Species exposure to climate extremes, 1941 to 2025: analysis code

This repository holds the R code behind "The 2025 Biodiversity Exposure Report" (Merow et al., in revision at BioScience). It computes, for 32,345 terrestrial vertebrate species and every year from 1941 to 2025, where in each species' range the climate was more extreme than the species had experienced during a historical baseline.

This is a fixed copy of the code for the paper. It is not under active development.

Related resources:

• Web dashboard: https://speciesexposure.github.io/
• Shiny app and the derived data: https://github.com/SpeciesExposure/exposureApp. Its file AI_HANDOFF.md documents the derived data in detail.

The method in brief

1. Climate. ERA5 monthly 2 m temperature and total precipitation are summarized by water year on a 0.25 degree grid. Water year Y runs from October of Y-1 to September of Y.

2. Variables. Six climate tables give eight exposure variables.

   Climate table	Meaning	Upper limit	Lower limit
   temp__12	Mean temperature of the water year	yes	yes
   temp__3__max	Mean temperature of the warmest 3 consecutive months	yes
   temp__3__min	Mean temperature of the coldest 3 consecutive months		yes
   precip__12	Total precipitation of the water year	yes	yes
   precip__3__max	Precipitation of the wettest 3 consecutive months	yes
   precip__3__min	Precipitation of the driest 3 consecutive months		yes

3. Ranges. Each species' range is the set of land cells it occupies. Mammals and birds use Area of Habitat maps. Amphibians and reptiles use IUCN range polygons.

4. Historical limit. For each species and variable, the 99th percentile across the baseline years is taken in every range cell, and the limit is the 99th percentile of those values across cells. Lower limits use the 1st percentile in the same way. The baseline is 1941 to 2022.

5. Exposure. A cell is exposed in a year when its value is beyond the limit. A species is exposed in a year when at least 25% of its range is exposed for at least one variable.

Requirements

• R 4.2 or later
• R packages: tidyverse, sf, terra, qs, qs2, gridExtra, scales, tidyterra, ggnewscale, patchwork, cowplot, magick, rnaturalearth, jsonlite
• A machine with several cores and tens of GB of free disk space. The exposure calculation writes one file per species and variable.

Input data

None of the input data is stored here. Download it and either place it under inputs/ as listed in config/paths.R, or create config/paths_local.R and point INPUTS and WORK_DIR to your own locations.

Input	Source
ERA5 monthly means, 1940 to 2025: 2 m temperature, total precipitation, land-sea mask	Copernicus Climate Data Store, https://cds.climate.copernicus.eu/datasets/reanalysis-era5-single-levels-monthly-means
Area of Habitat maps for mammals and birds	Lumbierres et al. (2022)
Range polygons for amphibians and reptiles	IUCN Red List spatial data, https://www.iucnredlist.org/resources/spatial-data-download
Assessment and taxonomy tables for all four groups	IUCN Red List
Ecoregions	RESOLVE Ecoregions 2017
Country borders for map overlays	TM World Borders 0.3

The ecoregion raster Ecoregions2017_v2.tif is the ecoregion id, ECO_ID, of the RESOLVE shapefile on the 0.25 degree grid. ecoregionMetadata.rds is the attribute table of the same shapefile.

Three small inputs are stored in data/:

File	Content
oni_v5_noaa_cpc_2026-09-15.txt	Oceanic Niño Index from the NOAA Climate Prediction Center, used to mark El Niño and La Niña years
body_mass_quartiles.csv	Body-mass quartile, 1 to 4, of each species within its group. The quartiles were estimated with a large language model. They are used only for the body-size panel of Fig. S1
species_missing_taxon_group_filled.csv	Taxonomic group of three species that have none in the IUCN tables. The groups were proposed by a large language model

Running the analysis

Run every script from the repository root. R/0_RunAll.r lists the steps in order.

Step	Script	What it does	Main outputs
1	R/1_Setup.r	Packages, run settings, directories
2	R/2_Functions.r	Functions for the exposure calculation
3	R/3b_ClimatePrep.r	Climate tables and the land grid	landTemplate.tif, six climate tables
4	R/3a_SpeciesRanges.r	Species ranges on the grid	one file of cell ids per species
5	R/4_Metadata.r	IUCN category, taxonomy, ecoregions and range size	spAttributes_v8.qs
6	R/5_CalcExposure.r	Historical limits and exposed cells for every species and variable	per-species files in variable_outputs/
7	R/6_MakeDF.r	Collects the results	singleMaxExposure_v1.qs, AllCellExposureSpXVar.qs, singleMaxExposure2025_v1.qs
8	R/8_AppData.r	Data files for the Shiny app	files in output/exposureApp_extdata/

singleMaxExposure_v1.qs has one row per species and year in which the species was exposed, with the variable that had the largest exposed range fraction. The counts in the paper come from this file. AllCellExposureSpXVar.qs has one row per species, variable, year and exposed cell.

Paper item	Script
Table 1, Table 2	tables/Table1_Table2.r
Table 3	tables/Table3_CohortByTaxonIUCN.r
Figure 1, Figure S1	figures/Fig1_FigS1_Time.r
Figure 2	figures/Fig2_Running.r
Figure 3	figures/Fig3_IUCN.r
Figure 4, Figure S5	figures/Fig4_FigS5_Maps.r
Figure 5	figures/Fig5_Stressor.r
Figure S2	figures/FigS2_TimeSensitivity.r
Figure S3	figures/FigS3_TimeThreshold.r
Figure S4	figures/FigS4_IUCN_Chronic.r
Figure S6	figures/FigS6_ProportionalExposure.r

figures/enso_strip.R holds the functions that mark El Niño and La Niña years. Run figures/Fig4_FigS5_Maps.r before figures/FigS6_ProportionalExposure.r, because the second uses a raster written by the first.

Not included: the example species figure in Box 2, the screenshots in Figure 6, and the taxon group column that was added to Table 1 for the paper.

Baseline sensitivity runs

Figure S2 compares the main run with two runs that end the historical baseline earlier. Each is a separate run of steps 6 and 7 with different settings:

END_HISTORICAL_YEAR <- 2015L; RUN_NAME <- "V8_2015"
source("R/1_Setup.r"); source("R/5_CalcExposure.r"); source("R/6_MakeDF.r")

Figure S3 compares percentile choices. It needs no extra run, because step 6 stores the limits for several percentiles.

Checks made on this copy

The scripts here were cleaned from the working code: notes, unused code and machine-specific paths were removed. Each script was then run and its output compared with the output used for the paper.

Script	Check	Result
3b_ClimatePrep.r	Land grid and climate tables rebuilt from the ERA5 files	Land grid and five tables identical. Warmest 3 months identical to within 0.00003 K, which is the rounding of the stored file
3a_SpeciesRanges.r	Ranges rebuilt for a sample of 69 species from all four groups	All 69 identical
4_Metadata.r	Species attributes rebuilt for all species	Group, IUCN category, taxonomy, range size and ecoregions identical for all 32,345 species
5_CalcExposure.r	Limits and exposed cells recomputed for 68 species and 8 variables	Identical for 7 variables. For the warmest 3 months the limits differ by at most 0.00003 K and no exposed-cell count changes
6_MakeDF.r	The three data frames rebuilt from the per-species outputs	All identical, including 75.5 million exposed cell records
8_AppData.r	App data rebuilt	Identical to the data shipped with the app
Tables 1 to 3	Rebuilt	Identical
Figures 1 to 4, S1 to S4, S6	Redrawn	Identical, pixel for pixel, to the existing figure files
Figure S5	Redrawn	Identical to the existing file before one correction. The chronically exposed species are now defined with the 25% of range rule, as in the text, which gives 1,584 species
Figure 5	Redrawn	The underlying tables are identical to the existing tables

Notes on reproducibility

• Run R in a UTF-8 locale, which is the default on most systems. figures/Fig5_Stressor.r uses an arrow character in its legend labels.

• R/4_Metadata.r fills missing taxonomy from the GBIF backbone and the Catalogue of Life through their web services. A rerun can return different values if those services have changed.

• Outputs are written in the qs2 format. Files in the older qs format can be read with the qread() function defined in R/1_Setup.r.

Citation

Merow, C., et al. The 2025 Biodiversity Exposure Report. In revision, BioScience.

License

MIT. See LICENSE.md.