Fertile or Barren: an open benchmark for testing geochemical fertility indices on mineralized and barren mafic-ultramafic intrusions.

Magmatic Ni-Cu-PGE sulfide deposits form through a sequence of processes (e.g., magma generation, transport, crustal interaction, sulfur saturation, and sulfide accumulation). These processes can modify whole-rock compositions in different ways, making it challenging to evaluate the geochemical fertility of mafic-ultramafic magmatic intrusions. Several whole-rock fertility indices have been proposed to identify intrusions capable of hosting magmatic sulfide mineralization, yet their performance is rarely assessed using a common dataset and consistent sampling criteria. We present Fertile or Barren, an open benchmark assembled from GEOROC 2.0 compilations and U.S. Geological Survey deposit databases to evaluate published geochemical fertility indices for magmatic Ni-Cu-PGE systems. We identified 31,886 mafic-ultramafic plutonic samples, including 13,025 samples from 514 named intrusions represented by at least five analyses from 521,961 analyses. Intrusions were classified as fertile, weakly mineralized, or barren, and individual samples were assigned to petrologic categories, including cumulate, chilled margin, fine-grained dike/sill, altered, and unresolved. We developed a public scoring workflow that evaluates each index using intrusion-weighted receiver operating characteristic (ROC) area-under-the-curve (AUC) statistics with bootstrap confidence intervals. Three indices (Ni/Cu, Ni depletion relative to MgO, and Cu/Zr) performed significantly below random expectation (AUC < 0.5), ranking barren intrusions above fertile ones. For depletion-based indices, this result is consistent with the predominance of sulfide-bearing cumulates in the fertile dataset, which comprise 65% of all fertile samples, rather than the sulfide-poor magma compositions these indices were originally designed to evaluate. This work provides a reproducible framework for evaluating existing geochemical fertility indices and identifying the geological and sampling conditions under which their application is most defensible by combining intrusion-level mineralization labels with sample-level petrologic classifications.