modEvA (version 3.33)

MESS: Multivariate Environmental Similarity Surfaces based on a data frame

Description

This function performs the MESS analysis of Elith et al. (2010) to determine the extent of the environmental differences between model training and model projection (extrapolation) data. It is applicable to variables in a matrix or data frame.

Usage

MESS(V, P, id.col = NULL, verbosity = 2)

Value

The function returns a data frame with the same column names as P, plus a column named TOTAL, quantifying the similarity between each point in the projection dataset and those in the reference dataset. Negative values indicate localities that are environmentally dissimilar from the reference region. The last column, MoD, indicates which of the column names of P corresponds to the most dissimilar variable, i.e., the limiting factor or the variable that drives the MESS in that locality (Elith et al. 2010).

Arguments

V

a matrix or data frame containing the variables (one in each column) in the training dataset.

P

a matrix or data frame containing the same variables in the area to which the model(s) will be projected. Variables (columns) must be in the same order as in V, and colnames(P) must exist.

id.col

optionally, the index number of a column containing the row identifiers in P. If provided, this column will be excluded from MESS calculations but included in the output.

verbosity

Integer number indicating the amount of messages to display while computing the results. The default is to display all messages. Set verbosity=0 for no messages.

Author

Alberto Jimenez-Valverde, A. Marcia Barbosa

Details

When model predictions are projected into regions, times or spatial resolutions not analysed in the training data, it may be important to measure the similarity between the new environments and those in the training sample (Elith et al. 2010), as models are not so reliable when predicting outside their domain (Barbosa et al. 2009). The Multivariate Environmental Similarity Surfaces (MESS) analysis measures the similarity in the analysed variables between any given locality in the projection dataset and the localities in the reference (training) dataset (Elith et al. 2010).

MESS analysis is implemented in the MAXENT software (Phillips et al. 2006) and in the dismo R package, but there it requires input variables in raster format. This implies not only the use of complex spatial data structures, but also that the units of analysis are rectangular pixels, whereas we often need to model distribution data recorded on less regular units (e.g. provinces, river basins), or on equal-area cells that are not necessarily rectangular (e.g. UTM cells, equal-area hexagons or other geometric shapes). The MESS function computes this analysis for variables in a data frame, where localities (in rows) may be of any size or shape.

References

Barbosa A.M., Real R. & Vargas J.M. (2009) Transferability of environmental favourability models in geographic space: the case of the Iberian desman (Galemys pyrenaicus) in Portugal and Spain. Ecological Modelling 220: 747-754

Elith J., Kearney M. & Phillips S. (2010) The art of modelling range-shifting species. Methods in Ecology and Evolution 1: 330-342

Mesgaran M.B., Cousens R.D. & Webber B.L. (2014) Here be dragons: a tool for quantifying novelty due to covariate range and correlation change when projecting species distribution models. Diversity and Distributions, 20: 1147-1159

Owens H.L., Campbell L.P., Dornak L.L., Saupe E.E., Barve N., Soberon J., Ingenloff K., Lira-Noriega A., Hensz C.M., Myers C.E. & Peterson A.T. (2013) Constraints on interpretation of ecological niche models by limited environmental ranges on calibration areas. Ecological Modelling, 263: 10-18

Phillips S.J., Anderson R.P. & Schapire R.E. (2006) Maximum entropy modeling of species geographic distributions. Ecological Modelling 190: 231-259

See Also

OA; mess in packages dismo and predicts; ecospat.climan in package ecospat; kuenm_mop and kuenm_mmop in package kuenm

Examples

if (FALSE) {
# load package 'fuzzySim' and its sample data:
require(fuzzySim)
data(rotif.env)


# add a column specifying the hemisphere:

unique(rotif.env$CONTINENT)

rotif.env$HEMISPHERE <- "Eastern"

rotif.env$HEMISPHERE[rotif.env$CONTINENT %in%
c("NORTHERN_AMERICA", "SOUTHERN_AMERICA")] <- "Western"

head(rotif.env)


# perform a MESS analysis
# suppose you'll extrapolate models from the Western hemisphere (Americas)
# to the Eastern hemisphere (rest of the world):

names(rotif.env)  # variables are in columns 5:17

west <- subset(rotif.env, HEMISPHERE == "Western", select = 5:17)
east <- subset(rotif.env, HEMISPHERE == "Eastern", select = 5:17)
east.with.ID <- subset(rotif.env, HEMISPHERE == "Eastern",
select = c(1, 5:17))

head(east)
head(east.with.ID)  # ID is in column 1

mess <- MESS(V = west, P = east)
mess.with.ID <- MESS(V = west, P = east.with.ID, id.col = 1)

head(mess)
head(mess.with.ID)

range(mess[ , "TOTAL"])
}