export.
So
here
we
believe
that
there's
an
urgent
need
for
Global
client
models
to
have
a
more
sophisticated
representation
of
multiple
Plankton
functional
time
in
the
Marine
ecosystem
models.
AC
So
we
are
currently
working
on
expanding
numerical
system
on
marble
with
the
intermediate
complexity,
so
that
we
can
use
this
model
to
do
future
cameras,
and
here
at
the
slide,
is
showing
the
latest
version
of
csm2
marine
consist
model.
It
includes
four
type
of
four
types
of
phytoplankton,
which
is
Pico
phytoplankton,
that's
a
child
calculus
first
and
diatoms,
and
they
only
include
one
generic
zopatin
group.
So
it
looks
pretty
simple
right.
AC
So
in
our
current
new
ecosystem
model,
which
is
ap14
model,
we
expand
the
Marine
ecosystem
for
phytokines.
We
we
have
eight
different
type
of
hydroplankton
and
for
and
also
have
four
different
typoso
content
for
the
Pico
size
phytoplankton.
We
have
protocolcos
cynical
caucus,
Pico
carriers
and
small
diasotrophs,
and
for
a
nanosense
phytoplankton
we
have
fire
synthesis,
calculated
supports
and
other
generic
nano-fetopentine
and
for
the
Microsoft
phytoplankton
we
have
the
classic.
AC
A
diagram
to
represent
to
represents
represent
that
group.
Beside
phytoplankton,
for
example,
we
have
four
different
types
of
plantains
and
includes
small
and
large
Microsoft
pentane
and
misozo
pentane
and
macrosophantine.
So
the
two
Microsoft
pentane
are
the
main
reasons
for
the
smaller
size
phytoplins,
and
it
also
that
is
the
direct
linkage
between
the
primary
producers
and
the
higher
traffic
levels
and
for
the
two
larger
sizes
zooplankton.
There
are
other
main
grazers
for
the
larger
size
phylopentin
and
they
are
also
on
the
main
food
source
of
the
Fisheries.
AC
AC
AC
The
model
captures
the
higher
value
for
the
fire
synthesis
and
is
is
the
same
with
the
Institute
data,
and
here
is
the
comparison
with
the
satellite
derived
phytoplankton
biomass
across
the
three
different
size,
phytopening
groups,
so
on
the
left
is
the
previous
version
of
csm2
model
in
the
middle.
Is
the
satellite
data
and
on
the
right,
is
our
AP
policy
model.
So
we
can
see
that
previous
model
didn't
really
capture
the
higher
value
for
the
Pico
phytoplankton
in
the
higher
latitude
and
our
ap4c
model.
AC
It
captures
the
higher
value
for
the
total
frequency
in
the
higher
latitude
and
also
our
total
Nano
C,
and
that
and
the
diet
and
C.
It
also
shows
similar
pattern
compared
with
the
satellite
products,
and
this
figure
shows
the
percentages
of
the
three
phytoplankton
groups
relative,
the
total
carbon
biomass.
So
we
can
see
that
both
model
and
the
satellite
shows
Pico
phytobenton
is
dominating
in
the
subtropical
region.
AC
It's
about
a
higher
than
50
of
the
total
biomass
and
the
Nano
phytoplankton,
and
micro
and
micro
phytoplins
is
co-dominating
in
the
higher
latitudes,
and
this
slide
shows
the
comparison
between
model
total
chlorophyll
and
the
satellite
chlorophyll.
So
we
can
see
our
model
are
captured,
really
well
of
the
space
pattern
and
the
magnitude
of
the
satellite
curve.
Fuel
and
the
biomass
is
within
the
reasonable
range
and
on
the
right
is
the
total
NPP
and
also
the
group
specific
MPP.
AC
You
can
see
that
total
MPP
is
in
the
reasonable
value.
It's
about
50
50
PG
per
year,
and
also
we
can
see
different.
The
MPP
from
different
group
is
also
showing
a
reasonable
pattern,
such
as
picofit
phytoplankton
is
a
it's.
The
dominant
dominant
species
of
the
total
MPP
is
about
60
of
the
Southern
MVP,
and
the
Nano
phytoplantic
and
microfit
is
kind
of
like
the
same.
AC
Similar
range
is
about
like
80
18
of
the
total
NPP,
and
this
figure
is
showing
the
amount
of
the
zolpentine
distribution
and,
and
we
can
see
that
the
model,
the
open
10,
shows
a
similar
special
pattern
compared
with
the
Meredith
Institute
biomass
and
also
the
depths
integrate
on
carbon.
Biomass
is
also
within
the
range
of
the
value
from
the
observations.
AC
So
here's
a
summary
of
this
project.
Currently
we
did,
we
are
developing
intermediate
capacity,
Marine
ecosystem
model
within
the
csm2
marble,
and
it
had
the
potential
to
do.
Clement
runs
and
the
current
AP
Sports
Z
Model
can
reproduce.
The
general
pattern
of
the
obser
observe
the
phytoplankton
groups
and
it's
showing
picofit
is
dominating
in
a
subtropical
region
and
a
nano
and
the
diatom
is
co-dominating
in
the
higher
latitudes
and
our
simulated
as
opentine
groups
is
also
showing
a
good
agreement
with
the
observations
and
finally,
is
the
implication
for
this
model.
AC
So
this
model
can
help
help
us
to
study
how
planting
size
and
also
Community
composition,
impact
the
carbon
export
and
the
CO2
flux
in
the
future,
and
also
this
model
can
help.
You
help
the
future
studies
to
incorporate
more
keep
missing
process
in
the
ecosystem,
such
as
low
pattern,
vertical
migration,
and
also
this
this
model
allows
for
the
future
ecosystem
extension
for
fisheries
and
also
other
higher
traffic
levels.
AC
AB
AC
AD
Phytoplankton
are
vital
for
marine
ecosystems
and
they're,
also
a
key
part
of
the
climate
system
due
to
their
role
in
the
carbon
cycle.
So
we
care
a
lot
about
these
spatio-temporal
distribution
of
phytoplankton
and
the
main
way
that
we
study
phytoplankton
populations
on
a
global
scale
is
through
remote
sensing
of
chlorophyll,
which
is
their
primary
photosynthetic
pigment
and
to
detect
chlorophyll.
We
rely
on
visible,
green
and
blue
wavelengths.
So
since
we're
using
the
visible
spectrum,
we
are
unable
to
detect
phytoplankton
in
the
absence
of
visible
light
or
beneath
the
cloud
cover.
AD
Unfortunately,
if
we
look
at
these
two
model,
climatologies
of
chlorophyll
and
clouds,
we
see
that
regions
with
high
chlorophyll
also
tend
to
correspond
to
high
cloud
cover.
So,
therefore,
we
are
unable
to
reliably
detect
phytoplankton
from
satellite
observations
in
the
regions
where
they
are
most
abundant.
AD
Here's
an
example
of
some
real
world
data.
The
missing
data
is
shown
in
white
here,
and
these
are
two
eight
day
composite
images.
You
can
see
that
even
after
eight
days
of
sampling,
there
are
still
significant
data
gaps
and
a
lot
of
these
data
gaps
are
due
to
clouds.
However,
there
are
other
factors
that
prevent
satellite
detection
as
well
in
the
high
latitude
Winters.
AD
We
see
that
there
are
no
observations
at
all,
and
this
is
because
there
just
isn't
enough
light
for
satellite
detection,
so
in
order
to
create
a
more
direct
comparison
between
the
full
field
model,
outputs
and
these
satellite
observations,
we
developed
a
satellite
emulator
for
chlorophyll
in
cesm.
So
now
the
model
generates
the
standard
total
chlorophyll
output.
In
addition
to
these
simulated
observations
of
chlorophyll,
so
we
can
use
these
simulated
observations
and
compare
them
to
the
total
chlorophyll
output
to
assess
potential
biases
due
to
missing
data.
AD
So
right
now
in
the
emulator,
we
have
included
several
factors
that
prevent
satellite
detection.
We
have
clouds
sunlight
and
sea
ice.
However,
as
I
mentioned
before,
there
are
other
factors
that
also
prevent
satellite
detection
that
are
not
included
yet
but
may
be
included
in
a
future
version
of
the
emulator.
AD
So
our
new
model
output
is
generated
within
pop,
but
we
needed
to
pass
in
several
other
variables
from
different
components.
So
here
we
use
Sea
ice
and
then
from
cam
we
passed
in
the
Solar
Zenith
angle
and
the
simulated
Cloud
observations
from
cost
and
then
using
these
new
variables
we
calculate
a
weight
for
each
grid
cell
in
pop,
which
which
is
the
total
fraction
of
a
grid
cell,
that
a
satellite
would
have
been
able
to
see.
Then
we
apply
these
weights
to
the
total
surface
chlorophyll
and
we
end
up
with
our
satellite
observed
chlorophyll.
AD
AD
AD
So
far.
What
I've
found
is
that
this
has
to
do
mainly
with
the
seasonal
phasing
of
of
clouds
and
chlorophyll,
so,
for
example,
in
the
North
Pacific
off
of
the
coasts
of
Japan
in
the
winter,
when
we
have
low
chlorophyll
values,
there's
High
cloud
coverage
and
then
in
the
spring
Bloom
when
chlorophyll
values
Peak
this
corresponds
to
low
cloud
cover,
so
we're
sampling
much
more
frequently
in
this
time
of
the
year,
leading
to
an
overall
overestimate
of
the
chlorophyll
there
and
in
the
blue
regions.
We
see
the
opposite
pattern.
AD
We
were
also
interested
in
looking
at
the
mean
seasonal
cycle
over
various
ocean
biomes,
and
you
can
see
that
the
impact
of
clouds
varies
between
these
different
biomes
and
it
not
only
affects
the
magnitude
of
the
seasonal
cycle,
but
it
can
also
impact
the
timing
of
the
peak
bloom
and
we
found
that
these
patterns
largely
emerge
due
to
the
spatial
correlation
within
biomes
between
chlorophyll
and
clouds.
So,
for
example,
in
the
Southern
Ocean
regions
that
have
high
chlorophyll
also
tend
to
correspond
to
low
cloud
cover.
AD
So
when
we
take
the
average
over
the
biome
we
end
up
with
an
overestimate
of
the
total
chlorophyll
and
I
also
wanted
to
note
here
that
you
can
see
how
the
Cloudy
chlorophyll
has
a
higher
variability
than
the
full
field.
Baseline
output,
all
right
so
so
far,
I've
talked
about
how
we've
used
this
tool
to
assess
biases
due
to
missing
data
from
cloud
cover.
But
there
are
several
other
applications
that
I
want
to
mention.
AD
The
first
one
is
model
tuning,
so
here
I'm,
comparing
the
model
climatology
to
our
real
world
observations
and
on
the
left
side,
I
used
the
total
full
field
model
output,
which
is
what
we
typically
use
to
compare
with
observations
and
then
on
the
right
side.
I
have
the
climatology
calculated
using
the
simulated
observations,
which
is
a
more
direct
comparison
to
our
real
world
data,
and
so
what
this
plot
suggests
is
that
the
actual
bias
in
our
model
may
be
greater
than
we
previously
thought.
AD
Another
application
that
we
envision
this
being
used
for
is
a
gap
filling
test
bed.
So
in
the
real
world
we
often
want
to
try
to
fill
in
the
gaps
in
our
satellite
data.
However,
we
don't
have
a
ground
truth
to
be
able
to
validate
these
results,
but
in
the
model
world,
what
we
can
do
is
fill
these
data
gaps
using
various
Gap
filling
techniques,
and
then
we
can
compare
this
interpolated
result
to
the
model.
Truth
in
here,
I've
done
a
linear
interpolation.
AD
But
ideally
we
would
want
to
test
this
with
many
different
methods
to
evaluate
which
one
does
the
best
and
then
hopefully
apply
that
method.
To
the
real
world
satellite
observations
in
order
to
improve
that
data
set
in
the
last
application,
that
I
want
to
mention
is
calculating
the
emergence
of
a
trend.
AD
AD
Right
I
also
may
have
missed
something,
but
this
schematic
here
suggests
that
you're
sort
of
sampling
swaths
within
the
model.
Do
you
actually
have
the
orbit
geometry
built
into
the
emulator?
We
do.
We
have
a
simplified
version
of
the
aqua,
Modis
sampling
pattern,
which
samples
of
1
30
pm,
and
it's
not
a
perfect
representation
of
the
orbital
pattern
because
we're
not
including
orbital
gaps
between
the
orbits.
AD
So
we
actually
tested
a
few
different
versions
of
this
and
the
results
that
I'm
showing
here
are
not
from
that
version.
But
we
are
planning
to
look
into
that
more
and
publish
that
in
our
paper
and
having
just
that,
orbital
geometry
thing
would
be
very
nice
for
using
this.
On
other,
you
know
other
sensors
yeah.
AD
AD
AD
We
have
trying
to
make
skillful
predictions
of
multiple
Marine
stressors
in
the
surface
and
subsurface
ocean
using
CSM
smile,
the
seasonal
to
multi-large
Ensemble
and
some
novel
observational
products,
so
I
think
we're
all
super
familiar
with
climate
change
in
the
ocean,
but
just
as
a
review,
three
stressors
we're
really
interested
in
is
heating.
The
ocean's
growing
warmer
decline
in
oxygen
stocks
are
on
the
ocean
and
then
acidification
in
the
ocean.
So
we
know
these
long-term
trends
and
we
have
the
a
lot
of
models
telling
us
about
these
long-term
trends
in
the
ocean.
AD
But
here
in
the
study,
we're
really
interested
in
trying
to
make
predictions
on
shorter
turn
time
scales
from
months
to
years
in
advance.
You
can
imagine
these
would
be
important
for
Fisheries
managers
who
are
trying
to
understand
how
they
should
manage
a
fishery
over
the
coming
months
and
years.
So
we're
trying
to
make
these
predictions
on
short-term
time
scales
and
we
want
them
to
be
relevant
for
Fisheries.
AD
So
where
are
we
actually
going
to
predict
these
stressors
or
try
and
predict
these
stressors
we're
going
to
focus
on
important
Fisheries
so
on
this
map
here
we're
showing
regions
where
fist
catch
is
greater
than
a
thousand
metric
tons
in
2015?
So
you
see
primarily
all
of
these
fisheries
and
fish
catches
happening
in
coastal
regions,
and
these
coastal
regions
are
really
captured
in
these
divisions
called
large
Moon
ecosystems,
which
I'm
sure
a
lot
of
us
are
also
familiar
with.
AD
AD
So
these
new
products
are
really
exciting,
and
one
limitation
of
these,
of
course,
is
that
the
observations
are
still
relatively
scarce,
even
with
machine
learning
algorithms.
So
if
we
want
to
look
at
high
resolution
regions
in
these
smaller
local
locales,
like
a
large
Marine
ecosystem,
we
want
to
look
at
regions
where
there
are
a
lot
of
observations
being
trained
into
these
machine
learning.
Algorithms.
So,
on
this
map,
we're
just
showing
regions
where
there's
a
lot
of
information
and
direct
observations
going
into
the
machine
learning
algorithms
and
these
data
products.
AD
Here
you
can
see
a
few
regions
pop
out
a
few
large
marine
ecosystems
sort
of
in
the
North
Pacific.
We
see
the
California
Current
and
the
Gulf
of
Alaska
and
we
also
see
the
croatio
current
along
with
a
few
other
regions
in
the
global
ocean.
But
for
this
talk
today
we're
going
to
primarily
focus
on
these
three
North
Pacific
Largemouth
ecosystems.
AD
How
are
we
actually
going
to
make
these
forecasts
we're
going
to
use
the
community
or
System
model
seasonal
to
multi
or
large
ensemble?
Briefly,
csmile
works
by
first
creating
a
historical
reconstruction
of
the
ocean
and
sea
ice
component
of
cesm
using
historical
atmosphere
and
terrestrial
forcings.
Here
we're
just
displaying
carbon
in
the
ocean,
the
surface
ocean
over
the
last
20
or
so
years
from
CSM
Fosse.
AD
You
can
see
it's
going
up
as
it
is
in
the
real
world
and
at
various
points
we
initialize
the
model
and
we
stop
giving
the
model
real
world
forcing
and
allow
it
to
split
up
as
a
fully
coupled
or
System
model
run,
and
each
of
these
run
at
the
beginning
of
each
of
these
initializations,
we
slightly
preserve
the
atmospheric
temperature
generating
an
ensemble
20
Ensemble
member.
Basically,
so
we
have
these
20
ensembles,
as
our
forecasts
and
our
most
reliable
forecast
is
going
to
be
the
average
of
all
these
Ensemble
members.
AD
As
with
other
forecasts,
and
then
we
can
take
these
forecasts
and
compare
them
to
observations
right
here,
we're
just
comparing
them
to
observations
and
from
the
fossee
Reconstruction.
But
we
could
also
compare
these
to
real
world
observations
to
determine
how
well
we're
predicting
things
in
the
near-term
future.
AD
So
looking
at
results,
we'll
just
start
by
looking
at
one
example-
and
this
is
SST
forecasts
in
the
Gulf
of
Alaska
on
the
y-axis,
we're
showing
forecast
skill,
just
the
anomaly
correlation
coefficient
or
Pearson's
r
value.
A
high
value
close
to
one
is
a
really
good
forecast.
Value
close
to
zero
is
not
a
very
good
forecast
and
on
the
X,
we're
showing
at
least
different
lead
times.
So
the
further
you
get
to
the
right,
the
farther
you
are
from
initialization.
AD
AD
AD
If
we
move
down
to
the
subsurface
ocean
and
compare
that
you
see,
we
have
slightly
lower
skills
right
at
the
first
moment
of
initialization,
but
over
time
there
really
is
no
skill
degradation
and
you
can
see
out
to
up
to
12
to
13
months.
We
have
really
really
high
skill
forecast
values
that
are
beating
the
persistence
forecast
really
handily
in
the
subsurface
ocean,
so
the
Gulf
of
Alaska
very,
very
exciting
results.
Let's
move
down
to
the
California
Current
California
current
slightly
different.
AD
AD
Finally,
we're
going
to
look
at
our
third
large
Marine
ecosystem,
the
croatio
current,
in
contrast
to
the
Eastern
Pacific,
where
we
have
a
lot
of
skill,
we
find
a
steep
drop-off
in
the
croatio
current
in
the
surface
ocean.
We
have
some
skill
for
temperature
and
carbon,
but
as
we
move
down
to
300
meters,
we
find
almost
no
skill.
AD
So
we've
looked
at
these
time
series,
but
in
a
little
more
digestible
format
what
forecasts
are
actually
useful?
How
many
months
of
these
in
one
year
do
we
actually
have
a
skillful
forecast
so
on
here
I've
shaded,
each
of
these
three
large
marine
ecosystems
that
we're
interested
in
by
how
many
months
we
have
a
skillful
forecast
and
a
skillful
forecast
here
we're
defining
as
an
R
value
above
0.5.
So
it's
just
generally
a
high
R
value
combined
with
beating
or
outperforming
a
persistence
forecast.
AD
So
we
have
these
two
things
that
it
needs
to
be
and
you'll
find
in
the
Western
Pacific
and
the
crochio
current,
except
for
surface
carbon.
We
have
very
very
low
months,
zero
to
one
months
of
forecasts,
skillful
forecasts,
but
as
we
move
into
the
Eastern
Pacific
again,
especially
for
temperature,
we
have
very
very
high
forecast
skill
with
up
to
10
months
of
useful
forecasts
in
each
the
Gulf
of
Alaska
and
the
California
Current,
and
then
for
carbon
and
oxygen.
AD
We
have
skillful
forecasts
up
to
10
months,
especially
in
the
Gulf
of
Alaska,
an
important
thing
to
note
here.
Why
is
the
skill
so
different
in
these
regions?
Well
I've
mentioned,
of
course,
some
of
the
dynamical
representation,
but
we
also
think
some
of
this
has
to
do
with
representations
of
inso
in
the
model,
and
so
it's
relatively
well
represented
and
has
a
lot
of
effects
and
impacts
on
the
Eastern
Pacific
that
it
doesn't
necessarily
have
on
the
Western
Pacific.
AD
So
in
conclusion,
novel
observational
products
are
allowing
us
to
validate
subsurface
predictions
of
biogeochemistry
for
the
very
first
time.
Csm
smile
is
accurately
predicting
surface
and
subsurface
Marine
stressors
in
a
variety
of
large
marine
ecosystems,
and
we
have
up
to
10
months
of
these
useful
forecasts
and
important
Largemouth
ecosystems,
and
with
that
I'll
open
up
any
questions.
We're
about
to
submit
this
paper
to
Earth's
future
and
would
love
any
feedback
thanks.
AD
Sam
any
questions,
please
great
presentation
on
your
previous
slide
in
the
top
right.
Yes,
what
do
you
think
is
going
on
with
oxygen
with
those
occasional
drops
and
then
and
then
it
then
it
goes
right
back
up
yeah.
So
that's
a
really
great
question
and
you
can
sort
of
see
a
pattern
where
it
almost
looks
like
there's
some
sort
of
effective
initialization.
We
have
four
annual
initializations
I,
didn't
really
mention,
but
we're
removing
seasonal
Cycles
from
all
of
our
analysis.
AD
Is
it
the
same
month
of
the
year,
regardless
of
which
ones
you're
starting
with
but
like
if
you,
if
it's
three
months
after
November
leave,
then
that
will
be
in
February
like
yes,
it's
it's
consistent
like
that
for
certain
months
there
is
a
consistent
drop-off
which
is
weird
yeah.
Is
there
any
sort
of
potential
like
bias
in
the
observations
that
are
that
are
present
from
those
months,
and
maybe
they
certainly
could
be
write?
These
observational
products
are
the
best
guests.
AD
AD
Okay,
thank
you.
I
have
another
question:
actually.
Is
there
any
sort
of
like
potential
physical
mechanisms
that
might
be
contributing
to
like
poor
model
skill
in
the
croatio
current
system
like?
Is
it
just?
Is
it
like
a
just
like
more
chaotic
system
physically?
That
could
be
you
know
affecting
it
or
from
my
understanding
it
is
just
the
representation
of
the
western
boundary.
Current
is
maybe
not
the
best
in
CSM,
especially
compared
to
the
eastern
part
of
the
Pacific,
where
we
have
the
upwelling.
That
is
pretty
well
represented
at
CSM
foreign.
AD
AD
I,
don't
you
know
I'm?
That's
my
characterization
and
I'm,
just
I,
just
you
know
open
the
floor
for
topics
or
comments
from
people
in
the
context
of
you
know
the
things
that
we
might
do
as
a
collective
to
advance.
You
know
to
advance
our
capacity
to
to
address
these
kinds
of
research
questions,
so
this
is
just
an
opportunity
to
to
engage
in
that
conversation
as
a
working
group.
I'll.
Just
add
to
that.
Also.
AD
The
thing
that
showed
up
in
every
talk
was
the
connection
between
observations
and
modeling
and
how
we
can
make
that
connection
stronger,
how
we
can
better
compare
observations
and
models
which
I
think
came
up
a
lot
in
the
land
model
working
group
biochemistry
part
this
morning,
whether
it's
by
having
an
observational,
emulator
or
by
other
means
yeah.
That's
a
great
point.
I
left
that
one
out.
AD
AD
Yeah,
okay,
okay,
the
status
of
marble
implementation
in
mom
six
is
ongoing.
I
think
we
can
safely
say
the
majority
of
the
development
work
is
done,
but
we're
sort
of
in
the
process
of
validating
the
port
by
comparing
marble
Solutions
in
mob6
to
those
those
to
a
reference
Solutions
in
in
pop
I'll,
also
say
that
some
recent
work
has
focused
on
enabling
a
single
column
version
of
marble
in
a
single
column.
AD
AD
Thanks
for
that
question,
Ann
Frank
I
just
turned
that
around
and
asked
this
working
group.
You
know
if
there
are
fundamental
limitations
in
the
representation
of
the
physical
state
that
have
been
holding
up
progress
in
BGC
that
you
know.
Please
engage
with
those
of
us
in
the
ocean
model
working
group
so
that
we
can
work
towards
a
resolution
of
those
problems,
especially
moving
towards
the
csm3
release,
we're
still
at
a
position
where
we
can
make
some
fairly
significant
changes
in
our
resolutions.
Configurations,
Etc
and
mom
gives
us
a
lot
more
flexibility
in
that
regard.
AD
Yeah,
that's
a
great
Point
go
ahead.
Yeah
I
have
a
question.
Following
up
on
that,
so
I
know
there
was
sort
of
I
think
if
it
was
forget
if
it
was
with
cvip6,
but
with
the
cesm2
model
there
were
some
like
subsurface
oxygen
issues
like
deep
oxygen
issues.
Have
those
been
resolved?
What's
the
status
of
that
they
have
not
been
resolved.
Okay,
that
model
still
simulates,
very
anemic,
North,
Pacific,
ventilation
and
I-
don't
know
Keith.
Do
you
want
to
talk
about
those
or
it's
too
painful.
AD
AD
Seeing
those
types
of
things
that's
still
in
the
sort
of
like
software,
coupling
phase
or
yeah
I,
don't
think
we
know
enough
about
the
circulation
of
the
coupled
model
with
Mom
to
know.
If
we
would
be
experiencing
similar
problems.
Okay.
Well,
we
do.
We
do
have
some
longer
runs
and
we
have
ideal
age.
We
have
CFCs,
we
have
other
tracers
that,
hopefully,
would
be
indicative
of
some
of
the
issues
we've
seen
in
part.
The
problem
is,
we
have
a
very
small
group
of
people
working
on
Mom,
six
development.
AD
We
need
more
eyes
looking
at
these
other
things,
something
you're
interested
in.
We
welcome
your
input
looking
and
help
and
looking
at
the
whole
spectrum
of
results,
we
have
yeah
so
I
think
historically,
some
of
the
key
biases
have
been
associated
with
thermocline
ventilation
and
oxygen,
minimum
Zone
ventilation
and
then
deep
ocean
Abyssal
flows
and
yeah.
We
need
to
be
evaluating
these
simulations
on
those
metrics.
We
have
tried
to
pull
these
tracers
into
the
model
development
process
at
an
earlier
stage,
so
we
don't
get
to
a
release
and
find
out.
AD
How
are
the
mixed
layer,
depth
biases,
looking
mom
said,
I'm
sick
Gustavo
will
show
that
tomorrow,
okay,
okay,
like
one
word,
yeah,
okay,
okay,
thank
you
again
too
deep,
yeah
wow!
Please
go
you
want
to
go
ahead.
Ask
your
question:
I
was
just
gonna
say
you
know!
For
the
thermocline
ventilation,
you
don't
have
to
wait
for
a
BGC
run.
If
you
know
you
could
just
look
at
the
CFCs
in
the
circulation
model,
if
we
can
match
those,
the
oxygen
will
will
look
good
yeah
agreed.
AD
Before
lunch,
which
is
from
this
community,
maybe
not
everyone
was
sitting
in
the
room
before
lunch
of
thoughts
about
what
would
be
useful
types
of
emission-driven
scenarios,
particularly
potentially
looking
at
climate
mitigation
type
scenarios.
What
would
this
community
find
interesting
in
that
space?
What
would
we
be
learning
from
them?
What
should
those
scenarios?
Look
like
good,
Nicola
2300
would
be
nice.
That's
that's
all.
AD
That's
going
to
be
mine,
well,
yeah,
longer
yeah,
just
generally
sort
of
like
yeah
I
would
like
to
see
emissions
driven
once
because
I
think
in
a
lot
of
ways.
Air
system
models
in
general,
especially
cesm,
are
probably
going
to
get
involved
with
a
lot
of
research
being
done
with
with
sort
of
nature-based
climate
Solutions,
and
things
like
that
and
I
think
emissions
based
scenarios
are
really
going
to
be
key
in
those.