Explosive, caldera forming "super-eruptions" (an eruption of VEI 8 or larger, resulting in 1000+ km³ of volcanic ejecta in ignimbrite sheets) are the single most destructive natural disaster native to Earth. Super-eruptions require three elements to occur: 1-crustal magmatic fluxes above background solidification rates, 2-growth of a batholith scale magma chamber, and 3-an eruption trigger. This study addresses these requirements with new petrographic and geochemical analyses and numerical simulations of crustal magma bodies. Crustal magmatic fluxes up to 10x steady-state arc rates are required to form volcanic provinces that host super-eruptions. Super-eruptions can occur in continental hot-spots or rift environments. Why arcs "flare-up" is the subject of active debate. Arcs may follow a regular cycle of lithospheric thickening, delamination, and asthenospheric upwelling (the Andean cycle); alternatively fertilized lithospheric mantle may undergo rapid melting. Targeted sampling (n = 165) of mapped but unsampled mafic and lamprophyric magmas in the San Juan magmatic locus of Colorado, an archetypical ignimbrite province, over three years identified both the lithospheric mantle reservoir and the most primitive San Juan magmas using optical petrography, whole rock geochemistry (n = 50) and Pb, Sr, and Nd isotope geochemistry (n = 32). These mafic magmas more closely resemble the continental lithosphere geochemically. Mixing models based on Energy Constrained Assimilation/Fractional-Crystallization (EC-AFC) indicate that the San Juan magmatism is the product of lithospheric melts and 30-40% crustal assimilation rather than asthenospheric upwelling. The Farallon flat-slab "pre-fluxed" and refrigerated the Colorado lithospheric mantle; removal of that slab at around 40 Ma triggered the SJVF "flare-up." Numerical simulations of crustal magma chamber growth indicate giant magma chambers form when high magma fluxes raise upper crustal temperatures to 300-400 °C at 5-10 km depth. These simulations focus on chamber growth, convection, and cooling at the expense of geometry or chamber mechanical failure with realistic sill-like geometry at the expense of thermal modeling. New 3D finite difference simulations emphasize the importance of geometry on chamber lifespan and crustal heating. A spherical chamber (i.e. model construct) requires 10x the cooling time of a 2km caldera footprint sill of same volume. Increasing sill thickness by 1km can double chamber longevity. Focused intrusions (i.e. 1D modeling) locally produce higher thermal gradients and preserve larger primary basalt volumes. Random intrusions in 3D yield basalt to crust ratios of 3-4:1 (required in the EC-AFC models). Random intrusion in 3D into the upper crust at "flare-up" fluxes ([greater than or equal to]10 km³ per k.y.) elevate average crustal geotherms by 10 °C / km, allowing for growth of batholithic scale magma chambers a wider footprint. Once situated in the upper crust, sub-caldera magma chambers cool inward forming moving crystallization and fluid saturation fronts. If the saturation front propagates faster than the crystallization front, nucleating fluid bubbles have the opportunity to grow, ascend, and collect at the chamber roof. New 2D finite difference models couple magma chamber cooling to fluid production to explore the conditions of fluid escape and collection. Less silicic magma composition, equant geometry, high ambient thermal gradient, and a stock all aid in fluid pocket growth by slowing the advance of the crystallization front (a fluid trap) and triggering saturation at lower fluid concentrations. Fluid pockets that grow to certain sizes ( > 500 m hemispherical bubble) have the potential to trigger an eruption by propagation of a fluid fracture to the surface. This mechanism possibly triggered the eruption of the 5000+ km³ Fish Canyon Tuff as well as smaller, recent eruptions (Pinatubo, El Chichón). Caldera forming super-eruptions occur in regions that meet these three requirements: 1-high magmatic flux, 2-rapid growth to batholithic size, and 3-a delayed eruption trigger. For the SJVF of Colorado melting of the "pre-fluxed" lithosphere provided the magmatic pulse which melted and heated the crust, forming a broad batholith. As magmatism peaked and began to wane, upper crustal magma chambers started to crystallize, exsolving fluids. These fluids ascended, collected, and fractured their way to the surface, triggering the Fish Canyon Tuff and other eruptions. / text

The mechanisms and timescales associated with the triggering of caldera-forming eruptions remain ambiguous and poorly constrained. Do such eruptions start vigorously, then escalate, or can there be episodicity? Are they triggered through internal processes (e.g. recharge, buoyancy), or can external modulations play an important role? Key to answering these questions is the ability to reconstruct the state of the magma body immediately prior to eruption. My dissertation research seeks to answer these questions through detailed investigation of four voluminous caldera-forming eruptions: (1) 650 km3, 0.767 Ma Bishop Tuff, Long Valley, (2) 530 km3, 25.4 ka Oruanui eruption, Taupo, (3) 2,500 km3, 2.08 Ma Huckleberry Ridge Tuff, Yellowstone and (4) 250 km3, 26.91 Ma Cebolla Creek Tuff, Colorado. The main techniques I applied integrated glass geochemistry (major, trace and volatile), diffusion modeling, and detailed field sampling. In chapters two, three, and four these methods are applied to the initial fall deposits of three supereruptions (Bishop, Oruanui and Huckleberry Ridge) that preserve field-evidence for different opening behaviors. These behaviors range from continuous deposition of fall deposits and ignimbrite (Bishop), to repetitive start/stop behavior, with time breaks between eruptive episodes on the order of weeks to months (Oruanui, Huckleberry Ridge). To reconstruct the timescales of opening activity and relate this to conduit processes, I used two methods that exploit diffusion of volatiles through minerals and melt, providing estimates for the rate at which magmas ascended to the surface. This knowledge is then integrated with the pre-eruptive configuration of the magma body, based on melt inclusion chemistry, to interpret what triggered these systems into unrest. Finally, in chapter five I take a different approach by integrating geochemical data for melt inclusions and phenocryst minerals to test whether the mechanism of heat and volatile recharge often called upon to trigger crystal-rich dacitic magmas (the so-called monotonous intermediates), is applicable to the Cebolla Creek Tuff. This dissertation includes both previously published and unpublished co-authored material, and three online supplementary excel files.