Bio GeoSpatial PDF - McGill University

24/02/2014

1

Asim BiswasAssistant Professor

Department of Natural Resource Sciences

Asim BiswasAssistant Professor

Department of Natural Resource Sciences

Quantifying soil spatial variability for

sustainable resource management

Quantifying soil spatial variability for

sustainable resource management

www.mcgill.ca

| 2

Background

Asim Biswas | Quantifying soil spatial variability

� B.Sc. (Agriculture) in Soil Science (East India; 2000 - 2004)

� M.Sc. (Agriculture) in Soil Science (South India; 2004 - 2006)-• Soil Pedology (soil formation and development)• Soil Mineralogy (soil mineralogical composition)

• Spatial Variability

� Ph.D. in Soil Science (Canada; May 2007 – June 2011)-• Soil Physics• Soil Hydrology (vadose-zone hydrology)

• Spatial Variability

• Pedometrics (Pedology + mathematics/statistics)

� Environmental Research Scientist (Post doc) (CSIRO; July 2011 – April 2013)

� Assistant Professor (McGill University; May 2013- Present)

www.mcgill.ca

Asim Biswas | Quantifying soil spatial variability | 3

Challenges� Change in -

• Population

• Environment• Weather and Climate• Biodiversity

• Land use and Land management

� Threats to -• Soil security (e.g. soil erosion, degradation )

• Water security (e.g. accessibility and quality)• Food security• Agricultural and natural resources

• Ecosystem and human health

www.mcgill.ca

| 4

Opportunities


• We can not control things beyond our reach (weather

events, changing climate).

• What ever lost is lost and will not get back.

• We should not let it go what we have.

• We can manage natural resources better (soil, water).

• Manage wisely and efficiently, more sustainably.

• Need quantitative information of our natural resources.

www.mcgill.ca

| 5

Soil and its formation


Climate

Organisms

Relief

Parent material

Time

SoilSoil propertiesProcesses

Soil properties vary from location to location

One of the 3 major natural resources

www.mcgill.ca

| 6

Characteristics of soil spatial variability


Distance

Soil

Pro

pert

y

Similar

Similar

Distance

Soil

Pro

pert

y

Cyclic

Distance

Soil

Pro

pert

y Different mean

and variance

Spatial similarity Periodicity

Nonstationarity

24/02/2014

2

www.mcgill.ca

| 7



Vegetation

Water storage

Topography

Water storage

Water storage

Vegetation & Topography

Nonlinearity

Non-additive

www.mcgill.ca

| 8



Vegetation

Water storage

Topography

Water storage

Water storage

Vegetation & Topography

Nonlinearity

Non-additive

“No single medicine for all diseases”

• Methods to better analyze soil spatial variability.

• Use of the information to infer underlying soil processes.

• What does soil variability inform about soil development?

• Using information of soil variability for attribute prediction.

www.mcgill.ca

| 9

Soil spatial variability


� What is the dominant scale of variation?

� Where do I sample?

� Where do I monitor?

� How do I untangle complexity to produce better

predictive relationship?

� How do I assess soil function at multiple scales?

� How do I meet user demand (farmers vs. catchment

managers)?

� What do I know on the underling processes and the

development of soil?

www.mcgill.ca

| 10

Relationship b/w soil properties


• Difficult to measure (e.g. Ks, water retention)

• Relatively easy to measure (e.g. particle sizes, OC)

• Predictive relationship (e.g. pedotransfer functions)

• Variability in soil properties

• Variability in the relationship

2008, 89, 473-488

0 50 100 150 200

www.mcgill.ca

| 11

Geostatistical analysis


Semivariogram of Ks Semivariogram of Sand

0 50 100 150 200

Lag Distance (m)

1.2

0.8

0.4

Sem

ivari

an

ce

Lag distance

Sem

iva

rian

ce

Sill

Nugget

Range

Spatially dependent Spatially independent

0 50 100 150 200

www.mcgill.ca

| 12

Geostatistical analysis


Semivariogram of Ks Semivariogram of Sand

0 50 100 150 200

Lag Distance (m)

1.2

0.8

0.4

Sem

ivari

an

ce

24/02/2014

3

0

30

60

90

120

150

180

210

240

270

300

330

0.0

0.2

0.4

0.6

0.8

General data

Nugget

h=3

h=20

h=120

Silt

BD

Clay

OCKs

Sand

BD

ClaySand

SiltKs

OC

Ks

OC

Clay

Silt

BD

Sand

Clay

OC Silt

Ks

BD

Sand

Ks

Clay

OC

SiltSand

BD

Principle component 1

Pri

nci

ple

com

pon

ent

2

www.mcgill.ca

| 13



Principle component analysis

0

30

60

90

120

150

180

210

240

270

300

330

0.0

0.2

0.4

0.6

0.8

General data

Nugget

h=3

h=20

h=120

Silt

BD

Clay

OCKs

Sand

BD

ClaySand

SiltKs

OC

Ks

OC

Clay

Silt

BD

Sand

Clay

OC Silt

Ks

BD

Sand

Ks

Clay

OC

SiltSand

BD

Principle component 1

Pri

nci

ple

com

pon

ent

2

www.mcgill.ca

| 14



Principle component analysis

What did we get?

Information on scales of variations in soil properties

The scale-varying relationship among soil properties

www.mcgill.ca

| 15

N-dynamics and landscape characteristics


www.mcgill.ca

| 16



N- Dynamics

- isotopes of nitrogen

- isotopic fractionation in physical, chemical and biological processes

- nitrification

- denitrification

- ammonia volatilization

- favour lighter 14N and depleted first

- N pool develops distinctly different δ15N signals

www.mcgill.ca

| 17



Central Saskatchewan, Canada

www.mcgill.ca

| 18



N- Dynamics



- nitrification

- denitrification




0

-2

-4

-6Rel. E

levation

(m)

9

8

7δ1

5N

(‰

)

0 50 100 150 200 250 300 350Distance (m)

Relative Elevation

δ15N

Central Saskatchewan, Canada Wavelet Transform

DistanceDistance

So

il p

rop

ert

y

24/02/2014

4

www.mcgill.ca

| 19



N- Dynamics



- nitrification

- denitrification




0

-2

-4

-6Rel. E

levation

(m)

9

8

7δ1

5N

(‰

)

0 50 100 150 200 250 300 350Distance (m)

Relative Elevation

δ15N

Central Saskatchewan, Canada Wavelet Transform

DistanceDistance

So

il p

rop

ert

y

Scaling

www.mcgill.ca

| 20

Wavelet transform


0

20

40

60

80

Location

Pro

pert

y

0 100 200 300 400 500

40

3020

10

0

Location0 100 200

Scale

80

60

40

20

0

1

.5

0

Sca

le

www.mcgill.ca

| 21



N- Dynamics

- natural abundance of 15N


- nitrification

- denitrification




0

-2

-4

-6Rel. E

levation

(m)

9

8

7δ1

5N

(‰

)

0 50 100 150 200 250 300 350Distance (m)

Scale

(m

)

12

24

48

960 60 120 180 240 300 360

Distance (m)

0

-2

-4

-6Rel. E

levation

(m)

9

8

7δ1

5N

(‰

)

Small variation

Large variation

Relative Elevation δ15N

www.mcgill.ca

| 22



N- Dynamics

- natural abundance of 15N


- nitrification

- denitrification




0

-2

-4

-6Rel. E

levation

(m)

9

8

7δ1

5N

(‰

)

0 50 100 150 200 250 300 350Distance (m)

Scale

(m

)

12

24

48

960 60 120 180 240 300 360

Distance (m)

0

-2

-4

-6Rel. E

levation

(m)

9

8

7δ1

5N

(‰

)

Small variation

Large variation

What it does?

It provides explicit information on scales and locations of variation.

Relative Elevationδ15N

www.mcgill.ca

| 23Asim Biswas | Quantifying soil spatial variability


r2 = 0

Traditional correlation analysis

Wavelet coherency analysis

Correlation at different scales and locations

0

-4

Rel. E

le.

(m)

Scale

(m

)

0 60 120 180 240 300 360Distance (m)

12

24

48

96

9

8

7

δ1

5N (‰

)

High correlation

Low correlation95%

significance

‘Left’Negative

‘Right’Positive

www.mcgill.ca



r2 = 0




0

-4

Rel. E

le.

(m)

Scale

(m

)

0 60 120 180 240 300 360Distance (m)

12

24

48

96

9

8

7

δ1

5N (‰

)

24/02/2014

5

www.mcgill.ca



r2 = 0




0

-4

Rel. E

le.

(m)

Scale

(m

)

0 60 120 180 240 300 360Distance (m)

12

24

48

96

9

8

7

δ1

5N (‰

)

High correlation

Low correlation95%

significance

‘Left’Negative

‘Right’Positive

What do we get?

Variability in environmental correlations.Traditional correlation may not identify existing relationship.

www.mcgill.ca

| 26

Similarity of the spatial pattern


www.mcgill.ca

| 27



� Similar relationship can be developed over time.

� Intra-season, inter-season, and inter-annual.

� The wet locations stay wet and dry locations stay dry over time.

� Identify the location with soil water storage stays close to field average.

555

Ele

vation (

m)

www.mcgill.ca

| 28



� Similar relationship can be developed over time.

� Intra-season, inter-season, and inter-annual.

� The wet locations stay wet and dry locations stay dry over time.

� Identify the location with soil water storage stays close to field average.

555

Ele

vation (

m)Such representative locations can be used to monitor or

validate remote sensing measurements.

?SSSAJ, 2011, 75: 1099-1109

www.mcgill.ca

| 29

Similarity between depths


?www.mcgill.ca

| 30



Distance (m)

Soil

wate

r sto

rage (

cm

)

0-20 cm

20-40 cm

40-60 cm

60-80 cm

80-100 cm

100-120 cm

120-140 cm

24/02/2014

6

?SSSAJ, 2011, 75: 1099-1109

www.mcgill.ca

| 31



Distance (m)

Soil

wate

r sto

rage (

cm

)0-20 cm

20-40 cm

40-60 cm

60-80 cm

80-100 cm

100-120 cm

120-140 cm

What does it enable?

Whole profile hydrological dynamics from surface measurements.

www.mcgill.ca

| 32

Separating scale-specific variations


�Scale-specific spatial variations

� Identify dominant scales and their relative

contribution to overall variability

Empirical mode decomposition

www.mcgill.ca

| 33

Empirical mode decomposition


IMF 1IMF 2IMF 3IMF 4IMF 5IMF 6Residue

www.mcgill.ca


Separating scale-specific variationsS

oil W

ate

r S

tora

ge (

cm

)

30

60 2 May 2008

-6

0

6 IMF 2-606 IMF 1

-606

IMF 3

-606 IMF 4

-6

06 IMF 5

Distance (m)

0 100 200 300 400 500

-606 IMF 6

Scale Contribution

~10 m ~6%

~40 m ~10%

~100 m ~41%

~180 m ~6%

~250 m ~5%

>280 m ~4%

>300 m ~28%

SSSAJ, 2011, 75: 1295-1306

www.mcgill.ca



IMF % Var. Cont.

Correlation

Re. Ele. Sand Silt Clay OC

1 6 0.00 0.02 -0.08 0.08 -0.01

2 10 -0.38** -0.11 0.10 0.03 0.38**

3 41 -0.70** -0.07 0.00 0.12 0.58**

4 6 -0.22* -0.26** 0.11 0.26** 0.31**

5 5 0.55** -0.59** 0.43** 0.36** 0.11

6 4 0.37** -0.57** 0.38** 0.39** 0.31**

* P < 0.05; ** P < 0.01

Var. Cont.- variance contribution, Re. Ele.-relative elevation,

OC- organic carbon

www.mcgill.ca



IMF % Var. Cont.

Correlation

Re. Ele. Sand Silt Clay OC

1 6 0.00 0.02 -0.08 0.08 -0.01

2 10 -0.38** -0.11 0.10 0.03 0.38**

3 41 -0.70** -0.07 0.00 0.12 0.58**

4 6 -0.22* -0.26** 0.11 0.26** 0.31**

5 5 0.55** -0.59** 0.43** 0.36** 0.11

6 4 0.37** -0.57** 0.38** 0.39** 0.31**

* P < 0.05; ** P < 0.01

Var. Cont.- variance contribution, Re. Ele.-relative elevation,

OC- organic carbon

What does it enable?

Scale-specific correlation.Dominant controlling factors at different scales.

24/02/2014

7

SSSAJ, 2013, 77: 1991-1995

www.mcgill.ca


Separating scale-specific variations in two dimensions

One property may vary at one scale, while others at other scales

Bi-dimensional empirical mode decomposition

www.mcgill.ca



DEMScale: 705 m

Var.: 2 %

Scale: 2896 mVar.: 7 %

Scale: 3905 mVar.: 14 %

Scale: 8839 mVar.: 77 %

Scale: 7509 m


Simmons Creek Catchment, NSW, Australia

SSSAJ, 2013, 77: 1991-1995

www.mcgill.ca



Bi-dimensional empirical

mode decomposition

DEMScale: 705 m

Var.: 2 %

Scale: 2896 mVar.: 7 %

Scale: 3905 mVar.: 14 %

Scale: 8839 mVar.: 77 %

Scale: 7509 m


Simmons Creek Catchment, NSW, AustraliaWhat does it enable?

Scale-specific predictionsMulti-scale digital soil mapping

www.mcgill.ca

| 40

Underlying soil processes


• Anisotropic soil spatial variations

www.mcgill.ca

| 41




www.mcgill.ca

| 42




What it indicated?

Development and modifications of soil in the landscape.

24/02/2014

8

www.mcgill.ca

| 43

Summary


�Optimize sampling strategy and experimental design

• Scales of hydrological processes

� Identify representative locations for monitoring

� Previously hidden predictive relationship

� Infer at depth from surface measurements

� Identify environmental controls at different scales

� Scale-specific prediction

� Multi-scale digital soil mapping Thank YouThank YouContactDepartment of Natural Resource Sciences

E-mail: [email protected]: (514) 398 7620, Fax: (514) 398 7990Website: http://nrs-staff.mcgill.ca/biswas

Documents

Bio GeoSpatial PDF - McGill University