------------------------------------------------------------
Species distribution modeling,
calibration and evaluation,
ensemble modeling
------------------------------------------------------------

https://biomodhub.github.io/biomod2/

Installation

• Stable version from cran :

install.packages("biomod2", dependencies = TRUE)

• Development version from biomodhub :

library(devtools)
devtools::install_github("biomodhub/biomod2", dependencies = TRUE)

All changes between versions are detailed in News.

biomod 4.3-4-1 - Abundance modelling, but better !

Please feel free to indicate if you notice some strange new behaviors !

What is changed ?

• Nothing for presence/absence (or presence-only) modelling.
• bm_RangeSize computes the difference between binary predictions, and is called by BIOMOD_RangeSize which can take as input BIOMOD.projection.out objects.

What is new ?

• You can now use non-binary data ! All the information can be found in the Abundance Vignette.
• 6 values can be given to data.type new parameter in BIOMOD_FormatingData : binary, count, multiclass, ordinal, relative, abundance.
• 2 new ensemble models have been added, to deal with qualitative data type : EMmode and EMfreq.
• DNN (Deep Neural Network) model has been added with the cito package. It can be used with all data types.
  Be sure to have a look at cito documentation before, especially the part about the torch installation.
• bm_ModelAnalysis provides classical graphics to explore the quality of your single models.
• Discover BIOMOD_Report, a new function to help you summarize all your modeling steps.

biomod 4.2-6 - Improved OptionsBigBoss and new model

What is changed ?

• To improve the models, we made some change in the options for OptionsBigboss. (This only affects the ANN, CTA and RF models.) You can check all your options with the get_options() function.
• To reduce the tuning calculation time, we update the tuning ranges for ANN, FDA and MARS models.

What is new ?

• biomod2 has a new model: RFd. It's a Random Forest model with a down-sampling method.
• You can now define seed.val for bm_PseudoAbsences() and BIOMOD_FormatingData().
• New fact.aggr argument, for pseudo-absences selection with the random and disk methods, allows to reduce the resolution of the environment.
• Possibility to give the same options for all datasets with "for_all_datasets" in bm_ModelingOptions().

biomod 4.2-5 - Modeling options & Tuning Update

What is changed ?

• modeling options are now automatically retrieved from single models functions, normally allowing the use of all arguments taken into account by these functions
• tuning has been cleaned up, but keep in mind that it is still a quite long running process
• in consequence, BIOMOD_ModelingOptions and BIOMOD_Tuning functions become secondary functions (bm_ModelingOptions and bm_Tuning), and modeling options can be directly built through BIOMOD_Modeling function

What is new ?

• ModelsTable and OptionsBigboss datasets (note that improvement of bigboss modeling options is planned in near future)
• 3 new vignettes have been created :
  • data preparation (questions you should ask yourself before modeling)
  • cross-validation (to prepare your own calibration / validation datasets)
  • modeling options (to help you navigate through the new way of parametrizing single models)

biomod 4.2 - Terra Update

What is changed ?

• biomod2 now relies on the new terra package that aims at replacing rasterand sp.
• biomod2 is still compatible with old format such as RasterStackand SpatialPointsDataFrame.
• biomod2 function will sometimes return SpatRaster from package terra that you can always convert into RasterStack using function stack in raster.

biomod 4.1 is now available

/!\ Package fresh start... meaning some changes in function names and parameters. We apologize for the trouble >{o.o}<
Sorry for the inconvenience, and please feel free to indicate if you notice some strange new behaviors !

What is changed ?

• all code functions have been cleaned, and old / unused functions have been removed
• function names have been standardized (BIOMOD_ for main functions, bm_ for secondary functions)
• parameter names have been standardized (same typo, same names for similar parameters across functions)
• all documentation and examples have been cleaned up

What is new ?

• plot functions have been re-written with ggplot2
• biomod2 website has been created, with proper roxygen2 documentation and help vignettes

But... why ?

• “For every minute spent on organizing, an hour is earned.” — Benjamin Franklin
  • better documentation, better formation, better help provided
  • new improvements to come (update of single models, implementation of abundance models, etc)