Aim: Recent studies in southern Africa identified past biome stability as an important predictor of biodiversity. We aimed to assess the extent to which past biome stability predicts present global biodiversity patterns, and the extent to which projected climatic changes may lead to eventual biome changes in areas with constant past biome.

Location: Global.

Taxon: Spermatophyta; terrestrial vertebrates.

Methods: Biome constancy was assessed and mapped using results from 89 dynamic global vegetation model simulations, driven by outputs of palaeoclimate experiments spanning the past 140 ka. We tested the hypothesis that terrestrial vertebrate diversity is predicted by biome constancy. We also simulated potential future vegetation, and hence potential future biome patterns, and quantified and mapped the extent of projected eventual future biome change in areas of past constant biome.

Results: Approximately 11% of global ice-free land had a constant biome since 140 ka. Aside from areas of constant Desert, many areas with constant biome support high species diversity. All terrestrial vertebrate groups show a strong positive relationship between biome constancy and vertebrate diversity in areas of greater diversity, but no relationship in less diverse areas. Climatic change projected by 2100 commits 46–66% of global ice-free land, and 34–52% of areas of past constant biome (excluding areas of constant Desert) to eventual biome change.

Main conclusions: Past biome stability strongly predicts vertebrate diversity in areas of higher diversity. Future climatic changes will lead to biome changes in many areas of past constant biome, with profound implications for biodiversity conservation. Some projected biome changes will result in substantial reductions in biospheric carbon sequestration and other ecosystem services.

The CMass and LAI files are derived from outputs produced by LPJ-GUESS simulations.  These simulations were driven using climates derived from a GCM run for pre-industrial conditions (000kDV_...) and four GCM runs driven by the changing greenhouse gas concentrations for 2050 (rcp4.55_... & rcp8.55_...) and 2100 (rcp4.57_... & rcp8.57_...) as defined by the RCP 4.5 and RCP 8.5 projections.  Full details of the GCM and relevant citations are given in Huntley et al. (2021) and Allen et al. (2020).  The LPJ-GUESS output files were opened (by JRMA) using Microsoft Excel, a header line added and, in the case of the CMass files, columns added giving the grid cell areas and ice-free land fractions, these being copied from the relevant columns of the Ice-free_land_fraction_89_time-slices file (see https://doi.org/10.5061/dryad.2fqz612mk for this file).  A second worksheet was then added to the CMass files and values for grid cell CMass calculated from the LPJ-GUESS CMass per unit area values output and the product of the grid cell area and ice-free land fraction values. The second worksheet was then saved in comma-delimited (.csv) format.

The Biome_assignments_V1.1_present_and_RCPs.csv file is derived from the primary output file generated by the FORTRAN program BiomiseLPJ_V1.1.  This program was written by BH and the source code is provided in the Supplementary Information to Allen et al. (2020).  The program output file was opened (by BH) using Microsoft Excel, a header line added, and the resulting worksheet saved in comma-delimited (.csv) format.

The Biome_extents_V1.1_present_and_RCPs and PFT_CMass_by_biome_V1.1_present_and_RCPs are also derived from output files generated by BiomiseLPJ_V1.1.  Minor editing was performed (by BH) using WordPad in order consistently to separate the blocks of data in the files and to ensure that all columns had labels in the header rows, thus rendering the files easier to open using software such as R.

The Biome_constancies_89_time-slices_R1_with_0k_ice-free_areas.csv and Biome_counts_89_time-slices_R1_with_number_of_biomes.csv files were derived from the Biome_assignments_V1.1_89_time-slices.csv file (see https://doi.org/10.5061/dryad.2fqz612mk for this file) associated with the results reported by Allen et al. (2020).  The latter file was processed using a FORTRAN program written by BH that calculates the % constancy of each biome for each grid cell, including in particular the % constancy of the biome inferred for the 'present', and also counts the number of biomes inferred for that grid cell across the 89 time slices.  Output files generated by the program were processed using Excel (by BH) and saved in comma-delimited (.csv) format.

See Readme_first.pdf