A data-driven approach to the petrological study of an ignimbrite flare-up in the central San Juan caldera cluster, Colorado

The last four ignimbrites of the central San Juan caldera cluster, Colorado, erupted at ~26.9 Ma in rapid succession (<80 k.y.), and recent studies provide a detailed petrologic understanding of these ignimbrites. Here, we apply hierarchical clustering to an extensive mineral chemistry dataset to test the ability of this approach to recognize useful petrologic aspects of these systems. With this approach, we objectively identify 3-6 unique chemical clusters per mineral (6 minerals total) and assess the chemical complexity of mineral profiles collected from core to rim based on these clusters. The relative proportions of clusters in each ignimbrite for each mineral provide information on the distribution and temporal evolution of temperature and pressure conditions within the magmatic system prior to eruption. This approach quantifies the complex architecture of crustal magmatic systems. Additionally, the fractions of compositional clusters and their distribution from core to rim of different minerals provides a truly unique fingerprint for each ignimbrite. This fingerprinting based on data clustering can be useful in tephrochronology, which currently uses glass major and trace elements for characterization and correlation of tephra layers. Data clustering provides a much more robust set of metrics with which to fingerprint a tephra layer, and this has far reaching implications in mapping, volcanology, tectonic geomorphology, basin development, archaeology, paleoclimatology, and paleoecology, among others.