Output from species distribution models (SDMs) with geographic constraints to estimate the spatial distribution of tree species at the global level at a 30-arc second resolution, presented in the publication "Regional uniqueness of tree species composition and response to forest loss and climate change". Data This file contains the results for 10,590 tree species. The results for each species are contained in a directory with the species name connected by an underscore. Each directory contains several .tif files that make up the tiles of the distribution maps for that species and a metadata file. The .tif files can be merged with the gdal_merge.py function to obtain a single .tif file per species, which will contain 9 bands that correspond to the predicted species distribution using climatic variables corresponding to various climate projections from Chelsa 2.1. Band order covariates_1981_2010: average of historical climate measurements from 1981 to 2010 covariates_2011_2040_ssp126: average future climate projection for 2011-2040 under shared socioeconomic pathway (SSP) 1.26 covariates_2011_2040_ssp370: average future climate projection for 2011-2040 under SSP 3.70 covariates_2011_2040_ssp585: average future climate projection for 2011-2040 under SSP 5.85 covariates_2041_2070_ssp126: average future climate projection for 2041-2070 under SSP 1.26 covariates_2041_2070_ssp370: average future climate projection for 2041-2070 under SSP 3.70 covariates_2041_2070_ssp585: average future climate projection for 2041-2070 under SSP 5.85 covariates_2071_2100_ssp126: average future climate projection for 2071-2100 under SSP 1.26 covariates_2071_2100_ssp370: average future climate projection for 2071-2100 under SSP 3.70 covariates_2071_2100_ssp585: average future climate projection for 2071-2100 under SSP 5.85 Metadata The metadata contains more information about the bands, as well as the following species-level properties: nobs: number of spatially distinct occurrence records used in model training precision: precision of binarised model output computed through 3-fold cross-validation threshold: threshold used to binarise probabilistic model output, determined as the threshold maximizing the true skill statistic (TSS) during 3-fold cross-validation f1: F1 score of binarised model output computed through 3-fold cross-validation auc: area under the ROC curve (AUC) of model output computed through 3-fold cross-validation prevalence: prevalence of presences (ie. occurrences records) throughout the training data which consisted of occurrence records and pseudo-absences tss: TSS of binarised model output computed through 3-fold cross-validation recall: recall of binarised model output computed through 3-fold cross-validation nativeness_info: indicates whether reported native countries were available for this species (possible values: "yes" or "no", should be "yes" for all species included) npa: number of pseudo-absences used in model training system:index: species name Merging example For example, the directory Abarema_barbouriana contains files Abarema_barbouriana_0.tif, Abarema_barbouriana_2.tif, ..., Abarema_barbouriana_9.tif and metadata.json. The tiles can be merged with the command "gdal_merge.py -o Abarema_barbouriana_merged.tif Abarema_barbouriana/Abarema_barbouriana_*.tif".