►
Description
This presentation provides an overview of the remote sensing methods employed by the Regen Network science team to measure soil organic carbon and other ecosystem health metrics at the field scale. In it, we provide a brief background on the power of remote sensing, highlight the methods used for grassland ecosystems, and present a case study walking through the quantification process.
Connect with our community 👇
• Website: https://cosmos.network/
• Twitter: https://twitter.com/cosmos
• Telegram: https://t.me/cosmosproject
• YouTube: https://www.youtube.com/c/CosmosProject
• Medium: https://blog.cosmos.network/
• GitHub: https://github.com/cosmos/cosmos-sdk
• Reddit: https://reddit.com/r/cosmosnetwork
A
Thanks
everyone,
so
much
for
joining
the
science
team
has
been
pretty
excited.
Putting
together
this
presentation.
I
know
it's
been
a
request
and
a
long
time
coming
to
really
talk
about
the
science
and
give
a
deep
dive
and
a
perspective
on
what
we've
been
working
on.
Gisele,
I
know
has
put
together
a
presentation
in
the
past
reviewing
some
of
this,
and
I
think
it's
just
really
good,
as
we
move
forward
with
this
methodology
and
upgrading
this
grasslands
methodology
to
really
give
a
background
and
understanding
of
what
we're
working
on.
A
A
A
A
Then
we'll
contextualize
this
with
an
example
of
wilmot
and
how
we
did
this
with
the
wilmot
farm.
We'll
briefly
touch
a
little
bit
on
the
co
benefits
and
the
methodology
we
use
there
to
calculate
and
quantify
the
co
benefits
we
align
with
each
project
and
then.
Finally,
as
sarah
mentioned,
we
will
have
gisele
talk
a
little
bit
about
the
frequently
asked
questions
and
then
really
provide
a
platform
to
just
ask
any
questions
you
have
of
us
in
terms
of
the
science
itself
in
terms
of
remote
sensing
or
the
carbon
plus
methodology.
A
So
that's
that's
the
overview
of
what
we're
going
to
talk
about
and
really
just
to
dive
in.
Like
I
mentioned,
I
think
it's
really
important
to
understand
some
of
the
basics
of
remote
sensing
before
we
dive
into
the
methodology
itself.
So
again,
this
is
a
very
brief
overview,
but
some
of
the
key
points,
at
least
I
consider
it
important,
as
we
just
discussed
remote
sensing
and
satellite
imagery.
So
on
the
left.
When
we're
talking
about
satellite
imagery,
there's
two
metrics
we
often
discuss
and
the
first
is
spatial
resolution
and
really
the
essence
of
this.
A
A
But
as
you
can
see,
the
more
detailed,
the
lower
the
the
pixel
size,
the
better
outputs
in
terms
of
reflecting,
what's
actually
happening
on
the
landscape.
So
that's
facial
resolution.
When
we
talk
about
spatial
resolution,
we're
just
talking
about
the
size
of
the
pixel
and
then
secondly,
temporal
resolution-
and
this
is
simply
just
the
time
between
images.
A
So
when
satellites
are
rotating
around
the
earth,
they're
taking
you
know
a
snapshot
of
a
certain
location
in
space
and
then
either
they
will
loop
back
around
or
a
dual
satellite
will
come
and
take
a
snapshot
of
that
same
location.
So
when
we're
talking
about,
say
temporal
resolution
of
five
days,
we're
essentially
saying
that
for
the
satellites
that
we
use
we're
going
to
be
getting
different
snapshots
every
five
days,
so
that's
kind
of
like
the
revisit
time.
A
So
that's
when
we're
talking
about
temporal
resolution,
that's
what
we're
referencing
and
then
we're
moving
over
to
the
right.
This
is
a
very
basic
breakdown
of
of
remote
sensing
and
we're
talking
when
we
talk
about
spectral
bands.
So
you
might
hear
us
when
sam
goes
through
our
methodology,
talking
about
aligning
different
spectral
bands
to
soil
organic
carbon
levels
and
when
we're
talking
about
spectral
bands,
what
we're
really
talking
about
is
the
alignment
to
different
wavelengths
within
the
electromagnetic
spectrum.
A
So
you
can
see
here
on
the
top
just
a
basic
infographic
on
the
electromagnetic
spectrum
and
then
down
below
you
can
see
sentinel
2
so
I'll
talk
about
this
in
the
next
slide,
but
that's
the
sensor,
the
satellite
we're
using
and
you
can
see
how
different
bands
are
aligning
to
different
wavelengths.
So
you
can
see
you
know
visible
light.
A
We
have
a
blue,
a
green
and
a
red
band
and
those
are
just
three
of
the
bands
that
sentinel
2
picks
up,
but
we
also
have
bands
outside
of
the
visible
light
spectrum,
so
you're
talking
about
near
infrared
or
short
wave
infrared,
so
we're
talking
about
different
bands.
You
know
we're
just
talking
about
these
different
ranges
within
the
electromagnetic
magnetic
spectrum
that
we're
able
to
pick
up.
A
So
that's
a
little
bit
of
background.
One
other
thing
I
wanted
to
touch
on
is
sentinel
2.
So
all
of
our
analysis,
thus
far
in
terms
of
our
carbon
plus
grassland
methodology,
is
using
sentinel
2
multi-spectral
instrument,
and
these
are
just
this-
is
a
quick
breakdown
of
some
of
the
specifications
of
this
satellite.
So,
first,
it's
a
european
space
agency
satellite
in
terms
of
data
set
availability,
we're
looking
at
2017
onward.
A
So
this
is
really
important
when
we
talk
about
onboarding
projects
with
historical
soil
sampling,
really
we
are
limited
in
terms
of
temporal
availability
by
the
data
set.
So
that's
why,
for
both
wilma
and
woodburn,
you
know,
even
though
they
may
have
samples
that
are
prior
to
2017,
we're
really
starting
with
2017
due
to
the
basis
of
that's
the
earliest,
we
can
get
sentinel
images.
A
So
that's
a
really
important
metric
as
we
onboard
projects
to
take
note
of
and
then
again
temporal
resolution
and
again,
that's
that
revisit
time
we're
looking
about
every
five
days,
spatial
resolution
we're
looking
10
to
60
meter
resolution.
Most
of
the
bands
we
use
are
actually
within
that
10
meter
resolution
framework
and
then
finally,
within
sentinel,
as
you
saw
in
that
previous
little
visual
there's
13
spectral
bands
that
it
takes
for
each
of
its
images.
A
A
I
think
a
great
example
of
this,
and
probably
the
most
famous
and
commonly
known
indices
is
ndvi,
and
this
stands
for
the
normalized
difference
vegetation
index,
it's
a
basic
index
on
vegetation
health,
and
I
think
it's
really
important
just
to
understand
kind
of
the
basics
of
what's
happening
here
so
on.
The
left
is
a
nice
little
infographic
and
you
can
see
that
in
terms
of
healthy
vegetation,
it
is
reflecting
about
50
percent
of
the
near
infrared
wavelength,
but
absorbing
a
lot
more
of
that
red
light
coming
in.
A
So
it's
not
reflecting
that
much
back,
but
for
a
stressed,
vegetation
or
a
stressed
plant,
you
can
see
there's
a
difference,
so
it's
actually
reflecting
back
more
of
that
red
light
and
absorbing
a
little
bit
more
of
that
near
infrared,
and
then
we
can
incorporate
this
into
an
equation.
You
can
see
down
at
the
bottom,
which
will
give
us
a
final
score
of
ndvi
and
moving
to
the
right.
A
You
can
see
how
that
reflects
in
terms
of
ranges,
so
you
can
see
on
the
right
if
you're
getting
kind
of
an
ndvi
value,
that's
0.661,
you
can
be
considering
that
pretty
healthy
vegetation
or
a
healthy
plant
and
then
some
of
the
lower
values
like
0
to
0.33
is
an
unhealthy
plant.
So
I
just
think
we
used
indices
not
only
in
some
of
our
quantifications
for
soil
carbon,
but
also
for
our
co-benefit
of
ecosystem
health.
So
I
figured
it
was
worth
touching
on
down
at
the
bottom.
A
I
just
listed
some
other
ecosystem
indicators
that
we
can
use
and
that
are
really
important,
so
you
can
look
at
fire
impacted
areas
or
you
can
look
at
the
water
content
held
in
the
soil.
You
know
bare
soil
indices
or
even
forest
health,
so
that's
just
a
basic
background
on
remote
sensing
indices
and
how
we
use
those.
A
And
finally,
the
last
thing
I
wanted
to
touch
on
before
we
really
dive
into
the
meat
of
our
actual
methodology
is
talking
about
the
benefit
of
remote
sensing,
so
really
the
basis
of
what
we're
doing
with
remote
sensing
and
tracking
changes
in
carbon
to
calculate
these
carbon
sequestration.
Metrics
is
based
on
time
series
assessment.
So
I
really
love
this
visual
on
the
right.
A
I
think
it
really
encompasses
the
benefit
and
the
advantage
of
using
remote
sensing
in
ecosystem
health,
and
you
know
carbon
change
analysis,
so
you
can
see
that
up
on
the
left,
we're
pulling
out
an
image
for
a
certain
region
and
within
that
image,
because
the
benefit
of
satellite
imagery
it's
being
revisited
every
so
often.
So
you
have
a
full
stack
of
images
that
you
can
then
look
across
time
or
look
back
in
time
to
look
at
the
change.
A
So
what
it's
showing
here
in
the
middle
is
just
looking
at
a
single
pixel
within
that
image,
and
you
can
see
how
that
changes
over
time
and
see
how
the
values
are
are
changing
to
reflect
changes
in
the
landscape
and
then
from
there.
What
you
can
actually
do
is
pull
out
these
spectral
values
and
then
look
at
how
again.
How
is
this
changing
over
time?
And
you
can
see
a
drop
here
and
relate
that
to
okay.
It
looks
like
between
you
know
the
years
of
2000
to
2005
something
happened.
A
We
can
look
into
there
and
see
how
that
changes.
It
can
detect
deforestation,
increased
vegetation,
just
one
of
the
big
advantages
of
remote
sensing
analysis
and
for
us
we're
looking
at
changes
in
percent
soil,
organic
carbon,
so
again,
looking
at
how
is
that
changing
over
time,
going
back
to
2017
at
each
pixel?
How
much
organic
carbon
are
we
picking
up
using
our
remote
sensing
indicators
and
again,
like
I
mentioned
we're,
also
using
this
to
look
at
changes
in
ecosystem
health?
A
So
that's
just
one
of
the
breakdowns
of
why
remote
sensing
is
so
important
why
we've
decided
to
use
it
and
how
it's
really
cutting
edge
in
terms
of
tracking
change
over
time
and
looking
at
changes
in
soil,
organic
carbon
so
now
I'll
pass
it
off.
I
believe,
oh
now,
we'll
pass
it
off
to
giselle
and
she's
going
to
talk
a
little
bit
about
the
state
of
the
art
of
remote
sensing
for
soiler
granite
carbon.
In
this
realm.
B
Yeah
thanks
sophia,
so
I
just
wanted
to
share
a
little
bit
of
context
here,
given
that
last
year
I
participated
in
a
work
with
dan,
kane
and
patrick
lawrence
for
the
the
ecosystem
services
market
consortium
and
it
was
targeted
to
analyze
the
different
methods
for
measuring
solar,
organic
carbon
and
this
the
state
of
the
art
of
of
all
the
tools
that
could
be
used,
and
we
created
a
report
specif
specifically
for
remote
sensing.
So
I
wanted
to
share
the
main
outcomes
here
briefly.
B
I
would
say
then,
on
the
right.
I
want
to
show
that
there's
also
a
coherence
of
this
related
to
the
increasing
amount
of
satellites
that
are
going
out
there
that
have
been
launched
and
are
also
going
to
be
increasing,
also
the
accuracy.
So
you
can
see
there
the
spatial
resolution
in
the
y-axis
and
you
can
see
that
has
increased.
I
mean
the
spatial
resolution
has
been
narrowed
down
into
a
few
meters
or
sub
meters
in
the
last
years.
B
Next,
okay-
and
here
I
wanted
to
show
first
on
the
left
you
can
see.
I
marked
in
red
the
sentinel
in
the
x
axis,
we
graphic
graphicated
the
resolution
and
then
in
the
y
you
have
the
error.
This
is
a
metric,
the
normalized
error,
from
from
the
measurements
on
soil,
organic
carbon,
from
all
the
papers
that
we
reviewed,
and
you
can
see
that
sentinel
is
quite
well
positioned
because
it
has
a
moderate
to
good
resolution.
B
You
know
the
the
how
well
people
are
formed
or
capable
of
getting
good
results
too,
because
this
is
a
very
nascent,
as
I
showed
before
in
asian
technology.
So
but
but
the
good
thing
is
that
you
can
see
from
the
purple
one
which
is
the
random
force,
that's
the
one
we
are
using.
If
you
are
good
at
it,
you
could
get
really
good
results,
and
this
is
this
is
again
based
on
reports
on
soil,
organic
carbon,
measured
by
remote
sensing.
It's
not
from
a
broader
spectrum.
B
C
So
there
are
two
key
players
in
quantifying
soil,
organic
carbon
stocks
and
those
would
be
the
land
manager
and
the
monitor
the
land
manager
is
going
to
be
providing
the
soil
samples
which
we
use
as
ground
truth
data.
And
then
the
monitor
is
going
to
take
those
soil
samples,
combine
them
with
remote
sensing
data,
mainly
satellite
imagery,
and
then
do
some
some
statistical
modeling
to
make
predictions
and
then
ultimately
turn
those
into
soil.
C
Organic
carbon
stocks,
so
kind
of
going
left
to
right
and
sophia
did
a
really
good
job
making
this
slide,
because
I
think
it
it
really.
You
know,
clarifies
how
we
go
about
doing
this,
so
we
start
with
the
soil
sampling.
So
when
we
adopt
a
project,
we
are
going
to
look
at
the
project
area,
which
is
highlighted
in
green.
This
is
an
example
of
wilmot
and,
depending
on
the
size
of
the
farm,
we
are
going
to
tell
the
land
manager.
C
This,
like
pre-processing
step,
would
be
kind
of
integration
of
remote
sensing
data
and
combining
that,
with
the
the
ground
truth
soil
carbon
data
to
make
a
data
set.
We
can
model
with
so
we're
mainly
focused
on
using
satellite
imagery.
We
will
take
a
set
of
satellite
images
taken
within
about
a
month
or
two
months
of
the
soil
sampling
date.
We'll
do
some
pre-processing
to
just
clip
those
satellite
images
to
our
study
area.
C
Make
sure
they're
make
sure
that
there's
no
clouds
or
atmospheric
irregularities
in
those
images
to
provide
like
the
the
best
data
set
for
analysis
and
then
we'll
combine
those
the
different
bands
into
a
data
set
with
soil
carbon.
So,
like
sophia
mentioned
earlier,
when
a
satellite
image
is
taken,
you're,
not
just
getting
the
the
visible
wavelengths
but
you're,
also
getting
infrared
or
gamma
wavelengths,
and
then
we
combine
all
of
those
different
what
we
call
bands
into
a
single
data
set.
C
This
is
also
a
section
where
we
might
incorporate
other
remote
sensing
data
such
as
percent
clay
or
percent
silt
found
in
the
soil.
We
can
also
incorporate
elevation
data
of
how
high
the
different
soil
samples
are,
or
things
like,
the
topographic
wetness
index,
which
might
tell
you
how
much
water
content
is
found
in
the
soil
at
different
areas,
and
so
we
kind.
C
I
think
gisele
mentioned
that
we
are
using
kind
of
two
different
statistical
modeling
methods.
The
first
is
linear
and
power
regression.
The
second
is
more
machine
learning
models.
So
in
our
original
methodology
we
were
focused
on
using
linear
and
power
regression
which
would
just
compare
soil
organic
carbon
to
a
single
band.
C
Let's
say
the
red
band
or
the
near-infrared
band
and
see
if
there
was
a
correlation
that
fit
the
data
really
well
and
if
there
was,
we
would
later
turn
that
into
percent
soil
organic
carbon
map,
but
recently,
and
especially
as
we
move
on
to
more
complex
project
areas
such
as
caban,
which
is
one
of
our
impact
ag
projects,
we're
starting
to
use
more
machine
learning
models
which
allow
us
to
incorporate
a
variety
of
data.
So,
rather
than
looking
at
one
band,
we
can
look
at
all
the
bands.
C
We
could
also
increase
incorporate
those
like
elevation
and
soil
health
metrics
into
that
as
well,
and
once
we
fit
a
model
that
we
find
to
be
accurate,
we'll
use
that
to
predict
soil
carbon.
I
think
you
can
go
to
the
next
slide
sophia,
so
to
provide
you
guys
with
a
more
concrete
example.
This
is
the
wilmot
project,
which
was
our
case
study
so
on
the
left,
you
can
see
the
soil
sample
locations
overlaid
on
the
project
area.
C
After
combining
those
with
the
satellite
images
we
created
these
percent
soil,
organic
carbon
maps,
the
percent
soil,
organic
carbon.
We
were
finding
range
from
about
one
percent
to
seven
percent
found
in
the
soil,
and
so
the
advantage
of
using
this
like
remote
sensing
data
is
that
for
each
one
of
these,
like
10
by
10,
meter
pixels,
we
can
predict
a
different
soil,
organic
carbon
value.
C
It's
important
to
note
that
we
only
predict
the
soil
depth
of
the
soil
samples.
So
if
the
farmer
collected
soil
samples
at
a
five
centimeter
depth,
we
would
predict
just
those
first
five
centimeters
of
soil,
but
then
typically
we're
doing
10
to
15.
If
they
are
able
to
collect
those
and
then
from
there
we
can
calculate
our
final
soil,
organic
carbon
stocks,
and
so
that
is
on
the
right
and
you
can
see
how
the
soil
carbons
varies
across
the
farm.
C
C
C
If
there's
a
net
gain,
then
those
are
creditable
carbon
stocks
before
we
generate
that
final
number,
we're
gonna,
reduce
the
animal
emissions
so
because
our
cattle
on
farm
emitting
methane
among
other
gases,
we're
gonna,
detect
that
from
our
carbon
credits
and
then
also
work
in
the
model
uncertainty.
So
all
these
statistical
models
that
we're
using
to
predict
soil
carbon
might
have
different
levels
of
uncertainty
and
we
want
to
make
sure
we
account
for
that,
because
we're
monetizing
like
these
soil
garden,
carbon
stocks
and
then
finally,
to
generate
our
final
number.
C
We
multiply
it
by
a
factor
of
3.67
because
there's
a
difference
between
measured
soil,
carbon
or
the
soil
carbon
found
in
the
soil
and
the
greenhouse
gas
equivalent
carbon,
which
is
by
this
factor
of
3.67
and
then
from
there
we'll
get
our
final
number
and
then
do
our
buffer
pool
and
other
deductions
to
get
our
final
number.
So
oh
yeah,
and
then
I
guess
what
also
distinguishes
our
methodology
from
other
methodologies
in
the
field.
C
Right
now
is
that
we
are
not
just
measuring
carbon,
but
we're
also
measuring
other
other,
like
ecosystem
health
indicators
and
metrics
of
of
the
kind
of
just
like
how
the
how
the
farm
is
being
managed.
So
the
three
that
we
focus
on
are
ecosystem
health
indicators,
which
helps
show
the
the
farm,
but
I
guess,
like
the
vigor
and
the
resilience
of
the
plants
and
just
landscape
of
the
farm,
the
soil
health
indicators
so
rather
than
measuring
just
soil
carbon.
C
So
you
can
go
to
the
next
slide,
so
I
guess
going
to
the
ecosystem
health
indicators.
The
one
of
the
main
ones
that
we
focus
on
would
be
ecosystem
resilience
and
ecosystem
vigor,
so
for
ecosystem
resilience.
We
want
to
make
sure
that
we're
actually
seeing
like
a
good
plant
coverage
across
these
grasslands
over
time.
So
what
we
do
is
we
compare
the
grass
cover
within
the
farm
to
the
area
outside
of
the
farm.
A
So
some
of
the
examples
is
we
have
a
whole
list
of
animal
welfare
metrics
to
look
at
so
say,
access
to
the
feed
and
water
or
risk
management
are
the
facilities
cleaned
and
you
know,
do
they
have
good
access
to
food
and
water?
Are
they
healthy?
So
what
we
do
is
we
look
at
all
these
different
metrics
and
discuss
with
the
land
owner,
which
ones
are
being
filled,
which
ones
might
need
improvement
and
again
are
getting
a
final
score
there.
A
So
those
are
the
two
other
ways,
the
two
other
co-benefits
that
we're
incorporating
into
the
final
reports
for
each
of
these
projects.
So
now
we'll
turn
it
over
to
gisele
who's
gonna.
Maybe
we're
gonna
move
out
a
little
bit
and
talk
about
some
of
the
frequently
asked
questions
and
then
open
the
floor
to
some
general
questions.
B
Yeah,
so
I
created
a
list
with
some
of
the
questions
that
have
been
asked
to
science
team
back
in
december
and
through
all
this
process,
but
I
will
tackle
maybe
first
three
and
then
open
for
more
questions.
If
you
will
to
review
this
and
see
if
you
want
to
ask
some
of
the
following
ones,
I
just
don't
want
to
spend
a
lot
a
whole
lot.
Talking
myself,
I
would
rather
listen
to
your
questions
now.
So
first,
I
will
answer
first
one.
B
B
So
here
I
have
like
a
schema
of
the
whole
process
starting
well,
and
the
resolution
is
not
very
well
very
good,
but
first
of
all,
we
estimate
the
uncertainty
from
the
solar,
organic
carbon
percentage
estimation
based
on
either
the
results
from
the
linear
or
power
regression,
as
sam
explained
that
we
were
using
until
now
or
based
on
the
estimated
accuracy
from
the
machine
learning
algorithms.
So
that's
one
source
of
uncertainty.
B
B
Any
uncertainty
from
the
lab
will
be
included
here,
just
because
this
is
like
the
overall
uncertainty
at
the
end
of
the
process,
where
you
compare
the
actual
values
with
the
with
the
values
from
the
map
that
is
created
and
then,
as
we
showed,
we
need
to
incorporate
the
bulk
density
in
order
to
estimate
the
final
uncertainty
for
the
carbon
stocks,
because
we
are
going
to
use
it
in
the
equation
for
the
final
stocks.
So
that's
another
source
of
uncertainty
and
the
way
we
are
doing
it
right
now.
B
For
instance,
we
were
using
the
transfer
functions
and
we
were
comparing
the
the
real
values
from
the
sample
years
to
the
estimated
values
from
the
the
petal
transfer
functions.
Pedal
transfer
functions
are
just
equations
based
on
by
geography
from
other
research
in
the
area
or
in
similar
ecoregions,
where
they
linked
the
solar,
organic
carbon
percentage.
B
Then
we
estimate
the
bulk
density
map,
based
on
the
the
things
that
we
do
have
from
the
years
that
we
we
don't.
We
just
don't
have
samples
from
from
the
soil
from
the
ground
with
the
valentine's
day.
Okay,
if
you
go
back
to
questions
yeah,
so
another
question
is
related
to
how
do
we
select
the
baseline?
B
And
this
is
not.
This
is
a
very
relevant
question.
I
would
say
that
so
far
it
has
been
given
that
we
all
only
run
wilmot
and
woodburn
fully
complete.
We
completed
only
those
studies.
The
way
we
were
doing
that
is
just
relying
on
the
information
we
had
from
the
grounds
and
we
were
looking
at
the
correlations
and
finding
and
measuring
the
accuracy,
and
if
the
method
was
good
enough,
then
we
could
use
that
ear
and
the
data
from
that
year
also.
B
We
are
only
going
back
in
time
until
2017,
because
that's
the
moment
where
the
sentinel
image,
the
first
images
are
available
because
they
launched
the
sentinel
satellite
that
year
we
could
be
using,
for
instance,
landsat
from
previous
years.
But
then
the
resolution
is
lower
and
the
spectral
resolution
also
and
that's
why
we
are
relying
on
the
sentinel
to
get
a
better
accuracy.
B
B
According
to
previous
information
from
historical
data-
and
they
are
often
only
developed
for
more
traditional,
not
regenerative
practices
but
more
traditional
and
in
the
case
of
comet
only
or
most
of
it
for
crops
cropland,
and
so
they
they
model
10
years
of
climate
of
climate.
They
run
the
models
for
that
those
variabilities
in
the
climate
and
with
that
they
get
the
stocks.
That
would
be
that
that
would
be
measured
in
that
piece
of
land
if
they
are
applying
certain
practice.
B
I
would,
I
would
say,
longer
term
improvement
of
the
methodology.
It's
not
simple
to
solve,
and
only
using
models.
I
wouldn't
recommend
that,
as
any
model
like
decent
has
at
least
a
0.5
uncertainty,
that's
like
the
best
uncertainty
you
can
find
in
bibliography
for
any
of
these
models,
based
on
the
comparison
of
the
parameters
at
regional
scales,
not
at
field
scale.