Long-term volcano evolution controlled by lateral collapse at Antuco volcano, southern Andes, Chile

Volcanic lateral (flank/sector) collapses are among the most destructive volcanic phenomena, capable of generating fast-moving debris avalanches that inundate large areas and trigger secondary hazards such as tsunamis. In addition to immediate impacts, collapses can rapidly decompress shallow magmatic and hydrothermal systems, potentially triggering blasts and altering subsequent eruptive behaviour. Over thousands of years, edifice unloading may reconfigure magma pathways, shift vents, and modify reservoir conditions, yet quantifying these long-term effects is challenging due to sparse eruption chronologies and limited integration of petrology with modelling. Antuco volcano (southern Andes, Chile) experienced a major lateral collapse at 6.9–7.2 ka BP that removed nearly 1 km of cone height and emplaced a ~6.4 km3 debris avalanche. Although Antuco was long considered dominantly basaltic in its post-glacial history, recent evidence of explosive andesites post-collapse suggests significant magmatic changes. This study investigates how unloading by lateral collapse affects magma storage and ascent over millennial timescales by integrating stratigraphically constrained petrology with magma ascent modelling across pre-, early post-, and late post-collapse stages.

Previous work documents post-collapse compositional shifts at several volcanoes, including broader, more evolved magma series after large (>1 km3) sector collapses (e.g., Ritter Island, St. Lucia, Pitons du Carbet, San Pedro, Fogo). In contrast, small collapses (0.2–1.0 km3) are less often followed by such changes, and some systems show no eruptive style change. Mechanical and numerical studies suggest edifice unloading can alter storage pressures, reconfigure pathways, and influence vent locations, but testing these links requires high-resolution chronologies and detailed petrological datasets. At Antuco, prior studies mapped long-term evolution, reported pre- and post-glacial compositions, identified a major mid-Holocene collapse and tephrostratigraphy, and inferred shallow storage zones in earlier stages. However, systematic, stratigraphically controlled sampling across pre- and post-collapse units with integrated textural, geochemical, and thermobarometric constraints, combined with conduit ascent modelling, remained limited. This study addresses that gap.

- Field sampling: Stratigraphically controlled collection during 2015–2021 across pre-collapse (Late Antuco; 17–7.1 ka), early post-collapse (~7.0–3.4 ka), and late post-collapse (<3.4 ka) units. Samples include 15 lavas from nine units and three pyroclastic units, complemented by published data for 13 pyroclastic units, covering 18 volcanic units total. - Bulk-rock geochemistry: Major oxides measured via ICP-OES (ActLabs) and XRF (Bruker S8 Tiger) on fused beads; Fe2O3 converted to FeO/Fe2O3; data normalized to anhydrous; literature data integrated. Loss on ignition <~2.06%. - Glass and mineral chemistry: EPMA on groundmass glass and crystals (plagioclase, clinopyroxene, olivine) using JXA-8530F and JXA-8230 under matched conditions; standards included plagioclase NMNH115900, Kakanui augite USNM 122142, and glass standards (VGA-99, NMNH 113716-1, VG-568). Beam 15 kV, 8–12 nA; defocused beam for glass; Na measured first with reduced times to limit volatilisation. Quantitative elemental maps used to locate glass in microlite-rich matrices. - Textural analysis: SEM-BSE imagery processed with JMicroVision to quantify phenocrysts, microlites, vesicles; crystal textures classified (sieve cores/rims, step/patchy, homogeneous); size populations distinguished (phenocrysts >0.3 mm, microphenocrysts 0.03–0.3 mm, microlites <0.03 mm). - Geothermobarometry and hygrometry: Mineral–liquid equilibrium tested (Kp criteria) for 148 pl–liq, 41 ol–liq, 39 cpx–liq, 20 ol–aug pairs. Applied Pl–liq hygrometer/thermometer (Waters & Lange, 2015), Putirka (2008) and Neave & Putirka (2017) calibrations for cpx–liq, ol–liq, and ol–aug thermometers; estimated uncertainties ~29–42 °C for thermometers and ~1.6–2.8 kbar for barometers. Pressures converted to depths using regional crustal density model. - Viscosity and density estimates: Crystal-free melt viscosity from Giordano et al. (2008) using measured compositions and temperatures; bulk magma viscosity with crystal fraction dependence (Vona et al., 2011); melt density via DensityX. - Unloading estimate: Overburden stress reduction from collapse computed by integrating ρ g over minimum removed thickness (~0.9 km) with ρ = 2000–2700 kg m−3, yielding ~0.18–0.24 kbar decompression. - Conduit ascent modelling: 1-D steady-state cylindrical conduit model (adapted from MAMMA) simulating fluid dynamics, volatile exsolution/expansion, outgassing (vertical), rheology (crystal-bearing suspension), wall friction (laminar/turbulent), and fragmentation criterion (Deborah number ≥0.01). Scenarios: • Pre-collapse: basalt (50.6 wt.% SiO2), 38 vol.% crystals, T=1130 °C, deep reservoir 4 kbar (~15 km), conduit length 15 km. • Early post-collapse: andesite (59.4 wt.% SiO2), 31 vol.% crystals, T=1000 °C, storage 1.5 kbar (~5 km), shortened conduit length 4.1 km (due to collapse). • Rebuilt edifice: same andesite at 1.5 kbar with conduit length restored to 5 km. Conduit radii 5 and 10 m tested (results shown for 10 m); initial H2O 2.5 wt.% (representative long-term shallow hydration). Gas phase single component (H2O).

- Stratigraphy and volumes: Antuco suffered a lateral collapse at 6.9–7.2 ka BP, removing ~0.9 km of summit height and emplacing a ~6.4 km3 debris avalanche. Pre-collapse Late Antuco lavas (17–7.1 ka) were exclusively basaltic. - Geochemical shift post-collapse: Early post-collapse products span basalts to high-silica andesites/trachyandesites, with glasses reaching dacite; lavas range 50.6–59.4 wt.% SiO2; tephras exhibit elevated SiO2, K2O, Na2O and decreased CaO relative to pre- and late post-collapse. Late post-collapse returns mainly to basalts/basaltic andesites. - Textures and mineral chemistry: Early post-collapse plagioclase shows An-depleted cores with An-enriched rims (max An <80), and olivine rims down to Fo~59, indicating cooler, more evolved magmas, in contrast to normal zoning and sieve textures prevalent pre- and late post-collapse. - Pre-eruptive H2O contents: Pre-collapse 2.5–3.2 wt.%, early post-collapse 1.4–5.1 wt.%, late post-collapse 0.11–2.3 wt.% (median/mean ~2.5 wt.% across whole dataset). - Temperatures (approximate): • Pre-collapse: pl–liq ~1098–1115 °C; ol–liq ~1070–1112 °C; cpx–liq ~1029–1070 °C; ol–aug ~1115–1167 °C. • Early post-collapse: plagioclase cores/rims ~1009–1040 °C; unzoned plagioclase/pyroxene ~1000–1020 °C. • Late post-collapse: high T again in plagioclase (rims ~1124–1128 °C; cores/zoned rims ~1128–1183 °C). - Storage pressures/depths: Pre- and late post-collapse crystallisation at ~3.2–6.2 kbar (~12–23 km). Early post-collapse shows shallow reservoirs at <0.8–3.9 kbar (0–3 km) and an intermediate range ~1.4–3.9 kbar (~5.3–14 km), indicating involvement of a shallow silicic reservoir. - Unloading magnitude: Estimated decompression of ~0.18–0.24 kbar from edifice removal, sufficient experimentally to drive H2O exsolution, plagioclase undercooling/crystallisation and viscosity increase. - Rheology: Melt viscosity increased from ~10^1–10^2 Pa s (pre/late post-collapse) to ~10^1–10^3 Pa s in early post-collapse; bulk magma viscosity (with crystals) increased from ~10^4 to ~10^6 Pa s in early post-collapse. - Conduit model results (10 m radius): • Basaltic pre-collapse scenario: ascent from 15 km in ~60 h; viscosity ~10^4–10^6 Pa s; effusive MER ~5.6×10^4 kg s−1; no fragmentation. • Early post-collapse andesite with shortened conduit (4.1 km): ascent in ~5 h; viscosity ~10^5–10^8 Pa s; explosive conditions attainable (Deborah number ≥0.01) with MER ~2×10^5 kg s−1. • Rebuilt edifice (5 km conduit) andesite: slow ascent (~0.05–0.1 m s−1), MER ~4×10^4 kg s−1, no fragmentation. - Interpretation: Lateral collapse unloading reactivated a shallow, cooler andesitic reservoir (<3 km), promoting more evolved, explosive eruptions immediately post-collapse. As the edifice rebuilt and overpressure conditions waned, eruptive behaviour shifted back to dominantly effusive basaltic activity, similar to pre-collapse conditions.

The study addresses how lateral collapse modulates long-term magmatic processes and eruptive behaviour. Petrological data demonstrate that pre-collapse eruptions tapped a hotter, deeper mafic reservoir (12–23 km), producing low-explosivity basaltic activity. The collapse reduced overburden pressure by ~0.18–0.24 kbar, reactivating a cooler, shallow silicic reservoir (<3 km) evidenced by more evolved bulk/glass compositions, lower anorthite/forsterite in crystals, lower temperatures, and shallow storage pressures. Conduit modelling shows that sudden unloading shortens effective conduit length and accelerates ascent, enabling fragmentation and higher mass eruption rates for andesitic magma. This mechanistically explains the observed early post-collapse increase in explosivity and compositional evolution. Over subsequent millennia, rapid edifice regeneration and exhaustion/degassing of the shallow evolved reservoir reduced decompression-driven ascent and explosivity, returning to effusive basaltic activity, matching the late post-collapse record. The findings suggest that many mafic stratovolcanoes with ephemeral shallow silicic reservoirs may experience similar post-collapse transitions, with implications for hazard assessments that must account for elevated explosivity and compositional variability on centennial–millennial timescales following large sector collapses.

Combining stratigraphically controlled petrology with conduit ascent modelling shows that Antuco’s mid-Holocene lateral collapse unloaded the magmatic system, reactivating a shallow, evolved (andesitic) reservoir and promoting a transient period of more explosive, silica-rich activity. As the edifice rebuilt and shallow overpressured magma was exhausted, eruptive conditions reverted to hotter, deeper mafic sources and dominantly effusive behaviour, restoring pre-collapse conditions within a few thousand years. The work demonstrates that sector collapses can exert long-lived controls on magma storage, ascent dynamics, and eruptive style, and that similar responses are plausible at other mafic stratovolcanoes hosting shallow silicic reservoirs. Future work should refine eruption chronologies, integrate multi-volatile (e.g., CO2–H2O) constraints, and employ coupled 2D/3D transient models to capture complex plumbing geometries and short-timescale dynamics to better quantify post-collapse hazard windows.

- Exposure limits and preservation: Older portions of Antuco’s history (<150 ka to 17 ka) are poorly exposed, limiting detailed pre-collapse trends. Some explosive deposits may be eroded or unpreserved. - Volatile assumptions: Modelling considered only H2O (no CO2 data available), potentially underrepresenting deep volatile behaviour and solubility effects. - Model simplifications: 1-D, steady-state conduit model cannot capture complex, time-dependent, multi-vent, or lateral outgassing dynamics; constant crystal fraction assumed; simplifications in gas escape (vertical only) and wall friction. - Thermobarometry uncertainties: Calibrations entail errors (~29–42 °C; ~1.6–2.8 kbar); cpx barometry resolution mainly distinguishes upper crust vs deeper storage; equilibrium tests and use of bulk vs glass compositions may introduce bias. - Parameter uncertainties: Conduit radius, initial conditions, and storage pressures/depths are inferred from petrology and regional models; actual plumbing geometries likely more complex. - Temporal resolution: Exact timing between collapse and subsequent explosive deposits is constrained stratigraphically but not to sub-annual precision.