This document discusses the process for selecting ecogeographic variables to create Ecogeographic Land Characterization (ELC) maps. There are three main components considered: bioclimatic, edaphic, and geophysic variables. For each component, relevant variables are identified and subjected to statistical analysis to avoid redundancy and ensure they are not correlated or collinear. Variables can be selected objectively based on the statistical analysis or subjectively based on expert knowledge. The final selected variables are then used to categorize ELC maps, which can be generalist maps covering broad species ranges or more specific species/genus maps. The process incorporates both statistical analysis and expert knowledge to determine the optimal variables for characterizing ecogeographic conditions
Ecogeographic land characterization for CWR diversity and gap analysis Workshop - presentation 2
1. Ecogeographic variable selection
For ELC maps
Mauricio Parra Quijano
Ecogeographic land characterization for CWR diversity and gap analysis
Training workshop
26–27 February 2014, Room UG08, Learning Centre, University of Birmingham
7. ELC map obtaining process
Variable selection
Bioclimatic variables
Geophysic variables
Edaphic variables
Cluster
analysis
Cluster
analysis
Cluster
analysis
Determining
optimal number
of groups
Determining
optimal number
of groups
Determining
optimal number
of groups
Combination
(N bioclimatic*N geophysic*N edaphic)
Categories
ELC MAP
Category description by statistics from input variables
8. What variables are included in
bioclimatic component?
-Precipitation
-Temperature
-Bioclimatic indexes
9. What variables are included in
edaphic component?
-Soil type
-pH
-CIC
-% organic carbon
-Depth
-% sand, silt and clay
.
.
10. What variables are included in
geophysic component?
-Slope
-Aspect
-Elevation
-Latitude/Longitude
-Solar irradiation
11. Types of ELC maps
According to the scope of the analysis, ELC maps can be
1. Generalist maps
Define major environments for great numbers of related/unrelated
species. For most of the species the ELC map should discriminate
different adaptive scenarios. An unadjusted relationship between ELC
category and adaptive traits in a minor group of species is expected
(see Parra-Quijano et al., 2012).
2. Species/Genus/Genepool maps
Define key environments for a particular species or a limited set of
genetically related species. An adjusted relationship between ELC
category and adaptive traits is expected.
12. Variable selection by type of ELC map
Generalist map
Most recognizable influencing variables on plant physiology
Variables which are known to determine vegetation zones within the work
frame
Variables that best summarize a group of variables (annual rather than
monthly, average rather than maximum-minimum)
Species/genus/genepool map
Most recognizable influencing variables on species/genus/genepool
distribution
Most recognizable influencing variables related to most important
biotic/abiotic adaptation traits for the species/genus/genepool
Particular interesting variables for the curator/breeder
13. But in all cases, there are rules to select
Avoid correlated variables, leaving only one per group of correlation (in each
component)
Avoid collinearity in selected variables
Avoid homogeneous variables (same value for the workframe)
Avoid introducing too many variables (more than ± five per component)
Do not over-represent variables about the same aspect in a single component if
the aim is to preserve the balance. Example:
Annual Precipitation + Precipitation of Wettest Quarter + Annual Mean Temperature
14. Statistical analysis (objective selection)
• Redundancy? Correlation? Collinearity?
x2 x3
x1
x2
x1
x3
• Bivariate correlation analysis, PCA, variance inflation factor VIF
• Significance. Through multiple regression analysis using as dependent variable
(adaptive variable such as plant height, 100 seed weight).
*Collinearity: refers to an exact or approximate linear relationship between two
explanatory variables.
15. Expert knowledge (subjective selection)
2012
To take advantage of the expertise knowledge to select the most important
variables , we can use two ways to obtain this valuable information:
1. References
2. Email/internet surveys