MaxEntModel: Model species distributions with MaxEnt using parallel...

In brshipley/megaSDM: Integrating Dispersal and Time-step Analyses into Species Distribution Models

Model species distributions with MaxEnt using parallel processing

Description

Takes occurrence points and background points of many species and models them using the MaxEnt algorithm, parallelizing the process across multiple computer cores.

Usage

MaxEntModel(
  occlist,
  bglist,
  model_output,
  ncores = 1,
  nrep = 1,
  categorical = NA,
  alloutputs = TRUE,
  reptype = "Subsample",
  test_percent = 20,
  features = c("linear", "quadratic", "product", "threshold", "hinge"),
  testsamples = FALSE,
  regularization = 1
)

Arguments

occlist	a list of .csv file names, each containing the occurrence points of a given taxon. The files should be named the same as the taxon of interest (e.g.,: ".../Canis_lupus.csv").
bglist	a list of .csv files corresponding to the background points provided for each taxon. This list of files should be ordered in the same way as the files given by occlist.
model_output	the directory where all output files will be placed.
ncores	the number of computer cores to parallelize the background point generation on. Default is 1; Using one fewer core than the computer has is usually optimal.
nrep	(integer) the number of replicates to run each species through.
categorical	(character). If categorical variables are used for modelling (e.g., soil type), they should be distinguished from the continuous data by a prefix (e.g., "C_soiltype.bil"). Provide the distinguishing prefix here so that MaxEnt can distinguish bewteen categorical and continuous environmental layers.
alloutputs	Should secondary outputs from MaxEnt be generated TRUE/FALSE? Including: 1: A raster showing the spatial distribution of clamping for each run. 2: A multidimensional environmental similarity surface (MESS) showing novel climates. 3: Files containing the parameters used to make the response curves. 4: Plots of the response curves for each parameter The final set of arguments are optional and used for tuning the maxent model and cross-validation:
reptype	Type of replication ("Crossvalidate", "Bootstrap", "Subsample"; see MAXENT manual). Default is "Subsample".
test_percent	(numeric): integer between 0 and 100: percentage of points "held back" for crossvalidation, Test AUC validation, etc. Default is 20.
features	(optional): a vector of the features for MaxEnt to model the species- environment relationships with. Options are one or more of "linear", "quadratic", "product", "threshold", "hinge". Refer to the MaxEnt help page for more information about each feature class If there are few occurrence points, hinge features are discouraged. Default is all feature classes.
testsamples	(optional) If cross-validation with a new set of occurrence points is required, this should be a list of full file paths corresponding to the validation occurrence points for each species. This will take presidence over the random test percentage given in test_percent. NOTE: if using null AUC validation, testsamples must be given!
regularization	(numeric) regularization parameter (penalizes complex models). A higher regularization means more weight given to simpler models. Default is 1.

Value

Provides the trained model for each replicate and species (.lambdas file), a summary of the outputs provided by the maxent.jar executable, a .csv file containing information on the AUC values, threshold values, variable importance, etc., and (as requested) all of the outputs given in the alloutputs description. A full summary of the output maxent.jar provides can be found the MaxEnt manual.

brshipley/megaSDM documentation built on Nov. 26, 2024, 6:08 a.m.