boyanangelov/sdmbench: v0.2.0

v0.2.0

This release modernizes the package by replacing all retired/deprecated dependencies with current alternatives, and updates the documentation and README to match.

Breaking changes

• mlr → mlr3/mlr3learners: Learner IDs have changed — classif.logreg → classif.log_reg, classif.randomForest → classif.ranger, classif.naiveBayes → classif.naive_bayes. benchmark_sdm() now returns an mlr3 BenchmarkResult object; use bmr$score() / bmr$aggregate() instead of mlr::getBMRPerformances().
• raster/sp → terra: All spatial operations now use terra. get_benchmarking_data() returns a SpatRaster instead of RasterBrick.

Removed dependencies

• scrubr (archived on CRAN) — replaced by CoordinateCleaner
• rgdal (retired) — replaced by terra
• maptools (retired) — replaced by terra
• sp — replaced by terra

Other changes

• geodata::worldclim_global() replaces raster::getData("worldclim") for current climate data
• terra::vect / terra::spatSample replace sp::SpatialPoints / dismo::randomPoints
• Fixed deprecated keras::predict_classes / predict_proba → predict()
• Updated ENMeval API to 2.0 (occs/bg/envs params, partitions arg, auc.val.avg column)
• partition_data() checkerboard methods refactored to remove duplicated code
• Improved roxygen documentation: PR-AUC vs ROC-AUC clarified, parameter types and cross-references added throughout
• README updated: all examples use mlr3 API, custom data format corrected, retired package references removed