Roeland Kindt senior ecologist World Agroforestry Centre (ICRAF) [email protected]
Tree Diversity Analysis: Application to sentinel landscape data Introduction to BiodiversityR
How to measure diversity?
• Richness: species number • Diversity: richness and evenness • Sample size effects (hierarchical
structure) • Differences in species composition:
ordination and cluster analysis
Diversity: richness and evenness Less diverse More diverse
richness
evenness
Rényi diversity profiles • One of several methods for
diversity ordering • Considers richness and evenness • Based on proportional abundance
for each species
α
α
α −=
∑=
1
)ln(1
S
iip
H
Rényi diversity profiles
S1 S2
Site A
Site D
S1 S2 S2
Site C
Site B
S3
S1 S2
S3S4
S5
S3
S1
S1
S1
S1 S1
S2
S3 S3
0 0.25 0.5 1 2 4 8 Inf
scale
0.0
0.5
1.0
1.5
Ren
yi d
iver
sity
Community ACommunity BCommunity CCommunity D
Scale parameter changes influence of richness and evenness on diversity measurement
Diversity profile ranks communities according to differences in diversity
B
A = D
C
EVENNESS RICHNESS
DIV
ERSITY
Rényi diversity profiles
S1 S2
Site A
Site D
S1 S2 S2
Site C
Site B
S3
S1 S2
S3S4
S5
S3
S1
S1
S1
S1 S1
S2
S3 S3
0 0.25 0.5 1 2 4 8 Inf
scale
0.0
0.5
1.0
1.5
Ren
yi d
iver
sity
Community ACommunity BCommunity CCommunity D
ln(richness)
Shannon ln(1/Simpson) ln(1/BergerParker)
B
A = D
C
Rényi diversity profiles
0 0.25 0.5 1 2 4 8 Inf
scale
0.8
1.0
1.2
1.4
1.6
1.8
2.0
Ren
yi d
iver
sity
Community ACommunity BCommunity C
1 2 3 4 50
5
10
15
20
25
30
35 33
29
5
28
5
Community A
1 2 3 4 5 60
5
10
15
20
25
30
35
40
45 42
Community B30
810
55
1 2 3 4 5 6 70
5
10
15
20
25
30
35 32
Community C21
12
16
4
96
Partial ordering: one community is richer but with lower evenness than a second community
Rényi diversity profiles
1 2 3 40
5
1010
Community A
6
1
4
1 2 3 4 5 6 7 80
5
1010
Community B
10
66
1
4 4
1
1 2 3 4 5 6 7 80
5
10
15
17
Community C
1 11 11 1 1
1 2 3 40
5
10
15
10
Community D
10 1010
1 2 3 4 5 6 7 80
5
10
15
20
25
30
3535
Community E
1 11 11 1 1
0 0.25 0.5 1 2 4 8 Inf
scale
0.0
0.5
1.0
1.5
2.0
Ren
yi d
iver
sity
Community ACommunity BCommunity CCommunity DCommunity E
0 0.1 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.5 3 10 100 inf.
alpha
0
1
2
3
4
5
6
H
boundaryfruitfirewoodmedicinetimberconstructionshadesoil fertilitybeveragecharcoalfodderornamentalall
1.0
1.6
2.7
4.5
7.4
12.2
20.1
33.1
54.6
90.0
148.4
244.7
403.4
EXP(
H)
Kindt R, Van Damme P, Simons AJ. 2006. Tree diversity in western Kenya: using profiles to characterise richness and evenness. Biodiversity and Conservation 15: 1253-1270.
0 0.1 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.5 3 10 100 inf.
alpha
0
1
2
3
4
5H
1.0
1.6
2.7
4.5
7.4
12.2
20.1
33.1
54.6
90.0
148.4
EXP(
H)
spicesfruitfirewoodmedicinestimuliconstructionshadesoil fertilitytoolsgumsfodderdrugsvegetablesall
0 0.1 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.5 3 10 100 inf.
alpha
0
1
2
3
4
5
6
H
boundaryfruitfirewoodmedicinetimberconstructionshadesoil fertilitybeveragecharcoalfodderornamentalall
1.0
1.6
2.7
4.5
7.4
12.2
20.1
33.1
54.6
90.0
148.4
244.7
403.4
EXP(
H)
0 0.1 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.5 3 10 100 inf.
alpha
0
1
2
3
4
5
6
H
1.0
1.6
2.7
4.5
7.4
12.2
20.1
33.1
54.6
90.0
148.4
244.7
403.4
EXP(
H)
boundaryfruitfirewoodmedicinetimberconstructionshadesoil fertilityleavescharcoalfodderornamentalstakesall
0 0.1 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.5 3 10 100 inf.
alpha
0
1
2
3
4
5
6
H
boundaryfruit/nutfirewoodmedicinetimberconstructionshadecashplant supportcharcoalfodderanimal trapsornamentalstool handlesall
1.0
1.6
2.7
4.5
7.4
12.2
20.1
33.1
54.6
90.0
148.4
244.7
403.4
EXP(
H)
Cameroon Mabira
Meru W-Kenya
Rényi evenness profiles
• Based on shape of diversity profiles: when richness is equal then differences in diversity reflect differences in evenness
• Expansion of measurement of evenness of Shannon index
• Transforms evenness defined for Hill diversity series
0ln HHE −= αα
Rényi evenness profiles
1 2 3 40
5
1010
Community A
6
1
4
1 2 3 4 5 6 7 80
5
1010
Community B
10
66
1
4 4
1
1 2 3 4 5 6 7 80
5
10
15
17
Community C
1 11 11 1 1
1 2 3 40
5
10
15
10
Community D
10 1010
1 2 3 4 5 6 7 80
5
10
15
20
25
30
3535
Community E
1 11 11 1 1
0 0.25 0.5 1 2 4 8 Inf
scale
-2.0
-1.5
-1.0
-0.5
0.0
Ren
yi e
venn
ess
Community ACommunity BCommunity CCommunity DCommunity E
0 0.25 0.5 1 2 4 8 Inf
scale
0.0
0.5
1.0
1.5
2.0
Ren
yi d
iver
sity
Community ACommunity BCommunity CCommunity DCommunity E
A = B
1 10 100
Number of farms
1
10
100
Num
ber o
f spe
cies
boundaryfruitfirewoodmedicinetimberconstructionshadesoil fertilitybeveragecharcoalfodderornamentalall
W-Kenya
Species accumulation curves
Kindt R, Van Damme P, Simons AJ. 2006. Patterns of Species Richness at Varying Scales in western Kenya: Planning for Agroecosystem Diversification. Biodiversity and Conservation 15: 3235-3249.
n = 35 n = 35
n = 35
Cameroon Mabira
W-Kenya Meru
Diversity accumulation surfaces
Analysis of differences in species composition
• Diversity only depends on frequencies, not on identities of the species
• Various ecological distance measures for pairwise differences in species composition
• Various ordination and cluster methods to summarize or analyze distances
Kindt R, Van Damme P, Simons AJ, Beeckman H. 2006. Planning tree species diversification in Kenya based on differences in tree species composition between farms. I. Study of Tree Uses. Agroforestry Systems 67: 215-228.
A B
C
Euclidean distance often problematic for species composition
A B C
BiodiversityR A free manual
and software for biodiversity and community ecology analysis
http://cran.r-project.org/web/packages/BiodiversityR/
http://cran.r-project.org/web/packages/vegan/
𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵.𝐵𝐵𝐴𝐴𝐴𝐴𝐵𝐵 = 𝜋𝜋 𝑟𝑟2
𝐶𝐶𝐴𝐴𝐶𝐶𝐶𝐶𝐶𝐶.𝐵𝐵𝐴𝐴𝐴𝐴𝐵𝐵
= 𝜋𝜋 𝐴𝐴1 +𝐴𝐴2
2
2
All trees, basal area
CA = DD: NO TREES
All trees, crown area
FIREWOOD
Frequencies calculated from basal area
CHARCOAL TIMBER
FODDER SHADE LIFE FENCE
FIREWOOD
Frequencies calculated from crown area
CHARCOAL TIMBER
FODDER SHADE LIFE FENCE
… short commercial break
AUC = 0.8775
Ensemble suitability modelling
∑∑=
mod
modmod )(w
PwPensemble
Ensemble = (weighted) average of different Modeling algorithms such as random forests (RF), Maximum Entropy (MAXENT) or Support Vector Machines (SVM, SVME)
VECEA: A higher resolution map for 7 countries in eastern Africa (Ethiopia, Kenya, Malawi, Rwanda, Uganda, Tanzania and Zambia)
http://www.vegetationmap4africa.org
Application domains: • Climate change modelling • Species selection • Seed source selection • (agro)ecological zonation • Protected area evaluation • Species distribution
mapping
http://www.vegetationmap4africa.org
Seed sources and seed zones http://www.vegetationmap4africa.org
Agroforestry Species Switchboard http://www.worldagroforestry.org/our_products/databases/switchboard