Pászto, V: Corine Land Cover dataset analysis with (geo)computational methods in GIS
1. Corine Land Cover dataset analysis with
(geo)computational methods in GIS
Vít Pászto
This presentation is co-financed by the
European Social Fund and the state
budget of the Czech Republic
2. "PLAYLIST"
● Introduction
○ Data used
■ Study Area
● Methods
○ Work-flow diagram
■ Results
● Conclusions
45 MIN.
First InDOG Doctoral Conference, 29th October - 1st November 2012, Olomouc
3. INTRODUCTION
● Computer capabilities used by landscape
ecologists
● Quantification of landscape patches
● Via various indexes and metrics
● Prerequisite to the study pattern-process
relationships (McGarigal and Marks, 1995)
● Progress faciliated by recent advances in
computer processing and GIT
First InDOG Doctoral Conference, 29th October - 1st November 2012, Olomouc
4. INTRODUCTION
● Shape and spatial metrics are exactly those
methods for quantitative description
● In combination with multivariate statistics, it
is possible to evaluate, classify and cluster
patches
● Available metrics were used (as many as
possible)
● Unusual approach in CLC analysis
First InDOG Doctoral Conference, 29th October - 1st November 2012, Olomouc
5. DATA
● Freely available CORINE Land Cover dataset:
○ 1990
○ 2000
○ 2006
● Level 1 of CLC - 5 classes:
○ Artificial surfaces
○ Agricultural areas ○ Wetlands
○ Forest and semi-natural areas ○ Water bodies
First InDOG Doctoral Conference, 29th October - 1st November 2012, Olomouc
6. STUDY AREA
● Olomouc region (800 km2) - 1/2 of London
● More than 944 patches analysed
First InDOG Doctoral Conference, 29th October - 1st November 2012, Olomouc
7. METHODS - Shape & spatial metrics
● Fundamentally based on patch area,
perimeter and shape
● Easy-to-obtain metrics & complex metrics
● Software used:
○ FRAGSTATS 4.1
○ Shape Metrics for ArcGIS for Desktop 10.x
● EXAMPLE/EXPLANATION
First InDOG Doctoral Conference, 29th October - 1st November 2012, Olomouc
8. METHODS - Shape & spatial metrics
● Fundamentally based on patch area,
perimeter and shape
● Easy-to-obtain metrics & complex metrics
● Software used:
○ FRAGSTATS 4.1
○ Shape Metrics for ArcGIS for Desktop 10.x
● EXAMPLE/EXPLANATION
First InDOG Doctoral Conference, 29th October - 1st November 2012, Olomouc
9. METHODS - Shape & spatial metrics
Shape and spatial metrics
Area index Contiguity index
Perimeter index (FRAGSTATS Core index
4.1)
Gyrate index Number of Core Areas
Perimeter-area ratio Core Area Index
Shape index Proximity index
Circumscribing index Normalized Proximity index
First InDOG Doctoral Conference, 29th October - 1st November 2012, Olomouc
10. METHODS - Shape & spatial metrics
Exchange index Girth index
Normalized Exchanged index Normalized Girth index
Spin index Dispersion index
Normalized Spin index Normalized Dispersion index
Perimeter index (Shape Metrics Toolbox) Range index
Normalized Perimeter index (Shape Metrics Normalized Range index
Toolbox)
Depth index Detour index
Normalized Depth index Normalized Detour index
First InDOG Doctoral Conference, 29th October - 1st November 2012, Olomouc
11. METHODS - Shape & spatial metrics
Exchange index Girth index
Normalized Exchanged index Normalized Girth index
Spin index Dispersion index
Normalized Spin index Normalized Dispersion index
Perimeter index (Shape Metrics Toolbox) Range index
Normalized Perimeter index (Shape Metrics Normalized Range index
Toolbox)
Depth index Detour index
Normalized Depth index Normalized Detour index
First InDOG Doctoral Conference, 29th October - 1st November 2012, Olomouc
12. METHODS - Multivariate statistics
● Principal Component Analysis (PCA) for
consequent clustering
● Cluster analysis:
○ DIvisive ANAlysis clustering (DIANA)
○ Partitioning Around Medoids (PAM)
● Software - Rstudio environment using R
programming language
First InDOG Doctoral Conference, 29th October - 1st November 2012, Olomouc
13. WORK-FLOW DIAGRAM
CLC (1990, 2000, 2006) DIANA
Metrics calculation
PAM
PCA Clustering
First InDOG Doctoral Conference, 29th October - 1st November 2012, Olomouc
14. RESULTS - DIANA clustering
Cluster number 1 2 3 4 5
Number of
patches
560 273 105 3 3
1 - Agriculture a. (49 %)
2 - Artificial s. (59 %)
3 - Artificial s. (42 %)
4, 5 - not so dominant
First InDOG Doctoral Conference, 29th October - 1st November 2012, Olomouc
15. RESULTS - PAM clustering
Cluster number 1 2 3 4 5
Number of
patches
191 255 210 182 6
1 - Artificial s. (43 %)
2 - Agriculture a. (45 %)
3 - Agriculture a.(51 %)
4 - Artificial s. (52 %)
5 - not so dominant
First InDOG Doctoral Conference, 29th October - 1st November 2012, Olomouc
16. CONCLUSIONS
● No significant grouping in Level 1 classes of
CLC nomenclature
● One original class does not create its own
specific class using metrics and clustering
● It is possible to group patches according to
their shape similarity
● Thus, it is needed to analyze patches
individually in more detailed level of CLC
First InDOG Doctoral Conference, 29th October - 1st November 2012, Olomouc
17. THE END
Vít Pászto
vit.paszto@gmail.com
Corine Land Cover dataset analysis with
(geo)computational methods in GIS
First InDOG Doctoral Conference, 29th October - 1st November 2012, Olomouc