Phenocryst or not? Using a stratovolcano’s crystal cargo to explore crustal-scale magmatic systems through time

Phenocrysts in magmas are often assumed to be the product of crystallization directly from a host magma. However, decades of research suggests that crystals carried in magmas (crystal cargo) are far more complex. We test the hypothesis that textural and chemical characteristics of phenocrysts (both native and disaggregated from crystal clots) in andesitic and dacitic lavas and tephras can be used to develop models for spatial arrangement of subcrustal magma systems and their evolution over time. To do this, we use geochemical fingerprinting of complex crystal cargo in a compositionally diverse set of lavas and tephras erupted over the last hundred thousand years at the very high-threat Mount Baker volcanic center in the northern Cascade Arc. These eruptive products, whose composition ranges from basaltic andesite to dacite, contain crystals of plagioclase (pl), clinopyroxene (cpx), orthopyroxene (opx), and sometimes olivine (ol). Multiple populations of each of these crystal types, even within a single thin section, show that they are not typical phenocrysts. Their textures and zoning profiles, especially within crystal clots, lead us to identify as many as five distinct co-crystallizing assemblages of pl-opx-cpx-ol that exist within each individual eruptive product. Careful work parsing out these assemblages has led us to infer that they are liquid-poor remnants of basaltic (B), basaltic andesite (BA), andesite (A), and dacite (D) mushes that exist in the subsurface. These crystal assemblages were first identified in mafic lavas that span ages of 110 ka to 9.8 ka, but we have also recently identified a subset of them in dacitic lava flows and a 6.7 ka dacitic tephra. The interpretation of these data is that multiple crystal mushes beneath Kulshan are tapped by multiple passing magmas that erupt at the surface at different times, producing the same array of crystal clot types/co-crystallizing assemblages in subsequent eruptive products for at least 100 ka. We use thermobarometry to constrain the depths of these mushes and textural/chemical characteristics to infer eruption triggers. We also show that a Holocene monogenetic flank eruption taps a very different mush system and has distinct eruption triggers. Our work shows that meticulous characterization of crystals is invaluable for development of conceptual models of individual volcanic centers to aid in hazard planning.