Data from: Species-specific responses to multiple climatic variables predict diverging locations of future climate change refugia

Rose, Miranda 1 ; Velazco, Santiago José Elías1 ; Thorne, James H.2; Regan, Helen3 ; Franklin, Janet1

Published Apr 13, 2026 on Dryad. https://doi.org/10.5061/dryad.mgqnk9994

Abstract

This dataset supports analyses of species-specific habitat suitability responses to projected climate change across 81 plant species in California. It includes pixel-level habitat suitability estimates from median consensus species distribution models under two general circulation models (CNRM-CM5 and HadGEM2-ES), alongside geographic features (latitude, distance to coast, elevation, and topographic heterogeneity) and projected changes in five hydroclimatic variables derived from the Basin Characterization Model (270 m resolution). Also included are species-level Spearman correlation values relating change in habitat suitability to each geographic and hydroclimatic variable. Data are provided as compressed CSV files alongside R scripts used for data preparation, processing, analysis, and figure generation. This dataset could and may be reused to explore climate-driven range shifts, refugia identification, and plant species responses to hydroclimatic change in Mediterranean-type ecosystems.

Description of the data and file structure

The data include 1) pixel level data for the 81 study species of current and future habitat suitability estimated by median consensus species distribution models, geographic features, and changes in hydroclimatic variables estimated by the Basin Characterization Model (270 m) and 2) the resulting spearman correlation values calculated between the change in habitat suitability and each of the geographic features and change in hydroclimatic variables. Code used to complete the correlation and linear regression analysis and create figures is included.

Files and variables

File: SI_species_information_clean.csv

Description: Species-level metadata for the 81 study species including growth form, primary habitat/vegetation community, SDM model performance metrics (AUC and TSS), spatial partitioning parameters, and block size used for model evaluation. This file contains an embedded data dictionary (Table 1, rows 1–14) followed by the data (Table 2, beginning row 19). Users reading this file in R or Python should skip the first 18 rows to load the data directly. cell_size is NA for species where latitudinal bands were used for spatial partitioning (n_part = 2), as block size is not applicable for those species.

File: SI_species_habitat_corr_data_clean.csv

Description: Spearman correlation results for relating habitat change to each geographic and change in hydroclimatic variables (species-level) for each of the general circulation models included (cnrm: CNRM-CM5 and hades: HadGEM2-ES). The file SI_species_habitat_corr_data.csv contains an embedded data dictionary (Table 1, rows 1–22) followed by the data (Table 2, beginning row 25). Users reading this file in R or Python should skip the first 24 rows to load the data directly.

File: Data_dictionary_species_habitat_grid.pdf

Description: Variable definitions for SI_species_habitat_grid_data are provided here.

File: SI_species_habitat_grid_data.gz

Description: Pixel-level data for current and future habitat suitability and ancillary geographic and hydroclimatic variables for each of the 81 study species under two general circulation models (CNRM-CM5 and HadGEM2-ES). Note: After decompression, this file is a CSV and can be read accordingly in R, Python, or any standard tabular data software. NAs reflect pixels where the source raster layers had no data coverage at those locations and are not applicable, rather than missing. All other columns are complete. Variable definitions are stored in a separate data dictionary file (Data_dictionary_species_habitat_grid.pdf).

Code/software

00_data_prep.R: R script for post-processing species distribution model outputs, including correction of extrapolation areas and preparation of current and future habitat suitability rasters for further analysis

01_data_processing.R: R script for processing and extracting pixel-level habitat suitability and ancillary geographic and hydroclimatic variables, and generating exploratory figures

02_data_analysis.R: R script for primary Spearman correlation and linear regression analyses relating habitat suitability change to geographic and hydroclimatic variables, and generating manuscript figures

03_Climate_change_fig.R: R script for generating maps of projected climate change in hydroclimatic variables across the study region

04_data_analysis2.R: R script for secondary analyses including random forest, PCA, clustering, and GAM models exploring species-level drivers of habitat suitability change

Packages used include: flexsdm, terra, tidyverse, broom, dotwhisker, ggplot2, viridis, landscapemetrics, data.table, GGally, ggcorrplot, patchwork, ggpubr, ggsci, randomForest, pheatmap, cluster, factoextra, mgcv, and car. These scripts are provided for transparency to document the full analytical workflow. The CSV files included in this repository are sufficient to reproduce the correlation analyses and figures presented in the manuscript. Note that not all upstream input data files (e.g. raw SDM outputs, BCM rasters) are included in this repository, and file paths are hardcoded to the original author's directory structure. As a result, the complete pipeline is not fully reproducible without access to the original data sources and path adjustments.