Professional Documents
Culture Documents
scales in an ecosystem?
Pierre Legendre
Département de sciences biologiques
Université de Montréal
Pierre.Legendre@umontreal.ca
http://www.bio.umontreal.ca/legendre/
Community
composition = [a] [b] [c]
data table Y
[d] = Residuals
Y X W
Community Environmental Spatial
composition variables base functions
data
z^ = f(X,Y) = b0 + b1X + b2Y + b3X2 + b4XY + b5Y2 + b6X3 + b7X2Y + b8XY2 + b9Y3
Small textbook example:
20 sampling sites in the Thau lagoon, southern France.
Response (Y): 2 types of aquatic heterotrophic bacteria (log-transf.)
Environmental (X): NH4, phaeopigments, bacterial production
Spatial (W): selected geographic monomials X2, X3, X2Y, XY2, Y3
In real-life studies, partitioning is carried out on larger data sets.
Spatial eigenfunctions
Second representation:
Distance-based Moran’s eigenvector maps
(dbMEM)
(formerly called Principal Coordinates of Neighbor Matrices, PCNM)
leading to multiscale analysis in variation partitioning
Borcard & Legendre 2002, 2004
Dray, Legendre & Peres-Neto 2006
Figure – Graphs of ten of the 49 dbMEM eigenfunctions that represent the spatial
variation along a transect with 50 equally-spaced points. Abscissa, from left to right:
sites 1 to 50. Ordinates: values along the dbMEM eigenfunctions.
Truncated matrix of Euclidean
Data Euclidean distances distances = neighbor matrix
Observed 1 2 3 4 5 ... ... n–1 1...max
variable 1 2 3 4 5 ... 1...max
1...max
y 1 2 3 4 5 ...
1...max Max x 4
1 2 3 4 5 ... 1...max
1 2 3 4 5 6 7 8 9 10 11
1 2 3 4 5 ... 1...max
x 1 2 3 4 5 1...max
(spatial coordinates) 1 2 3 4 1...max
1...max
1 2 3
1...max
1 2
1
1
Y X (+) + 0 –
Truncation distance ≥ length of the longest link. Longest link here: D(7, 8) = 3.0414
Technical notes on MEM eigenfunctions
Dependent variable
10 g) Data (100%)
8
6
4 Simulated
2
data
.
0
−2
−4
−6
0 10 20 30 40 50 60 70 80 90 100
8 h) Detrended data
4 Step 1:
0
−4
detrending
−8
1,5 4
j) Broad-scale submodel (R2 = 0.058) k) Intermediate-scale submodel (R2 = 0.246)
3
1,0 PCNM #2 PCNM #6, 8, 14
2
0,5
1
0,0 0
−0,5 −1
−1,0 −2
1Peres-Neto, P. R., P. Legendre, S. Dray and D. Borcard. 2006. Variation partitioning of species data
matrices: estimation and comparison of fractions. Ecology 87: 2614-2625.
2 Blanchet F. G., P. Legendre and D. Borcard. 2008. Forward selection of explanatory variables.
Ecology 89: 2623-2632.
Example 1
Regular one-dimensional transect in upper Amazonia1
Data: abundance of the fern Adiantum tomentosum in quadrats.
Sampling design: 260 adjacent, square (5 m x 5 m) subplots forming a
transect in the region of Nauta, Peru.
Questions
• At what spatial scales is the abundance of this species structured?
• Are these scales related to those of the environmental variables?
Pre-treatment
• The abundances were square-root transformed
• and detrended (significant linear trend: R2 = 0.102, p = 0.001)
1Data from Tuomisto & Poulsen 2000, reanalysed in Borcard, Legendre, Avois-Jacquet &
Tuomisto 2004.
3
(a) Data R2 = 0.815
2
Forward selection 1
50 dbMEM eigenfunctions –1
1
(c) Medium-scale submodel, 12 PCNMs, R2 = 0.126
(periodogram analysis): -1
-1,5
1 21 41 61 81 101 121 141 161 181 201 221 241 260
Legendre, P., X. Mi, H. Ren, K. Ma, M. Yu, I. F. Sun, and F. He. 2009. Partitioning beta
diversity in a subtropical broad-leaved forest of China. Ecology 90: 663-674.
1The Gutianshan forest plot is a member of the Center for Tropical Forest Science (CTFS).
Details on the plot available at http://www.ctfs.si.edu/site/Gutianshan/.
Example 3
Gutianshan forest plot in China
Questions
• How much of the variation in species composition among sites (beta
diversity) is spatially structured?
• Of that, how much is related to the environmental variables?
Variation in
species data Y = [a] = [b] = [c] =
(beta diversity) 0.029 0.278 0.348
1 Peres-Neto, P. R. and P. Legendre. 2010. Estimating and controlling for spatial structure in
the study of ecological communities. Global Ecology and Biogeography 19: 174-184.
Stéphane Dray: MEM analysis
Eigen-decomposition
of a spatial weighting matrix W
B = 0/1 Hadamard
product A = edge
W= connectivity weighting
matrix * matrix
among sites
Dray, Legendre and Peres-Neto (2006); Legendre and Legendre (2012, Chapter 14).
Difference between classical PCNM and dbMEM
A site is connected
to itself (D = 0)
A site is not
connected to itself
(D = 4 × threshold)
• Replace A by some other weights, e.g. resistance of the landscape.
W=
B = 0/1
connectivity
matrix
among sites
Hadamard
product
*
A = edge
weighting
matrix
Guillaume Blanchet: AEM analysis
Link
Spatial asymmetry
1 = presence of a link
0 = absence of a link
# Link number
# Site number
Constructing asymmetric eigenvector maps
(AEM eigenfunctions)
Link
Spatial asymmetry
1 = presence of a link
0 = absence of a link
# Link number
# Site number
Constructing asymmetric eigenvector maps
(AEM eigenfunctions)
Link
Spatial asymmetry
1 = presence of a link
0 = absence of a link
1 2 3 4 5 6 7 8 9 10 11 12 13 14 … 57
Site 8
0 1 1 1 0 0 0 0 0 1 0 1 0 1 0 0
# Link number
# Site number
Matrix of edges E
Matrix of edges E
AEM 1 AEM 2 AEM 3
[...]
Negative values
Positive values
Three applications of AEM analysis in the following paper –
t d ir ecti on
Cu r r e n
AEM model: R2adj = 51.4% using 4 selected AEMs
Example 2 – Dominique Monti, UAG
Atya innocous