Mapping and revealing the tree biodiversity of the Brazilian Cerrado through biome-wide sampling efforts

raw_data_prj_bdg_Giles_2026.csv Filtered tree-inventory dataset derived from the Brazilian National Forest Inventory (NFI/IFN) for the Brazilian Cerrado. The data include only plots located in native savanna and forest formations, as determined by field-based physiognomic classification, and retain woody individuals with DBH ≥ 10 cm sampled between 2018 and 2020. Plots affected by access impediments, non-native vegetation, high taxonomic uncertainty (>15% unidentified stems), or exotic species were excluded following the methodological workflow described in the manuscript. This dataset represents the standardized input used for all diversity calculations and spatial analyses. data_diversity_index_density_cerrado.csv Derived dataset containing plot-level tree diversity and density metrics calculated from the filtered NFI data. Variables include Fisher’s alpha, species richness standardized per hectare (Sha), stem density, and related diversity indicators used in statistical analyses and modeling. These values were computed after correcting for variable sampling area and were directly used in the spatial interpolation models and in the analyses of environmental, climatic, and anthropogenic drivers of diversity. cerrado_map_sha.tif Spatial raster representing the predicted distribution of tree species richness per hectare (Sha) across the Brazilian Cerrado. Predictions were generated using LOESS regression models fitted separately for savanna and forest formations, based solely on geographic coordinates (latitude, longitude, and their interaction), following the modeling framework described in the manuscript. The map is restricted to areas within the original Cerrado vegetation extent and does not extrapolate beyond the sampled environmental space. cerrado_error_map_sha.tif Spatial raster of prediction errors associated with the LOESS-based tree species richness model. The error surface represents the difference between observed and predicted values and was used to assess model uncertainty, spatial performance, and residual structure across the Cerrado. This layer supports the evaluation of prediction reliability and complements the richness map.