Experiment vs nature: Using amphiboles to test models of magma storage and pre-eruptive magma dynamics preceding the 2006 eruption of Augustine Volcano, Alaska

This study investigates magmatic conditions preceding the 2006 eruption of Augustine Volcano through the use of amphibole compositions and textures. Due to their restricted stability region and common occurrence in calc-alkaline magmas, amphiboles are important for investigating pre-eruptive magmatic conditions at subduction zone volcanoes. Chapter 1 presents a study into geochemical and textural variations of natural amphibole phenocrysts in the erupted magmas. Magnesiohornblendes in the high- and low-silica andesites exhibit limited compositional variability. Intermediate-silica andesites and quenched mafic enclaves contain amphiboles that vary in composition and classification (magnesiohornblende-magnesiohastinsitetschermakite). Compositional variations are controlled by temperature-dependent substitutions. Both high-and low-silica andesites represent magmas that were stored in the shallow crust at 4-8 km depth, remaining distinct due to a complex sub-surface plumbing system. Intermediate-silica andesites and quenched mafic inclusions represent newly formed hybrids of resident high- and low-silica andesite magmas and an intruding basalt. Chapter 2 presents the results of a phase equilibria study the refines the model for high-silica andesite storage. The natural phase assemblage was reproduced between 860-880°C and 120-200 MPa. Experimental plagioclase and groundmass glass compositions most closely replicate natural samples at ∼130-140 MPa. Estimated storage conditions fall within the ranges suggested by natural petrological data and modeled storage depths from geodetic data. The high temperature stability of experimental quartz and biotite (not identified in natural samples) may reflect the high f 02 of the Augustine system as well as the rapid kinetics associated with the crystal-poor sintered starting material of some experiments. Chapter 3 presents results of the first experimental study to target heating-induced amphibole reaction rim formation. Experiments show that reaction rims form on remarkably short timescales. They share mineralogical and textural features with natural reaction rims previously thought to represent decompression processes. Reaction rims cannot be simply classified on the basis of semi-quantitative observations. Rather, in-depth data collection (e.g. X-rap mapping), and the calculation of kinetic parameters (e.g. crystal nucleation rates), is necessary. Chapter 4 presents a new MATLAB® based program that performs mineral formula recalculations and the associated propagation of analytical uncertainty.