sdm-hotspot-spatial-effort

v0.2.0 — refocused to the Iberian-hotspot replication.

This release tests whether Phillips et al. (2009)'s target-group-background correction for sample selection bias can restore biodiversity-hotspot identity, for Iberian breeding birds on a HEALPix-NESTED grid, benchmarked against the EU Article 12 expert-rangemap gold standard.

Result (Contradicted): target-group-background SDM richness hotspots disagree with the Article 12 gold standard at 97% (museum) / 100% (all-observations) symmetric set non-overlap at the ~25 km reference scale — worse than uncorrected raw GBIF richness, and far from the Hurlbert & Jetz reference range. Robust to the MaxEnt feature set. Effort bias in hotspot identity is a spatial sampling-location problem that per-species modelling cannot fix, even though it improves per-species AUC.

Changes since v0.1.0: the Phillips AUC reproduction (formerly an embedded "Arm A") is split into its own repository, sdm-phillips-reproduction (10.5281/zenodo.20473156, Validated), which validates the MaxEnt + target-group-background implementation used here. This repository is now the replication study only.

Reference paper: Phillips et al. (2009), Ecological Applications 19(1):181-197. https://doi.org/10.1890/07-2153.1 Builds on sibling chains: 10.5281/zenodo.20363556 and 10.5281/zenodo.20451519.

Cite via CITATION.cff.