►
From YouTube: CTSM Tutorial - Thursday
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
Tutorials
we've
had,
you
know,
20
to
15
to
20.
You
know
30
to
60
minute
long
presentations
and
we're
collapsing
all
of
those
or
some
of
those
into
these
five-minute
lightning
talks
today.
So
what
you'll
get
is
kind
of
necessarily
short,
but
hopefully
give
you
enough
information
about.
A
You
know
different
components,
different
features
of
the
model
that
you
can
figure
out
where
to
learn
more
on
your
own,
so
we'll
spend
the
first
hour
together
going
through
the
series
of
lightning
talks,
then
have
a
quick
break
and
then
spend
the
last
couple
hours
working
on
a
couple
of
different
options
that
are
both
single
point:
different
flavors
of
single
point
simulations
before
before
coming
back
together
and
closing
for,
what's
been
a
very
short
tutorial.
A
Before
we
get
started,
does
anyone
have
questions,
or
are
we
good
to
good
to
have
peter
go
for
presenters
who
are
just
kind
of
logging
on
I'll?
Have
you
share
your
screen?
You
can
present
your
own
slides
and
then
maybe
as
we're
transitioning.
If
folks
have
questions
during
that
transition
between
between
speakers,
we
can.
We
can
ask
questions.
Then.
A
Great
well
that
no
one
has
no
one,
has
questions
now
I'll
turn
it
over
and
let
peter
talk
about
land
use
and
land
cover
change.
B
Thanks
well,
let's
get
my
screen
sharing
up
okay,
so
I'm
peter
lawrence,
I'm
in
the
terrestrial
sciences
section
at
encar.
One
of
my
chief
things
I
do
is
land
use
and
land
cover
change
so
putting
together
all
the
data
sets
that
run
behind
ctsm
and
in
the
cmf6
experiments
all
of
those
components.
B
So
what
I'm
going
to
do
is
quickly
take
you-
and
I
was
very
fortunate-
dave-
gave
a
great
background
and
introduction
and
sort
of
motivation
for
why
we
do
land
use
and
land
cover
change
from
a
carbon
cycle
and
also
from
a
biogeophysics
perspective,
but
because
we
only
got
short
time.
I'm
gonna
quickly
get
into
this.
B
So,
as
dave
said,
you
know
the
way
we
represent
the
the
the
heterogeneity
of
the
landscape
of
the
of
the
world
is
by
taking
this
gridded
approach
and
then
doing
some,
what
we
call
subgrid
tiling.
So
you
go
from
the
the
grid
cell,
which
is
whatever
resolution
that
you
decided
to
run
your
model
at
and
you
can
break
it
down
into
the
the
sub
components
of
that
at
a
land
unit
level.
So
we
have
vegetation
lakes,
urban
glaciers,
crops
and
one
of
the
nice
things
about.
B
That
is
that,
rather
than
just
being
static,
we
can
change
it
on
an
annual
basis.
So
what
happens?
Is
that
composition
of
what
is
natural
vegetation?
What
is
crops
and
now
like
looking
going
forward,
we're
looking
at
lakes
and
urban
and
glaciers
all
being
dynamically
able
to
change
as
well
on
the
on
that
sort
of
time,
step
period
so
below
the
below
the
land
unit
level?
We
have
columns.
So
this
is
basically
where
all
the
soils
are
represented
in
terms
of
hydrology
and
silicon,
but
we
can
basically
change
anything.
B
We
can
change
things
at
the
landing
level
and
also
at
the
plant,
functional
type
and
functional
type
level
below
so.
On
top
of
the
changes
in
land
use
and
land
cover
change,
we
can
actually
do
other
other
things,
such
as
wood
harvest,
so
we
can
go
into
all
the
plant,
functional
types
and
all
the
tree
types
and
we
can
say:
okay,
let's
take
out
carbon
in
terms
of
wood
carbon
that
comes
out,
and
so
we
can
do
this
forestry
process.
B
We
also
do
agriculture,
so
we
have
to
prescribe
the
amount
of
irrigation
the
amount
of
fertilizer
and
the
types
of
crops
that
we
have
within
the
crop
model.
So
to
do
that.
For
these
time
series
we
start
off
with
the
current
day
and
we
have
a
whole
set
of
tools
that
allow
us
to
go
from
current
day
satellite
and
inventory
project
products,
and
we
pull
them
together
and
we
can
produce
a
snapshot
of
the
world
as
it
is
every
year
and
I
haven't
got
very
much
time
to
go
through
this
process.
B
But
what
you
can
see
is
there's
a
whole
number
of
products
that
come
together
and
we
produce
up
here
in
this
top
right.
These
time
series
and
we've
got
trees
on
the
top
and
then
crop
area
on
the
bottom,
and
this
goes
back
to
6000
years
before
the
common
era,
so
about
8
000
years
ago,
and
we
have
these
pathways
forwards
through.
So
the
shared
socioeconomic
pathways
which
have
different
land
use-
and
you
can
see
there
are-
there-
are
five
of
those
ssps
here.
B
This
is
just
a
little.
The
motivation
that
dave
was
talking
about
before
was
well.
You
know
land
use,
even
though
current
day
it
isn't
like
such
a
massive
amount
of
carbon.
That's
coming
out.
If
we
look
historically,
it's
not
until
about
1950s
that
actually
fossil
fuel
emissions
exceed
land
use
emissions,
and
you
can
see
that
in
terms
of
a
third
of
all
of
the
carbon,
that's
actually
in
the
atmosphere
from
human
activities
comes
from.
Land
use
and
land
cover
change
just
quickly.
B
When
we
represent
land
use
and
land
cover
change,
we
can
actually
then
run
the
model
in
different
ways,
so
we
can
do
a
a
first.
We
can
do
a
full
land
use,
land
cover,
change
versus
a
no
land
use
land
cover
change,
and
that
allows
us
to
do
the
carbon
analysis.
I'm
going
to
call
it
there,
because
that's
my
three
minutes.
So
there
we
go
and
I'll
stop.
Sharing.
C
Okay,
so
I'm
gordon
bonan,
a
senior
scientist
in
the
terrestrial
sciences
section
I'm
going
to
talk
about
the
surface,
energy
balance,
canopy
fluxes
portion
of
clm-
and
this
is
fundamental
to
clm,
regardless
of
all
the
bells
and
whistles
and
everything
clm
does.
It
has
to
provide,
what's
called
the
sort
of
the
lower
boundary
conditions
required
by
the
atmosphere
model,
and
this
is
albedo,
long
wave
radiation
and
then
the
turbulent
fluxes
of
momentum
sensibly,
latent
heat
and
water
vapor.
C
The
way
that
this
is
done
is
actually
by
writing
out
equations
for
the
surface
energy
balance
and
then
solving
for
the
temperature
that
satisfies
the
energy
balance
and
normally
what
this
is
is
it's
the
absorb
solar
and
absorb
long
wave
radiation
the
it
must
be
balanced
by
the
emitted,
long
wave
radiation,
the
sensible
heat
flux,
the
latent
heat
flux
and
soil,
heat
storage,
the
emitted,
long
wave
radiation,
sensible
heat
latent
heat
and
soil
heat
storage
can
be
written
as
a
function
of
a
surface
temperature
and
what
the
clm
does
is
normally
solves
for
a
surface
temperature
that
balances
this
energy
budget.
C
C
C
The
clm
is
solving
for
the
vegetated
temperature
and
the
ground
temperature
that
and
the
sort
of
surface
air
temperature
that
or
canopy
air
temperature
that
satisfies
these
equations.
On
the
canopy
side,
it's
a
big
leaf,
canopy
and
there's
no
vertical
structure
to
it.
All
it's
just
sort
of
a
a
a
leaf.
C
It
can
be
divided,
it's
divided
into
sunlit
and
shaded
components,
but
there
is
no
vertical
structure
and
that
really
contrasts
with
the
soil,
which
has
20
layers
going
to
eight
and
a
half
meters,
five
bedrock
layers
down
to
50
meters
and
12
snow
layers.
So
the
soil
processes
are
resolved
vertically,
but
the
canopy
processes
aren't
and
I'll
come
back
to
that.
In
a
minute,
the
standard
way
of
testing
these
models
is
to
compare
to
eddie
covariance
flux
tower
data.
C
This
is
a
really
old
study,
one
of
the
first
studies
that
actually
did
this
in
boreal
forest
back
in
1997,
and
I
put
this
up
just
to
sort
of
say.
I
really
think
the
development
of
these
models
has
been
advanced
in
parallel
with
eddie
covariance
flux
tower
data
we
couldn't
have
done
it
without
those
data.
Sets
there
the
the
code
that
we
used
to
do
the
surface,
fluxes
and
clm
goes
back
30
years
ago,
or
so
to
bob
dickinson's
biosphere
atmosphere,
transfer
scheme.
There
are
other
ways
of
doing
things
more
advanced
ways.
C
This
is
a
multi-layered
canopy
that
I've
been
working
on
that
actually
does
resolve
profiles
through
the
canopy,
and
the
key
point
I
want
to
say
here
is:
I
really
think
the
physics
and
physiology
of
these
this
model
is
simpler
and
more
consistent
with
theory
than
is
the
clm
big
leaf
canopy.
So
you
can
think
of
this.
As
in
terms
of
is
this
a
complexity,
adding
complexity
or
is
it
actually
reducing
complexity?
I'd
say
it's
reducing
complexity,
and
the
final
thing
I
want
to
end
with
is
this
slide.
C
That
dave
lawrence
showed
yesterday
about
clm
is
just
the
starting
point
for
your
science.
It's
not
the
science
itself.
You
know
dave
did
a
really
good
job
of
sort
of
summarizing
this,
particularly
this
idea
of
easy
to
run
the
model
and
get
an
answer,
but
it's
much
harder
to
understand
why
you
get
this
answer.
I
think
that
this
bullet
he
left
off
and
I
think
it's
an
important
one
just
because
it's
in
the
model
doesn't
mean
it's
correct.
C
We
like
to
think
we
know
what
we're
doing,
but
we
don't
always
get
things
right
and
you
really
need
to
be
critical
about
what's
in
the
model,
just
because
it's
there
doesn't
mean
we're
actually
doing
it
right,
it
doesn't
mean
that
it
can't
be
corrected,
and
I
think
surface
fluxes
really
point
to
that
and
I'll
leave
it.
There.
E
A
E
C
Yeah,
that's
the
diffuse
radiation
penetrates
more
deeply
into
the
canopy,
so
you're,
seeing
the
shaded
leaves
being
more
illuminated.
Also,
I
do
want
to
point
out
that
the
sunlit
and
shaded
leaves
are
only
used
to
calculate
photosynthesis
and
stomata
conductance.
C
The
energy
balance
calculations
are
still
done
for
a
single
big
leaf,
so
in
effect,
what
you're
doing
is
you're
getting
a
sunlit
and
shaded
conduct,
stomata
conductance
that
gets
integrated
to
a
canopy
conductance
and
that
single
canopy
conductance
is
used
to
calculate
fluxes
for
a
single
big
leaf.
So
it's
not
a
full
big
sunlit
shaded
model,
it's
a
partial,
sunlit
shaded
model,
but
alan
the
the
point
you
make
about
diffuse
radiation
is
is
correct
and
that
will
show
up
through
the
photosynthesis
and
stomata
conductance
of
the
sunline
and
shaded
leaves.
C
A
Great
so
we'll
quickly
jump
to
biogeochemistry
and
I
kind
of
think
of
clm
as
having
two
big
components,
and
so
one
is
by
geophysics,
which
you
know
the
hydrology
and
canopy
fluxes.
That
gordon
was
talking
about
are
part
of
the
biogeophysics
and
then
subsequently,
we
think
about
biogeochemistry
and
the
reason
for
thinking
about
biogeochemistry
is
kind
of
evident
by
a
plot
that
many
of
you
have
seen
one
flavor
of
before.
A
So
this
is
actually
data
from
not
from
mauna
loa
and
the
keeling
curve
that
that
many
of
us
are
used
to
seeing,
but
also
from
actually
from
nywat
ridge,
which
is
just
up
the
road
from
where
we
are
in
boulder,
and
I
got
to
drag
my
family
up
there
a
couple
of
winters
ago
and
it's
you
know
one
of
the
it's
one
of
these,
like
you
know
to
me:
it's
seminal
places
where
science
happens
and
it's
like
a
shack
at
12
000
feet.
A
So
this
is
where
the
data
are
collected
from
in
this
picture,
but
it's
after
manilow,
it's
one
of
the
longest
co2
records
that
we
have
and
dave
showed
this
figure
yesterday.
But
I
feel
like
it's
helpful
for
me
as
a
biogeochemist
to
think
about
how
new
this
science
is,
and
so
things
like
the
carbon
cycle
really
only
came
into
earth
into
large
scale.
Earth
system
models
really
around
the
turn
of
this
century.
A
So
since
my
professional
career
started
and
then
things
like
nutrients
came
on
a
little
bit
after
that,
and
so
even
though
biochemistry
and
carbon
cycle
science
are
old,
they're
still
relatively
new
in
terms
of
their
science,
and
so
this
is
the
picture.
You
know
kind
of
a
blown
up
version
of
what
biogeochemistry
is
trying
to
do
in
clm,
and
so
as
opposed
to
trying
to
go
over
all
of
this,
which
is
impossible
in
three
to
five
minutes.
A
What
I
did
was
in
this
talk
linked
the
kind
of
highlights
of
what
we're
doing
in
biogeochemistry
to
parts
of
the
tech
note,
and
so
you
can
I'll
share
a
link
to
this
talk
in
just
a
minute
in
the
chat.
But
you
can
follow
these
links
to
look
at
you
know
what
are
we
doing
for
photosynthesis?
What
are
we
doing
for
allocation?
A
You
know,
but
basically
there's
kind
of
four
to
five
parts
of
what
we're
doing
for
biogeochemistry.
So
photosynthesis
is
obviously
the
first
one.
Then
after
plants
have
some
gpp
to
to
deal
with,
they
have
to
allocate
it
to
different
components.
A
You
know
to
leaves
or
to
find
roots
or
to
wood
and
all
those
different
that
allocation
has
different
stoichiometry,
that's
associated
with
it,
that's
specific
for
each
pft,
based
on
what
kind
of
tissue
you
have
and
how
much
nitrogen
is
in
it
that
determines
the
autotrophic
respiration
rates
and
then
and
then
phonology
is
kind
of
a
big
lever
in
the
model
and
we'll
talk
a
little
bit
more
about
phonology
in
the
practical
today
and
then
there's
also
mortality
and
turnover.
A
A
If
you
want
to
dig
a
little
bit
deeper,
then
the
tech
note,
the
tech
notes
are
good
kind
of
high
level
overview,
but
I
actually
want
to
look
at
the
code.
Here's
some
of
the
modules
that
you
can
look
at
and
these
are
in
the
source
code,
which
is
under
src
and
then
there's
biogeophysics,
where
photosynthesis
is
handled
by
geochemistry,
where
all
the
carbon
and
nitrogen
stuff
is
handled,
and
then
soil
biogeocam
is
its
own
directory.
That's
where
all
the
below
ground
stuff
happens,
all
the
action,
so
that's
it
for
biogeochemistry.
G
Great
okay,
so
I'll
give
you
a
quick
overview
of
crops
and
salem
five.
Really.
I
want
to
start
with
just
emphasizing
that
agriculture
changes
the
land
surface.
It
changes
both
how
it
looks
which
is
really
important
for
physical
impacts,
and
it
also
changes
how
it
functions,
which
is
really
important
for
biogeochemical
impacts,
and
for
these
reasons
it's
really
important
to
understand
and
to
stimulate
the
role
of
agriculture
and
agricultural
management
within
the
earth
system,
and
that's
part
of
the
motivation
for
us,
including
this
in
cesm.
G
Currently,
we
have,
I
guess,
what
I'll
call
eight
crop
types
there
are
some
additional
crop
types
and
that
we
differentiate
between
temperate
and
tropical
corn
and
soybean
we
do
have
bioenergy
crops
are
new.
We
also
have
winter
and
spring
wheat
in
clm5,
so
we
have
a
lot
of
different
kinds
of
crop,
a
lot
of
the
major
kinds
of
crops.
G
G
G
Crops
are
a
little
bit
different
than
native
vegetation
in
that
they
have
active
phenology,
and
currently
we
have
four
phenological
phases.
The
first
is
planting.
The
second
is
leaf
emergence.
The
third
is
grain
fill
and
the
fourth
is
harvest.
So
each
of
these
is
a
trigger
for
a
new
phenological
phase.
G
They're
all
temperature
based
triggers
so
keep
that
in
mind,
there's
no
soil,
moisture
or
nutrient
stress
or
anything,
and
I
want
to
highlight
two
forthcoming
developments,
because
they're
gonna
allow
for
a
little,
hopefully
better
representation
of
phenology
in
clm5.
The
first
is
work
led
by
bin
peng,
which
will
is
integrating
that
apsim
model
into
clm.
G
We
have
two:
two
types
of
management
are
included.
Currently,
the
first
that
I'll
highlight
is
fertilization.
Fertilization
happens
by
crop
type
and
by
year
and
is
again
based
on
the
luh2
data
it
is
applied
when
the
leaves
emerge
we
have
manure.
That
is
constant,
is
a
constant
application
through
time
to
grams
of
nitrogen
per
meter
squared
per
year,
and
then
we
have
chemical
fertilizer,
which
is
prescribed
by
the
crop
type
the
year
and
the
country
we
do
have
some
development
needs,
but
this
is
a
decent
representation
of
fertilizer
for
the
time
being.
G
The
other
other
management
method
that
I'll
highlight
here
is
irrigation,
so
irrigation
is
applied
daily.
If
soil
moisture
falls
below
a
fixed
threshold,
we
currently
have
a
couple
different
application
methods.
Drip
is
available
by
default.
Sean
swenson,
also
developed
a
method
for
sprinkler,
irrigation
and
yang
is
developing
a
flood
irrigation
method
that
is,
that
we're
hoping
to
incorporate
into
clm
there's
also
multiple
water
sources
by
default.
The
water
comes
from
rivers
or
surface
water,
but
there's
also
groundwater
and
reservoirs
that
are
available.
G
The
reservoirs
is
under
development
by
ena
van
der
kellen
I'll.
Just
give
you
a
quick
highlight
of
model
performance
first
is
that
we
see
the
global
crop
yields
increase
through
time,
so
the
black
line
is
clm,
the
red
line
is
unfa-o,
and
so
these
are
global,
crop
builds
and
what
you
can.
G
What
you'll
notice
here
is
that,
from
you
know
the
when
data
were
first
available
through
about
the
1990s
clm
does
a
decent
job
of
predicting
global
crop
yields,
but
then
it
levels
off
and
that's
because
we
don't
include
processes
related
to
agricultural
intensification.
G
You
know
technological
advances,
planting
advances,
things
like
that
I'll
also
just
highlight
really
quickly:
corn
yields
compared
to
an
fao,
earthstat
dataset
that
we
put
together-
and
this
is
this-
this
is
probably
the
crop
that
we
do
the
worst
on
in
terms
of
looking
at
spatial
patterns
of
yield-
and
I
want
to
highlight
two
things:
one
is
that
crops,
corn
yields
are
overestimated
or
sorry
underestimated,
and
temperate
latitudes
and
they're
overestimated
in
tropical
latitudes.
G
So
I'll
I'll
stop
there.
B
D
Hold
on
one
second,
so
yeah
I'll
be
talking
about
fates,
the
functionally
assembled
terrestrial
ecosystem
simulator,
and
I
want
to
take
a
minute
before
I
start
to
acknowledge
the
global
fates
team.
This
wouldn't
be
possible
without
an
incredible
number
of
people
coming
together
to
collaborate
and
work
on
this,
in
particular,
charlie,
coven,
rosie,
fisher
and
ryan
knox.
D
Okay,
it's
me
every
time,
and
so
you
heard
a
lot
about
yesterday.
What
terrestrial
ecology
is
in
earth
system
models
and
dave
probably
mentioned
this,
but
the
way
that
land
is
implemented.
The
way
that
vegetation
heterogeneity
is
implemented
ignores
complex
vegetation
structure
and
so
we're
missing
a
lot
of
detail,
and
so
within
most
earth
system
models
you
have
a
big
leaf
and
so
a
pool
of
carbon,
and
then
those
pools
are
separated
by
a
given
area.
D
Cohort
models
are
the
in-between
solution,
and
so
what
they
do
is
they
capture
different
sizes
of
different
types
of
vegetation,
using
a
density
for
each
size,
and
so
you
get
more
information
with,
hopefully
improved
computational
costs
over
an
individual
based
model,
and
so
fates
is
a
cohort
model.
D
It's
a
module
that
functions
within
the
community
land
model
and
it
replaces
the
traditional
big
leaf
pooled
vegetation
with
more
realistic
vegetation.
This
allows
us
to
get
plant
physiology
competition,
dynamic
assembly
and
vegetation
distribution,
which
is
still
cutting
edge
for
these
types
of
models
and
so
within
fades.
The
fades
portion
handles
everything
in
this
yellow
box.
Fates
handles
all
the
vegetation
structure,
regeneration
seeds
as
well
as
physiology,
and
then
it
interacts
with
the
host
land
model
and
it
interacts
with
clm.
D
It
can
also
interact
with
elm,
and
so
those
interactions
have
different
layers
of
coupling
in
terms
of
how
frequently
those
that
information
is
updated,
updated
with
each
time
step.
Something
like
photosynthesis
is
tightly
coupled
and
updated
multiple
times
within
a
time
step,
and
so
within
fates.
We
have
multiple
patches
where
a
big
leaf
model
has
different
pools
of
vegetation.
D
We
have
multiple
patches
that
track
time
sense
disturbance,
and
so
you
can
have
a
newly
disturbed
forest
within
a
grid
cell
and
then
on
a
different
patch.
Within
that
grid
cell
you
can
have
a
mature
forest
or
an
old
growth
forest,
and
so
what
this
means
is
that,
if
you
think
back
to
the
difference
between
the
host
land
model
patch
versus
a
fades
patch
within
the
host
land
model,
you're
going
to
have
a
single
pool,
giving
you
information
about
whatever
process
you're
interested
in
these
coupled
processes,
but
within
fates.
D
We
have
a
heterogeneous
patch
and
so
that
information
is
depending
on
the
process.
You're
talking
about
is
going
to
be
an
emergent
emergent
from
the
structure
on
that
patch
and
so
fundamentally,
if
you're
thinking
about
canopy
turbulence,
you're
going
to
get
a
different
picture
when
you
consider
this
heterogeneous
canopy,
this
is
area
that
is
in
need
of
development.
So
if
you're
interested
in
this
please
get
in
touch,
there
are
a
lot
of
instances.
Instances
in
reality,
where
structure
is
essential.
D
Hydrodynamics
we
have
a
fates,
hydro,
module
that
tracks
a
lot
of
detailed
information
to
get
to
capture
more
details
about
the
structure
nutrients
we
have
parte,
which
is
the
nutrient-enabled
model
fire.
I'm
not
going
to
read
the
details
of
these
you'll.
Have
access
to
this
talk?
Size
structure
is
essential
to
mortality
and
regeneration
processes
associated
with
fire.
You
get
variability
and
snow
occlusion
pests,
pests,
differentially
attack,
different
sizes
of
trees,
different
types
of
trees,
harvest.
D
We
all
know
that
harvest
you
don't
go
and
harvest
everything
you
want
to
choose
the
trees
or
the
sizes
of
trees.
You
want.
I
talked
about
canopy
turbulence.
There
are
many
processes
where
structure
is
an
important
component
of
it,
and
so
we
feel
that
fates
gives
us
more
information
than
what
we
have
with
the
big
leaf
model.
D
A
A
Looks
like
hannah
asked
a
question
in
the
chat.
Looking
at
the
outputs
I
figured
the
level
of
pft
given
in
fates
has
dimensions
of
12.,
so
that
means
fates
can
only
go
up
to
12
pfts.
B
H
Okay,
I'm
just
gonna
give
a
quick
overview
here
of
this
hills:
hydrology
model
that
we've
been
developing.
H
And
my
slide
won't
go
forward:
okay,
okay!
So,
as
was
mentioned
yesterday
in
other
talks,
clm
or
ctsm
uses
this
hierarchical
data
structure
and
on
the
vegetated
land
unit.
The
normal
configuration
consists
of
a
a
single
column,
sort
of
describing
the
the
soil,
and
then
inside
that
column
there
can
be
different
kinds
of
vegetation
arranged
in
patches
and
that's
typically
how
the
models
run
and
has
been
run
in
the
past.
Although
you
know,
as
you
know,
crops
for
example
can
exist
on
their
own
columns.
H
H
Another
way
that
you
might
do
that
is
to
actually
connect
those
soil
columns
via
the
subsurface
lateral
flow.
Some
kind
of
called
drainage
called
base
flow,
and
so
then
these
columns
can
be
connected
within
the
model
and
they
can
pass
water
back
and
forth.
H
Terms
of
their
vertical
processes
with
it
and
then
the
water,
from
the
lateral
flow
capacity
and
columns,
and
so
the
reasons
that
we're
interested
in
doing
this
is
is
the
typical
model
has
no
real
notion
of
typography
within
it.
If
you
look
across
the
landscape-
and
there
are
obvious
correlations
and
covariances
between
patterns
in,
for
example,
vegetation
and
topography.
H
So
these
are
some
examples
of
places
where
you
see
these,
in
some
cases,
sharp
gradients
and
vegetation,
and
these
are
related
to
the
topography
and
the
soil
moisture
within
them,
and
so
what
this
might
allow
us
to
do.
That
is
look
at
some
of
the
spatial
covariances
between
things
like
vegetation
and
position,
them
hill,
slope
and
soil,
moisture
and
position
in
the
slope
and
start
to
look
at
why
some
of
these
patterns
emerge.
H
There's
different
ways
to
connect
these
columns
so
on
the
left
shows
a
standard.
You
know
single
column
connected
to
the
routing
network.
You
could
do
you
could
do
serial
or
parallel
connectivity
or
you
could
do
some
sort
of
hybrid
connectivity
and
all
these
things
can
be
specified
on
your
input.
Data
set.
H
What
sort
of
outputs
look
like
are?
Oh
sorry,
I
should
also
say
we
can
also
do
downscaling
of
of
meteorology
based
on
the
characteristics
of
the
hill
slopes,
and
so
we
now
can
look
at
the
impacts
on,
for
example,
insulation,
incoming
radiation
due
to
sloping
aspects.
So
here's
an
example
of
the
western
us
comparing
mean
annual
solar
radiation,
the
top
left
would
be
a
north
facing
and
the
bottom
left
would
be
south
facing,
and
so
you
can
see
these
sort
of
large
differences
in
the
annual
mean
in
terms
of
incoming
solar
radiation.
H
Due
to
these
topographic
effects,
we
can
also
do
precipitation
downscaling.
So
both
the
amount
and
the
partitioning
between
rain
and
snow
are
affected
by
elevation,
gradients
and
again
just
an
example
of
showing
the
evolution
of
the
snowpack
for
a
low
elevation
at
the
top
left
and
a
high
elevation
at
the
bottom
right.
So
you
can
see
that
the
snowpack
persists
longer,
due
to
both
differences
in
temperature
and
differences
in
this
partitioning
in
the
amount
of
of
precipitation
and
then
finally,
just
an
example
of
what
this
might
look
like.
H
So
here
this
would
be
a
a
grid
cell
that
would
have
20
columns
within
it,
and
so
you
get
this
essentially
a
distribution
of
soy
moisture
state.
So
these
are
all
soil,
moisture
depth
and
time,
and
you
can
have
a
bunch
of
columns
which
are
relatively
dry
and
then
you
can
see
the
impact
of
the
convergence
of
subsurface
lateral
flow,
and
you
also
have
some
columns
in
which
they're
much
wetter
and
an
active
water
table
exists.
H
I
Yeah,
can
you
hear
me,
I
think,
I'm
on
yep
cool
all
right.
Let's
see
if
I
can
share
my
screen.
I
All
right
are
we
good?
Okay,
so
yeah,
I
am
brett
raska,
I'm
part
of
the
data
assimilation,
research
section
here
at
ncar,
and
we
have
developed
an
ensemble
da
system
that
you
can
couple
with
clm
called
the
data
assimilation,
your
research
testbed,
and
so
you
know,
data
simulation
as
many.
You
could
be
aware.
It's
when
you're
fusing
information
from
your
model
forecast
with
observations
to
help
give
you
the
best
idea.
I
What's
going
on
as
possible
and
so
to
kind
of
put
this
in
context
of
what
you've
learned
this
week
in
different
modes
of
clms,
you
have
clm
bgc,
so
the
full
biogeochemistry,
which
has
really
no
external
constraints,
relying
on
all
these
model
mechanisms
and
developments
and
parameterizations.
To
give
you
an
internally
prognostic
simulation
of
what
the
land
surface
is
doing.
I
I
I
Clm
dart,
when
they're
coupled
together
is
kind
of
the
next
step
from
there,
where
you
can
basically
bring
in
any
observation,
preferably
something
that's
simulated
or
closely
simulated
by
clm,
and
so
you
can
kind
of
constrain
behavior
from
multiple
directions.
It
also
provides.
You
uncertainty
estimates
on
top
of
that,
and
so
how
do
we
do
that
so
really
quickly?
I
The
ensemble
da
system
uses
a
bayesian
approach,
and
so
the
best
model
forecast
of
the
posterior
in
blue
is
a
combination
of
the
forecast
generated
from
clm.
So
it's
like
the
prior
estimate
combined
with
observations,
and
so
we
fuse
those
two
things
together,
taking
into
account
uncertainty
to
come
up
with
the
best
solution
for
what
the
the
land
surface
is
doing
and
so
on
the
bottom
part
of
the
screen.
I
Here
I'm
trying
to
illustrate
that
a
bit
further,
so
we're
we're
literally
generating
an
ensemble
of
c
ln,
runs
and
multi-instance
run
and
we're
getting
that
through
sampling,
the
uncertainty
in
cam,
so
the
atmospheric
model
and
cesm
and
those
are
driving
different
clm
instances
and
sampling
the
uncertainty
in
the
model
forecast.
We
run
that
forward
in
time
and
then
we
come
to
a
point
where
we
have
a
series
of
observations
and
then
we
estimate
what
the
model
thinks.
I
Those
observations
should
be
going
up
to
step
three
and
we
compare
them
to
the
true
observations
and
then
adjust
those
more
closely
such
that
the
model
is
representing
reality
of
what
we
observe
and
then
the
final
step.
Is
we
take
that
information?
We
can
apply
it
to
the
rest
of
the
complete
model
state
of
the
system,
and
so
just
I'm
just
trying
to
show
an
overview
of
different
types
of
work
that
we've
been
involved
in
so
assimilating
leaf
area.
I
So
you
can
imagine
if
you
get
a
better
idea
of
what
hydrology
is
doing,
that
has
implications
for
the
carbon
cycle,
but
it
also
has
implications
for
the
surface
albedo
in
the
surface
energy
balance
as
well,
and
so
really
quickly.
Just
wanted
to
advertise.
Some
of
the
tutorials
I
mean
generally
dart
is
recommended
for
more
advanced
users,
but
it's
a
good
point
to
introduce
this
to
all
like
new
users
as
well.
I
So
you
can
go
and
use
a
toy
model
to
test
these
concepts
and
when
you're
familiar
with
that,
you
can
go
into
the
clm
5
dart
tutorial,
which
I've
designed
and
just
is
just
giving
a
quick
overview
of
how
it's
organized,
and
so
I
kind
of
included
13
steps
and
the
left
right
there
so
going
through
like
detailed
instructions
on
the
downloading
the
setup
running
and
the
diagnosis
of
your
assimilation
run
for
a
simple
example
and
just
kind
of
an
example
on
the
right.
I
I
took
excerpts
of
what
kind
of
documentation
what
the
documentation
looks
like
so
instructions
detailed
examples
on
what
the
scripts
should
look
like
in
a
bunch
of
definitions,
so
you
can
understand
what's
going
on
there,
this
is
so
the
clm5.tutorial
isn't
containerized.
It
does
assume
that
you
have
a
count
on
cheyenne
and
generally,
when
you're
using
ensemble
systems.
It
is
recommended
that
you,
you
have
a
high
performance
computer,
but
most
of
the
initial
tutorials
can
be
done
on
your
local
machine,
and
so
that's
it.
A
Some
thanks
brett,
I
think
keith
you're
up
next.
If
brett
you
can
stop
sharing
so
we're
kind
of
transitioning
from
more
of
the
model.
A
I
don't
know
direct
model
science
and
code
to
like
other
features,
so
things
like
dart
and
then
diagnostics
that
keith
will
talk
about
here
and
I
can
see
there's
a
bunch
going
on
in
the
chat
that
I'm
not
totally
able
to
follow.
So
I'm
hoping
that
other
folks
are
able
to
to
kind
of
follow.
What's
going
on
in
the
chat
that
can
help
out
here.
A
A
E
Know
if
you
have
a
microphone
yeah
peter's
saying
I
have
low
resources
or
something
like
that.
A
E
E
Yeah
some
people
said
within
the
terrestrial
sciences,
section
they've
kind
of
covered
islam.
Yesterday's
talk,
but
thailand
is
the
international
landmark
benchmarking.
A
E
Yeah,
so
this
is
a
fair
amount
of
documentation
on
ilam
there's
a
paper
in
james.
It
describes
the
design,
the
theory
and
implementation.
E
E
So
I
thought
I'd
do
is
take
a
look
at
some
of
the
example
outputs.
Assuming
this
this
will
work.
So
after
the
package
is
run,
I
can
copy
the
results
over
to
a
web
server.
E
A
E
D
Hey
keith
we're
we're
seeing
the
we're
seeing
your
perspective
of
the
powerpoint.
We
are
not
seeing
the
web
page
like
it
seems
like
you
need
to
swap
screens
there
you
go
now.
It's
now
you're
on
that's
right!
That's
right!.
E
Yeah,
so
this
is
the
main
page.
So
in
this
column
here
come
five,
it's
the
first
between
c115
and
ctsm51,
and
then
these
are
essentially
a
summary
score
page.
Where
these
kind
of
orange
orange
values
means
the
model
is
doing
worse
than
the
other
model,
and
the
blue.
E
The
blue
colors
are
one
of
the
models
doing
better
and
you
see
these
categories
so
there's
like
ecosystem
carbon
cycle
category,
hydrology
cycle,
category
radiation,
energy
cycle,
forcing
forcings
and
then
relationships
relationships
are
pretty
interesting
so
especially
like
how
evapotranspiration
varies
with
methory
index
and
observations
and
in
the
model
it's
a
number
of
those
relationships.
E
E
So
this
is
the
kind
of
the
mean
state
page,
there's
a
table
here
of
statistics
for
the
two
models
and
then
and
then
the
scores
there's
like
a
bias,
score
spatial
distribution,
score
and
an
overall
score
and
there's
lots
of
observations.
The
benchmark
mean
model
mean
prized
by
a
score.
It's
a
nice
tailored
diagram
down
here
compared
to
models.
E
D
A
And
so
dave
and
keith
kind
of
both
mentioned
this,
but
we
use
a
diagnostic
package
and
ilam
regularly
as
we're
doing
model
development
just
to
see
what
what
parts
of
the
model
are.
Looking
better
with
different
code
changes
that
come
in
and
which
ones
are
looking
worse.
F
Yes,
thank
you.
Dave
yeah,
just
didn't
realize
that
okay
awesome,
so
I'm
gonna
talk
a
little
bit
about
software
engineering
for
clm,
and
my
main
main
thing
I'm
gonna
talk
about
is
why
you
should
care
about
software
engineering.
So
that's
my
next
slide.
F
Let's
see,
why
do
you
care
fundamentally,
ctsm
is
science
that's
expressed
in
terms
of
software,
so
fundamentally,
ctsm
is
a
software
project
and
ctsm
is
a
scientific
instrument
and,
like
other
precision,
scientific
instruments
needs
to
be
managed
in
a
way
that
allows
for
the
science
done
with
it
to
flourish,
and
so
we
do
need
to
pay
attention
to
the
software
engineering.
F
Using
this
is
the
part
that
why
you
should
care
is
using
good
software
engineering
practices
will
help
the
development
of
your
science.
So
there's
some
things
that
I'm
going
to
talk
about
that
are
really
going
to
help
you
to
manage
your
science.
You
know
get
your
science
changes
in
by
following.
You
know,
good
practices,
and
the
other
thing
is
that
following
good
practices
will
make
it
easier
for
your
work
to
come
back
into
ctsm.
F
So,
if
you're
doing
something
that
you,
you
think,
would
be
useful
for
the
community
and
you
want
to
see
it
come
back
into
ctsm,
the
better.
Your
practices
are
the
more
likely
that
is
to
happen.
F
The
the
main
thing
and
everybody
finds
this
in
their
own
way
is
that
you
need
to
save
off
working
versions.
So
I
in
the
beginning
of
my
career,
I
did
not
use
version
control,
but
I
did
find
very
quickly
that,
having
having
saved
versions
that
worked
was
really
important
because
you
can,
you
can
make
us,
make
a
change
and
then
something
breaks
and
then
finding
that
finding
what
broke
is
really
really
difficult.
F
If
you
don't
have
saved
versions
that
you
know
work
so
because
the
truth
is
that
debugging
is
always
very
hard,
but
comparing
code
to
working
versions
can
be,
can
be
much
easier
way
to
find
problems
and
usually
what
you
will
be
doing
will
be
changing
changing
the
model,
changing
the
existing
model
in
way
in
ways
that
you
know
it
wasn't
designed
for
and
so
you're
going
to
find
problems,
and
by
being
able
to
compare
to
working
versions,
you're
going
to
be
able
to
find
those
problems,
much
easier.
F
So
what
that
above
means
is
that
you
really
need
to
use
version
control
of
some
sort
now,
sometimes
what
people
do
is
that
they,
you
know,
save
up
for
inversions
with
just
copying
just
copying
files
around
which
isn't
as
efficient
as
using
a
version
control
system
and
in
the
good
tutorial
we've
already
kind
of
introduced
you
to
get,
and
we
recommend
that
you
use
that.
F
That
system
that
you
use
you
you
take
advantage
of,
get
it'll,
make
it
easier
for
your
development
and
easy
for
your
science
to
to
be
worked
out.
So
next
question
is:
how
do
you
interact
with
us?
F
We
have
limited
resources,
but
we
do
want
you
to
send
reports
on
problems
you
run
into
with
issues
in
github,
for
example,
and
if
something
you're
doing
may
be
of
general
interest,
you
should
create
an
issue
in
github
as
well,
and
if
the
land
model
working
group
is
interested
in
your
work,
you'll
likely
cycle
through
from
an
issue
from
having
an
issue
to
then
having
a
pull
request
and
then
finally,
your
feature
becoming
part
of
the
model
and
another
way
to
interact
with
us
is
to
interact
with
the
community
in
in
the
discuss
csm
forums,
which
I'm
going
to
talk
about
a
little
bit
and
actually
in
the
chat.
F
We
were
talking
about
the
discuss,
csm
forums
where
there's
a
question
that
was
on
the
forums
that
was
being
answered
and
the
advantage
of
the
discussed
csm
forums
is
that
you
interact
with
not
only
those
of
us
at
ncar
that
are
working
on
the
model,
but
you
work
with
the
whole
community,
that's
working
on
csm
and
clm.
F
So
that's
a
great
way
to
do
it
so
and
then.
Lastly,
some
websites
for
help.
So
the
cts
on
wiki
page
is
here.
F
There's
a
development
wiki
that
shows
how
to
shows
the
development
cycle
and
what
you
should
do
in
terms
of
get
and
setting
up
branches
and
that
sort
of
thing,
and
then
they
discuss
csm
forums
and
I'm
just
going
to
quickly.
Go
there
see
if
this
works.
It
worked
for
keith,
so
maybe
it'll
work
for
me,
okay
looks
like
it
does
okay,
so
this
is
the
github
page
and
the
wiki.
F
F
Development
workflow
is
a
good
one
to
go.
Go
through
that
talks
about
testing
there's.
Also,
this
quick
start
to
ctsm
development
with
git,
and
that's
really
the
main
one.
It
talks
about
learning,
git
setting
up,
get
creating
a
fork,
creating
a
branch,
etc.
F
A
Yeah
that
looks
better
and
can
you
scroll
to
the
top
it
was.
I
just
had
a
hard
time
kind
of
following,
so
you
know
a
lot
of
times.
A
There's
you
know
erica,
like
eric
said
he
always
goes
to
code,
because
that's
where
he's
working
all
the
time,
I
always
go
to
the
wiki
until
there's
that
kind
of
wiki
book
icon,
because
there's
just
an
unbelievable
amount
of
resources
there,
but
those
little
tabs
at
the
top.
That
sort
of
look
innocuous
are
unbelievably
helpful
in
navigating
through
a
github
page,
if
you're
not
used
to
doing
that
or
if
you
have
a
thousand
browser
windows
open
like
eric
does
exactly
I
do
so.
A
I
just
wanted
to
jump
in
because
for
users
that
are
new
to
github,
navigating
the
buttons
on
the
top,
I
think,
is
one
of
the
more
important
features
that
is
helpful
to
know
about.
F
Yeah
and
yeah,
so
the
code
is
in
and
you
can
browse
the
code
this
way
like
it.
I
talked
about
you
know,
making
issues,
and
so
you
could
make
a
new
issue
here.
You
could
look
through
existing
issues.
Sometimes
you,
you
might
find
a
problem
in
the
model
and
you
wonder
about
has
somebody
else
run
into
this
and
you
can
you
can
search
and
see
if
somebody
else
has
found
the
same
problem
again?
F
If
you
have
work,
that's
going
to
come
into
the
main
model,
that's
going
to
be
done
through
a
poll
request
so
that
you
might
look
at
that.
We
do
have
some
projects
that
might
be
useful
to
look
at
and
see.
You
know
where
big
projects
are
and
then
of
course,
the
wiki
which
has
a
bunch
of
resources
on
on
helping
to
work
with
the
model.
F
So,
let's
see
oh
and
the
last
thing
I
wanted,
I
just
wanted
to
show
the
discuss
csm
forum
and
this,
so
the
main
forum
is
for
all
of
csm
and
there's
a
forum
for
ctsm
clm.
Those
are
rtm,
and
that's
this
one,
and
it
looks
like
the
question
that
was
asked
in
the
chat,
was
here
and
adriana
answered
that
question
14
minutes
ago.
So
this
is
another
great
resource
to
work
with
the
community
on
asking
questions
on
how
you
do
something
or
finding
problems
and
solving
issues.
So
all
right,
that's
it
for
me.
A
Let's
go
ahead
and
take
a
10
minute
break
now
and
then
we'll
come
back
and
kind
of
regroup,
get
ready
for
the
practical
session.
So
come
back.
Get
we'll
give
you
a
12
minute
break,
come
back
at
10
after
the
hour.
That's
all
right!.