Changelog • elements

elements 0.8.1 (01/06/2026)

Changing elements::ExamplePlot to elements::ExamplePlots, which now includes a set of vegetation plots not in the EVA and which weren’t included in the training data.
Altering taxa in TestModels to include all 26 taxa now present in the elements::ExamplePlots and Stellaria graminea + Silene flos-cuculi.
Adding additional example scenarios which includes Climate Change + Nutrient + Grazing permutations.
Altering elements::TaxonomicBackbone, now based on GBIF taxonomy (so as to include lichen taxa, otherwise absent in EuroSL).
Ensuring that elements::ModellingTaxaLookup only includes taxon_code entries for taxa in elements::ModelledTaxaCodes.

Now modelling taxa at the species and species aggregate ranks only, aggregating subspecies and varieties into aggregates, see elements::ModellingTaxaLookup.
Adding object elements::MaximumProbabilities containing the maximum predicted probability for each taxon using its training data.
Adding True Skill Statistic to elements::PerformanceMetrics.
Changing the name of the elements::VariableLookup object to elements::VariableNamesLookup.
Re-training models to using plots which were previously exluded due to a lack of coordinate precision value, but which are specified in their respective documentation as having a precision of less than 1000m.
Now using EuroSL as the taxonomic backbone, however this does not include lichens - to be revised!
Moving calculation of HOAData to elementsWorkflow.
Updating vignettes with new scenario data.
Updating documentation to reflect use of ClimateEU.

Adding models argument to elements::predict_occ_taxon.
Adding elements::ModellingTaxaLookup object.
Adding object elements::ModelledTaxaCodes which contains the taxon codes for modelled taxa.
Changing filehash database format to RDS.
Re-fitting models. Key changes include:
- Decreasing the minimum presence threshold to 30.
- Changing the source of climate data to ClimateEU.
- Altering the hyper-parameter tuning ranges.
- Changing spatio-temporal cross-validation technique to blockCV.

Adding function elements::plot_be which creates a bivariate probability/habitat suitability plot.
Adding function elements::calc_overlap which calculates pairwise niche overlap values for a set of taxa.
Adding function elements::extract_svs which extracts the support vectors from a model. PROTOTYPE.
Ensuring the full variable names are passed to the x axes in elements::plot_me.

Adding object elements::Gradients which contains sequences of values between the minima and maxima for each environmental variable in elements::VariableNames.
Adding object elements::HOAData which contains predictions for each variable along its gradient (see elements::Gradients) produced using the elements::predict_occ function with the ‘holdopt’ argument set to all other variables.
Incorporating the elements::HOAData object into elements::plot_me, which can be selected for plotting by setting thr ‘me_type’ argument to “hoa”.
Ensuring that the X axis limits in the elements::plot_me extend to the minima and maxima of the gradients as present in elements::Gradients.

Fixing issue where elements::env_filter does not work when there are no taxa-predictor combinations screened out.

Adding two functions elements::envelope_filter and elements::envelope_filter_taxon which uses the distributional data present in elements::NicheWidths to apply envelope filters. These functions are integrated into the elements::env_filter function as a pre-screening option.
Separating out the calculation of euclidean distance in the elements::env_filter function into two new functions: elements::calc_distance_once and elements::calc_distance.
Ensuring that all objects have data for the taxa present in the ENMs.
Changing the ‘append_predictors’ argument to ‘append’ in elements::predict_occ_taxon, elements::predict_occ, elements::calc_distance_once, and elements::calc_distance, this argument now controls which columns are appended to the results, namely: “all”: all columns supplied in the predictors dataframe; “predictors”: all predictor variable columns (elements::VariableNames); or “ids”: all non-predictor variable columns.
Adding three new objects elements::EUNISDiagnosticTaxa, elements::EUNISConstantTaxa, and elements::EUNISDominantTaxa, derived from https://doi.org/10.1111/avsc.12519.
Creating new pkgdown site article “Environmental Filtering”.

Full model training re-run after addressing outstanding taxonomic alignment issues. Now at 6631 taxa.
Bundling imbalance information in elements::Imbalances
Re-creating elements::TaxonomicBackbone object, now using GBIF as EuroSL does not include lichen or hybrid taxa. I will also align this taxonomic backbone to the EIVE1.5 “Euro+Med augmented” taxonomy upon release of EIVE1.5.
Changing pkgdown site theme to sandstone.
Improving elements::startup, which now creates a new environment (elementsEnv) in the parent environment which contains the ENMs, rather than simply loading the ENMs directly into the global environment.

Removing superfluous code from elements::predict_occ.
Including example code running scenario c in the README.
Adding Zenodo README badge.
Adding function elements::env_filter, which filters a given set of taxa using a set of predictor variables to retain the most suitable taxa only.

Bundling two additional example objects elements::ExampleScenarios and elements::ExamplePlot

Including test models (Silene flos-cuculi and Stellaria graminea) to ensure the package is functional (for these two taxa). These models are present in the “./inst/testdata/TestModels” object which is loaded if “./inst/extdata/Models” is not present, or if specified in the elements::startup function.

Adding the most recent iteration of the models, which we re-trained on the EVA data and data from the GBNVPD not yet incorporated into the EVA.
Adding the ‘holdopt’ and ‘limit’ arguments to elements::predict_occ_taxon and elements::predict_occ.
Removing the elements::ModelTaxa object.
Adding the elements::TaxonomicBackbone object.
Changing the name of the elements::VariableNames object to elements::VariableLookup and adding a new object elements::VariableNames which contains a vector of the model variable names only.

Adding the ability to plot multiple taxa in elements::plot_me function.
Adding the argument “normalise” to the elements::plot_me function, which re-scales the y axis values in PDP plots using min-max scaling.