Ispra, 26/2/2008 1/21 SOIL Action Meeting, 04/03/2008 The OSACA Web Application A B D(A;B) Concepts and applications.

Ispra, 26/2/2008 1/21

SOIL Action Meeting, 04/03/2008

The OSACA Web Application

A BD(A;B)

Concepts and applications

Ispra, 26/2/2008 2/21

SOIL Action Meeting, 04/03/2008

OSACA finalities

To classify soil observations into soil types (ST) or clusters (soil classification tool)

Soil observations

Soil types

To allocate soil observations according to soil types provided by the user

Soil observations

To classify soil observation horizons into horizon types (id. )

User’s soil types

+

Range of ST

required

- Table of distances between observations and ST

- Description of ST

- Table of distances between ST

- Quality of the classification

- Table of distances between observations and ST

- Table of distances between ST

inputs outputs

outputsinputs

Ispra, 26/2/2008 3/21

SOIL Action Meeting, 04/03/2008

The classification process

Min. # Soil Type

centroids

Calculation of

distances

Calculation of new

centroids

Dintra/Dinter ratio

Min+1. # Soil Type centroids

Max. # Soil Type

centroids

Initialisation of cluster centroids

Observation allocation

Centroid readjustment

Optimisation process

optimization

K-means techniques

Ispra, 26/2/2008 4/21

SOIL Action Meeting, 04/03/2008

Initialisation of centroids

NC Soil Types or Clusters being required by the user, we fix γ an initial large distance

Pick up randomly 2NC observations

Calculate the distance between the 2NC observations

If distance < γ, choose other observations and go to , otherwise go to

Calculate distance between the whole set of observations and the 2NC observations, and allocate the corresponding observations (those with dmin)

Sort by decreasing order of population the 2NC clusters

If more than NC clusters are empty, γ = γ/2 and go to , otherwise, choose the NC first ones and consider the corresponding centroids as initial centroids

Ispra, 26/2/2008 5/21

SOIL Action Meeting, 04/03/2008

The classification process

Min. # Soil Type

centroids

Calculation of

distances

Calculation of new

centroids

Dintra/Dinter ratio

Min+1. # Soil Type centroids

Max. # Soil Type

centroids

Initialisation of cluster centroids

Observation allocation

Centroid readjustment

Optimisation process

optimization

K-means techniques

Ispra, 26/2/2008 6/21

SOIL Action Meeting, 04/03/2008

Calculation of distances

OSACA provides 3 kinds of distances useful for different purposes

Pedological distance for emphasizing the different succession of horizons whatever their thickness

Utilitarian or environmental distance for emphasizing the successions of horizons for which the thickness is important

Joint or normalised distance is a mixed between the two previous distances

Ispra, 26/2/2008 7/21

SOIL Action Meeting, 04/03/2008

D(x,A)

D(y,B)

D(z,C)

0 cm

20 cm

60 cm

120 cm

A

B

C

Centroid

Pedological distance

+

+

+

+

D

(# D)

Calculation of distances

x

y

z

0 cm

30 cm

80 cm

100 cm

Observation

w

130 cm

D(w,C)

Ispra, 26/2/2008 8/21

SOIL Action Meeting, 04/03/2008

D*e

Max depth

D(x,A)

D(y,B)

D(z,C)

0 cm

20 cm

60 cm

120 cm

A

B

C

Centroid

Utilitarian distance

+

+

+

+

Calculation of distances

x

x

y

z

0 cm

30 cm

80 cm

100 cm

Observation

w

130 cm

D(w,C)

D(x,B)

D(y,C)

D(w,?)

+

+

Ispra, 26/2/2008 9/21

SOIL Action Meeting, 04/03/2008

D(x,A)

D(y,B)

D(z,C)

0 cm

20 cm

60 cm

120 cm

A

B

C

Centroid

Joint distance

+

+

+

D(w,C)

x

y

z

0 cm

30 cm

80 cm

100 cm

Observation

w

130 cm

Calculation of distances

scaled depths

x

y

z

w

A

B

C

No scaled depths

+

+D(y,C)

D(x,B)

0

1

0

1

D*e

Max depth

Ispra, 26/2/2008 10/21

SOIL Action Meeting, 04/03/2008

The classification process

Min. # Soil Type

centroids

Calculation of

distances

Calculation of new

centroids

Dintra/Dinter ratio

Min+1. # Soil Type centroids

Max. # Soil Type

centroids

Initialisation of cluster centroids

Observation allocation

Centroid readjustment

Optimisation process

optimization

K-means techniques

Ispra, 26/2/2008 11/21

SOIL Action Meeting, 04/03/2008

Observation allocation and centroid readjustment

Distance table

1 0.7 1.30.1 0.3

2

3

4

5

A B C D

2.5

3.0

0.4

1.20.1

0.4

0.8

0.6

0.61.5

0.2

0.3

0.0

0.1

1.2

1.9

B

ST

A

C

B

B

0.1

dmin

0.1

0.2

0.3

0.1

Soil observations

Distances to centroids Allocation process

New centroid for B cluster

= observation

for which dintra B cluster minimal

Centroid readjustment

process

Distance calculation iteration stopped when centroids ite n= centroids ite n-1

Ispra, 26/2/2008 12/21

SOIL Action Meeting, 04/03/2008

OSACA interface

! Access at http://eusoils.jrc.ec.europa.eu/osaca !

Enter the observation table

If needed, enter the soil types table you want to use

Ispra, 26/2/2008 13/21

SOIL Action Meeting, 04/03/2008

Enter the number of horizon classes

Enter the number of soil classes you want

Ispra, 26/2/2008 14/21

SOIL Action Meeting, 04/03/2008

Get the results (tables)

Ispra, 26/2/2008 15/21

SOIL Action Meeting, 04/03/2008

OSACA Applications

Soil taxonomy purpose

Available Water Capacity prediction of soil profiles

Focusing on

The number of classes (level of taxonomic detail)

The classes robustness (sensitivity of the distances)

OSACA has been used for testing

Look at:

Carré & Jacobson, 2008. Numerical classification of soil profile data using distance metrics. Geoderma (available on line)