Skip to content

1.2.2 updates

Andrew Schwenke requested to merge patches-CC into main

1.2.2

  • Data split improving handling
    • Improving and adding handling for models when using data split and repeats
      • For now this applies only to biomod2 models, and is a work in progress
      • Model evaluation tables saved in eval_tables (e.g. Loss-function, Evluation-statistics, etc) will now show values for each individual model that was generated
      • Two new plots in eval including a mean and standard deviation plot and a boxplot of the evaluation statistics for comparison of all models that were generated
  • Categorical predictor changes
    • Enable CTA to run with categorical predictors
    • Enable unconstrained projection for experiments that have categorical predictors
      • edit: reverted this in the short term because it can be compute intensive (workshops etc.) but the functionality exits
      • This was previously disabled due to the "new levels" issue but functions have been added that compare levels used in the fitted model to the new environment, and levels that didn't appear in the model are dropped from the new environment.
    • Added tryCatch for unconstrained projections
  • MaxEnt changes
    • MaxEnt overhauled // in progress
    • Updated MaxEnt to handle replicates
    • Updated MaxEnt to use maximum sensitivity and specificity threshold (cloglog) to generate binary map (habitat suitability)
    • MaxEnt predictions with replicates will output GeoTIFFs containing one layer per model, and a mean layer (pending pipeline changes TBD)
    • Removed output of MaxEnt model objects for cloglog and logistic as these were identical to the raw object - transformation occurs for prediction
    • Needs much refactoring going forward but for now is stuck in places between the old (dismo/raster) and new (terra) functions
    • Some changes will have to wait for added front end inputs to support more complicated workflows and choosing which models and thresholds etc to use for evaluations, or how results for multiple predictions should be summarised and reported.
  • Bug fixes
    • Fix bug in bioclim model causing projection to fail
    • Fix bug EC_projecting_ensemble - function wasn't exported
  • Raster plots
    • Rolled out the updated raster plots for most models now, replacing base R raster plots with ggplot versions.

1.2.1 - 2023-03-20

  • Fix bug in dummy raster creation affecting re-sampling
  • Fix bug when constraining raster causing NaNs in values over 255 due to terra assuming 8 bit integer format
  • Add EC_save_projection_generic to update old format raster plots.
    • Rolling out starting with bioclim model.
  • Refactoring of EC_projecting_ensemble

Merge request reports