TERN Ecosystem Surveillance Plots Kakadu National Park
Raphael Viscarra Rossel_Mapping Australian soils and their condition
1. Mapping
Australian
soil:
the
TERN-‐soil
approach
TERN
Symposium,
18
–
21
February
2013
PresentaEon
by
Dr
Raphael
VISCARRA
ROSSEL
2. What
soil
informa.on
do
we
currently
have?
•
Few
spaEal
(and
temporal)
3&89&:;<!54+'!=
0(12!3!4&
*+#,-)!
.
data
5*6!3!!
• Incomplete
and
scaQered
1#-20+)"!
.
!"#$%!
.
&'()"'!
.
coverage
3("4+-("!
.
5+)0"##+!
.
6"40/7#)"!
.
! &"# "#! $%#!! '()*+,-,./
./0+#$!
.
• IndividualisEc
approaches
to
survey
• Inconsistent
methods
• IncompaEble
soil
point
and
spaEal
databases
3&:;&8<=!>?5!=<<<9
• Disparate
informaEon
4+56!7!$8"3$ *+#,-)!
.
9-:!7!&$
management
89
1#-20+)"!
.
• …
!"#$%
3$
.
!
&'()"'!
.
3("4+-("!
.
5+)0"##+!
.
6"40/7#)"!
.
! ()! "#! $%&'! *+,-./0/12
./0+#$!
.
3. The
TERN
soil
and
landscape
facility
Produce
soil
informaEon
and
data
infrastructures
for
Australia
that:
•
is
consistent,
reliable
and
comprehensive
• is
at
an
appropriate
scale
and
resoluEon
• describes
soil
variability
in
x,y,z
(i.e.
soil
volume)
• integrates
with
earth
observaEon,
climate,
biogeochemical
and
ecological
models,
etc.
• provides
esEmates
of
uncertainty
• furthers
our
understanding
of
Australian
soils
and
the
Improving
the
quality
of
environment
ecosystem
research
4. Components
of
the
TERN-‐soil
facility
DATA
COVARIATES
CollaEon
DISSAGREGATION
of
naEonal
Improved
soil
data.
topographic
NATIONAL
An
improved
MAPPING
aQributes.
method
to
make
QUANTIFYING
New
beQer
use
of
spectroscopic
New
covariates
New
fine
UNCERTAINTY
historical
soil
COMMUNICATION
measurement.
resoluEon
based
on
soil
informaEon.
maps
of
soil
Assessment
of
spectra.
A
soYware
aQributes
with
uncertainEes
in
An
improved,
A
soYware
tool
soil
mapping.
consistent
soil
tool
for
New
covariates
measures
of
to
perform
the
informaEon
Improved
from
remote
uncertainty.
dissagregaEon.
An
approach
to
systems
that
is
inference.
sensing.
combine
the
readily
accessible.
A
soYware
tool
to
perform
the
dissagregated
and
property
mapping.
maps.
5. Components
of
the
TERN
soil
facility
DATA
COVARIATES
CollaEon
DISSAGREGATION
of
naEonal
Improved
soil
data.
topographic
NATIONAL
An
improved
aQributes.
method
to
make
MAPPING
QUANTIFYING
New
beQer
use
of
spectroscopic
New
covariates
New
fine
UNCERTAINTY
historical
soil
COMMUNICATION
measurement.
based
on
soil
resoluEon
informaEon.
spectra.
maps
of
soil
Assessment
of
A
soYware
aQributes
with
uncertainEes
in
An
improved,
A
soYware
tool
soil
mapping.
consistent
soil
tool
for
New
covariates
measures
of
to
perform
the
informaEon
Improved
from
remote
uncertainty.
dissagregaEon.
An
approach
to
systems
that
is
inference.
sensing.
combine
the
readily
accessible.
A
soYware
tool
to
perform
the
dissagregated
and
property
mapping.
maps.
6. TERN
soils
and
the
work
being
presented
DATA
COVARIATES
CollaEon
of
naEonal
Improved
DISSAGREGATION
soil
data.
topographic
NATIONAL
ApplicaEon
aQributes
Development
of
a
MAPPING
and
the
need.
3
&
1
arc
sec
tool
to
make
use
QUANTIFYING
of
historical
soil
New
fine
UNCERTAINTY
informaEon
resoluEon
COMMUNICATION
Ross
Searle
mapping
of
UncertainEes
John
Gallant
total
P
in
soil
InformaEon
and
regolith
measurement
systems
and
Nathan
Odgers
RVR
(me)
and
mapping
delivery
thru
Mark
Thomas
TERN
and
DAP
David
Clifford
Peter
Wilson
David
Jacquier
Facility
director:
Mike
GRUNDY
Project
co-‐leaders:
Raphael
VISCARRA
ROSSEL
&
Ross
SEARLE
8. Specifica.ons
for
na.onal
soil
mapping
We
are
aiming
for
the
following
properEes:
1.
Total
P
(%)
2.
Total
N
(%)
3.
Organic
Carbon
(%)
4.
Bulk
Density
(Mg/m3)
5.
Sand,
Silt
and
Clay
(%)
6.
pH
7.
ECEC
(caEons
and
exchangeable
acidity
mmol(+)/kg)
8.
EC
(Electrical
conducEvity
mS/m)
9.
Available
Water
Capacity
(mm/m)
7.
Depth
to
restricEng
layer
and
depth
of
regolith
(cm)
NaEonal
maps
with
spaEal
resoluEon
An
inference
system
(SINFERS)
will
then
be
used
3
arc
sec
(around
90
m)
to
derive
other
properEes
to
parameterize
ecosystem
models
9. Mapping
total
P:
the
dataset
Total
53,805
data
from
around
7000
sites
All
ASC
orders
represented
Mean
=
0.03
St.
Dev.
=
0.05
Minimum
=
0
Median
=
0.02
Maximum
=
1
Data
needs
to
be
harmonised
to
standard
depths
10. Harmonisa.on
to
standard
depths
• Splines
(Bishop
et
al.,
1999)
• ‘Infill’
simulaEons
approach
data
with
only
2
depths
0-‐5
cm
5-‐15
cm
15-‐30
cm
30-‐60
cm
60-‐100
cm
100-‐200
cm
11. The
data
at
the
standard
depths
0−5 cm 5−15 cm 15−30 cm
med.
=
0.026
med.
=
0.025
med.
=
0.021
800 1000
1000
-‐ Skewed
distribuEons
1000
600
500
-‐ Total
P
generally
500
400
200
decreasing
with
0
0
0
depth
−200
-‐ ConcentraEons
are
0.0 0.1 0.2 0.3 0.4 0.5 0.0 0.1 0.2 0.3 0.4 0.5 0.0 0.1 0.2 0.3 0.4 0.5
consistent
with
what
30−60 cm 60−100 cm 100−200 cm
we
know
of
med.
=
0.018
med.
=
0.016
med.
=
0.016
1500
1500
1500
Australian
soils
–
1000
1000
1000
contain
small
amounts
of
P
500
500
500
0
0
0
−500
−500
−500
0.0 0.1 0.2 0.3 0.4 0.5 0.0 0.1 0.2 0.3 0.4 0.5 0.0 0.2 0.4 0.6
Total P
12. Predic.ve
spa.al
modelling
Conceptual
model
and
spaEal
predictors
(covariates):
Total
P
=
f([Pveg],
[PliQer],
[POM],
[Pmin],
[Prock],…)
C3
C4
Temp
Press
Cpre-‐eur
PC1
ASC
NPP
GPP
SRad
DEM
Kaol
LU
Litho
+
others….
13. Spa.al
modelling
with
model
trees
Different
models
with
0–5
cm
different
predictors
used
in
each
ruleset
sets
14. Maps
of
total
P
for
the
standard
depths
Total
P
/%
0.6
0–5
cm
5–15
cm
15–30
cm
0
30–60
cm
60–100
cm
100–200
cm
18. Comparing
to
what
we
currently
have
TERN
soils
ASRIS
NLWR
Audit
(2001)
Australian
natural
resource
atlas
Raupach
et
al.
(2001)
19. Conclusions
The
TERN-‐soil
facility
will
provide
a
new,
updatable
spaEal
soil
data
infrastructure
that
will:
• provide
current
baselines
of
soil
condiEon
• further
our
understanding
of
soil
and
the
environment.
• integrate
with
modelling
to
provide
ecosystem
services
and
help
devise
soluEons
to
issues
that
we
are
facing
(food,
water,
energy
securiEes,
climate
change,
soil
degradataEon)
20. Thank
you
CSIRO
Land
&
Water
Raphael
VISCARRA
ROSSEL
Principal
Research
ScienEsts
v t
+61
2
6246
5945
v e
raphael.viscarra-‐rossel@csiro.au
v w
www.csiro.au/