This document describes tools and methods for eco-geographical land characterization (ELC) mapping and characterization of plant genetic resource collecting sites. It discusses how ELC maps are developed by selecting important bioclimatic, geophysical, and edaphic variables through statistical analysis and clustering. It also describes how the Ecogeo tool can be used to characterize collecting sites based on geographic and environmental variables extracted from GIS data to build a characterization matrix for the sites.
Presentation 4 - SelecVar, ELCmapas and ECOGEO tools
1. Mauricio Parra Quijano
International Treaty on Plant Genetic Resources
for Food and Agriculture
CAPFITOGEN Program Coordinator
http://www.capfitogen.net
2. ELC maps
It allows the user to create eco-geographical land
characterization maps (ELC), that reflect adaptive
scenarios for a given species (or species groups) and a
specific country or region
4. Variable selection
Geophysical variables
Cluster analysis
Determination of
optimal number
of groups
Combination
(N bioclimatic*N geophysical*N edaphic)
Categories
MAP
Description of categories using original variables
Edaphic variables
Cluster analysis
Determination of
optimal number
of groups
Bioclimatic variables
Cluster analysis
Determination of
optimal number
of groups
How an ELC map is developed?
5. Expert opinion / knowledge
• Experts on target species are a valuable source of information
• Surveys are an efficient way to gather information from expert knowledge
(internet/email, meetings, workshops, etc.).
• Variable lists are made by components, with details on the nature of the variables
(explanation of codes, variable units, source, etc..). Then a value is assigned based
on the importance that a given variable has regarding the adaptation of the
species.
Bibliography search on major factors in the adaptation of target species
Variable selection – subjective/objective
Subjective option
6. • Redundancy? Correlation? Collinearity? Importance on species adaptation?
• Bivariate correlations analysis, Principal Component Analysis
• Importance of each variable analysis
x1
x2
x1
x1
x1
Variable selection – subjective/objective
Objective option:
Easiest way: Use SelecVar
7. What type of map you need?
Depending on the approach of the analysis, the ELC map can be :
1. Generalist map
2. Map by species / gene pool / group of related Sp
(Specific map)
It defines the major environments for a large number of species
(related or not). For most of these species, the ELC map should
discriminate different adaptive scenarios in a given target area. It is
expected to find unadjusted relationships between adaptive
characteristic of a smaller group of species and the resulting map (see
Parra-Quijano et al., 2012).
They define in more detail the key environments for a particular
species or a limited set of genetically related species. A good fit
between the map and the adaptive characteristics of the target
species is expected.
8. ELC mapas tool results
• Maps (which can be opened with DIVA-GIS) and tables describing each category.
9. SelecVar
It allows to select the most important and non-
redundant ecogeographical variables for ELC maps
from the objective point of view
SelecVar
10. Why this plant/population is here…
And why when you translocate this plants and provide “better
conditions” they …
11. What underlying or obvious abiotic factors
are controlling adaptation?
CAPFITOGEN tools include
105 ecogeographical variables:
67 bioclimatic
7 geophysic
31 edaphic
12. Why to select a set of most important
variables?
To obtain reliable maps showing different
ecogeographic scenarios
To obtain accurate species distribution
models
13. How to select a set of most important
variables?
What variables are the most important to create groups which represent similar plant
adaptation scenarios?
• Clustvarsel
• Random Forest
Precipitation1
Temperature12
Soil3
Landscape3
Groups
1
2
3
4
5
Speciespresencedata
14. How to select a set of most important
variables?
What variables are providing different information and have the most discriminatory
ability?
• Principal Component Analysis (PCA)
Precipitation1
Temperature12
Soil3
Landscape3
CS2
CS3
CS1
tmax11
bio1
bio3
tmin2
15. How to select a set of most important
variables?
What variables are related to others introducing redundancy?
• Bivariate correlation analysis
Precipitation1
Precipitation2
Precipitation3
Precipitation12
Temperature1
Temperature5
Annual temp
Soil2
Soil3
Landscape3
P12
P1
P3
P2
S2
S1
L1
PRECIPITATION
TEMPERATURE
SOIL
landscape
PRECIPITATION
TEMPERATURE
SOIL
LANDSCAPE
16. ECOGEO
It allows to perform eco-geographical characterization
of the geo-referenced collecting sites
17. 0 cm
5 cm
10 cm
Internodes
length
= 5.56 cm
1 2 3
1 0 1
0 1 0
= present = 1
= absent = 0
NOT of the
germplasm
but of the
collecting site
ECOGEO is a characterization
18. Process of ecogeographical characterization
Characterization
matrix :
Rows: Germplasm
identifier
Columns:
Ecogreographical
descriptors
passport
Data (including
coordinates)
GIS
Elevation
Average Annual Temp
Soil Organic Carbon
Soil pH
….
….
Y
X
19. Point or radial extraction?
2 4 3
1 3 2
1 3 2
1
1
3
1 1 3 4
Ecogeografical variable X
NA
NA
NA
NA
1 1 3 4NA
ACCENUMB VARIABLE
a NA
b NA
c 2
2 4 3
1 3 2
1 3 2
1
1
3
1 1 3 4
NA
NA
NA
NA
1 1 3 4NA
a
b
c
Distribution of
passport data entries
2 4 3
1 3 2
1 3 2
1
1
3
1 1 3 4
NA
NA
NA
NA
1 1 3 4NA
GIS overlap Extraction results
ACCENUMB VARIABLE
a NA (1)
b 1
c 3
a
b
c
True location
a=68
b=65
c=50
GEOQUAL
uncertainty
Radius
Radial extraction
20. ACCENUMB CAPTURED
VALUES
AVERAGE
a NA,1,1 1
b NA,1,1 1
c 3,2,1,3,2,
3
2.333
GIS overlap
Results of radial extraction
ACCENUMB VARIABLE
a 1
b 1
c 3
Correct extraction
ACCENUMB VARIABLE
a NA
b NA
c 2
Point extraction
1
1
2.333
Radial extraction
2 4 3
1 3 2
1 3 2
1
1
3
1 1 3 4
NA
NA
NA
NA
1 1 3 4NA