sdmModelling: A function to carry out the sdm using random forest.

In charliem2003/sdmProfiling: A Spatially-explicit Species Distribution Model Evaluation Tool

A function to carry out the sdm using random forest.

Description

This function uses random forest, but you could replace it with your own if you want to use a different algorithm/model/ensemble method. Anything can be used as long as the output is a single raster object of the predicted probability of occurrences.

Usage

sdmModelling(samples, envStack, modFormula, ntrees, plot = FALSE)

Arguments

samples	data frame of the samples for model building (e.g. samples taken from species generated through create_sp). The data frame needs to have columns 'x' and 'y' for the coordinates, and 'presence' with the presence (1) and absence (0) of the species
envStack	raster stack of environmental variables used for the model (e.g output from create_env_nsets)
modFormula	model formula to be applied
ntrees	number of trees for generating the random forest
plot	TRUE or FALSE whether to plot the predicted distribution. For the plot the probability of occurrence map is converted to presence-absence by thresholding at 0.5 which is a sensible guess for random forests (i.e. 50 also plotted (absence = black, presences = red)

Value

A raster of predicted probability of occurrences

Author(s)

Charlie Marsh ([e-mail]) & Yoni Gavish

Examples

### generate a set of environmental variables
set.seed(9999)
envSet <- create_env_nsets(cellDims = c(100, 100),
                           sets     = c(4, 4, 3, 1),
                           model    = "Sph",
                           psill    = 1.5,
                           dep1     = 1,
                           rangeFun = function() exp(runif(1, 1, 6)),
                           propSamp = 0.25)

### generate a virtual species from the variables
sp <- create_sp(envStack = envSet,
                spFun    = "x[1] * x[5] * x[9]",
                spModel  = "Sph",
                spPsill  = 1,
                spRange  = 500,
                propSamp = 0.5,
                prev     = 0.1)

### an initial 'sample' of the species (assuming perfect detection)
sampPts <- data.frame(sampleRandom(sp$presence, 50, na.rm = TRUE, xy = TRUE))

### a formula to fit to random forest (additive for all vars + quadratics)
(form <- paste0("presence ~ ", paste(names(envSet), collapse = " + "), " + I(",
                paste(names(envSet), collapse = " ^ 2) + I("), " ^ 2)"))

### run the model
spMod <- sdmModelling(samples = sampPts,
                      envStack = envSet,
                      modFormula = form,
                      ntrees = 500,
                      plot = TRUE)

### compare the model against the true species distribution
par(mfrow = c(1, 3), mar = c(0.5, 0.5, 3, 3))
plot(sp$prob, axes = FALSE, main = "True prob. of occ.")
plot(sp$presence, axes = FALSE, main = "True distribution")
plot(spMod, axes = FALSE, main = "SDM prediction")

charliem2003/sdmProfiling documentation built on June 13, 2022, 4:43 a.m.