- Orihuela-Rivero, Raúl
- Jay-García, Louis S.
- Reyes-Betancort, J. Alfredo
- Charlier, Elisa
- Patiño, Jairo
[Description of methods used for collection/generation of data] Taxonomic, functional and biogeographical information for vascular plants was compiled from multiple open biodiversity databases and literature sources. The modeling dataset includes species already recorded in the Canary Islands, classified as invasive or non-invasive. Predictors include taxonomy (family and genus), life form, dispersal mechanism and biogeographical range attributes. A Bayesian Additive Regression Trees (BART) model was trained to predict invasion risk based on these predictors. The model was validated using stratified five-fold cross-validation and subsequently applied to predict invasion probabilities for plant species not yet recorded in the Canary Islands., [Methods for processing the data] Data were cleaned and standardized using R software. Steps included: taxonomic standardization using POWO, removal of hybrids and unresolved taxa, filtering species lacking essential trait or distribution data, transformation and encoding of categorical predictors, model training using the dbarts package, cross-validation and threshold optimization using TSS metrics., This database compiles taxonomic, biological, and biogeographical predictors together with invasion risk predictions for vascular plant species relevant to the Canary Islands. The dataset is organized into several tables. Table S1 contains the complete compiled dataset of vascular plant species used for model development, including taxonomic identity, biological traits, and distributional variables. Table S2 lists species flagged as potentially misclassified invasive taxa following model reassessment and their evaluation using the WG-WRA system. Tables S3–S5 summarize model interpretation outputs, including permutation-based predictor importance and split-frequency importance of predictors in the BART model. Tables S6–S8 provide invasion risk predictions for plant species not currently recorded in the Canary Islands, identifying potential candidate invaders and distinguishing between high-confidence predictions and species with higher predictive uncertainty. The database also includes references and methodological metadata describing data sources, processing procedures, and modeling approaches used to generate invasion risk predictions. The dataset supports proactive identification of potentially invasive plant species and provides a resource for invasion risk assessment and biosecurity planning in island ecosystems., This research was supported by the Spanish Ministry of Science and Innovation (MICINN) project (ASTERALIEN-PID2019-110538GA-I00 and DecodAdapt-PID2023-147122NB- I00). R.O.-R. was supported by the University Teacher Training (FPU) program (FPU21/02585) funded by the Ministry of Universities and iMOVE 2025 program (iMOVE25152) funded by Spanish National Research Council (CSIC) and MICINN. LSJ-G was co-funded by the Academia Canaria de Investigación, innovación y Sociedad de la Información de la Consejería de Universidades, Ciencia e Innovación y Cultura and by European Social FundPlus (ESF+) (FPI 2021 Fellowship TESIS2021010101). J.P. was supported by the MICINN through the Ramón y Cajal Program (RYC-2016-20506)., Plant_invasion_risk_predictors_and_BART_model_predictions_Canary_Islands_flora.xlsx The spreadsheet contains the following sheets: Table S1-Complete compiled dataset of vascular plant species used in the study including taxonomic, biological and biogeographical variables Table S2-Species flagged as potentially misclassified invasive taxa during model reassessment Table S3-Summary of permutation-based predictor importance from the BART model Table S4-Frequency of use of one-hot encoded predictors in BART tree splits Table S5-Aggregated predictor importance grouped by original variables Table S6-Model predictions of invasion risk for plant species absent from the Canary Islands Table S7-High-confidence candidate invasive species (low predictive uncertainty) Table S8-Candidate invasive species with higher predictive uncertainty (watchlist candidates), Peer reviewed