goeveg (version 0.2.0)

ordiselect: Species selection for ordination plots

Description

This function simplifies the selection of relevant species in ordination diagrams. It works with result objects from vegan package. The selection can be based upon cover abundances, frequency values and/or species fit to multivariate analysis. The result is a vector of names of the selected species and can be used for the select argument in ordination plots.

Usage

ordiselect(matrix, ord, ablim = 1, fitlim = 1, choices = c(1, 2), method = "axes", env, p.max = 0.05, freq = FALSE)

Arguments

matrix

Community data, a matrix-like object with samples in rows and species in columns.

ord

vegan ordination result object (e.g. from decorana, cca, rda or metaMDS).

ablim

Proportion of species with highest abundances to be displayed. Value between 0 and 1.

fitlim

Proportion of species with best fit to be displayed. Value between 0 and 1.

choices

Axes shown.

method

The species fit method: "axes" or "vars". See details for methods.

env

Fitted environmental variabes (result object of envfit). Only used if method = "vars".

p.max

Significance limit for variables used in method = "vars".

freq

Whether to use cover abundances (= default) or frequencies of matrix. If TRUE, frequencies of species are used.

Details

Two methods for species fit are implemented.

• In method = "vars" environmental variables are used for selecting best fitting species. This is a distance-based approach for showing the species with best species-environment-correlation in ordination diagrams. Therefore Euclidean distances between species and environment variable centroids are calculated. Only high-responding species with very close or very far distances are considered.

If method = "vars" is used, the environmental variables need to be fitted with envfit and the result of this function must be provided to the env argument. The p.max argument allows selection of only significant variables, default is p.max = 0.05. The two described methods work well both in eigenvalue-based and in distance-based ordinations. But note, that the distance-based approach for species fit is recommended for distance-based methods (e.g. NMDS), in which axes are arbitrary. If axes fit should be applied on distance-based ordination, species scores need to be calculated during the analysis, e.g. by selecting wascores = TRUE in metaMDS. On the other hand, distance calculation may be meaningless in Eigenvalue-based approaches. However, both methods provide good option of objective reduction of visible species in ordination plot for better interpretation issues. The default for matrix input is a cover-abundance-matrix. This matrix should also be used for ordination. If no limit is defined for one of the arguments ablim, fitlim, all species are displayed.

Examples

## Calculate DCA
library(vegan)
scheden.dca <- decorana(schedenveg)

## Select the 30% most abundant species and call the result
limited <- ordiselect(schedenveg, scheden.dca, ablim = 0.3)
limited

## Use the result in plotting
plot(scheden.dca, display="n")
points(scheden.dca, display="sites")
points(scheden.dca, display="species",
   select = limited, pch=3, col="red", cex=0.7)
ordipointlabel(scheden.dca, display="species",
   select = limited, col="red", cex=0.7, add = TRUE)

## Select the 30% most frequent species with 50% best axis fit
limited <- ordiselect(schedenveg, scheden.dca, ablim = 0.3,
   fitlim = 0.5, freq = TRUE)

## Select the 30% most abundant species with 60% best environmental fit
## in NDMS for axes 1 & 3
nmds <- metaMDS(schedenveg, k = 3)   # run NMDS
env13 <- envfit(nmds, schedenenv[,2:10], choices=c(1,3))
limited13 <- ordiselect(schedenveg, nmds, ablim = 0.3, fitlim = 0.6,
   choices = c(1,3), method = "vars", env = env13)