►
From YouTube: CESM Workshop: Land Model Working Group
Description
The 26th Annual CESM Workshop will be a virtual workshop with a modified schedule on its already scheduled date. Specifically, the virtual Workshop will begin with a full-day schedule on 14 June 2021 with presentations on the state of the CESM; by the award recipients; and three invited speakers in the morning, followed by order 15-minute highlight and progress presentations from each of the CESM Working Groups (WG) in the afternoon.
On 15-17 June 2021, working groups and cross working groups have half-day sessions, some with presentations and some that are discussion only.
B
B
B
B
B
B
Yeah,
probably
just
the
pdf,
the
one
that's
linked
on
the
website,
thanks,
we
can
go
and
get
started,
and
I'd
like
to
welcome
all
of
you
to
the
land
model
working
group
session
at
the
26th
annual
cesm
workshop
and
kudos
to
dave
for
passing
the
torch
on
being
co-chair,
because
this
is
my
second
workshop
as
co-chair
and
second
virtual
one.
I
look
forward
to
getting
to
see
you
all
in
person,
hopefully
before
too
long
and
then,
as
our
meeting
gets
started.
B
Just
please
remember
the
code
of
conduct
that
we
all
agreed
to
when
registering
for
the
meeting.
That's
shown
here
on
the
screen,
especially
in
the
discussion
session,
I'll
kind
of
go
over
this
again,
but
but
make
sure
that
we
offer
constructive
feedback
and
share
the
air
and
encourage
kind
of
discussion
as
we
as
we
talked
about
some
of
these
ideas
so,
along
with
rosie
fisher
who's.
B
The
new
land
model
working
group
external
co-chair,
I'd
like
to
to
present
some
of
the
work
that
we've
been
doing
as
a
larger
group
before
moving
on
to
the
larger
agenda,
and
so
today
we'll
have
a
conversation
about
kind
of
where
we
are
and
where
we're
heading
as
a
land
model
working
group.
Until
about
noon
this
afternoon,
jackie
schuman
and
others
will
be
having
a
discussion
on
fire
in
cesm
which
I'd
encourage
you
all
to
go
to
and
then
from
3
20
till
almost
5
o'clock.
B
Colorado
time,
we'll
be
also
having
a
virtual
poster
session.
So
there's
roughly
10
posters
that
were
submitted
to
the
land
model.
Working
group
and
that'll
be
a
virtual
session
where
you
can
kind
of
drop
into
a
breakout
room
and
talk
with
people
kind
of
in
a
more
intimate
setting
about
the
work
they're
doing.
B
So
tomorrow
from
8
to
8
30,
there
will
be
a
zoom
link
sent
out
and
again
similar
to
the
the
land
model.
Working
group
meeting
this
this
spring
will
just
be
an
opportunity
to
drop
into
breakout
rooms
and
talk
with
whoever's
there.
So
if
there's
somebody
that
you're
dying
to
to
chat
with
maybe
send
them
an
email
I'll
be
there,
a
number
of
us
from
dss
will
be
there.
B
B
B
Bug
fixes
some
major
and
minor
grip
code
changes
and
then
some
features
that
that
I'll
be
highlighting
today,
a
lot
of
what's
gone
into
the
cpsm51
release
is
now
being
used
for
the
ppe,
the
parameter,
perturbation
experiment
that
I'll
also
spend
a
little
bit
of
time
talking
about
which
is
kind
of
the
precursor
to
what
will
be.
You
know,
cts,
m52
and
then
eventually,
cts
m6,
so
kind
of
to.
B
B
So
one
of
the
highlights
that
I
want
to
talk
about
is
the
biomass
heat
storage
additions
that
sean
swenson
has
added
to
cts
m51
and
the
the
feature
here
that's
relevant
both
for
the
coupled
climate
and
for
the
land
model
is
that
by
warming
up
vegetation,
the
air
temperature
is
cooler.
So
this
is
just
a
plot
of
two
meter
annual
air
temperature.
B
B
B
So
there's
a
lot
of
work
to
be
done
here.
There's
some
aside
from
adding
the
feature
of
biomass
heat
storage.
This
this
code
change
has
also
removed
the
stability
cap
that
was
used
before
so
so
some
more
investigations
kind
of
warranted,
especially
in
the
couple
of
models
to
see
what
the
effects
of
biomasses
storage
are
on
daily
temperature
ranges,
as
well
as
seasonal
temperatures
in
different
regions.
B
B
The
plot
is
a
little
granular,
but
it's
pretty
clear
to
see
that
you
get
a
better
match
to
observations
for
lai
with
these
architectural
phenology
changes
that
leah
burch
added
that
has
impacts
on
on
carbon
fluxes
so
that
ecosystem
exchange
has
a
higher
peak-to-trough
amplitude,
given
this
hell
a
higher
lai
in
the
region.
So
thanks
to
leah
for
these
contributions
and
then
to
keith
for
providing
the
simulations.
B
Another
feature:
that's
in
cts151:
that's
really
critical
for
being
able
to
do
the
parameter.
Perturbation
experiment
is
the
matrix,
and
so
chris
luo
and
chris
liu
and
ichi
luo
were
the
two
contributors
who
added
this
and
then
eric
klusek
did
a
ton
of
work
to
kind
of
work
with
chris
and
nietzsche
up
until
even
yesterday,
there's
been
challenges
or
not
challenges.
Small
issues
with
the
matrix
we've
had
to
figure
out,
but
the
take
home
here
is
that
spin
up
for
bgc
is
about
10
times
faster,
using
this
matrix
approach.
B
B
So
some
of
the
ongoing
efforts
that
we
have
you
know
there's
a
lot
here
and
I'll
come
back
to
this
slide.
There's
efforts
to
look
at
change,
dust
and
bboc
emissions
in
the
atmosphere.
Hydrology
naoki
is
going
to
present
about
miser
out
in
a
little
bit
and
shawn's,
also
going
to
talk
about
representative
hill
slopes.
B
So
the
ppe
I'm
going
to
kind
of
go
through
this
quickly.
The
the
big
question
here
is
kind
of
trying
to
look
at
parametric
for
structural
uncertainty
in
the
model,
because
we
know
that
the
model
entered
d
general
is
high,
especially
for
carbon
flexes
in
land
models,
and
so
the
goals
for
the
ppe
are
really
three-fold.
They're
to
complete
comprehensive
parameter,
uncertainty
assessment
and
calibration
with
cts
m51.
B
Forest
hoffman
has
helped
us
create
a
sparse
grid
to
to
to
kind
of
reduce
the
number
of
points
that
we're
trying
to
simulate,
and
then
there's
been
a
bunch
of
scripting
infrastructure
to
try
to
automate
this
process
so
phase
one
is
done
now,
which
is
kind
of
these
one.
At
a
time
simulations
checking
for
reasonableness
and
sensitivity
of
the
parameters.
B
So
this
is
just
a
quick
plot
that
daniel
provided
of
the
kind
of
top
ten
parameter
effects,
maybe
top
eight
parameter
effects
on
mean
global
gpp.
So
things
like
photosynthesis,
not
surprising
jv
max,
are
important,
both
for
the
mean
and
the
inner
annual
variability.
You
can
see
that
some
things
are
more
important
for
the
some
parameters
are
more
important
for
the
variability
than
they
are
the
global
mean
so
kudos
to
daniel
and
katie
for
all
the
work
that
they've
done
here,
and
it
should
be
pretty
exciting
to
see
this
moving
forward.
B
Another
talk
that
we'll
hear
from
today
is
from
meg
fowler
who's
working
with
scam.
I'm
not
going
to
say
a
lot
more
about
that
here,
other
than
there's
some
new
capabilities
that
are
enabling
scam
to
run
the
ctsm,
restart
files
to
get
initial
conditions,
and
then
fates.
There's
also
that
you
know
the
fates
community
is
large
and
and
ever
growing,
and
it's
exciting
to
to
kind
of
have
this
coupling
between
ctsm
and
fates
happening
in
more
real
time
and
and
closer.
B
So
again,
this
is
largely
largely
credits
to
to
greg
and
ryan
the
software
engineers
who
we've
been
working
closely
with
as
well.
As
you
know,
the
rest
of
the
kind
of
fates
leadership,
team,
rosie,
jackie
charlie
and
then
also
adriana
foster,
is
going
to
be
a
new
project.
Scientist
we'll
be
moderating
the
discussion
later
this
morning
and
joining
us
at
tss
later
on
this
summer.
B
But
the
idea
here
is
to
be
able
to
kind
of
turn
on
and
off
different
features
within
fates
to
isolate
you
know:
what's
the
effect
of
competition
in
the
model,
what's
the
effect
of
just
you
know,
photosynthesis
alone
in
in
in
these
simulations,
so
we'll
hear
we'll
hear
a
lot
more
about
fates.
Moving
forward,
we'll
also
hear
a
bunch
from
sean
talking
about
the
hill
slope
model.
B
So
I'm
not
gonna
go
through
a
ton
here,
so
I'm
just
gonna
go
ahead
and
skip
the
slide
just
to
say
that
it's
it's
exciting
to
see
that
there's.
This
kind
of
our
ability
to
capture
heterogeneity
across
the
land
surface
is
kind
of
greatly
increasing,
and
so
what's
shown
here
is
from
lowland
to
upland
columns
in
this
idealized
case
and
then
north
and
south
facing
hill
slopes
within
a
grid
cell.
That
would,
you
know,
be
flat
effectively
in
clm
5.
B
with
the
hill
slope
model.
You
get
these
features
of
high
snow
and
low
snow
areas
based
on
elevation
and
aspect,
so
high
elevation
and
north
facing
slopes
tend
to
have
more
snow
than
low
elevation
and
south
facing
slopes
in
in
the
model,
which
then
translates
down
into
into
soil,
moisture
conditions.
B
B
So
this
will
be
within
cesm
lab,
which
is
just
a
containerized
version
of
cesm
kind
of
with
a
jupiter
that
has
jupiter
notebook,
enabled
and
a
pengio
stack
of
analysis
of
of
tools
to
help
us
with
the
model
evaluation
and
also-
and-
and
I
think
this
will
kind
of
change-
the
way
that
we're
able
to
do
things
like
tutorials
and
and
kind
of
get
a
new
community
of
people
using
ctsm
simulations.
So
I'm
happy
to
talk
about
that.
We'll
have
a
little
bit
of
time
at
the
end
here
and
then.
B
Lastly,
I
don't
know
how
many
of
you
were
able
to
to
stick
it
out
until
almost
five
yesterday,
but
I
do
want
to
encourage
the
working
group
to
start
thinking
about
using
the
ensembles
of
model
runs
that
are
being
produced
now,
and
so
this
is
just
a
teaser
for
looking
at
snow
water,
equivalent
and
runoff
for
a
grid
cell.
That
includes
places
like
crested,
butte
and
aspen,
and
I
found
this
plot
helpful
this
winter
when
I
was
debating.
B
It's
like
yep,
you
should
go
because
next,
you
know
next
snow
is
is
decreasing
over
time
and
so
so
daniel
and
I
have
been
and
have
been
working
with
these
quite
a
bit,
and
we
have
some
scripts
that
we
can
make
available
if
people
are
interested
in
and
so
the
cesm2
large
ensemble
is
a
hundred
member
ensemble.
It
goes
from
1850
to
2100
under
ssp370
and
there's
a
lot
of
high
frequency
output.
So
what's
shown
here
is
actually
daily
sui
and
and
runoff
from
the
model.
B
There's
also
the
smile
ensemble
that
yaga
talked
about
yesterday,
which
is
a
series
of
two
years
initialized
forecasts
that
start
at
four
different
times
during
the
year
right
now
they
go
from
about
1960
through
2019,
but
they
want
to
get
those
up
through
kind
of
present
day
as
soon
as
possible,
and
then
lastly,
they'll
there
will
be
a
new
cesm
cmip5,
forcing
and
so
with.
The
combination
of
the
ces
m1,
large
ensemble,
cesm2,
large
ensemble,
and
then
the
csm2
cmip5
ensemble.
B
It
will
let
us
start
looking
at
both
forcing
uncertainty
and
model
n
certain
d
across
these
generations
of
ces
m2.
So
again,
just
encouraging
members,
yep
you're
good
for
two
minutes.
It's
time,
almost
all
right,
good
I'll
finish
with
this
slide
again,
and
I
think
we
probably
won't
have
a
lot
of
time,
because
I
want
to
be
able
to
move
on
to
the
kind
of
set
the
stage
for
the
discussion.
If
you
have
questions
for
me,
please
put
them
in
the
chat
and
I'll
get
to
them
at
some
point.
B
F
F
So
here
I
just
plotted-
or
I
included
an
example
of
agricultural,
nitrogen
fertilization
and
cesm2,
and
I'm
illustrating
here
that
it
increases
the
seasonal
carbon
dioxide
cycle,
and
so
this
is
just
a
direct
effect
of
agricultural
management
on
the
carbon
cycle,
and
so
this
is
part
of
the
reason
why
agriculture
is
really
important
component
of
the
earth
system.
So
in
the
next
few
minutes,
I'll
highlight
a
few
updates
about
the
agriculture
model
because
they
know
that
I've
been
presenting
on
what
the
agriculture
model
can
do
for
many
years
now.
F
F
F
F
The
next
thing
that
I
want
to
highlight
is
that
we
do
have
a
few
new
crop
types,
so
we
made
two
new
biofuel
crop
types
miscanthus
and
switch
grass.
These
were
developed
by
janna
and
ten.
I
also
wanted
to
highlight-
because
I
don't
think
I've
done
this
well
in
the
past-
that
we
do
have
winter
wheat
code
available,
it's
just
with
winter,
wheat
and
also
with
miscanthus
and
switch
grass.
F
We
don't
have
spatial
distributions
figured
out
very
well
for
those
crop
types,
and
so,
if
you
want
to
use
any
of
these
crop
types,
the
code
is
available,
but
the
distributions
are
not
yet
on
the
land,
use
and
land
surface
data
sets.
So,
but
we
do,
we
do
have
more
new
crop
tips
available,
which
is
exciting
next
slide.
F
So
one
of
the
bigger
developments
that
we're
working
on
is
this
clm
access
and
the
lead
developer
here
is
bin
peng,
and
this
is
a
new
module
for
agriculture
and
clm.
It
focuses
on
updating
crop,
phenology
and
allocation,
so
the
way
that
it's
set
up
right
now
is
that
it's
it's
modularized,
so
that
there's
you
know
what
we're
calling
here
a
host
land
model.
This
is
a
diagram
of
clm
and
then
the
axis.
F
What
axis
crop
model
which
is
effectively
apsim
crop
model
put
into
clm,
is
just
the
the
different
phenol
phenology
phases
and
the
different
allocation
to
different
plant
organs,
so
that
modular
structure
allows
it
to
be
connected
to
different
host
models,
and
I'm
really
hoping
that
we
can
get
this
functioning.
You
know
it's
we're
on
our
way
with
clm,
but
I'm
hoping
we
can
get
it
functioning
with
fates
at
some
point
and
so
that's
a
hook.
That
needs
to
happen
next
slide.
F
Some
of
the
benefits
of
clm
access
is,
it
adds
a
lot
of
unique
new
features,
including
a
representation
of
over
30
crop
types.
We
also
can
simulate
cultivars
through
changes
to
crop
parameters,
crops
crop
type,
specific
parameters.
It
allows
for
flexible
and
realistic
phenology,
basically
allowing
phenology
phases
to
vary
by
crop
types.
So
there's
no
set
number
of
phase
like
there
currently
is
in
clm.
F
Okay,
next
slide
you're
at
five.
Just
to
give
you
an
update,
okay,
progress.
So
far,
you
know:
we've
implemented
some
of
these
things
and
we're
currently
working
to
hook
it
to
to
connect
to
clm.
Okay,
next
slide
I'll
just
go
through
these
real
quickly.
We
have
some
updates
in
irrigation
management,
so
some
new
application
methods
being
developed
and
some
multiple
water
sources
next
slide,
there's
also
tillage
and
residue
management
that
has
been
developed.
I
haven't
pulled
this
over
into
clm,
but
the
code
does
exist
so
there's
there.
F
F
D
G
D
G
G
G
G
So
we
look
into
major
land
use.
Land
covers
such
as
agricultural
pasture
grazing
and
secondary
vegetation
cover
for
agriculture.
We
have
three
categories:
conventional
tillage,
no-till
and
animal
plowing,
and
for
grazing
we
have
light
median
and
heavy
intensities.
So
in
doing
so
we
will
be
able
to
apply
different
language
impacts
through
the
human
history.
G
G
So
after
constructing
these
models,
we
now
are
able
to
apply
them
globally,
using
available
global
data
sets
and
the
environment
mental
variables
then
based
on
the
land
you
stand
covered
and
so
relations
that
your
hand
I
created
for
the
past
1
000
years.
We
will
be
able
to
apply
varying
language
and
cover
impacts
to
four
types
of
soils
and
estimate
the
soil
property
change
over
time.
G
So
we
look
into
each
soil
poverty
individually,
because
those
are
generally
are
the
input
for
the
earth
system
models.
So
with
those
parameters,
they
will
be
able
to
calculate
the
hydraulic
properties
of
the
soil,
and
then
we
can
create
a
time
series
of
global
soil
properties
how
those
properties
change
over
time
to
be
read
into
our
system
models.
G
So
here
to
do
a
simple
demonstration:
we
compute
the
soil
water
holding
capacity
using
petal
transfer,
functions,
use,
base,
angles,
soil
property
changes,
and
then
we
can
compute,
so
your
water
cooling
capacity
under
undisturbed
soil
and
and
disturbed
conditions.
So
the
bottom
figure
shows
the
difference
between
disturbed
and
undisturbed
soil.
So
you
can
see
a
large
difference
in
water
holding
capacity
change
in
the
north,
europe
and
north
asia
and
west
indi
indonesia,
so
most
of
them
are
due
to
soil
compaction.
G
G
G
So
this
is
just
a
simple
test
and
we
already
can
see
an
impact
so
the
net
the
future
work.
We
hope
to
increase
our
erosion
in
these
estimates
and
we
want
to
after
accessing
the
source,
thinking
exchange.
We
want
to
create
a
time
series
of
soil
property
in
the
past
1000
years.
Then
we
can
test
it
in
our
cmc
model,
hopefully-
and
maybe
include
a
few
years
to
test
the
impact
and
first
just
to
see
how
much
your
work
caught.
You
would
change
the
hydraulic
water
cycle
and
energy
cycle
and.
G
D
F
Yeah,
I
guess
that's
really
interesting,
so
I've
been
working
on
some
a
disturbance,
synthesis
for
the
boreal
region
and
where
you
know
we
have
some
colleagues
who
are
working
with
seismic
lines
and
oil
and
gas
wells
extraction
in
like
canada,
mostly
and
they're,
actually
very
impactful
on
soil
and
hydrology
and
they're
pretty
extensive,
at
least
in
canada.
So
I'm
wondering
if
that's
something
you're
you
thought
about
at
all.
H
Yeah
I
apologize
my
dog
has
somehow
gotten
outside
and
is
barking
but
anyway
good
morning.
Everyone
just
on
the
theme
of
how
do
we
maintain
food
production
while
securing
ecosystem
services
in
a
changing
world.
I
thought
I'd
talk
about
a
few
different
activities
that
are
connected
to
this
and
hopes
that
in
in
the
future,
we
could
work
toward
connecting
some
of
these
thoughts.
H
So
next
slide
for
about
the
past
10
years
have
been
well
even
more,
been
doing
work
with
water
management
groups,
focusing
both
on
prediction
and
long
long
range
production
and
there's
tremendous
interest
in
this.
In
in
the
kind
of
scientific
services
we
can
get
through
better
models,
data
and
methods.
You
know
whether
it's
in
small
places
like
this
little
diversion
dam
up
in
the
in
the
in
the
bottom
right
or
in
lake
mead,
and
this
is
work
that
other
people
at
ncar,
such
as
flavio,
have
engaged
in
as
well
next
slide.
H
H
We
also
have
also
done
work
at
the
global
level
where
the
world
meteorological
organization
wmo
is
trying
to
stand
up
a
global
monitoring
system,
monitoring
and
prediction
system
using
model
outputs
for
hydrology
and
climate,
and
this
is
mostly
working
with
climate
model,
forecast
groups
next
slide
and
then
so
on
the
longer
range.
We
also
have
done
work
over
many
years
using
these
models,
these
land
models
and
hydrology
models
to
project
water
security
into
the
future.
H
A
notable
feature
of
all
this
work,
however,
is
that
we
don't
use
hydrology
fields
from
the
land
models
in
coupled
modeling
like
cesm.
Instead,
we
use
the
climate
model
outputs
and
we
take
it
through
a
certain
cascade
of
methods.
That's
often
somewhat
arduous,
taking
climate
model,
outputs
and
then
downscaling
them
with
mesoscale,
atmospheric
or
regional
models,
putting
them
through
calibrated
offline
hydrology
models
to
get
the
projection
information.
This
is
starting
to
change
a
little
bit.
H
This
example
on
the
left
of
a
paper
with
blog
you
later,
where
we
analyze
the
runoff
response
and
cement
models
at
the
regional
scale.
So
this
is
for
the
upper
colorado
basin
and
if
the
black
line
is
kind
of
the
observed
precipitation
to
runoff
relationship,
you
can
see
that
in
the
cmip
class
models,
the
sensitivities
of
runoff
to
rainfall
are
all
over
the
map
and
on
the
right,
you
have
the
typical
kind
of
land
model
evaluation
through
ilam,
where
you
know
runoff
and
other
hydrologic
variables
are
scrutinized,
but
they're.
H
Just
a
small
part
of
a
much
larger
analysis
that
focuses
more
on
fluxes
to
the
atmosphere,
whether
heat
or
moisture
next
slide-
and
this
is
my
final
slide.
So
I
think
we
have
and
we'll
hear
about
in
this
workshop,
a
lot
of
activities
that
will
help
us
move
toward
maybe
connecting
these
dots
of
the
work
that
we
do
for
water
agencies
on
their
water
security
problems,
with
the
work
that
we
do
with
ctsm
and
cesm
in
the
earth
system.
H
Science
context
we'll
hear
about
the
hill
slopes,
which
are
get
down
toward
finer
scale,
hydrologic
sensitivities
and
the
parameter
estimation
work
in
parameter,
sensitivity,
work,
I
think
also
the
smile
predictions
are
are
probably
a
nice
ingredient,
but
the
idea
is
that
if
we
can
start
to
inch
toward
connecting
some
of
these
dots,
we
may
be
able
to
need
collect
again.
It
looks
like
there's.
H
D
D
H
Yeah,
the
motivation
for
this
project
is
they're,
starting
that
there's
starting
to
be
papers
in
which
people
are
taking
runoff
in
other
fields
directly
from
couple
dsms,
often
with
fairly,
I
guess,
mediocre
validation
of
those
fields
and
then
writing
papers
discussing
water
security
at
the
regional
scale,
whether
for
california
or
miss,
mississippi
river
basin.
I
mean
these
aren't
small
small
watersheds,
but
still
so
flavio,
and
I
looked
at
these
these
land
surface
sensitivities
to
see
if
we,
if
they
would
be
credible
for
projecting
change
into
the
future.
H
D
I
I
I
So
this
shows
the
difference
between
river
victoria
and
network
and
then
grid
network.
You
can
see
very
different
same
locations
and
the
blue
is
reverse
segments
and
then
you
can.
Hopefully
you
can
see
blockchain
lines,
the
catchments,
so
it's
you
can
notice
using
a
very
few
few,
a
number
of
elements
you
can
depict
accurate,
like
a
liver,
realistic,
liver
drainage
system
in
a
vector
event
compared
to
greedy,
so
that'll
make
it
easier
to
connect
river
segments
and
catchment
to
other
hydraulic
features,
regulates.
I
Next,
please,
so
we
completely
implement
implementing
river
miserables
in
the
ctsm
in
the
one-way
coupling
mode.
Yes,
and
then
we
prepare
some
two
global
scale.
River
network
data
sets
hdmi,
which
is
cost
resolutions
and
merit.
Hydro
is
very
fine
resolution
and
I
also
need
we
needed
to
pre-compute
like
esmfs.
I
We
greeted
weight
data
for
each
network
data
with
a
couple
of
random
model
grids
and
then,
in
terms
of
river
miserable
development.
We
completed
implementing
like
a
parallel
computing
scheme
and
they
also
implemented
like
routing
scheme
for
natural
lake
and
also
we
can
now
extract
or
inject
prescribed
water
from
each
river
segments
or
lake.
I
Yes,
please
so
I
started
some
evaluating
some
miserables,
mostly
compared
to
mozart,
and
the
left
is
showing
a
time
series.
I
mean
seasonally
30
years
means
for
six
big
river
in
the
world
and
then
blue
is
muzzled.
Unembed
on
the
orange
is
misroad
running
on
the
merit
title,
so
miserable
has
two
routing
physics,
so
showing
three
lines
so
and
then
right
is
showing
some
comparison
with
dye
data,
so
die
data
has
a
drainage,
drainage
area
and
then
discharge
monthly
scale.
I
I
So
I
I
started
some
testing
some
lake
routine
and
then
just
showing
like
a
annual
phone
and
seasonality
for
three,
a
three
big
river
blue.
This
is
only
miserable,
so
red
is
running
running
in
israel
and
then
also
turning
on
a
lake
and
then
blue
is
on
the
river.
So
you
can
see
seasonality
you
can
pick.
I
I
So
another
thing
is:
we
are
actively
working
on
the
river
management
scheme
in
the
missed
routes.
The
biggest
challenge
may
be
the
showing
right
in
the
picture,
so
irrigated
area
which
is
coming
from
crm
and
then
and
then
late
that
surprise.
Water
is
not
located
same
place,
so
we
have
to
somehow
we
have
to
collect
irrigated
area
and
entire
area
and
objectively
so
and
also
working
on
a
somewhere
like
visible
model
itself.
So
and
then
next,
okay,
so
that's
my
end,
so
I
I'd
like
to
acknowledge
some
many
collaborators
who
actually
doing
work.
B
B
I
D
J
I
I
Oh
okay,
so
yeah.
So
what
I'm
showing
here
is
just
like
a
natural
lake.
So
it's
a
very
simple
equation:
it's
an
energy
coming.
Some
discharge
is
based
on
how
much
water
in
the
rates
that
make
it
make
it
to
some
suppressed
peak
flow.
But
yes,
if
you,
if
you
include
hydro
rakes,
I
mean
I
don't
know
how
many
amazon
ones
in
particular
basin,
but
it's
it's
gonna
be
usually
like.
If
you
include
rate,
it's
picture
is
more
like
flow
is
attenuated.
A
A
A
When
we
combine
these
two
impact,
if
we
compare
with
other
similar
5
cmf6
model,
you
see
there
is
a
variety
of
the
range
even
in
terms
of
the
sign.
They
do
not
agree
on
the
water
availability
site.
When
we
go
to
the
leaf
area
index
that
is
shown
on
the
bottom
panel,
you
see
the
clm
has
higher
sensitivity
in
terms
of
the
response
of
the
lai.
Compared
to
the
other
model,
I
know
will
has
done
some
work,
so
he
will
elaborate
it
later
on
we'll
go
to
the
next
slide.
A
Please,
okay,
so
in
terms
of
this
is
coming
back
to
to
a
different
problem.
What
we're
finding
a
key
finding
here
is:
the
groundwater
is
declining,
even
in
the
in
the
the
wet
region
like
a
southeastern
united
states.
This
is
a
work
that
I
did
with
with
the
geological
science
collaboration
and
what
was
surprised
to
me
was
you
see
the
irrigation
in
the
left-hand
plot.
Those
are
dot,
dot,
dress
on
with
a
different
saturation
density
and
those
irrigation
extent
are
not
determined
by
the
climate,
but
what
is
happening
underground?
A
What
is
how
the
equifax
distributed
and
which
aquifer
are
more
productive
compared
to
the
other
equipment?
One
thing
I
learned
there
is
a
fall
line
and
there's
no
no
irrigation
above
the
fault
line,
because
there
is
no
productivity
for
of
the
fall
line,
so
what's
happening
underground
and
how
we
represent.
Those
dynamics
are
really
important.
A
Okay,
we
can
go
to
the
next
slide
going
on
the
sensory
time
scale.
How
does
these
kind
of
mega
drought
or
drought
risk
are
changing,
so
we
have
done
a
study,
we're
using
the
csm
large
ensemble
and
what
we
are
finding
that
enso
influence
is
increasing.
That
is
shown
by
the
one
of
those
two
green
bar
here
as
we
go
into
the
future.
But
what
is
contrastingly
also
effect
happening
is
the
landscape
is
becoming
drier.
A
The
soil
moisture
memory
is
decreasing
that
is
shown
by
this
yeah
second
green
line,
but
when
you
combine
both
soil
moisture
memory
and
then
so
effects,
you
see
a
greater
signal
to
noise
ratio,
but
a
effect
of
the
decreasing
soil.
Moisture
memory
is
shown
in
this
kind
of
frequency
distribution.
So
what
we
are
finding
that
some
of
the
low
frequency
variability
is
increasing.
You
see
the
power
spectra
with
five
five
different
30
years
period
from
blue
to
red,
going
into
the
future,
and
you
see,
but
there
are
more
power.
A
Those
are
coming
in
the
higher
frequencies,
so
water
is
becoming
more
variable,
whereas
low
frequency
power
is
decreasing.
That
means
there
is.
There
is
likely
to
be
less
mega
drought,
but
there
is
likely
to
be
more
shorter
term
drought.
We
know
there
are
uncertainty
with
respect
to
model
enzo
projection,
even
with
respect
to
the
soil
and
memory
yeah.
When
we
compare
two
different
model
it
it
looks
very
different.
So
how
do
we
represent
those
uncertainty
into
the
future
climate
and
we'll
go
to
my
last
slide?
A
Okay,
so
how
do
we
address
all
these
challenges
going
into
the
future?
So
this
is
my
take.
One
is
multi-disciplinary
collaboration,
so
we
will
already
talked
about
that
trend
card
is
developing
collaboration
with
the
neon.
I'm
very
excited
about
that
collaboration
model
development
and
car
is
already
leader
in
developing
the
community
earth
system
model
we
which
continue
developing
and
putting
the
resources
in
the
model.
Development
model
development
is
a
key.
That's
how
one
scientist
talk
with
each
other
and
finally,
the
machine
learning.
A
This
is
a
new
area
and
people
different
scientist
has
different
thought.
This
is
my
take
it's
building
the
machine
learning
in
a
complementary
approach
where
let
climate
model
do
their
best.
We
should
continue
developing
a
better
model
and
let
big
data
to
the
rest
means
there
are
few
things
that
at
the
climate
model,
it's
very
difficult
to
understand.
One
of
the
thing
illustrated
in
this
bottom
final
is
the
bottom.
Left
final:
is
the
uncertainty
in
the
national
water
model
forecast?
So
this
is
very
small
reason.
A
I
picked
up
alabama
and
some
part
of
the
georgia
and
vs
the
uncertainty
you
see.
There
are
different
colors
that
are
representing
different
uncertainty
and
we
can
do
few
things
like
we
can
categorize
these
uncertainties
for
the
land
use
change
category
and
yes,
we
see
like
if
we
developed
area
is
showing
more
uncertainty
compared
to
the
agriculture
and
world,
but
there
can
be
many
other
structure
that
as
a
land
modeler
or
as
a
kind
of
we
develop
the
model
for
single
column
which,
if
we
try
to
understand
those,
are
little
beyond.
A
Like
one
person
can
do
it,
so
can
the
big
data
help
there
to
to
understand
or
model
those
uncertainty
and
improve
the
forecast
assessment
or
the
uncertainty
in
the
forecast
at
the
engage
basin?
That's
one
approach,
or
one
way
to
engage
the
machine
learning
part
into
what
we
do
the
water
forecast.
So
that's
it.
D
D
C
D
I
C
M
M
M
Okay,
so
I
guess
I'm
not
going
to
be
talking
about
model
developments,
but
how
we
can
sort
of
leverage
these
model
developments
to
answer
particular
scientific
questions
and
focusing
give
some
a
couple
examples
about
maintaining
ecosystem
services
and
how
much
ecological
resolution
do.
We
really
need
to
address
which
particular
questions,
and
it
will
definitely
be
a
forest
biased
set
of
questions
all
right
next
slide.
M
So
I
think
we
all
probably
recognize
that
it's
not
just
climate
change,
there's
a
lot
of
social
change.
There's
land
use,
change,
there's
adaptation
and
management
mitigation
efforts
going
undergoing
and
carrying
capacity
of
ecosystems
and
the
entire
planet
is
changing,
as
well
as
the
desired
conditions.
What
are
we
really
aiming
for?
What
do
we
want?
So
I
think
we
have
to
keep
all
of
these
things
in
in
mind
as
we're
applying
the
models
to
answer
questions
next
slide.
M
M
But
the
next
slide,
please
there's
a
lot
of
missing
processes
and
missing
potential
outcomes.
And
how
important
are
these?
I
don't
know
because
we
haven't
redone
this
analysis
where
these
are
incorporated.
So,
for
example,
insect
outbreaks
which
are
a
major
source
of
mortality
in
the
western
u.s
were
not
incorporated
in
this
or
nor
were
pft
specific
disturbance
effects
other
than
their
response
to
drought.
But
response
to
fire
was,
you
know
there
were
really
no
differences
across
pft.
M
M
You
know
things
have
no
trouble
regrowing
in
clm,
and
so
we
know
that
that's
becoming
a
really
critical
point
for
forests
is
as
climate's
changing
and
there's
large-scale
disturbance.
What
is
really
how
how
successful
will
the
regeneration
be,
which
leads
to
well?
We
have
novel
ecosystems
and
so
running
with
static.
You
know
with
no
without
dynamic
pfts.
M
We
could
really
miss
some
really
big
changes
that
are
coming
down
the
pike,
and
so
these
questions
point
some
of
them
point
to
either
needing
different
implementations
in
clm
or
needing
to
use
fades.
So
next
slide.
Please-
and
I
think,
if
we're
interested
in
looking
at
how
adaptation
is
going
to
influence
forests
in
particular,
we
really
do
need
vegetation
demography,
and
this
is
just
showing
an
example
from
the
nature
conservancy
about
how
fire
may
affect
a
forest.
M
M
D
M
Well,
I
I
guess
just
you
know,
bringing
in
turning
on
hydro
as
well.
You
know
bringing
that
into
this
mix
of
reduced
complexity
is
another
really
promising,
and
then
so
all
of
these
new
developments
that
are
so
exciting
of
you
know
the
hill
slope
model
and
just
there's
a
lot
of
really
interesting,
exciting
developments
that
I
think
we're
really
close
to
being
able
to
to
really
understand
what
the
level
of
simulated
ecological
resolution
that
we
need.
D
M
D
D
Okay,
yes,
and
that's
good,
to
remind
everyone
that
you
know
these
five-minute
talks
with
the
two
minutes
for
questions
is
really
just
to
address
each
individual
talk,
but
then
we
can
come
back.
We're
gonna
have
everyone
together
for
the
open
discussion
in
just
a
few
just
a
little
bit
or
was
that
the
final
talk
for
the
session
will.
B
C
Okay,
yeah,
so
andy
raises
an
interesting
point
about
complexity
and
uncertainty,
and
this
is
actually
coming
up
with
my
multi-layer.
Canopy
too.
I
actually
believe
that
the
multi-layer
canopy
is
not
adding
complexity
to
the
model.
It's
actually
making
the
model
simpler.
It
has
fewer
parameters,
it
has
sort
of
a
better
sort
of
theoretical
basis
and
therefore
I
feel
like
you're
sort
of
reducing
complexity
and
you're
reducing
uncertainty.
C
C
B
And
gordon
just
to
clarify:
do
you
mean
because
it's
resolving
some
process
better
like
so,
for
example,
the
big
leaf
model
isn't
resolving
all
the
processes
that
the
multi-layer
canopy
model
is,
and
so
because
of
that
the
it's
effectively
reducing
uncertainty,
because
it's
it's
better
able
to
capture
reality.
Somehow.
C
Yeah,
it's
actually
directly
resolving
processes,
but
it
also
has
like
a
theoretical
basis
to
it.
For
example,
like
we
know
like
to
me,
the
fundamental
concept
of
fates
is,
you
know,
essentially
sort
of
like
natural
selection.
I
guess
you
know
that
we
know
the
world
operates
off
of
individual
plants.
There's
no
such
idea
of
organization
that
the
ecosystem
scale
there's
nothing,
that
sort
of
says.
Oh
carbon
flows
from
foliage
to
stems
to
roots
to
and
drops
to
the
ground
that
we
can
mathematically.
Do
that,
but
that's
not
how
the
real
world
operates
at
all.
C
It
operates
at
the
individual
plant
level,
and
so
fates
is
sort
of
accepting
that
as
the
sort
of
the
basis.
So
you
have
realism
right
there
and
to
me,
that's
a
reduction
in
uncertainty
because
we're
actually
modeling
what
is
actually
happening
in
the
real
world,
not
a
mathematical
abstraction.
The
same
is
true
of
the
multi-layer
canopy
we're
resolving
what
happens
in
the
canopy.
C
You
know
there's
a
question
of
how
good
we
do
that,
but
there's
also
a
question
of
how
good
does
the
big
leaf
model
work
with
with
all
its
sort
of
so-called
simplifications,
yet
tuning
to
actually
get
it
to
work
right.
So
to
me,
this
is
like
the
big
sort
of
philosophical
question
we
face
with
yes,
clm
and
now
moving
into
ctsm
is:
what
do
we
mean
by
complexity,
yeah
and
uncertainty.
C
C
So,
while
you
have
this
increased
possibility
of
evaluation
closer
to
the
real
world,
you
know
with
data
sets
that
are
more
readily
yeah,
they're,
less
abstract,
they're,
more
kind
of
realistic
data
sets.
I
think
the
complexity
still
is
real
because
of
the
high
non-linearity
and
the
responses
within
these
models.
So
thinking
specifically
of
faiths
when
you
do
have
many
different
processes
contributing
to
outcomes
that
didn't
exist
in
the
model
before
really
understanding
what
is
driving,
you
know
what
is
sort
of
responsible
for
the
outcomes
and
and
tracing
you
know.
C
Problems
with
the
model
representation
does
become
more
difficult.
So
I
think,
having
more
realistic
processes,
I
guess
does
introduce
a
level
of
a
need
for
a
level
of
model,
understanding
that
is
kind
of
quite
deep,
and
so
I
I
yeah
just
to
say
I
think
it's
that
both
are
there's
this
opportunity,
but
there's
a
challenge
that
comes
with
it
and
not
to
be
brushed
over.
So.
B
Yeah,
I
can
definitely
second
that
I
think
at
least
within
tss
or
within
within
ncar.
We
kind
of
feel
like
oh,
my
gosh,
there's
all
this
stuff
happening
in
fates
and
we
need
more
experts
that
understand
it
in-house.
You
know-
and
I
think
similarly
other
features
like
hill
slip
model
or
or
the
ag
model,
how?
How
do
we
kind
of
stay
on
top
of
all
of
the
new
exciting
developments
that
people
are
contributing
joshua?
Do
you
wanna
unmute
yourself.
C
Yeah,
so
I
mean
mine
is
a
very
specific
point
related
to
the
the
pfts
that
you
know.
Paulie
was
talking
about
and
the
fact
that
you
know,
as
some
of
us
have
been
looking
at
pfts
and
fates
and
trying
to
calibrate
them.
You
know
versus
real
data.
There's
a
lot
of
things
you
start
doing
and
you're
like
at
first
you're
really
excited,
and
then
you
find
there's
a
lot
of
challenges
to
that
process,
and
is
that
increasing
uncertainty?
C
So
I
don't
think
that
that's
really
creating
new
uncertainty,
it's
uncovering
uncertainty
and
it
doesn't
mean
that
there
are
doesn't
present
huge
challenges
because
you
now
have
the
capacity
with
the
added
realism
of
the
structure
of
the
model
presents
the
challenge
of
matching
reality
in
a
different
way
and
that's
something
which
I've
seen.
But
I
don't
know
if
that's
true
across
all
aspects
of
fates
or
some
of
the
other
model
improvements
that
have
happened.
I
don't
know
that
all
of
them
are.
I
think
there
probably
are
some
cases
where
they're
potentially
increasing
complexity.
B
It
does
raise
some
interesting
questions
like
how
how
much
of
a
historical
contingencies
do
we
need
to
know
about,
for
you
know
like
failing,
was
talking
about
the
soil
conditions
and,
and
danica
mentioned
the
shift
in
cultivation.
How
much
of
that
history
do
we
need
to
get
right
kind
of
moving
forward
once
we
now
that
we're
building
a
model
that
seems
to
care
about
it,
charlie,
go
ahead.
C
He
says
just
a
shift
from
you
know
from
the
canopy
to
the
individual
is
the
you
know,
fundamental
organization,
and
so
that
in
itself
I
agree
is
kind
of
reduction
in
complexity.
You
know
there's
a
shift
off
of
from.
You
know
some
parameters
to
other
parameters
and
we
have
to
think
about
olamitry,
but
we
don't
have
to
think
about
these
sort
of
artificial
allocation
parameters,
but
then
yeah,
then
all
these
other
things
that
come
along
with
it.
C
All
these
you
know
pft
distribution,
you
know
maps
and
things
like
that
which
have
an
enormous
amount
of
information
that
we're
feeding
into
the
model
that
instead,
when
you
know
when
we
activate
the
dgvm
and
fades,
then
you
know
we
ask
the
model
to
predict
that
itself
and
so
in
a
sense,
the
model
is
then
doing
a
lot
more
with
a
lot
less
information
in
in
principle,
you
know
et
cetera,
et
cetera,
and
then
you
know
what
do
we?
How
do
we
define
our
pfts?
C
Do
we
you
know,
do
we
what
what
axes
of
trade
officers?
So
I
guess
I
guess
I
agree
with
gordon's
point
you
know
is
initially
you
know
it's
a
basic
level
of
faith
versus
big
leaf,
but
then
you
know
it
does
bring
in
allow
us
to
bring
in
all
these
extra
things
which
may
end
up
increasing
the
complexity
of
the
model.
You
know
possibly
quite.
H
Oh
yeah,
I
guess
on
this.
I
generally
agree
that
if
you
can,
if
you
have
the
observations
to
estimate
parameters
with
some
confident
in
these
more
complex
formerizations,
that
this
complexity
is
bringing
you
closer
to
the
real
world,
I
wonder
if
people
worry
a
little
bit
that
in
places
where
you
don't
really
have
good
estimates
of
parameters
and
you're
kind
of
maybe
clueing
together,
some
numbers
to
get
the
model
to
run
because
it
needs
them.
H
I
think
that's
my
fear.
I
see
models
being
set
up
all
the
time
that
are
highly
complex
and
then
you
look
at
certain
input,
parameter
fields,
and
you
know
that
they're
not
based
on
observations
and
then
the
models
are
very
difficult
to
get
to
represent
real
variability.
So
I
don't
know
if
that's
a
concern
with
fates
or
not.
B
Yeah
sorry
mbb,
you
mean,
like
I,
don't
know,
a
lot
of
the
work
with
fates
has
been.
You
know
I'll
pick
on
paulie,
for
example
like
focus
over
california
forest
and
then
all
of
a
sudden,
we
un,
we
take
our
insights
from
california
forest
and
somehow
apply
it
globally
and
we're
going
to
be
stuck
in
a
in
a
parameter
space
that
that
kind
of
constrains
the
outcomes
that
we
might
get.
H
Yeah,
maybe
the
model
becomes
a
little
bit
over
constrained
and
not
as
flexible
to
respond
to
different
situations,
and
we
see
this
all
the
time
in
land
modeling.
You
know
we
lay
down
these
super
high
resolution
models
with
a
billion
separate
parameters,
all
guesstimated,
and
it
really
leads
the
model
into
one
place
in
terms
of
its
responses
and
it's
linkages
between
different
variables
and
that
may
not
be
realistic
and
are
more
realistic
than
men.
So
I
guess
I
see
it
as
a
trade-off
like
this
idea
of
justifiable
complexity
that
you
can.
M
Yeah,
I
think
exactly
to
that
point
that
it's
it's
critical
to
do
something
when
we
move
phase
to
a
new
ecosystem.
You
know
it's
used
a
lot
in
the
tropics
and
in
boreal
and
that
we
really
need
to
do
a
parameter
sensitivity
testing
to
exactly
what
you
said,
andy
to
make
sure
that
we
get
the
emergent
properties
that
we
are
expecting
given
the
conditions
that
we're
forcing
the
model
with,
and
I
think
that
can
also
go
back
to
answering
the
earlier
question
about.
M
Does
it
increase
uncertainty
and
I
think
we
can
actually
like
bound
that
uncertainty.
You
know
we
don't
have
to
run
with
just
one
parameter
set.
We
can
say
this
is
we
can
bound
our
simulations
with
you
know,
learning
from
you
know,
incorporating
like
the
ppe
that
dave
and
katie
and
daniel
are
working
on
and
we've
been
doing
that
with.
M
We
need
to
understand
why
we're
getting
these
responses,
and
so
it
takes
a
lot
of
a
lot
of
work
I
think
to
trace
through
and
that's
where
I
think
what
rosie
you've
been
doing
with
the
reduced
complexity
is
so
critical
because
we
can
then
trace
back
through
and
say.
Is
it
you
know?
What's
driving
this
response
and
I
think
it's
a
really
necessary
time-consuming
part
of.
L
I
guess
I
guess
I
guess
my
point
is
that
is
that
our
approach
in
clm
is
to
have
lots
and
lots
of
domain
experts
contribute
to
the
model.
Who
are
you
know
knowledgeable
in
one
particular
field
right
model
itself
is
lots
of
little
models
stuck
together
right
and
the
problem
we
face
is
that
when
we
stick
all
the
little
models
together,
they
all
kind
of
interfere
with
each
other,
and
so
I
really.
L
I
really
do
increasingly
think
that
the
key
is
to
actually
pick
apart
the
little
models
and
have
them
all
be
able
to
run
in
isolation
from
each
other,
and
so
people
who
care
about
particular
things.
Photosynthesis,
fire,
soil,
hydrology,
biochemistry
and
so
on-
can
really
look
at
those
things
and-
and
you
know
properly
understand
and
look
at
the
uncertainty
and
calibrate
them
and
so
on
and
so
forth.
So
yeah.
L
I
I've
been
really
quite
hard,
the
last
few
days
about
how
to
actually
take
the
reduced
complexity
thing
further
and
have
like
smaller,
smaller
chunks
of
model
that
we
can
calibrate.
So
we're
going
to
be
pushing
on
that
quite
a
lot.
So
I
don't
think
I
don't
think
the
answer
is,
is
simple.
Models
are
complex
models.
I
think
the
answer
is
a
smart
modeling
system.
B
A
So
this
is
this
is
very
interesting
topic
like
how
do
we
kind
of
reconcile
that
parameter
overfitting
for
a
given
site
and
conceptually
clm
is
a
single
column
model
so
and
some
of
the
work
that
gordon
is
doing
with
respect
to
the
neon.
So
can
we
develop
a
kind
of
a
kind
of
reduced
order
test
paid
that
will
allow
us
to
test
some
of
these
parameterization
across
different
eco
climatic
setting?
B
B
B
C
K
C
And
what
eddie,
what
was
bringing
up
was
exactly
about
the
overfitting
and
also
we'll
talk
about
this-
that
if
we
look
specifically
at
one
side
and
tune
the
model
to
fit
the
operations
at
this
one
side,
we
will
get
the
something
like
which
is
called,
I
think,
look
elsa
effect.
So
as
soon
as
we
run
the
model
on
other
sides,
we
might
get
worse
response
that.
C
K
K
K
K
This
is
a
realistic,
you
know.
Is
this
something?
That's
still
a
year
away,
two
years
away
six
months,
naoki's
work
looks
great,
but
we
can't
couple
it
to
the
ocean.
We
don't
have
a
plan
for
that
sean's
work
which
we
haven't
heard
about.
Yet
we
don't
know
how
to
do
land
use
change
with
it.
Fates
we
don't
know
how
to
run
the
crop
model
with
it.
You
know
every
single.
F
F
I
wonder
if
it's
worth
this
group
thinking
or
trying
to
think
about
ways
in
which
we
can
reduce
complexity
but
still
keep
the
important
components
of
you
know
the
earth
system.
For
example,
you
know
I
could
see
reducing
complexity
would
mean
turning
off
crops,
but
crops
actually
do
have
quite
a
large
impact
on
climate,
and
so
then
you're
we're
starting
to
get
out
of
those.
So
how
can
we
reduce
complexity
enough,
but
still
maintain
the
the
impacts
on
climate,
but
also
understanding
the
ecosystem,
services
and
those
feedbacks?
F
And
so
you
know,
I
think,
there's
we.
We
just
need
to
be
strategic
in
thinking
about
how
we
move
forward
with
some
of
this,
instead
of
just
adding
continually
adding
complexity.
Thinking
about
how
to
maybe
add
these
new
features,
but
in
a
reduced
complexity
way-
and
I
I
guess
I
just
I'm
wondering
if
this
group
can
help
to
think
about
pathways
forward-
how
we
might
do
that.
B
K
C
You
know
the:
how
does
the
land
influence
climate,
but
these
sort
of
questions
are
also
really
heavily
relevant
on
shorter
time
scales
for
sort
of
a
forecasting
time
scale
I
mean
the
today's
the
water
talk
today
was
a
great
example
of
that.
So
to
what
extent
can
the
work
that
you've
been
doing
with
ctsm
to
sort
of
combine
the
land
models
that
make
the
land
model
relevant
to
both
numerical
weather
prediction
on
these
very
short
time
scales
and
climate
simulation?
K
K
K
That
would
go
into
a
fully
coupled
or
system
model
simulation
for
the
next
generation
model,
and
you
know
maybe
one
way
of
discriminating
on
that
is
to
say
just
think
about
that
really
interesting,
csm
simulations
that
we
heard
about
yesterday.
You
know
that
the
large
ensemble,
the
decatur
prediction,
the
single,
forcing
experiments
you
know
the
regional
refined.
K
K
B
D
Well
and
this
this
goes
to
a
lot
of
the
comments
that
are
going
on
in
the
chat,
and
so,
if
you
have
a
minute,
you
know
scroll
through.
A
lot
of
these
comments
are
exactly
what
dave
was
saying.
How
do
we
prioritize
these
things?
Which
things
do
we
turn
on,
so
that
we
can
run
it
coupled,
and
so
the
discussion
is
really
running
in
parallel.
F
Thanks
yeah,
I
guess
I
was
wondering
like
in
the
context
of
some
of
the
like
ecosystem
services
related
editions
that
we
want
to
add
or
that
you
know
we're
working
on
like
how.
How
do
we
see
those
in
the
context
of
the
climate
justice
talk
that
we
heard
yesterday
and
like?
How
can
we
better
interface
with
you
know,
people
that
are
going
to
be
ultimately
like
very
negatively
affected
and,
more
so
than
other
popular
demographics?
D
B
H
They
first
just
ran
for
forecasting
short
term,
and
now
they
run
for
climate
and
more
groups
are
trying
to
do
that
up
in
canada
as
well-
and
I
think
that's
really.
The
challenge
in
front
of
us
now
is:
how
can
we
make
these
land
models
good
enough?
That
we
can
do
regional,
hydrology
incredibly
and
then
even
push
lower
scales
than
regional?
H
And
so
I
think
I
agree
with
dave
that
ctsm
is
not
only
used
in
a
global
climate
projection
context,
but
I
still
feel
like
we
have
a
long
way
to
go
to
know
how
best
to
run
ctsm
within
this
coupled
applications-oriented
context.
You
know,
for
instance,
there's
a
comment
about
maybe
turning
off
fates,
so
that
would
be
you
know,
more
a
faster
model
or
a
lighter
model,
but
we
don't
really
even
know
if,
if
you
did
that,
would
that
make
things
worse?
H
E
Thanks
this
is
great
in
terms
of
the
the
club,
the
ecosystem
services.
I
was
thinking
about
climate
interventions
such
as
carbon
dioxide
removal,
reforestation,
deforestation
or
looking
at
agricultural
practices,
and
these
are
areas
where
you
know
we
could
really
focus
on,
because
they're
becoming
really
important
in
terms
of
informed
decision
making
and
looking
at
different
ways
of
looking
at
climate
type
solutions.
E
E
But
how
do
we
ensure
that
the
management
we're
doing
is
really
policy
relevant
so
that
we
can
make
sure
that
we're
part
of
that
debate
that
discussion
that's
going
on,
and
I
think
you
know
that
area
is
really
still
really
fertile
for
us
to
work
on
and
working
how
you
make
faiths
and
the
crop
model
everything
else
sort
of
respond
into
that
sort
of
that
space
of
research
is
really
important.
So
I'd
say
that
would
be
a
priority.
For
my
my
research
at
least.
D
D
But
so
just
keep
in
mind
that
we
have
a
few
more
minutes
of
this
discussion
and
then
we're
going
to
have
another
set
of
talks
followed
by
more
discussion,
and
so
don't
see
it
as
the
end,
but
there's
some
really
interesting
stuff
going
on
in
the
chat
about
prioritizing
different
features.
Specifically
what
takes
priority?
D
A
Yeah,
I
I
want
to
make
a
comment.
So
so,
while
we
post
this
question
regarding
the
food
security,
water
security,
so
as
a
result
of
this,
like
this
workshop,
like
what
is
our
excel
item
or
are
we
like
so
like
yeah?
Are
we
putting
getting
more
resources
from
the
nsf
to
to
hire
a
postdoc
or
someone
at
the
end
car
who
can
do
more
like
clm
application
or
the
environmental
justice
part?
A
B
D
F
Sure
you
can
jump
in,
it
seems
it
seems
like
there's.
The
integrating
the
crop
model
with
fates
is
a
top
priority,
both
from
the
feet
side
of
things
and
also
from
clm
side
of
things.
So
that's,
I
guess,
that's
the
discussion
that's
going
going
on
and
then
we
need
to
think
about
how
how
we're
going
to
do
this
moving
forward.
F
F
So
you
know
that's
just
a
thing
that
I
think
we
need
to
keep
in
mind
as
we
are
moving
forward,
and
I
don't
you
know
this-
is
it's
going
to
require
a
lot
of
help,
at
least
from
my
perspective,
from
the
software
engineers
that
are
you
know
that
know
both
clm
and
vapes
in
order
to
move
this
forward?
So
that
is,
you
know
we
just
have
to
have
to
decide
when
to
start
and
how
to
start.
E
I
guess
this:
this
issue
comes
up
in
fates
like
having
the
shared
soil
column
for
all
patches
and
everything
else,
because
the
host
land
model
thing-
maybe
some
work-
needs
to
be
thought
about.
How
do
we
integrate
soil
columns
and
patches
in
fates
so
that
we
do
this
transition
backwards
and
forth
between
crops
and
and
also
being
able
to
do
the
you
know
the
hillside,
hydrology
and
trying
to
understand
how
that
integrates
across
so
rather
than
divorcing
out
the
soils
from
the
ecosystem
component
of
fates?
C
I
feel
a
big
capacity
change
of
the
typical
crop
sec
is
represented
like
if
you
look
at
the
wheat,
for
example,
for
for
the
last
100
years,
the
yield
and
the
zone
changed
a
lot
and
also
the
co2
yeah
capture
egg
yeah,
all
the
properties
of
the
crops.
So,
if
you
run
with
weights
in
this
perspective
and
a
crop
model
yeah,
I
just
wonder
how
this
couldn't
be
implemented.
B
C
B
C
B
C
F
B
And
while
folks
are
tuning
back
in
our
second
set
of
talks
is
focused
on
how
to
header
heterogeneity
and
model
complexity
affect
land
atmospheric
feedbacks
kind
of
a
variety
of
speakers
in
this
session,
which
will
be
fun
to
fun
to
see
how
the
ideas
kind
of
mesh
together,
but
I'd
like
to.
I
was
considering
this
the
conversation
that
we've
been
having
kind
of
build
on
it,
and
I
do
like
the
idea
of
of
thinking.
If
this
could
be.
You
know,
kind
of
an
opinion
like
could
like
is
there?
B
Is
there
enough
here
to
turn
this
into
an
opinion
piece
or
a
perspectives
piece
and
something
like
eos?
It's
got
it'd,
be,
I
think,
it'd
be
kind
of
fun
to
think
about
anyway.
That's
that's
what
I'm
thinking
about
right
now,
so
I'll
go
ahead
and
pull
up
sean's
presentation,
sean!
You
can
go
ahead
and
unmute
yourself
and
come
to
the
floor.
B
B
O
O
O
It
might
look
like
this,
so
this
is
say
order.
10,
20
points
so
now
at
any
given
grid
cell
we're
now
modeling
sort
of
10
10
order
order,
10
instances
of
ctsm
and
getting
a
slightly
different
simulation
at
each
one
of
those
points
due
to
differences
in
meteorology
and
surface
properties
and
vegetation
types,
etc.
O
One
thing
that
we've
been
looking
at
recently,
though,
is
we,
since
we
constructed
this
reduced
form
from
the
higher
higher
dimensional
data,
we
can
actually
go
back
and
so,
for
example,
any
one
of
these
given
points
here
corresponds
to
the
order.
In
the
previous
example,
10
000
points
in
that
grid,
and
so
we
know
what
those
points
are,
and
so
we
can.
We
can
then
map
these
data
again.
Each
one
of
these
would
be
sort
of
a
separate
simulation.
We
can
map
those
whatever
variable.
O
So
this
is
an
example
of
that.
These
two
panels
here
on
the
right
are
there's
aspect
so
direction,
given
a
value
one
through
four
and
the
the
essentially
the
height
of
the
hill
slope,
the
height
above
near
strange.
Also
given
a
value
and
when
you
convolve
those
together,
you
get
this
map
on
the
left,
which
has
a
value
of,
in
this
case,
1
to
16
corresponding
to
the
16
columns
within
this
particular
simulation,
and
so
every
one
of
those
columns.
O
We
can
then
map
back
at
a
higher
resolution
using
this
kind
of
mapping
procedure
next
slide,
and
so
what
that
might
look
like
here
and
I've
changed
the
color
scheme.
Sorry,
but
again,
this
would
be
the
height,
so
this
kind
of
tells
where
the
topographic
highs
and
lows
are
and
the
aspect
telling
you
whether
it's
northeast
south
or
west,
and
so
you
can
see
how
that
would
impact
the
solar
radiation.
O
That
was
felt
throughout
the
the
grid
cell
and-
and
you
can
see
here
that
this
in
the
annual
mean
results
in
something
like
a
30
watt
per
meter,
squared
difference,
and
so
we
can
start
to
to
see
how
this
corresponds
to.
You
know
actual
observations
at
various
locations
within
this
grid
cell
to
see
how
good
this
representative,
hill
soap
idea
might
might
be
doing
next
slide.
O
Similarly,
you
can
do
this
with
something
like
snowpack,
so
here
you
can
see
the
estimates
of
snowpack.
I
think
this
is
just
a
snapshot,
but
again
you
can
see
where,
in
these
hot
regions
of
high
elevation,
you
tend
to
have
higher
snowpack
in
regions
of
low
elevation.
You
have
a
lower
snowpack
and
then
within
a
given
sort
of
elevation
band.
You
can
see
there's
differences,
this
gradation
within
the
area
of
high
snowpack
and
that's
coming
from
the
aspect
maps.
O
So
you
can
see
the
impact
of
these
various
things
in
those
kind
of
maps.
But
one
question
is:
is
this
actually
giving
us?
You
know
useful
information,
or
is
this
in
some
way,
maybe
more
realistic
than
just
the
simple,
one-dimensional
view
next
slide?
So
in
this
case,
I
I
went
ahead
and
looked
at
a
couple,
different
observational
estimates
based
on
satellite.
The
reason
I
did
these
two
there's
landsat
and
modis.
O
So
what
this
is,
what
we
wanted
to
do
with
this
instead
was
see.
If
maybe
we
can
get
a
better
handle
on
this
correspondence
between
the
meteorological
forcing
and
the
surface
properties
and
the
vegetation.
If
we
went
to
these
irregular
grids,
and
so
that's
what
I'm
working
on
right
now
and
since
that's
fine,
this
is
a
little
stop
right
there,
but
yeah.
D
O
Well,
this
slide
was
just
going
to
show
so
so
in
the
previous
slide,
you
can
see
all
these
catchments
that
we're
using
and
that's
straightforward.
The
challenge
is
coming
up
with
this.
These
meshes
like
an
esf
smf
mesh
that
you
can
then
use,
and
so
that's
what's
been
taking
a
while
with
this,
but
you
can
see
how
this
might
be
interesting.
You
know
you
could
you
could
either
align
your
your?
O
D
L
B
D
K
B
N
N
Cool
thanks
well
so
today
I
wanted
to
talk
just
really
high
level
on
some
of
the
efforts
that
were
going
on
to
figure
out
how
we
can
leverage
sub-grade
heterogeneity
to
hopefully
improve
mind
atmosphere,
interactions
which
is
work.
I've
been
doing
with
dave
lawrence,
virginia
and
nate
cheney
attitude
next
slide.
N
So
I
think,
you've
all
seen
this
figure
in
particular
quite
a
few
times,
but
when
we
think
about
the
levels
of
separate
heterogeneity
that
exists
in
cesm,
I
want
to
focus
on
what
exists
in
clm
and
cam
and
on
the
land
side
within
a
single
grid
cell.
What's
really
nice
is
that
you
have
all
these
different
land
surface
types,
so
you
can
have
urban
area
and
lake
and
different
types
of
pfts
on
vegetated
columns,
all
kind
of
coexisting,
and
you
can
imagine
that
each
of
those
patches
has
different
temperatures
and
humidities
associated
with
them.
N
If
we
advance
to
the
next,
cam
also
has
representations
of
separate
heterogeneity
in
cam
6.
This
is
mostly
through
the
club
parameterization,
I'm
not
going
to
go
through
all
of
how
club
works,
but
ultimately
it's
a
way
of
representing
a
sub
grid
variability
in
vertical
velocity,
which
is
w
temperature
potential
temperature,
in
this
case,
at
l
and
humidity
q
sub
t,
and
it
does
that
by
assuming
a
double
gaussian.
And
so
what
I'm
showing
here
in
these
two
panels
below
is
in
two
different
grid
cells.
N
If
you
want
to
understand
the
separate
distribution
of
vertical
velocity,
you
can
imagine,
there's
some
skewness
and
some
variants
associated
with
that.
That
comes
with
these
probability
density
functions.
So
you
have
some
estimate
of
the
level
of
turbulence
how
strong
the
updrafts
are,
etc
in
the
atmospheric
model,
if
you
advance
again,
what
that
leaves
us
with
is
a
question
of
how
we
can
actually
leverage
separate
heterogeneity
that
exists
in
both
of
these
model
components
the
land
in
the
atmosphere,
to
alter
the
way
that
the
land
and
atmosphere
interact
in
cesm
so
next
slide.
N
N
How
we
actually
do
that,
but
I
do
want
to
highlight
there
are
two
options
for
how
you
represent
the
surface,
so
one
of
which
is
a
homogeneous
option,
sort
of
a
business
as
usual
approach,
where
we
just
tell
the
atmosphere
about
the
grid
cell
mean
values
that
are
coming
out
of
the
land.
On
the
flip
side,
we
have
a
heterogeneous
approach
now
that
can
convey
some
information
about
how
much
heterogeneity
there
is
at
the
surface,
and
so
I
feel
advanced
again.
N
N
What's
the
mean
diamond
cycle
of
two
meter,
temperature
and
humidity,
and
when
we
have
a
heterogeneous
surface,
which
is
in
blue,
you
wind
up
with
a
warmer
and
wetter
surface
than
assuming
a
homogeneous
surface
which
is
in
red,
and
so
there
are
these
differences
that
are
coming
through,
which
is
reassuring
next
slide
on
the
perfect
on
the
flip
side
of
class.
We're
also
looking
at
how
you
can
take
the
heterogeneity,
that's
in
club
and
actually
hopefully,
impact
the
surface
flux
diocese
in
particular,
so
not
just
land
atmosphere,
heterogeneity
but
atmospheric
heterogeneous
plants.
D
N
N
Yeah,
that's
a
great
question,
so
what
we're
doing
is
kind
of
taking
advantage
of
the
fact
that
club
has
those
and
and
skews
attached
to
it
and
we're
computing
those
within
the
land
side.
Now
so
we're
looking
at
the
variances
of
temperature
and
humidity
in
particular,
and
how
that
can
be
passed
up
through
the
club
parameters.
N
D
Yeah,
I
will
time
myself
so
I'll,
be
talking
about
capturing
or
the
complexity
of
capturing
vegetation
structure
next
slide,
and
so
we
talked
a
little
bit
about
this
based
on
polly's
talk,
but
with
fates.
The
functionally
assembled
terrestrial
ecosystem
simulator
it
is
it
replaces
it
runs
within
ctsm-
replaces
the
big
leaf
vegetation
with
more
realistic
size,
structured
vegetation.
So
we
can
capture
physiology
competition,
dynamic
ecosystem
assembly
next
slide.
D
And
so
one
of
the
challenges
of
this
complexity
is
how
do
you
test
this
model,
and
so
I
have
a
project
in
boreal
canada,
where
we're
looking
at
post-fire
recovery
and
so
within
canada.
The
uv-fmu
model
operates
there
and
it
captures
forest
dynamics,
and
so
you
can
see
a
biomass
and
species
distribution
for
a
black,
spruce,
a
north
facing
slope
and
a
south
facing
slope,
and
this
model
has
been
evaluated
at
multiple
sites
across
alaska
and
canada,
and
so
it
performs
really
well
next
slide.
D
And
so
it
offers
us
an
opportunity
to
do
detailed
benchmarking,
but
the
models
themselves
have
similarity,
but
some
vast
differences
fates
is
daily
and
changes
lamentary
based
on
photosynthesis,
whereas
uv
fme
is
annual
and
uses
diameter
growth
to
change.
Allowmentry
fates
has
many
complex
parameters.
Uvfme
has
simple
parameters:
fates
is
aggregated,
the
patches
are
not
independent,
whereas
uv-fme
is
independent
and
fates
has
shared
soils
whereas
evf
and
me
as
plot-specific
soil
dynamics.
D
We
have
fates
which
uses
mortality
rates
and
you
can
see
that
for
the
two
different
standages,
the
young
stand
on
the
left
and
then
the
mature
stand
there's
a
little
variation
in
carbon
starvation
mortality
which
is
shown
in
purple
and
so
as
the
trees
get
taller
they're
still
dying
at
the
same.
The
balance
of
mortality
is
similar,
whereas
uv
fme
on
the
right,
which
uses
stochastic
mortality
as
trees,
get
taller
shade
mortality
which
can
be
thought
of
as
light
starvation,
changes
for
these
larger
sizes
and
between
standages
and
then,
if
we
next
slide.
D
Feedbacks
a
forest
that
has
a
large
dominance
of
small
tiny
trees
is
going
to
have
a
very
different
feedback
on
the
atmosphere
than
a
forest
that
has
a
mix
of
stands
of
different
sizes,
and
so,
even
though
it
so
palmor
craft
and
model
evaluations
evaluate
models
with
as
many
hoops
as
possible,
and
this
is
absolutely
true
for
fates.
We
can't
just
we
have
to
look
at
as
much
data
as
possible
next
slide,
and
so
the
benchmarking
is
really
an
essential
part
of
evaluating
fates.
As
we
move
forward.
D
Different
models,
such
as
gap
models,
can
act
as
field
data
surrogates.
The
data
needs
for
validation
across
large
regions
can
be
prohibitive,
and
we
can
use
this
to
highlight
processes
that
drive
vegetation
dynamics
specifically
for
fades,
and
I'm
going
to
run
out
of
time.
So
I'll
go
into
my
questions.
D
We
need
to
look
more
at
mortality,
these
rates
versus
a
stochastic
mortality,
and
then
we
need
questions.
How
else
can
we
test
and
challenge
fates
to
really
so
that
we
can
really
trust
what's
happening?
How
do
we
define
these
functional?
How
do
we
use
functional
relationships
defined
by
data
and
observations,
to
help
us
capture
these
essential
processes
and
improve
them,
and
with
that
we
have
a
little
bit
of
time
for
questions.
L
D
F
Right,
so
we
were,
we
wanted
to
kind
of
eliminate
the
differences
between
you
know
like
the
fate
slash,
clm
kind
of
like
environmental
driver
simulations,
so
we
just
wanted
to
look
at
only
like
if
the
trees
or
the
cohorts
are
seeing
the
same
environment.
How
do
they
differ
in
terms
of
how
they
respond
to
that
environment?.
L
Okay,
good,
can
you
hear
me
all
right
yeah?
So
I
guess
we
talked
quite
a
lot
already
about
the
the
fates
sp
progress
in
the
discussion.
Yeah-
and
I
just
I
don't
want
to
thank
charlie
and
particularly
greg
who's
done
loads
of
work
on
getting
fates
sp
nearly
up
to
the
the
fates
main
branch
next
slide.
L
So
yeah
so
I
mean
so
we
have
this
kind
of
overall
hazardous
of
some
lots
of
effort
on,
I
guess
increasing
model
complexity,
adding
in
processes
we
think
are
important
to
controlling
ecosystem
responses
to
climate,
including
nutrients
crown
damage.
Jesse
neves
been
commenting
in
the
model
plant
hydraulics.
Obviously,
that's
kind
of
a
long
term
project
of
ours,
land
use,
change,
talking
to
peter's
point
earlier
and
fire,
obviously
jackie.
Others
are
working
on
that.
L
I'm
important
to
photocommoners
processes
that
are
not
not
central.
We
feed
in
observations
to
drive
those
things
externally,
and
this
is
the
figure
from
last
challenge
paper
introducing
this
sort
of
hypothetical
set
of
modes
in
which
we
could
run
land
surface
models
in
general,
not
just
fates
right.
So
I
think
that
speaks
to
our
our
sort
of
concept
for
this
as
an
overall
strategy
and
so
for
the
next
slide.
L
So
this
this
is
where
we're
at
in
fates
at
the
moment
with
introducing
reduced
complexity
modes,
we
have
five
reduced
complexity
modes
and
the
full
fates
mode,
in
addition
to
which
we
obviously
have
switches
for
turning
on
the
fire
and
the
hydraulics
and
the
nutrients.
But
and
so
the
moment
we
have
the
satellite
phenology
mode
and
at
the
top
left
there.
L
I'm
gonna
tell
you
about
any
minute
where
we
just
run
the
things
in
green,
the
gas
exchange
and
leak
area
index,
and
then
we
have
static
stun
structure
which
is
similar,
but
instead
of
prescribing
leaky,
arrange
the
model.
We
prescribe
a
standard
structure
from
observations
and
that
also
tests
the
fast
times
of
processes.
L
Then
we
have
this
mode,
the
fixed
biography
mode,
where
we
allow
the
sound
structure
to
evolve,
but
we
prescribe
plant
functional
types
to
the
model
and
we
can.
We
can
run
that
in
a
mode
where
the
vegetation
we
allow,
in
a
particular
grid
cell
competes
with
competes
with
each
other
or
where
it
doesn't,
they
all
just
sit
on
their
own
patches
and
then.
Lastly,
we
have
prescribed
physiology
mode,
which
is
testing
kind
of
things.
L
L
L
Great
right,
so
just
just
to
kind
of
zoom
in
on
this,
so
yeah
satellite
phenology
mode.
You
only
do
the
things
in
green
and
we
turn
off
all
the
things
in
pink
and
that
allows
us
to
isolate
the
fact
that
the
model
only
some
parameters
in
in
salem
fates,
control,
gas
exchange
and,
of
course
that
includes
everything
upstream
of
gpp.
L
So
it
includes
everything
to
do
with
hydrology
everything
to
do
with
all
the
plant
hydraulics,
if
that's
on
everything
to
do
with
photosynthesis,
but
that
is
still
nevertheless
a
much
smaller
subset
of
model
components
than
we
would
be
tackling
if
we
tried
to
calculate
the
entire
model
all
at
once.
You
know,
as
we
kind
of
attempted
to
do
in
the
past
next
slide,.
L
And
so
I've
been
kind
of
like
playing
with
this,
this
this
comparison
between
the
clm,
big
leaf
mode,
sorry,
the
crm,
big
leaf
satellite
phenology
mode
and
the
fates
satellite
phenology
mode
and
trying
to
get
to
a
point
where
I
can
compare
them
directly
and
that
so
far
has
actually
been
super
useful
from
the
perspective
of
like
cleaning
up
what's
happening
inside
the
fast
times
closes
in
the
fates.
So
I
found
an
issue
in
mindfulness
of
crops
in
sp
mode.
L
We
found
that
the
impact
of
day
length
adjustment,
which
is
not
happening
in
fate,
was
quite
large,
had
a
bug
in
the
handling
of
cellular
index.
Without
fixing,
we
found
an
issue
to
do
with
how
fate's
radiation
transfer
layering
system
impacts.
The
answers
and
we've
found
that
it's
a
strange
sort
of
impact
of
the
collapse,
smoothing
parameters
that
we'd
originally
taken
out
in
fates
on
water
use,
efficiency,
which
we
didn't
actually
look
at
when
we
were
taking
them
out.
L
So
that's
actually
changed
the
water
using
the
model
quite
substantially,
and
I'm
going
to
go
back
and
look
at
it.
But
we
wouldn't
have
known
about
that.
If
we
hadn't
had
an
sp
mode
to
fix
it,
and
that's
the
kind
of
thing
that
I
think
will
increase
confidence
in
our
model
as
we
go
forward
like
through
this
cascade
of
different
sp
motion
to
like
nail
down
the
behavior
so
next
slide,
I
think
it's
my
last
one
I
know
yeah,
and
so
the
next
next
stage
with
this
is
to
do
this
calibration
effort.
L
This
is
the
kind
of
error
space
on
on
gpp.
If
you
change
stomatal
slope
and
vc
racks,
which
are
obviously
the
two
biggest
monitors
inside
of
sp
mode
next
slide,
and
so
yeah
I
would
like
is-
is
closing
down
the
structural
configuration
of
sp
modules
for
calibration,
and
then
I
guess
I
guess
I
kind
of
envisaged
that
and
I
have
like
an
orderly
transition
to
ctsm,
where
you
have
like
fates.
L
Sp
comes
in
first,
followed
by
no
cop
mode,
followed
by
our
fixed
biography
mode
and
then
plates
full
mode,
as
they're
all
pulled
into
shape,
and
I
think
that'll
be
a
really
nice
way
of
making
ourselves.
You
know
having
having
more
sanity
about
the
process
of
calibrating
the
model
that
we
had
before.
L
You
know
the
problem,
the
problem
before
when
we
can't
wait
to
see
five
is
that
we're
trying
to
do
all
the
things
all
at
the
same
time,
and
I
think,
if
we
kind
of
go
in
with
this
deliberate
strategy
of
adding
one
thing
at
a
time,
it
will
it'll
be
really
helpful
and
that's
actually.
In
the
last
couple
of
days,
I've
actually
had
a
few
ideas
for
things
that
go
in
between
face
sp
and
no
comp
to
think
about.
You
know
calibrating
the
processes
that
affect
lai
and
biomass
and
isolation.
L
L
Oh
yeah,
just
I
wanted
to
thank
the
software
engineers
for
making
this
experimental
expanding
test
suite,
which
catches
places
where
the
new
modes
break
existing
code.
Lastly,
we've
been
working
through
and
that's
kind
of
essential
to
keeping
this
code
to
a
single
version,
so
thank
you
particularly
to
ryan
and
greg
mcnabbin
and
eric
and
bill,
and
the
other
genius
software
engineers
right.
C
L
C
C
P
Yeah
thanks
well,
thanks
for
the
introduction,
yeah,
I'm
happy
to
be
part
of
ncar
now,
and
can
you
hear
me:
okay,
yep,
okay,
cool
and
although
it's
a
bit
weird
joining
in
the
middle
of
the
pandemic,
but
hope
to
be
there
soon,
so
I've
collaborated
through
the
university
of
utah
before
in
using
clm,
4.5
and
5,
and
also
collaborated
with
the
dart
group
using
clm,
5
and
dart.
So
I'll
talk
a
little
bit
about
that
today.
Next.
P
P
P
And
so
some
of
the
work
that
I've
been
doing
now
this
this
paper
on
the
left
was
just
recently
accepted,
and
so
here
we're
ingesting
a
remotely
sensed
above
ground,
biomass
and
leaf
area
observations
across
the
western
united
states,
and
I
should
say
one
of
the
reasons
why
we're
doing
this
is
you
know,
obviously
the
western
united
states.
The
complex
terrain
makes
it
very
difficult
to
simulate
this.
P
Do
the
high
heterogeneity
and
one
of
the
reasons
why
it's
exciting
to
see
the
hill
slope
hiltel
hydrology
being
developed
as
well,
but
if
you're
using
clm
5.0
one
of
one
way
to
do
this
is
through
you
know,
adjusting
the
model
state,
and
so
what
we
have
done
so
far
is
adjusted.
The
biomass
related
variables,
carbon
and
nitrogen,
and
leaf
area
across
the
western
united
states
directly
through
the
system,
and
then
this
influences
other
related
variables
like
soil,
carbon,
soil,
water,
subsurface,
soil,
water
and,
of
course,
the
carbon
fluxes.
P
So
if
you
get
a
better
idea
of
what
the
state
of
the
system
is,
you
can
get
a
better
representation
of
the
fluxes
and
so
where
we
hope
to
be
going
with
this.
Of
course,
areas
like
the
western
united
states
are
water,
limited
and
so,
with
more
remotely
sense.
Observations
like
solar,
induced
fluorescence.
P
Because
of
that
strong
linkage
to
to
the
carbon
cycle,
and
these
these
snow
teles
snotel
sites
are
point
level,
but
the
way
that
the
system
works,
of
course,
is
that
they
have
kind
of
a
range
of
influence
horizontally,
so
that
you
can
use
even
point
estimates
to
give
you
a
regional
update,
and
there
are
other
data
products
out
there
I'll
get
more
into
that
later.
Like
somos
map
that
have
you
know,
can
give
you
an
idea
of
global
soil,
moisture
or
snow
cover.
P
We
use
that
relationship
with
the
unobserved
variables
on
the
y-axis,
so
things
like
the
specific
pfts
or
the
crop
functional
types
we
linearly
regress
that
update
onto
all
those
variables
columns
as
well
to
update
the
entire
system.
So
we
need
clm
that
covariance
between
what
is
observed
and
what
is
unobserved
to
make
that
correction
in
the
future.
The
bottom
row
with
you
know
in
improving
earth
system
measurements
like
this
land
cover
product,
for
example,
high
resolution.
P
We
can
actually
you
know,
add
more
metadata
to
perhaps
a
coarse
biomass
observation
data
set
and
only
only
correct
for
the
specific
pft
responsible
for
that.
For
that
observation,
so
you
know
in
the
dart
the
data
assimilation
world
is
called
localizing
on
the
specific
state
space
next
slide,
and
so
I
just
wanted
to
emphasize
the
advancement
both
in
models
and
observations
are
happening
together
and
so
on
the
top
row.
We
have
an
expanding
satellite
network,
expanding
land,
surface
properties
that
we're
also
monitoring.
So
it's
a
kind
of
an
exciting
time.
H
P
This,
and
so
I
mean
the
figure
just
essentially
shows
that
there's
increasing
coverage,
increasing
spatial
resolution
and
temporal
resolution
and
at
the
same
time
keep
in
mind
that
you
know
clm
is
always
a
moving
target
of
more
complexity.
So
I
worked
on
you
know,
versions
of
4.5,
which
was
just
a
simple,
hey.
P
I'll
try
to
hurry
up,
but
anyway,
thank
you.
So
the
fact
that
we're
increasing
model
complexity
we're
basically
bringing
bringing
the
model
closer
to
what
we
can
observe
things
like
leaf
water
potential
and
sif.
That's
either
in
leaf
or
in
the
canopy
next
slide,
and
so
again,
just
more
emerging
satellite
data,
and
I
just
wanted
to
emphasize
that
that
dart
is
designed
to
add
on
to
these
these
new
data
products
either
either
single
data
products.
P
A
lot
of
these
traits
are
inferred
from
from
multiple
observations,
so
we
have
observation
converters
that
take
the
raw
data
format
and
convert
it
into
the
assimilations
so
that
the
simulation
system
can
understand
it
and
things
like
the
forward
operators.
I
mean
adaptive.
Inflation
can
take
into
account
that
sometimes
there's
not
a
direct
analog
between
what
is
observed
and
what
is
actually
simulated
by
clm
and
adaptive.
P
Inflation
particularly
takes
into
account
the
fact
that
sometimes
there's
systemic
errors
or
biases
that
are
not
quantified
for,
but
you
can
still
have
a
successful
update
or
simulation
next
slide
and
then
really
quick.
Finally-
and
this
just
gets
on
the
discussion
from
the
previous
theme-
parameter
estimation
generally,
when
we
think
of
ensembl
common
filters,
we
update
the
states,
the
seal
and
model
states,
and
so
that's
what
the
top
row
is
doing
there,
but
also
you
know
if
you're
working
with
the
model
parameters
are
really
important,
and
so
these
are
the
type.
P
That's
generally
doing
parameter
estimation
with
mcmc
techniques
are
really
computationally
heavy,
and
so
an
alternative
perhaps
is
to
use
an
ensemble
common
filter
like
dart
and
so
using
literature
or
plant
trait
databases
to
create
a
prior
distribution
and
then
in
a
similar
way,
using
that
to
create
the
spread
in
clm,
adding
more
observations
and
use
that
to
estimate
the
parameters.
So
that's
something
else
that
the
system
is
capable
of.
So
I'm
done.
Thank
you.
You're.
P
J
So
if
we
think
about
some
change
in
the
land
surface,
like
a
change
in
vegetation,
then
that
can
pretty
directly
drive
a
local
change
in
the
atmosphere
like
changes
in
your
air
temperature
or
your
humidity.
Just
by
changing
your
surface
plexus
of
water
and
energy
and
those
local
changes
in
the
atmosphere
can
also
include
things
like
near
surface.
Wind
speeds
that
can
end
up
triggering
these
sort
of
larger
responses
in
the
atmosphere.
J
That
can
be
either
right
next
door
or
really
far
away,
and
I
think
there's
two
big
questions
that
we
don't
have
particularly
good
answers
to.
Yet
with
respect
to
this
general
topic
and
that's
the
first
one
being
how
big
of
a
kick
do
you
need
to
make
to
the
land
surface
in
order
for
the
atmosphere
to
actually
respond
in
any
significant
way
and,
secondly,
how
and
where
do
clouds
in
particular
respond
to
these
changes
in
the
land
surface.
J
So
starting
off
with
that
first
question
next
slide:
you
can
think
about
this
in
sort
of
two
extremes
and
most
land
surface
changes
would
be
falling
somewhere
between
these
two
extremes.
So
on
the
left,
we
have
continental
or
global
scale.
Changes
in
vegetation-
and
many
studies
have
shown
this
over
the
past
several
decades-
that
if
you
have
a
really
drastic
perturbation
to
your
land
circus,
you
can
trigger
very
drastic
large
scale.
J
Atmospheric
responses
in
particular
what
the
figure
is
showing
you
is
the
change
in
precipitation
in
summertime
in
a
model
simulation
where
you
have
basically
totally
dry
soil
on
the
bottom
and
totally
wet
soil
on
the
top.
So
you
can
see
you
get
very
different:
atmospheric
precipitation
patterns
over
land
and
even
to
some
extent
over
the
ocean,
depending
on
this
drastic
kick
you've
made
to
the
land
surface.
J
On
the
other
hand,
if
you
look
on
the
right,
what
you're
seeing
is
on
the
top
a
sort
of
false
color
image
of
a
forest
where
the
green
is
trees
and
the
brown
is
dirt
and
on
the
bottom,
an
infrared
image
of
that
same
forest,
where
the
purple
colors
are
very
cool
and
the
yellow
colors
are
very
hot.
So
then
you
could
imagine.
J
That
way,
and
those
are
you
know,
memory.
So
that's
five
minutes.
Okay,
super
and
those
are
very
opposite
responses,
but
they
can
both
lead
to
an
increase
in
cloud
cover.
Can
you
skip
ahead,
two
slides
and
this
change
in
cloud
cover
matters
because
it
feeds
back
on
the
land
surface
so
on
the
left.
This
is
a
change
in
temperature
in
response
to
decreasing
land,
evaporation
without
a
response
in
clouds
and
the
same
change
in
temperature.
If
you
let
the
atmosphere
respond
next
slide,
I
promise
I'll
be
quick.
J
I
just
want
to
point
out
that
this
matters
not
only
for
physical
climate,
but
also
for
the
carbon
cycle,
because
changes
in
cloud
cover
change
your
partitioning
between
direct
and
diffuse
radiation
and
that
matters
a
lot
for
photosynthesis
and
next
slide.
That's
one
you've
already
seen
and
that's
just
to
wrap
up
that
these
are
sort
of.
I
think
two
big
questions
that
matter
with
respect
to
our
coupling
between
the
land
surface
and
the
atmosphere
and
meg
already
addressed
some
of
these
heterogeneities
in
that
they
change
how
your
atmosphere
responds.
B
F
F
Yeah,
I'm
happy
to
happy
to
clarify
here.
I
I
guess
I
just
was
thinking
you
know.
The
grids
are
very
irregular,
and
so
you
know
that,
given
that
the
atmosphere
is
used
to
seeing
a
square
grid,
you
know
I
guess
how
I
I
was
just
curious-
how
that
would
work
and
if
it
would
change
things
and
then
you
know,
meg
followed
up.
Yes
with
her
sort
of
coupling
land
atmosphere,
coupling
subgrid
stuff-
and
you
know
just
thinking
about
those
kinds
of
connections
and
how
that
would
work
within
that
current
gridded
system.
O
O
Real
apples
to
apples
way
to
to
make
that
kind
of
comparison,
because
everything's
going
to
be,
I
mean
the
land
surface,
will
be
different.
It'll
be
responding
differently
because
it'll
be
seeing
a
different,
slightly
different
atmosphere
as
well,
just
due
to
how
the
interpolation
will
be
different
based
on
the
gridding
and
stuff.
So.
K
Different,
you
know
that
your
irregular
grid
size
isn't
too
much
different
to
the
atmosphere.
It's
probably
going
to
be
fine.
I
think
megs
will
work
as
well.
Although
you'll
start
averaging,
your
moments
will
get
averaged
in
the
coupler
and
sent
to
club,
so
it
does
become
a
little
bit
harder
to
interpret,
I
think
would
be.
That
would
be
the
biggest
challenge,
but
there
would
be
no
technical
barriers
to
it.
Just
working.
K
So
yeah
I
mean
I
think
it's
going
to
be.
We
have
obviously
haven't
decided
if
we're
going
to
go
to
regular
grids.
That
would
be
a
big
big
change
as
a
default
way
of
running,
but
it's
going
to
be
interesting
to
start
actually
exploring
this
feature.
We've
had
the
capability
for
a
long
time.
I've
never
tried
to
exploit
it,
and
if
we
find
that
it
makes
a
pretty
big
difference
to
the
hydrology
simulations,
then
we'll
you
know
have
some
decisions
to
make
about.
P
O
Right,
no,
these
also
they're.
Not
they
don't
have
any
spatial
awareness
of
where
they
are
within
the
grid
cell,
but
they
just
have
the
spatial
awareness
of
that.
They
have
neighbors
sort
of
an
uphill
and
downhill
directions,
and
things
like
that,
so
they
are
connected
in
that
way,
but
yeah
not
within
a
specific
point,
but
the
remapping
was
sort
of
saying
that
you
could
go.
You
could
then
go
back
and
say
you
know.
This
element
of
my
hill
slope
corresponds
to
these.
O
F
O
O
I
guess
it's
mo
yeah,
I
mean
given
information.
Like
I
said
we
can,
we
can
definitely
apply
the
information
differently
across
the
hill
slope.
You
know,
so
you
could
imagine
that
lowlands
versus
uplands
might
might
vary
differently,
and
so
I
think
that
that
would
be
sort
of
a
research
question.
You
know
based
on
taking
the
data
like
johan
and
his
group's
data.
If
you
can
digest
that
in
some
way
and
to
see
them,
you
could
definitely
do
that.
C
D
J
I
had
a
question
for
meg,
so
I
don't
want
to
detract
from
the
soil
stuff,
but
people
still
have
questions
about
that,
but
for
for
meg
I
was
just
wondering
if
you
could
clarify
in
your
subgrid
heterogeneity
fluxes
that
are
getting
passed
to
club.
Do
they
then
get
merged
into
a
single
flux
that
club
sees
or
can
club
handle
getting
past
like
several
different
neighboring
fluxes
in
a
single
cam
grid
cell.
N
Yeah,
that's
a
great
question.
So
what
we're
doing
is
just
passing
a
single
flux,
so
within
clm
you
basically
calculate
club's
moments
per
patch,
then
there's
a
way
of
asking
them
together
that
accounts
for
the
fact
that
they're,
you
know
different,
and
then
you
pass
that
grid
cell
mean
up
to
up
to
club.
So
it's
not
you're
not
directly
telling
it
the
heterogeneity,
you're
sort
of
accounting
for
it
right
in
a
different.
J
K
We'll
just
add
that
the
other
can
kind
of
add
on
that
that
the
nc
thing
that
meg
didn't
get
to
tell
you
about.
Was
that
part
of
what
we're
learning
is
that
club
was
already
you
know,
treating
the
land
surface
differently,
we're
calculating
the
these.
You
know
monocot
stuff
differently
than
we
were
doing
in
clm,
so
there
was
an
inconsistency
between
cam
and
clm
that
meg
uncovered
by
going
through
this
process
and
a
nice
side
benefit
of
all.
This
is
now
we're.
K
A
So
if
there
is
some
common,
these
two
are
very
different
philosophy
in,
in
the
other
case,
you
basically
parameterize
the
heterogeneity
using
a
statical
function.
In
other
case,
you
are
modeling
each
at
the
individual
column,
the
heterogeneity
that
that's
kind
of
very
yeah
way.
The
way
the
clm
has
been
built.
So
is
there
some
way
between?
We
can
combine
these
two
to
come
up
with
the
to
come
to
parameterize
the
heterogeneity
at
the
land
surface
or
like,
like.
There
is
a
concept
of
the
big
model
where
you
parameterize
the
different
infiltration
capacity.
A
Yeah,
so
I
this
was
a
question
with
respect
to
how
do
we
parameterize
this
heterogeneity?
There
is
one
concept
like
you
represent
every
heterogeneity
as
a
single
column.
Then
you
model
it
another
concept:
is
you
parameterize
using
the
a
curve,
a
statical
function
and
that's
what
like
I
got
from
the
megan
talk
like
that's
how
they
parameterize
in
that
atmospheric
model.
So
if
that
some
concept
that
can
be
useful
for
the
land
modeling
community
to
parameterize
the
heterogeneity.
B
I
know
at
oak
ridge
people
working
with
peter
thornton's
group
have
done
that
they
took
pft
like
a
distribution
of
parameters
for
pfts
that
are
kind
of
informed
by
tri
or
some
other
observations.
I
don't
know
if
that's
something
that
could
work
into
dart,
where
you
give
kind
of
a
distribution,
a
pft
parameters
that
are
vary
by
latitude
or
something.
P
Yeah
you're,
I'm
just
jumping.
Well
yeah
I
mean
you.
Could
you
could
basically
do
that?
I
mean
you,
know
pf
or
parameters
within
clm.
Is
it
treated
like
a
special
state
right?
So
you
know
it's
fixed
to
you
know
parameters
typically
fixed
to
a
certain
pft,
so
yeah.
If
you
had
a
good
prior
estimate,
I
mean
you
can
use
that
to
sample
the
parameter
space,
to
give
you
your
your
ensemble
and
get
a
posterior.
B
E
I
guess
this
is
following
up
on
the
idea
of
horizontal
and
spatial
heterogeneity
and
quite
often
we
sort
of
like
you
know
we
have
these
pfts
or
patches
or
whatever
else,
and
we
say
a
certain
percentage
of
the
grizzle.
Is
there
since
this
and
sort
of
comes
back
to
what
we're
saying
about
like
the
spatial
organization
is
really
important.
So
if
you've
got
a
savannah
and
you've
got
say,
10,
trees
and
90
grasses,
that's
very
different
than
if
you
had
a
patch
of
forest
and
then
90
of
the
grid
cells,
grassland.
E
And
do
we
need
to
start
thinking
about
how
we
organize,
especially
within
the
grid
style,
because
that
level
of
spatial
organization
is
really
kind
of
important,
especially
if
you're
starting
to
look
at
what
meg's
doing
in
terms
of
trying
to
understand
mesoscale
circulations
within
these
grid
cells?
Looking
at
various
levels
of
you
know,
convective
versus
like
what's
happening
in
terms
of
the
boundary
layer
processes.
E
So
maybe
that's
an
area
that
we
should
sort
of
return
to
and
try
and
think
of
and
I've
always
thought
of
that
as
like
you
know,
you've
got
the
land
unit
and
then
you've
got
the
land
units
broken
into
plant,
functional
types
or
it
seems
to
be
there's
a
way.
We
could
start
thinking
about
describing
the
the
organization
within
the
grid
cell.
I
guess
it
comes
back
to
sean's
hill
slope
hydrology
as
well
yeah.
That's
just
what
I
thought.
K
Okay,
I'll
just
respond
to
that
real
quick
bill.
I
see
you're
about
to
talk
just
say
that
within
class
peter
we
are
that
we
are
considering
that
and
like
a
representative
length
scales
that
you
can
calculate,
you
know
before
the
run.
That
would
help
you
understand.
You
know
how
things
are
connected
to
each
other,
whether
that
lakes
are
all
right
next
to
each
other,
whether
the
lakes
are
distributed
across
the
grid
cell
and
so
nate
cheney
who's.
K
C
Bill
go
ahead,
yeah
I
was
actually
going
to
say
similar
to
peter
and
so
I'll
just
say
add
something
additional
that
I've
been
thinking
about,
which
is
you
know
we
have.
We
have
now
all
these
different
processes
and
getting
more
and
more
that
are
that
are
capturing
different
elements
of
the
subgrid
heterogeneity
and
something
that
I
find
kind
of
hard
to
wrap.
My
mind
around
is
is
and
tying
in
with.
What
dave
just
said
is
what
is
that?
C
What
are
the
different
length
scales
of
these
different
representations
of
subgrid,
heterogeneity
and
so
which
ones
which
ones
do
you
want
sort
of
turned
on
together?
Which
ones
are
kind
of
mutually
exclusive,
like
just
taking
one
example
like
hill
slope,
hydrology
and
our
subgrid
or
sub-column
snow
parameterization
like
how
do
those
two
work
at
similar
or
different
scales
of
heterogeneity
or
length
scales,
and
do
we
need
to
adjust
the
sub
sean?
You
probably
have
a
have
an
intuitive
sense
of
this,
but
but
it's
sometimes
hard
for
me
like
do.
C
We
need
to
adjust
the
sub
columns
snow
parameterization
if
we
turn
on
hill
slopes,
because
some
of
that
sub
grid
heterogeneity
is
now
is
now
explicitly
accounted
for
and
then
and
then
yeah,
how
about?
How
does
that
tie
in
with
what
meg's
doing
and
the
scales
of
heterogeneity
that
that's
supposed
to
be
representing?
H
C
So
I
was
looking
into
this
with
respect
to
start
and
end
of
the
growing
season
before
based
on
metrological
data,
and
what
I
found
with
all
the
latitude
dependent
parameterizations
of
these
was
that
there
was
a
pretty
good
match,
for
example
in
north
america,
where
it
might
have
been
developed
and
no
match
at
all
in
northern
europe.
Because
of
the
different.
B
D
D
B
No,
I
mean,
I
think,
like
you
know
right
now,
we
don't
really
know
how
we
would
do
land
use
land
cover
change
with
the
hillswood
model.
We
don't
know
how
we
would
do
yeah.
How
do
we
sorry?
How
would
we
do
land
use
plant
cover
change
on
the
hill
slope
model?
How
we
would
do
fates
on
the
hill
slope
model.
B
F
F
F
B
G
B
But
it
is
interesting
to
think
about,
especially
thinking
about
the
shifting
cultivation
and
fates,
the
soil
degradation
and
fates
and
shifting
cultivation
like
all
of
those
things
together.
I
think
it
makes
it
it
it's
a
lot
of
information
that
we're
kind
of
black
boxing
I'll
say
this
as
the
soil
person
like
we're
black
boxing
the
soils
again-
and
I
don't
know
how
long
we
can
continue
doing
that.
E
E
C
B
D
B
M
K
B
O
C
K
So
we're
getting
some.
You
know
some
bang
for
our
buck
from
yao
scout,
so
that
will
come
and
that'll
pose
its
whole
new
set
of
challenges
right.
If
we
get
our
genius
information
from
the
atmosphere
but
yeah
coming
down
the
question
of
priorities,
you
know
that
this
is.
We
could
ask
the
question:
is
it
or
also
realistic
in
terms
of
what
we're
gonna
have
for
the
next
version
for
csm3?
You
could
ask
one
possibility
we
say
we're
gonna,
we
don't
think
fates
is
gonna.
K
Get
there
all
the
phase,
people,
don't
you
know,
shoot
me,
but
we
don't
think
we're.
Gonna
have
a
global
fates
carbon
cycle
simulation
that
will
just
transiently
that
we
can
trust
in
csm3
as
our
default
configuration.
So
if
that's
the
case,
should
we
spend
some
time
trying
to
get
a
hillslope
version
and
possibly
with
meg's
clasp
work
working
in
a
transient
sense
as
a
big
new
feature
of
the
model
of
sub
good?
You
know
sub
good
hydrogenate,
so
you
know
I
don't
know
what
the
answer
that.
L
C
K
Well,
no
either
I
mean
in
terms
of
like
what
what
would
potentially
work
with
anything
else.
I
don't
think
there's
any
like
things
that
saying
we
can't
do
it.
It's
just
you
know
with
the
to
do
faiths,
we
have
to
get
faiths
working
while
also
running
the
crop
model
at
the
same
time,
and
that
is
one
set
of
technical
coding
work
that
needs
to
happen
to
make
that
possible
and
then
on
the
hillslope.
B
K
Need
to
be
able
to
decide
how
we're
going
to
make
decisions
about
where
to
deforest.
When
you
have
these
multiple
columns,
you
know
to
usually
force
the
north
facing
slope
the
south
facing
slope
evenly
and
then
also,
if
we're
assuming
that
cropland
is
on
flat
land
and
not
on
hill
slopes.
You
know:
there's
there's
like
decisions
that
need
to
be
made,
so
both
of
those
projects
would
require
significant
software
engineering
work
which
aren't
probably
maybe
don't
build
on
each
other.
K
K
Well,
you
know
it's
interesting,
like
gokan
didn't
mention
csm3
yesterday
at
all
last
last
year
he
was
mentioning
csm3
starting
this
summer,
so
now
it
doesn't
even
have
a
mention.
So
I
don't
think
we
know
unless
we've
heard
something
I
we'll,
but
it's
not
long.
You
know,
I
think
my
my
my
interpretation
is
as
soon
as
I
get
mom
six
working
they're
gonna
start
pushing
really
hard
for
csm3
as
long
as
kokon's
in
charge.
K
C
B
Yeah,
I
mean
it's
kind
of
interesting.
I
guess
it's
not
a
totally
post
cmip
world,
but
you
know
we're
not
we're,
not
necessarily
we're
not
planning
on
releasing
csm3
for
cmip
x,
which
makes
it
interesting
to
think
about
what
what
you
know.
What
are
the
experiments
that
we
want
to
be
able
to
run
if
it's
not
the
deck,
the
deck
plus
mips.
K
You
know,
I
think
what
came
up
in
the
chats
this
morning
was.
Was
these
questions
about
you
know,
mitigation
and
adaptation
related
to
the
carbon
cycle
could
easily
be.
You
know
we
could
just
say:
that's
that's
our
target.
We
want
to
understand
the
potential
of
force
as
a
mitigation
tool
and
agricultural
management
as
a
mitigation
tool
yeah.
That
would
be.
That
would
be
one
pretty
compelling
argument
and
that
would
that
would
lean
towards
the
fates.
L
K
L
K
K
B
F
I
guess
on
that
note
of
kind
of
trying
to
narrow
down
our
spatial
heterogeneity.
What
other
you
know,
sort
of
suite
of
observations.
Can
we
be
leveraging
to
confirm
that
that
these
new
fates
modes
are
you
know
that
we
haven't
already
been
using,
so
I
mean
I'm
thinking
like
jedi
and
icesat,
and
you
know
how
can
we
kind
of
you
know
come
together
because
that
you
know
that
is
like
kind
of
a
huge
undertaking.
But
how
can
we
come
together
to
facilitate
those
kinds
of
observations,
observation
model,
comparisons.
L
L
F
F
I
guess
either
growth
or
regeneration,
and
I
was
just
wondering
if
anyone
else
had
thoughts
on
that.
So
I
mean
this
would
potentially
be
something
in
addition
to
what
is
going
on
with
fate's
hydro
and
the
nutrients
like.
Would
there
be
other
things
that
we
could
limit
or
or
prefer
one
pft
or
over
another
in
terms
of
the
soil,
texture
or
the.
B
One
of
the
things
that
changes
most
with
land
use
is
soil
structure
which
is
not
really
captured
at
all
by
measurements.
Of
that
we're
using
for
the
pedal
transfer
functions
that
control
hydrology
now
so
it's
a
it's
a
separate,
totally
separate
kind
of
can
of
worms
that
we're
not
really
thinking
about.
F
E
Yeah,
so
I
just
following
up
on
that:
well,
we
currently
use
prescribed
soil
carbon
right
for
looking
at
the
the
actual
pillar
transfer
functions,
so
if
we
could
somehow
have
a
way
of
having
you
know
actually,
instead
of
going
prescriptive,
we
actually
modeled
off
what
the
carbon
cycle
was
doing
in
terms
of
soil
carbon.
That
might
be
a
start,
but
really
no,
I
just
don't
think
that
way.
You
know,
we've
had
soil
carbon
quite
a
while
now.
B
And
we've
started
some
discussions
to
get
updated
data
sets
that
we
could
use,
and
I
don't
think
those
would
be
transient.
So
we're
we've
been
still
using
the
igbt
soils
data
so
getting
soils
grid
is
you
know
a
21st
century
approximation
of
that
map
and
then
either
making
it
transient
in
a
prognostic
sense
or
a
diagnostic
sense
could
be
nice.
K
Response
to
peter
we
have
tried
that
jeffy.
You
did
that.
I
tried
it
once
too,
but
unfortunately
it's
nice,
except
for
in
the
arctic,
where
it
goes
run
away,
and
so
that's
that's
the
problem.
You
have
to
deal
with
a
spin-up
problem,
and
so
it's
essentially
we
can't
do
it
right
now,
unless
somebody
can
solve
and-
and
one
hypothesis
is,
that
you
need
to
have
fire
where
there's
some
return,
animal
that
the
peat
eventually
burns
or
the
organic
amount
of
burns
on
some
thousand-year
times
time
scale.
K
B
F
Yeah,
I
was
just
gonna
add
more
more.
Another
can
of
forms
would
be
like
other
types
of
soil
sort
of
like
thermo
karst
and
like
drainage
events
and
those
lumps
to
that
arctic
related
soil
land
use
land
cover
change.
I
don't
know
if
anyone's
working
on
that
for
the
model,
but
I'm
not
even
really
sure
how
you
would
do
that.
F
B
B
And
I
do
know
I,
I
am
still
kind
of
intrigued
by
sanjiv's
challenge
like
what's
going
to
come
out
of
this,
and-
and
I
like
the
idea
of
you
know
as
a
group-
maybe
addressing
lara's
question
like
what
experiments
do
we
want
to
run
and
what
are
the
like
thinking
about
what
we
need
to
to
get
there,
and
so,
whether
that's
you
know
continuing
this
discussion
internally,
whether
that's
writing
a
paper.
You
know
that
kind
of
sets
the
stage
for
you
know,
here's
why
we
need
more
support
to
do
x,
y
and
z.
B
I
think
these
are
all
kind
of
great
aspirations
as
with
anything
they
take
a
champion.
So
if
there
is
a
champion
that
wants
to
kind
of
step
forward
to
you
know
we
need
we
need
to
keep
continuing
the
conversation.
If
we
want
it
to
be
more
formal
than
that,
I
think
that
I'd
be
super
supportive
of
that.
B
Yeah-
and
I
want
to
thank
again
everyone
for
their
contributions-
we
I
think
we
need
to
come
back
one
for
those
of
us
that
are
interested
in
talking
about
fire
and
and
the
format
there
I
think,
will
be
similar
with
some
some
quick
presentations
from
a
pretty
diverse
crowd
and
then
and
then
more
discussions
to
follow.
But
thank
thanks.
I
want
to
thank
everyone
again
for
coming
this
morning
and
hopefully
we'll
see
each
other
in
person
before
too
long.