Access this dataset on Dryad

This dataset supports the study published in Ecology and Evolution
(Romer et al. 2025; https://doi.org/10.1002/ece3.72334), which evaluated climatic and anthropogenic drivers of habitat suitability for the invasive Cuban knight anole (Anolis equestris) across its native (Cuba) and invasive (Florida, USA) ranges.
Replicated species-distribution models (SDMs) were constructed using eight algorithms and ten independent pseudo-absence sets, with regional stratification and 10-fold cross-validation.
A Random Forest ensemble showed the highest performance and was validated with a 15-year independent monitoring dataset from South Florida.

Descriptions

ECE-2025-02-00429.R

• Type: R script
• Purpose: Implements the full SDM workflow, including data import, preprocessing, model training (eight algorithms × 10 replicates), ensemble generation, evaluation, and figure export.
• Key packages: biomod2, terra, sf, randomForest, ggplot2, glmmTMB, emmeans.
• Output: Performance metrics, variable-importance plots, and suitability rasters.

independent_survey_data.csv

• Type: Presence/absence dataset (2008–2023).
• Source: long-term (2008–2023) monitoring effort for invasive ectotherms in South Florida.
• Columns:
  • SiteID: Survey location code.
  • Latitude, Longitude: decimal degrees (WGS 84).
  • Presence: 1 = present, 0 = absent.

ThreatenedInvertDataObscured.csv

• Type: Occurrence data for three protected invertebrates
  (Liguus fasciatus, Papilio aristodemus ponceanus, Cicindelidia floridana).
• Coordinates: Randomly displaced ≤ 10 km to protect sensitive sites.
• Columns:
  • Species: Scientific name.
  • Latitude, Longitude: decimal degrees (WGS 84).
  • Source: Originating agency (FWC or FDEP).

Algo_Selection.png

• Content: Comparison of algorithm-level and ensemble-level performance using True Skill Statistic (TSS) and Boyce Index (BI).

Variable_Importance.png

• Content: Regional variable-importance scores for the final Random Forest ensemble.

Cuba_Map.jpg and Florida_Map.jpg

• Content: Habitat-suitability rasters (0–1 scale) for each region.
• Resolution: 30 arc-sec (~1 km²).

invert_cosuitability_annotated_plot.png

• Content: Predicted suitability of A. equestris at locations of the three threatened invertebrates.

SFig1.jpg – SFig4.jpg

• Content: Individual exports of supplementary figures (e.g., correlation matrix, diagnostics).

Supplementary_Figures_ECE-2025-02-00429.pdf

• Content: Combined supplemental figures

Supplementary_Tables_ECE-2025-02-00429.xlsx

• Sheets:
  • Model_Performance: TSS and BI for all algorithms × regions.
  • Variable_Importance: Scaled importance scores per predictor.
  • Predictor_Correlations: Pairwise Pearson r values (|r| > 0.7 flagged).

Key Variables

• Latitude / Longitude: decimal degrees (WGS 84)
• Presence: 1 = present, 0 = absent
• TSS, BI: unitless model-performance metrics (0–1)
• Suitability: predicted habitat suitability (0–1 continuous scale)
• BIO1 – BIO19: WorldClim bioclimatic predictors (°C or mm)
• Elevation: meters above sea level (m)
• NDVI, EVI: vegetation indices (unitless −1 to 1)
• Settlement Model Grid, Built-up Surface, Population Density: urbanization metrics (GHSL; fraction or people per km²)
• iNaturalist Effort: log-transformed observation density ln(records per km² + 1)

Key Information Sources

Environmental and ancillary data were obtained from publicly available repositories:

• WorldClim v2.1 Bioclimatic Variables: https://www.worldclim.org/data/worldclim21.html
• Global Multi-resolution Terrain Elevation Data (GMTED2010): https://topotools.cr.usgs.gov/gmted/
• Global Human Settlement Layers (GHSL): https://ghsl.jrc.ec.europa.eu/
• MODIS EVI and NDVI Products: https://modis.gsfc.nasa.gov/data/dataprod/
• Florida Administrative Boundaries (FL DOT): https://hub.arcgis.com/datasets/519e0a0ed5984bedba53695e1f56c1ee
• Cuba Administrative Boundaries (OCHA): https://data.humdata.org/dataset/cod-ab-cub
• GBIF Occurrence Records for Anolis equestris: https://www.gbif.org/species/2458613
• iNaturalist Observation Data (Phylum Chordata): https://www.inaturalist.org/

All listed sources are open-access and compatible with CC0. Derived layers (e.g., scaled rasters, effort surfaces) were authored by the dataset creators and are likewise released under CC0.

Code / Software

R (version 4.3.1 or later) is required to run ECE-2025-02-00429.R.
The script is fully annotated and self-contained:

1. Library loading
2. Data import and cleaning
3. Model fitting and ensemble construction
4. Validation and evaluation
5. Figure and table export

Hardware used: Windows 11 OS, Intel i7 CPU, 32 GB RAM.

Licensing and Attribution

• Data: Released under CC0 Public Domain Dedication (https://creativecommons.org/publicdomain/zero/1.0/).
• Creative works and figures ( Zenodo supplements ): Licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/).
• All data were generated by the authors or derived from sources permitting unrestricted reuse; no copyrighted or restricted materials are included.