Data and code for selection-function inversion for archive-aware biodiversity inference from global Anura GBIF records

This record contains the reproducible data and code release associated with the study “Selection-function inversion for archive-aware biodiversity inference from federated occurrence records”.

The release supports an archive-aware analysis of global Anura occurrence records from the Global Biodiversity Information Facility. The study treats GBIF records as outputs of multiple archives with distinct observation processes rather than as a single homogeneous survey. Occurrence records were standardised into a 50-km cell-year-species-archive data structure, archive overlap was used to identify connected observation systems, environmental and accessibility covariates were matched to cell-year-species strata, and an environment-informed latent biodiversity surface was estimated together with archive-specific selection functions.

The archive contains four components. The code folder includes numbered Python scripts used for GBIF download, preprocessing, archive-overlap analysis, selection-function inversion, validation, inference-shift analysis, archive ablation, archive-subset ensembles, hidden-priority profiling and spatial clustering tests. The processed-input folder includes derived grid, species, archive and cell-year-species-archive tables. The model-output folder includes final latent-field and archive-selection outputs. The manuscript-result-table folder includes summary tables used to support the main results, supplementary results and figures.

Raw GBIF occurrence downloads are not redistributed in this record. They are publicly available from GBIF and can be retrieved through the GBIF download information associated with the study. This release is intended to support reproducibility, inspection of derived results and reuse of the archive-aware biodiversity inference workflow.