Changes in v.0.2.4 / 24 September 2025

* Added functionality
  
  1. Hurdle models now also work with data with generated zeroes.

  2. Hurdle models now can read plotdata also from the workspace.

  3. Cutoffs for presence-part in hurdle models are now flexible via
     weighting the relationship between sensitivity and specificity or
     by setting the values for sensitivity or specificity.

* Bug fixes
  
  1. All terra::writeRaster provided with NAflag = -9999. Reason: Some
     terra versions use nan for NODATA_value which renders them
     unreadable; this is avoided by setting the NAflag.

  2. data2SDMaps2 does not crash any longer under Windows when
     argument outdir is not defined. This happened due to an escape
     character in the path to a temporary file in directory C:\Users.
  
----------------------------------------------------------------------  

Changes in v.0.2.3 / 24 March 2025

* Bug fixes

  1. Bug fixed in making partial plots: wrong column header in variable
     contributions was addressed

  2. Bug fixed in data2SDMaps2: When n was not defined in argument
     gen.zer.options no zeroes were generated instead of setting the
     number of zeroes to the number of observations

* Other changes

  1. Informative error message added for the case when there are no
     grids for environmental covariates that are present in the
     plotdata.

  2. Ranger bias correction is not used any longer in the presence
     part of a hurdle model.
  
----------------------------------------------------------------------
  
Changes in v0.2.2 / 9 Dec 2024

* Added functionality

  Logical argument calc.pvalues added to function SDMapsSummary2:
  computational intensive calculation of p-values can be toggled

* Bug fix

  Bug fixed in making partial plots: wrong column header in variable
  contributions was addressed
  
----------------------------------------------------------------------
  
Changes in v0.2.1 / 30 Okt 2024

* Added functionality

  sdmdata for the 2nd part of a hurdle model do not overwrite the
  original sdmdata in directory OBSERVATIONS but are stored in
  directory OBSERVATIONS.hurdle2.
  
* Bug fixes

  1. Bug fixed in on-screen success messages

  2. Bug fixed in combining Hurdle parts 1 and 2  

----------------------------------------------------------------------
  
Changes in v0.2.0 / 11 Okt 2024

Added functionality

  SDMaps2 now also can make hurdle models.
  
----------------------------------------------------------------------

Changes in v0.1.0 / 5 Sept 2024
  
* Added functionality

  1. model type ranger also yields maps of standard errors and maps of
     predictions +/- standard errors

  2. Variable importance for ranger also includes significance levels
  
  3. Argument added in SDMaps2 to choose between pdf and cairo

* Bug fixes  
  
  1. Bug fixed in adding interpolated residuals: Hard lower limit (0) for
     count and density predictions
  
  2. Bug fixed in location plot: now all contours are black
  
  3. Bug fixed in residual interpolation; is now made with help of
     package stars instead of sf.

----------------------------------------------------------------------

Changes in v0.0.02 / 14 Nov 2023

All remaining package sp code is replaced by terra code.
  
----------------------------------------------------------------------

Changes in v0.0.1 / 3 Nov 2023

* Added functionality

  1. Function dataExploration now also works with generated zeroes

  2. Hyperparameter tuning implemented for xgboost count and density
     models

  3. Bias correction for ranger predictions also implemented in
     crossvalidation

  4. Crossvalidation added for xgboost models
  
* Bug fix: No probabilities of <0 and >1 any longer in presence models

* Other changes

  1. Screen messages are unified

  2. Bias correction for ranger now silent and without bulky screen
     messages
    
----------------------------------------------------------------------

v0.0.0 / 1 Sept 2023

SDMaps2 is a condensed version of SDMaps with functionality reduced to
frequently used standard procedures. The most important differences
are:

  1. Only random forests (via algorithm ranger) and boosted regression
     trees (via algorithm XGBoost) are used as modelling techniques.

  2. All commands used from packages maptools, sp and raster have been
     replaced by commands from sf and terra as far as possible.
