Submarine cores record magma evolution toward a catastrophic eruption at Kikai Caldera

Catastrophic caldera-forming eruptions evacuate large volumes of magma, generate widespread pyroclastic flows and ash, and can recur on 10^4–10^5 year timescales. Understanding recharge, differentiation, and accumulation processes in crustal reservoirs during the build-up (incubation–maturation–fermentation) stages of a caldera cycle is critical but often hindered by incomplete proximal records as younger deposits bury older ones. Distal tephra cores can capture continuous temporal changes, and submarine settings typically preserve primary stratigraphy better than subaerial environments. Kikai Caldera (southern Japan) experienced major eruptions at ~140 ka, 95 ka (Kikai-Tozurahara, K-Tz; VEI 7), and 7.3 ka (Kikai-Akahoya, K-Ah; VEI 7), with post-K-Ah mafic and rhyolitic activity on and near the caldera rim. While K-Ah deposits include minor mafic clasts suggesting mafic–felsic interaction, K-Tz subaerial deposits are predominantly rhyolitic. Inter-caldera activity is poorly constrained; the altered Komoriko Tephra (9–16 ka) indicates pre-K-Ah volcanism but lacks geochemical data. This study analyzes a submarine sediment core near Kikai to reconstruct volcanism during the K-Tz event and the subsequent inter-caldera phase leading to K-Ah, addressing whether mafic magmas interacted with felsic reservoirs and how magma compositions evolved toward the next catastrophic eruption.

Prior work outlines generalized caldera cycles with phases of incubation, maturation, and fermentation and documents recurrence at numerous calderas. Marine and distal coring studies have proven effective for reconstructing explosive volcanism histories where proximal preservation is limited. At Kikai, K-Ah is the largest Holocene eruption in Japan, dispersing ash >1000 km and depositing thick proximal pyroclastic flows; minor scoria and banded pumice in K-Ah deposits imply mafic–felsic mixing. K-Tz produced widespread rhyolitic tephra, with proximal thick pumice flow units. Post-K-Ah construction includes basaltic cones (e.g., Inamura-dake) and rhyolitic edifices (Iwo-dake, Showa-Iwojima) and a submarine rhyolitic lava dome in the caldera interior. Komoriko Tephra (9–16 ka) records subaerial inter-caldera eruptions but is chemically altered, leaving a gap in geochemical constraints. These studies motivate a continuous, better-preserved submarine record to resolve magma evolution between catastrophic eruptions at Kikai.

Coring and site: Hydraulic Piston Coring System (HPCS) and Short HPCS on D/V Chikyu (Expedition 912 Leg-2) collected cores at Site C9036, Hole B (30°51.202′N, 130°26.998′E), 233 m water depth, 4.3 km NE of Takeshima, outside the submarine Kikai Caldera. Maximum depth was 95 m below seafloor (mbsf). Seismic data indicated gently dipping, outward-younging sedimentary layers. The site is not directly downwind of prevailing westerlies, so some eruptive layers may be absent. Stratigraphy: Three volcaniclastic units were identified: (1) AK unit (19.1–22.4 mbsf), reddish–orange, interbedded poorly sorted pumice lapilli and well-sorted ash, glass-dominant with minor plagioclase, orthopyroxene, clinopyroxene; correlated to 7.3 ka K-Ah tephra by seismic and glass chemistry. (2) BW unit (22.4–33.5 and 39.3–42.5 mbsf, with coring gaps), dark gray, coarse ash to lapilli glass and mineral fragments with foraminifera and shell pieces; mostly massive in upper section with local lamination and grading at 29.6–30.4 mbsf; lower section shows interbedded sorted/unsorted layers; represents inter-caldera volcanism with mixed contributions from multiple eruptions and secondary transport by submarine currents. Apparent sedimentation rate in upper BW: ~0.6 m kyr⁻¹. (3) TZ unit (~48.5–78.0 mbsf, with gaps), white to light gray ash and pumice lapilli lacking bioclasts, containing glass, plagioclase, quartz, pyroxenes, Fe-Ti oxides; correlated to 95 ka K-Tz eruption; internal alternations of sorted/unsorted layers indicate repeated deposition by submarine pyroclastic/turbidity currents. Sample selection and chronology: Twelve samples were analyzed: AK-1 (AK unit); TZ-1/2/3 from lower/central/upper TZ; BW-1 to BW-8 from BW. Radiocarbon dating (14C) on foraminifera and shell fragments in BW-1 to BW-8 provided ages monotonically younging from ~43 ka (BW-1) to ~11 ka (BW-8). No datable bioclasts were present in AK or TZ. Ages were calibrated using OxCal v4.4 with the Marine20 calibration curve. Sample preparation: Core sediments were wet-sieved into >500, 250–500, 120–250, and 62–120 µm fractions. Bioclasts for 14C were handpicked from >500 µm. Remaining grains (volcanic glass and minerals) were mounted in acrylic resin, polished, and ultrasonically cleaned prior to analysis. Geochemical analyses: Major and trace elements of individual glass fragments and minerals (plagioclase, clinopyroxene, orthopyroxene, Fe-Ti oxides, quartz) were measured by laser-ablation ICP-MS at JAMSTEC using a 200 nm femtosecond laser (OK-Fs2000K), ~30 µm spot and ~20 µm depth, coupled to a Thermo Fisher Element XR sector-field ICP-MS with He carrier gas. Standards: BCR-2G for normalization (measured after every five unknowns); BHVO-2G and GSD-1G for QC; repeat analyses assessed reproducibility. Approximately 50 grains per sample and size fraction were analyzed, avoiding vesicles/phenocrysts where possible; some microlite-bearing glass was analyzed as bulk if microlites were small relative to the spot. Alteration screening: To mitigate alteration effects, glass analyses anomalously depleted in K2O relative to SiO2 were excluded using [K2O (wt%)] < 0.1 × [SiO2 (wt%)] − 5.2; samples with [P2O5] > 0.35 wt% or elevated Pb and U relative to Nb ([Pb (ppm)] > 4 × Nb + 8; [U (ppm)] > 6) were removed to avoid secondary mineral precipitation effects. Interpretation emphasizes trends of data populations rather than individual outliers. Mineral chemistry: Plagioclase anorthite content (An#) and trace elements (e.g., Ce, Ba) were used to distinguish populations; pyroxene Mg# and CaO used to infer magma affinity. Subaerial K-Tz rhyolite plagioclase was also analyzed for comparison. Sedimentary interpretation: Upper BW samples (BW-3, BW-5 to BW-8) are treated as composites reflecting eruptions near their 14C ages; BW-4 (laminated) likely represents an individual turbidity event; lower BW (BW-1, BW-2) may include material erupted prior to their depositional ages due to secondary transport.

- Submarine core C9036B records Kikai’s 95 ka (K-Tz) and 7.3 ka (K-Ah) eruptions plus inter-caldera activity between ~43 and ~11 ka at an apparent sedimentation rate of ~0.6 m kyr⁻¹ for the upper BW unit. - Newly identified mafic glass fragments and high-An plagioclase (An-rich population distinct from low-An# felsic plagioclase) occur within the K-Tz (TZ) unit, demonstrating involvement of mafic magma in the predominantly felsic K-Tz caldera-forming eruption—evidence not recognized in subaerial outcrops. - Geochemical contrasts between K-Tz and K-Ah felsic magmas: K-Tz felsic glass is more evolved (higher SiO2, K2O; lower CaO, Al2O3, FeO) and shows strong depletions in Sr, P, Ti, Eu, and MREEs, consistent with extensive fractionation of plagioclase and Fe-Ti oxides and MREE-depleting phases; K-Ah magmas (both mafic and felsic) have elevated TiO2 and P2O5 for a given SiO2. - Inter-caldera BW glasses (BW-2 to BW-8) commonly exhibit elevated TiO2 and P2O5 and incompatible/REE patterns similar to K-Ah, indicating long-lived (>35 kyr) supply of K-Ah-like magmas into the system after K-Tz; BW compositions do not overlap with the TZ (K-Tz) felsic field, suggesting the disappearance of K-Tz-like felsic magma by ≥43 ka. - Temporal compositional evolution within BW: BW-1 is mafic-dominated; BW-2 shows comparable mafic and felsic components; BW-3 to BW-5 are felsic-dominated with SiO2 modes at 73–76 wt%; BW-6 to BW-8 shift toward lower SiO2 with a relative scarcity of highly felsic (73–76 wt% SiO2) glass, implying decreased eruption of the most evolved melt in the final ~9 kyr prior to K-Ah. - Mineral chemistry supports dual magma inputs during K-Tz (mafic high-Mg# clinopyroxene; high-An plagioclase) alongside felsic assemblages; BW pyroxenes are comparatively lower Mg#, consistent with more evolved inter-caldera eruptives. - Sedimentary structures indicate repeated depositional pulses during K-Tz (alternating sorted/unsorted layers) and largely undisturbed accumulation during the inter-caldera interval at the core site, with local mixing by submarine currents.

The submarine core provides a more continuous and less reworked record than proximal subaerial sequences, enabling reconstruction of magma system evolution across a full inter-caldera interval. Mafic glass and high-An plagioclase within the K-Tz unit demonstrate that mafic inputs interacted with the felsic reservoir before or during the 95 ka caldera-forming eruption, potentially acting as a trigger via heating, volatile supply, or replenishment. After K-Tz, volcanism resumed by at least ~43 ka with both mafic and felsic eruptions; over time, felsic contributions increased, consistent with maturation of the caldera cycle. However, during the final ~9 kyr prior to K-Ah, the most evolved (73–76 wt% SiO2) felsic melt appears less frequently in erupted materials, while overall BW glass chemistry remains K-Ah-like. This pattern is interpreted as storage and accumulation of the highly evolved rhyolitic melt in the reservoir rather than its eruption, building the large-volume magma body that was evacuated at 7.3 ka. The absence of TZ-like felsic signatures in BW indicates a fundamental reorganization of the magma system after K-Tz. Overall, the results align with caldera-cycle models involving incubation, maturation, and storage of silicic melt, with local specifics at Kikai governed by long-term inputs of both mafic and felsic magmas from a similar source and their shallow-level fractionation and mixing.

Submarine coring outside Kikai Caldera reveals previously unrecognized mafic involvement in the 95 ka K-Tz caldera-forming eruption and documents inter-caldera magma evolution leading to the 7.3 ka K-Ah event. Inter-caldera deposits show a long-lived (>35 kyr) supply of K-Ah-like magmas and a shift from mafic- to felsic-dominated eruptions, followed by a late-stage reduction in the eruption of highly evolved melts—interpreted as reservoir accumulation of rhyolitic melt culminating in the K-Ah supereruption. This work demonstrates the value of marine cores for capturing continuous proximal volcanic records and resolving preparatory processes toward catastrophic eruptions. Future research should integrate additional cores around the caldera, high-precision geochronology to resolve individual eruptive events within the BW unit, petrological modeling of storage conditions and timescales, and isotopic constraints to quantify mantle versus crustal contributions and mixing histories.

- Potential alteration of glass compositions: despite screening criteria (e.g., K2O–SiO2 filtering; exclusion of high P2O5, Pb, U), minor to moderate alteration and microlite effects may broaden geochemical trends. - Sedimentary mixing and secondary transport: BW unit materials are composites of multiple eruptions and may reflect reworking by submarine currents; lower BW (BW-1, BW-2) may include material erupted prior to their 14C depositional ages, limiting precise event attribution. - Stratigraphic gaps: coring vacancies (e.g., 33.5–38.5, 51.7–62.0, 64.0–74.0 mbsf) and site positioning not directly downwind of prevailing westerlies may result in missing tephra horizons. - Lack of bioclasts in catastrophic units: no direct 14C ages for AK and TZ units; ages rely on literature correlations. - Source and plumbing constraints: the study infers but does not directly resolve magma sources, storage conditions, or absolute volumes; isotopic and thermal-chronometry constraints were limited in this dataset.