Repetitive Duality of Rhyolite Compositions, Timescales, and Storage and Extraction Conditions for Pleistocene Caldera-forming Eruptions, Hokkaido, Japan

Abstract

Abstract During the Early Pleistocene, numerous caldera-forming eruptions occurred in the southernmost Kurile arc (central Hokkaido, Japan), building an extensive pyroclastic plateau with an area >1600 km2. The arc remains active today, and proximity to populations and infrastructure makes understanding these magmatic systems a critical endeavor. We investigate three major caldera-forming ignimbrite eruptions: Biei (c. 2·0 Ma), Tokachi (c. 1·2 Ma), and Tokachi–Mitsumata (c. 1·0 Ma), with an emphasis on constraining the pressures of magma extraction and storage and the timescales of crystallization. Although all pumice glass compositions from the three eruptions are high-silica rhyolites (77–78 wt% SiO2), hierarchical clustering analysis of major and trace element glass data indicates that the Tokachi and Tokachi–Mistumata ignimbrites each have two distinct pumice populations (Type-1F and Type-2F). We find that these two distinct pumice types record pre-eruptive temperatures, extraction pressures, and crystallization timescales that are strikingly similar between the two eruptions. Using the rhyolite-MELTS geobarometer, we estimate that although all magma types from all three eruptions had storage pressures of 50–150 MPa (∼2–6 km), Type-1F magma was extracted from a deeper mush reservoir (200–450 MPa) compared with Type-2F (100–200 MPa). Pre-eruptive temperatures, constrained by plagioclase–liquid equilibration thermometry and rhyolite-MELTS, suggest that Type-1F magma in both eruptions was hotter (800–820 °C) compared with Type-2F (780–800 °C), but that both reached thermal equilibrium upon eruption (760–780 °C). Because zircon is observed only as inclusions and rarely in contact with glass, we conclude that all magmas were zircon-undersaturated, and thus zircon saturation temperatures, which are 60–100 °C lower than those estimated by the other three independent thermometers, underestimate magmatic temperatures. Using these temperatures as minimum estimates, diffusional relaxation times of Ti zonation in quartz, as revealed by cathodoluminescence (CL), give absolute maximum quartz residence times of <1800 years for Type-2F samples and <600 years for Type-1F for all samples; residence times are <300 years for all samples if the more reasonable Fe–Ti oxide temperature is used instead (∼770 °C). Our modelling therefore suggests that the melt-dominated rhyolite magmas that fed these caldera-forming eruptions were ephemeral features that crystallized within the shallow crust for centuries to several millennia. Rapid rim growth occurred in all magma types in all three eruptions, with a majority of quartz rims (10–200 µm) having grown in less than 2 years. Using plagioclase textures and major and trace element data, we conclude that the bright-CL rims of quartz resulted from decompression and subsequent rapid growth, rather than by a recharge-driven heating event. Thus, decompression occurred within 2 years prior to eruption. Remarkably, the two distinct magma types are statistically similar in terms of composition, crystallization timescales, magma storage conditions, and extraction depths, despite being from eruptions that occurred 240 kyr apart, and from calderas that are separated by 35 km. This suggests magma assembly and storage processes that are spatiotemporally repetitive in this region of Hokkaido.