Caldera-forming eruptions are some of the largest eruptions on Earth. Although infrequent, they can produce >5,000 km3 of pyroclastic flows and fall deposits and can generate collapse features ~100 km in diameter. Understanding how such large bodies of magma are assembled and stored in the crust is an important parameter in our future ability to forecast eruption but this requires a well-understood foundational framework of the architecture of the magma reservoir. Caldera-forming magma reservoirs are broadly composed of an interconnected network of smaller melt bodies of variable crystallinity that are fed by a large crystal mush zone. Recent studies have shown these large magma reservoirs can persist on timescales of 104–105 years, and temporal and geochemical evidence suggests that long-lived bodies chiefly consist of crystal-rich magmas in a “mush” system, where melt exists within the interstices of a crystalline framework, and any melt-dominated bodies are ephemeral features existing only for a short time prior to eruption. Some well-studied systems record simple mineral growth histories which can lead to a clear understanding of how a magma evolved prior to eruption (e.g. Yellowstone), and others preserve stratigraphic relationships that can reveal simple zoning patterns within magma reservoirs prior to eruption (e.g. Bishop Tuff). The Okataina Volcanic Center, Taupō Volcanic Zone, New Zealand is a highly complicated system that reveals neither simple growth histories nor simple zoning patterns and which requires a more complicated explanation for magma assembly and storage pathways. My work focuses on understanding changes within the magma reservoir across a caldera cycle at the Okataina Volcanic Center. The Okataina Volcanic Center has hosted at least two caldera-forming eruptions in its modern history, the most recent of which is the Rotoiti eruption. The Rotoiti eruption occurred around ~45 ka and evacuated 100 km3 of magma which produced pyroclastic flow and fall deposits. The Rotoiti eruption was immediately followed by the Earthquake Flat eruption, which erupted days to weeks after the Rotoiti eruption producing ~10 km3 of pyroclastic flow and fall deposits. The Earthquake Flat eruption is broadly considered syn-eruptive with the Rotoiti eruption. Pre-caldera dome growth was dominant ~50–150 kyrs prior to the Rotoiti eruption. An apparent lapse in volcanism between the dome growth period (~96 ka) and the Rotoiti eruption (~45 ka) may suggest less magmatic activity during that time, with the caveat that the lack of 45–96 ka eruptive deposits in the rock record does not preclude their existence; caldera-forming pyroclastic flow deposits have been documented to destroy and/or cover preceding older eruptive deposits. At least twenty-three post-caldera eruptions follow the Rotoiti and Earthquake Flat eruptions, and are dominated by smaller volume, crystal-poor to crystal-moderate eruptions. This work examines the petrologic, geochemical, and geochronological signatures from deposits across a caldera-forming cycle (pre-caldera Kakapiko Dome; caldera-forming eruption Rotoiti, syn-caldera eruption Earthquake Flat; and post-caldera eruption Hauparu) to better understand the overall variations in the architecture of the magma system throughout the caldera-forming cycle. Throughout this study, I investigate the intensive parameters that govern petrologic storages conditions, and geochemical, thermal, and geochronological variation across a caldera-forming cycle to understand the broad architecture of the system. In chapter one, I present results of a phase-equilibrium experimental study to constrain the pressure, temperature, and oxygen fugacity of the Earthquake Flat magma body immediately prior to eruption. I present results of H2O-saturated phase equilibrium experiments buffered at NNO on a glass concentrate of the Earthquake Flat rhyolitic tuff. Based on co-saturation of the melt concentrate with quartz and plagioclase, our experiments show that the Earthquake Flat rhyolitic magma was most likely stored at crystal-rich conditions between 75–200 MPa, and 710–760 °C, saturated with H2O-rich magmatic vapor. In particular, we find co-saturation of quartz, plagioclase, and biotite at 140 MPa, and 755 ºC. Prior experimental work on the Rotoiti rhyolite estimated its pre-eruptive temperature and pressure by cummingtonite stability to ≤750 °C and ≤300 MPa if H2O-saturated. Although the two eruptions vented only ~20 km apart, connections between the two magma bodies are unknown. Despite similar shallow, low-temperature storage conditions of the Earthquake Flat and Rotoiti magma bodies, the higher K2O concentration of the former accounts for the presence of biotite versus cummingtonite (and/or orthopyroxene + hornblende) in the latter. The high crystallinity of the Earthquake Flat tuff, combined with abundant poly- and mono-mineralic clusters, suggests that the Earthquake Flat magma body was a shallow, mushy proto-pluton prior to mobilization and eruption. In chapter two, I present trace element compositional data and 238U-230Th ages measured in zircon to constrain the chemical, thermal, and temporal variation across a caldera-forming cycle in the Okataina Volcanic Center. In-situ U-series ages in zircon surfaces and interiors reveal ages from secular equilibrium to within error of eruption for all eruption stages throughout the caldera-forming cycle. Despite this range in ages, there are no systematic patterns in Ti-in-zircon thermometry or common geochemical indices (U/Th, Yb/Gd, (ΣREE + Y)/P, [Hf]) over time. Rather, signatures in zircon reveal that co-crystallization of accessory and major phases complicate simple progressive evolution histories. I also present a zircon-melt model that uses zircon chemistries to predict the initial host melt from which that zircon crystallized. The results show that zircon crystallized from a much wider range of melt compositions than the host glass, which suggests that geochemically diverse melts must have been sustained throughout the zircon crystallization history. In chapter three, I present a new workflow to decrease impurities in bulk mineral separates in order to accurately 238U-230Th bulk mineral ages for major phases. Due to the low concentrations of U and Th in major phases (e.g. plagioclase, biotite, oxides), I measure U and Th concentrations and isotopic compositions by bulk dissolution of hundreds to thousands of grains, which homogenizes any inclusions present within the grains. My results show that even at very low abundances, inclusions of zircon and glass dominate the U and Th budgets of the bulk sample, resulting in an inaccurate (238U)/(230Th) age. While there was some success with the Earthquake Flat and Rotoiti plagioclase separates, the results yield isochrons with large uncertainties that do not constrain the bulk crystallization age. I interpret these results to suggest that the bulk mineral separates are still dominated by mineral inclusions of glass and in some cases, zircon. The complexity of these results compared to other systems is consistent with complex and protracted crystallization and residence of major minerals within a crystal mush. In chapter four, I use trace element diffusion chronometry to constrain the timescales on which plagioclase crystals resided at magmatic temperatures prior to eruption. Using Sr diffusion in plagioclase modeled at a range of temperatures (600–850 ºC), my results show that Sr diffusion time scales are short for the crystal-moderate and crystal poor-eruptions (100–101 years) but only slightly longer for the crystal-rich eruptions (101–102 years). The bulk of the Okataina Volcanic Center plagioclase contain flat Sr profiles, which indicate that either the plagioclase grew rapidly enough such that Sr did not have time to diffuse, or the plagioclase grew in the presence of a compositionally changing magma that supplied continuous Sr contents such that there was no progressive depletion in the Sr composition of the melt. These results provide a point of comparison with previous work modeling plagioclase diffusion in magmas from the Taupō Volcanic Center. Collectively, these chapters help to understand the pressure and temperature regimes at which magmas are stored in the shallow crust, the timescales that minerals are stored prior to eruption, and the different chemical environments in which those minerals are stored. These lines of evidence reveal that there are no systematic variations in the compositional, thermal, and temporal diversity within or between eruptive bodies across a caldera-forming cycle at the Okataina Volcanic Center. These observations are distinct from trends found in other caldera-forming systems and are important characteristics of the architecture of the shallow plumbing system beneath the Okataina Volcanic Center. This work has important implications for caldera cyclicity and hazard mitigation. The results of this study are consistent with others from the Okataina Volcanic Center that support models of caldera cyclicity. This work suggests that the Okataina Volcanic Center magma reservoir is in the early- to mid-maturation phase of caldera growth, which can last hundreds of thousands of years, but which is typically followed by caldera-forming eruption. Furthermore, this work documents how large eruptions do not always occur from crystal-poor magmas within the caldera margins but can also form from crystal-rich magmas that occur at extra-caldera locations. This observation suggests that geophysical monitoring efforts should be expanded to include extra-caldera monitoring.