Similarity maps for partition groups

Maps environmental similarity of reference partitions (occurrences or background) to all cells with values in the predictor variable rasters. This function uses raster data, and thus cannot map similarity values using only tables of environmental values f or occurrences or background. Further, this function does not calculate similarity for categorical variables.

Usage

evalplot.envSim.map(
  e = NULL,
  envs,
  occs.z = NULL,
  bg.z = NULL,
  occs.grp = NULL,
  bg.grp = NULL,
  ref.data = "occs",
  sim.type = "mess",
  categoricals = NULL,
  envs.vars = NULL,
  bb.buf = 0,
  occs.testing.z = NULL,
  plot.bg.pts = FALSE,
  sim.palette = NULL,
  pts.size = 1.5,
  gradient.colors = c("red", "white", "blue"),
  na.color = "gray",
  return.tbl = FALSE,
  return.ras = FALSE,
  quiet = FALSE
)

Arguments

e

ENMevaluation object (optional)

envs

RasterStack: environmental predictor variables used to build the models in "e"; categorical variables should be removed before input or identified with the argument "categoricals", as they cannot be used to calculate MESS

occs.z

data frame: longitude, latitude, and environmental predictor variable values for occurrence records, in that order (optional); the first two columns must be named "longitude" and "latitude"

bg.z

data frame: longitude, latitude, and environmental predictor variable values for background records, in that order (optional); the first two columns must be named "longitude" and "latitude"

occs.grp

numeric vector: partition groups for occurrence records (optional)

bg.grp

numeric vector: partition groups for background records (optional)

ref.data

character: the reference to calculate MESS based on occurrences ("occs") or background ("bg"), with default "occs"

sim.type

character: either "mess" for Multivariate Environmental Similarity Surface, "most_diff" for most different variable, or "most_sim" for most similar variable; uses similarity function from package rmaxent

categoricals

character vector: names of categorical variables in input RasterStack or data frames to be removed from the analysis; these must be specified as this function was intended for use with continuous data only

envs.vars

character vector: names of a predictor variable to plot similarities for; if left NULL, calculations are done with respect to all variables (optional)

bb.buf

numeric: distance used to buffer (extend) the mapping extent in map units; for latitude/longitude, this is in degrees (optional)

occs.testing.z

data frame: longitude, latitude, and environmental predictor variable values for fully withheld testing records, in that order; this is for use only with the "testing" partition option when an ENMevaluation object is not input (optional)

plot.bg.pts

boolean: if TRUE, plot background points when using ref.data = "bg"

sim.palette

character: RColorBrewer palette name to use for plotting discrete variables; if NULL, default is "Set1"

pts.size

numeric: custom point size for ggplot

gradient.colors

character vector: colors used for ggplot2::scale_fill_gradient2

na.color

character: color used for NA values

return.tbl

boolean: if TRUE, return the data frames of similarity values used to make the ggplot instead of the plot itself

return.ras

boolean: if TRUE, return the RasterStack of similarity values used to make the ggplot instead of the plot itself

quiet

boolean: if TRUE, silence all function messages (but not errors)

Value

A ggplot of environmental similarities between the occurrence or background data for each partition and all predictor variable values in the extent.

Details

When fully withheld testing groups are used, make sure to input either an ENMevaluation object or the argument occs.testing.z. In the resulting plot, partition 1 refers to the training data, while partition 2 refers to the fully withheld testing group.

Rasters are plotted showing the environmental similarity estimates for each partition group. The similarity between environmental values associated with the validation occurrence or background records per partition group and those associated with the entire study extent (specified by the extent of the input RasterStack "envs") are calculated, and the minimum similarity per grid is returned. For option "mess", higher negative values indicate greater environmental difference between the validation occurrences and the study extent, and higher positive values indicate greater similarity. This function uses the `similarity()` function from the package `rmaxent` (https://github.com/johnbaums/rmaxent/) to calculate the similarities. Please see the below reference for details on MESS.

References

Baumgartner J, Wilson P (2021). _rmaxent: Tools for working with Maxent in R_. R package version 0.8.5.9000, <URL: https://github.com/johnbaums/rmaxent>. Elith, J., Kearney, M., and Phillips, S. (2010) The art of modelling range-shifting species. Methods in Ecology and Evolution, 1: 330-342. doi:10.1111/j.2041-210X.2010.00036.x