Caldera-forming volcanic eruptions are low-frequency, high-impact events capable of discharging tens to thousands of cubic kilometres of magma explosively on timescales of hours to days, with ...devastating effects on local and global scales. Because no such eruption has been monitored during its long build-up phase, the precursor phenomena are not well understood. Geophysical signals obtained during recent episodes of unrest at calderas such as Yellowstone, USA, and Campi Flegrei, Italy, are difficult to interpret, and the conditions necessary for large eruptions are poorly constrained. Here we present a study of pre-eruptive magmatic processes and their timescales using chemically zoned crystals from the 'Minoan' caldera-forming eruption of Santorini volcano, Greece, which occurred in the late 1600s BC. The results provide insights into how rapidly large silicic systems may pass from a quiescent state to one on the edge of eruption. Despite the large volume of erupted magma (40-60 cubic kilometres), and the 18,000-year gestation period between the Minoan eruption and the previous major eruption, most crystals in the Minoan magma record processes that occurred less than about 100 years before the eruption. Recharge of the magma reservoir by large volumes of silicic magma (and some mafic magma) occurred during the century before eruption, and mixing between different silicic magma batches was still taking place during the final months. Final assembly of large silicic magma reservoirs may occur on timescales that are geologically very short by comparison with the preceding repose period, with major growth phases immediately before eruption. These observations have implications for the monitoring of long-dormant, but potentially active, caldera systems.
We use the tephrostratigraphic framework along the Aegean Volcanic Arc established in Part 1 of this contribution to determine hemipelagic sedimentation rates, calculate new tephra ages, and ...constrain the minimum magnitudes of (sub)plinian eruptions of the last 200 kyrs. Hemipelagic sedimentation rates range from ∼0.5 cm/kyr up to ∼40 cm/kyr and vary laterally as well as over time. Interpolation between dated tephras yields an eruption age of ∼37 ka for the Firiplaka tephra, showing that explosive volcanism on Milos is ∼24 kyrs younger than previously thought. The four marine Nisyros tephras (N1 to N4) identified in Part 1 (including the Upper (N1) and Lower (N4) Pumice) have ages of ∼57 ka, ∼63 ka, ∼69 ka, and ∼76 ka, respectively. Eruption ages for the Yali‐1 and Yali‐2 tephras are ∼55 ka and ∼34 ka, respectively. The Yali‐2 tephra comprises two geochemically and laterally distinct marine facies. The southern facies is identical to the Yali‐2 fall deposit on land but the western facies has slightly less evolved glass compositions. Overall, erupted plinian and co‐ignimbrite fall tephra volumes range from <1 to 56 km3 (excluding possible caldera fillings and ignimbrite volumes), and 80% of the eruptions had magnitude 5.5 < M ≤ 7.2 (M = log(m)‐7; m = erupted magma mass in kg). Twenty percent of the tephras represent 3.2 < M < 5.5 eruptions. The long‐term average tephra magma mass flux through highly explosive eruptions of Santorini is estimated at ∼40 kg/s. The analogous data for the Kos‐Yali‐Nisyros volcanic complex is less‐well constrained but similar to Santorini.
Plain Language Summary
Sediment cores from the seafloor of the eastern Aegean Sea contain numerous ash layers from (sub)plinian eruptions from the Aegean Volcanic Arc that were correlated in Part 1. These correlations facilitate determination of sedimentation rates of ∼0.5–∼40 cm/kyr within the hemipelagic sediment bracketing the dated tephras. Sedimentation rates show temporal and lateral variations in the context of climate changes, and regional tectonics. Exceptionally high hemipelagic sedimentation rates within the last 4 kyrs, are linked to the 3.6 ka Minoan and the 1650 AD Kolumbo eruptions that emplaced abundant erodible tephra. Using the sedimentation rates we additionally determine the ages of hitherto undated tephras. We deduce an age of ∼37 ka for a Milos eruption, as well as ∼57 ka to ∼76 ka for marine Nisyros and ∼34 and ∼55 ka for Yali tephras, for which previous dating attempts yielded controversial ages. The ash distribution in the marine realm of up to 105 km2 represents a major fraction of the erupted tephra volumes that range from <1 to 56 km3, placing 60% of the investigated eruptions into magnitude category M6, 20% into M7, and 20% into M3 to M5 classes. Over the past ∼200,000 years, Santorini discharged magmas at an average rate of ∼40 kg/s.
Key Points
Tephrochronology for the Aegean Arc eruptions
Sedimentation rate variability in the Aegean Sea
Eruptive volumes and masses for the major Aegean Arc eruptions
The rhyodacitic magma discharged during the 30–80 km
3
DRE (dense rock equivalent) Late Bronze Age (LBA; also called ‘Minoan’) eruption of Santorini caldera is known from previous studies to have had ...a complex history of polybaric ascent and storage prior to eruption. We refine the timescales of these processes by modelling Mg–Fe diffusion profiles in orthopyroxene and clinopyroxene crystals. The data are integrated with previously published information on the LBA eruption (phase equilibria studies, melt inclusion volatile barometry, Mg-in-plagioclase diffusion chronometry), as well as new plagioclase crystal size distributions and the established pre-LBA history of the volcano, to reconstruct the events that led up to the assembly and discharge of the LBA magma chamber. Orthopyroxene, clinopyroxene and plagioclase crystals in the rhyodacite have compositionally distinct rims, overgrowing relict, probably source-derived, more magnesian (or calcic) cores, and record one or more crystallization (plag ≫ opx > cpx) events during the few centuries to years prior to eruption. The crystallization event(s) can be explained by the rapid transfer of rhyodacitic melt from a dioritic/gabbroic region of the subcaldera pluton (mostly in the 8–12 km depth range), followed by injection, cooling and mixing in a large melt lens at 4–6 km depth (the pre-eruptive magma chamber). Since crystals from all eruptive phases yield similar timescales, the melt transfer event(s), the last of which took place less than 2 years before the eruption, must have involved most of the magma that subsequently erupted. The data are consistent with a model in which prolonged generation, storage and segregation of silicic melts were followed by gravitational instability in the subcaldera pluton, causing the rapid interconnection and amalgamation of melt-rich domains. The melts then drained to the top of the pluton, at fluxes of up to 0.1–1 km
3
year
− 1
, where steep vertical gradients of density and rheology probably caused them to inject laterally, forming a short-lived holding chamber prior to eruption. This interpretation is consistent with growing evidence that some large silicic magma chambers are transient features on geological timescales. A similar process preceded at least one earlier caldera-forming eruption on Santorini, suggesting that it may be a general feature of this rift-hosted magmatic system.
The late-seventeenth century BC Minoan eruption of Santorini discharged 30–60 km
3
of magma, and caldera collapse deepened and widened the existing 22 ka caldera. A study of juvenile, cognate, and ...accidental components in the eruption products provides new constraints on vent development during the five eruptive phases, and on the processes that initiated the eruption. The eruption began with subplinian (phase 0) and plinian (phase 1) phases from a vent on a NE–SW fault line that bisects the volcanic field. During phase 1, the magma fragmentation level dropped from the surface to the level of subvolcanic basement and magmatic intrusions. The fragmentation level shallowed again, and the vent migrated northwards (during phase 2) into the flooded 22 ka caldera. The eruption then became strongly phreatomagmatic and discharged low-temperature ignimbrite containing abundant fragments of post-22 ka, pre-Minoan intracaldera lavas (phase 3). Phase 4 discharged hot, fluidized pyroclastic flows from subaerial vents and constructed three main ignimbrite fans (northwestern, eastern, and southern) around the volcano. The first phase-4 flows were discharged from a vent, or vents, in the northern half of the volcanic field, and laid down lithic-block-rich ignimbrite and lag breccias across much of the NW fan. About a tenth of the lithic debris in these flows was subvolcanic basement. New subaerial vents then opened up, probably across much of the volcanic field, and finer-grained ignimbrite was discharged to form the E and S fans. If major caldera collapse took place during the eruption, it probably occurred during phase 4. Three juvenile components were discharged during the eruption—a volumetrically dominant rhyodacitic pumice and two andesitic components: microphenocryst-rich andesitic pumices and quenched andesitic enclaves. The microphenocryst-rich pumices form a textural, mineralogical, chemical, and thermal continuum with co-erupted hornblende diorite nodules, and together they are interpreted as the contents of a small, variably crystallized intrusion that was fragmented and discharged during the eruption, mostly during phases 0 and 1. The microphenocryst-rich pumices, hornblende diorite, andesitic enclaves, and fragments of pre-Minoan intracaldera andesitic lava together form a chemically distinct suite of Ba-rich, Zr-poor andesites that is unique in the products of Santorini since 530 ka. Once the Minoan magma reservoir was primed for eruption by recharge-generated pressurization, the rhyodacite moved upwards by exploiting the plane of weakness offered by the pre-existing andesite–diorite intrusion, dragging some of the crystal-rich contents of the intrusion with it.
The Milos, Christiana‐Santorini‐Kolumbo (CSK) and Kos‐Yali‐Nisyros (KYN) volcanic complexes of the Aegean Volcanic Arc have repeatedly produced highly explosive eruptions from at least ∼360 ka into ...historic times and still show recent unrest. We present the marine tephra record from an array of 50, up to 7.4 m long, sediment cores along the arc collected in 2017 during RV Poseidon cruise POS513, which complements earlier work on distal to ultra‐distal eastern Mediterranean sediment cores. A unique set of glass‐shard trace element (LA‐ICPMS) compositions complements our major element (EMP) data on 220 primary ash layers and 40 terrestrial samples to support geochemical fingerprinting for correlations with 19 known tephras from all three volcanic complexes and with the 39 ka Campanian Ignimbrite from the Campi Flegrei, Italy. The correlations include 11 eruptions from CSK (Kameni, Kolumbo 1650, Minoan, Cape Riva, Cape Tripiti, Upper Scoriae 1 and 2, Middle Pumice, Cape Thera, Lower Pumice, Cape Therma 3). We identify a previously unknown widespread tephra from a plinian eruption on Milos (Firiplaka Tephra). Near the KYN we correlate marine tephras with the Kos Plateau Tuff, the Yali 1 and Yali 2 tephras, and the Upper and Lower Pumice on Nisyros. Between these two major tephras, we found two tephras from Nisyros not yet observed on land. The four Nisyros tephras form a systematic trend toward more evolved magma compositions. In the companion paper we use the tephrostratigraphic framework established here to constrain new eruption ages and magnitudes as a contribution to volcanic hazard assessment.
Plain Language Summary
The Aegean Volcanic Arc comprises the Milos, Christiana‐Santorini‐Kolumbo and Kos‐Yali‐Nisyros volcanic complexes that present particularly high threats for humans and economy due to abundant highly explosive eruptions in the past. The systematic catalog of how eruption products are dispersed on the seafloor (marine tephras) with time provides information on the number and recurrence of eruptions, on their size, and intensities and is thus essential to quantitatively assess future volcanic hazards and risks. During RV Poseidon cruise POS513 in the Eastern Aegean Sea we recovered 50 sediment cores up to 7.4 m long. More than 220 tephra deposits (e.g., volcanic glass shards) from these eruptions were identified. Glass shard compositions from all layers were used for subsequent geochemical fingerprinting to correlate them with 19 known onshore Aegean eruptions as well as with the 39 ka Campanian Ignimbrite eruption from the Campi Flegrei, Italy. Correlations with 11 eruptions from Christiana‐Santorini‐Kolumbo are established. We identify a previously unknown widespread tephra from an eruption on Milos (Firiplaka Tephra). At the eastern region of the arc, we correlate 7 marine tephras with the Kos‐Yali‐Nisyros volcanic complex.
Key Points
Marine tephrostratigraphy for the Aegean Arc
Chemical fingerprinting to correlate on and offshore tephras
The formation of shallow, caldera-sized reservoirs of crystal-poor silicic magma requires the generation of large volumes of silicic melt, followed by the segregation of that melt and its ...accumulation in the upper crust. The 21.8 ± 0.4-ka Cape Riva eruption of Santorini discharged >10 km
3
of crystal-poor dacitic magma, along with <<1 km
3
of hybrid andesite, and collapsed a pre-existing lava shield. We have carried out a field, petrological, chemical, and high-resolution
40
Ar/
39
Ar chronological study of a sequence of lavas discharged prior to the Cape Riva eruption to constrain the crustal residence time of the Cape Riva magma reservoir. The lavas were erupted between 39 and 25 ka, forming a ∼2-km
3
complex of dacitic flows, coulées and domes up to 200 m thick (Therasia dome complex). The Therasia dacites show little chemical variation with time, suggesting derivation from one or more thermally buffered reservoirs. Minor pyroclastic layers occur intercalated within the lava succession, particularly near the top. A prominent pumice fall deposit correlates with the 26-ka Y-4 ash layer found in deep-sea sediments SE of Santorini. One of the last Therasia lavas to be discharged was a hybrid andesite formed by the mixing of dacite and basalt. The Cape Riva eruption occurred no more than 2,800 ± 1,400 years after the final Therasia activity. The Cape Riva dacite is similar in major element composition to the Therasia dacites, but is poorer in K and most incompatible trace elements (e.g. Rb, Zr and LREE). The same chemical differences are observed between the Cape Riva and Therasia hybrid andesites, and between the calculated basaltic mixing end-members of each series. The Therasia and Cape Riva dacites are distinct silicic magma batches and are not related by shallow processes of crystal fractionation or assimilation. The Therasia lavas were therefore not simply precursory leaks from the growing Cape Riva magma reservoir. The change 21.8 ky ago from a magma series richer in incompatible elements to one poorer in those elements is one step in the well documented decrease with time of incompatibles in Santorini magmas over the last 530 ky. The two dacitic magma batches are interpreted to have been emplaced sequentially into the upper crust beneath the summit of the volcano, the first (Therasia) then being partially, or wholly, flushed out by the arrival of the second (Cape Riva). This constrains the upper-crustal residence time of the Cape Riva reservoir to less than 2,800 ± 1,400 years, and the associated time-averaged magma accumulation rate to >0.004 km
3
year
-1
. Rapid ascent and accumulation of the Cape Riva dacite may have been caused by an increased flux of mantle-derived basalt into the crust, explaining the occurrence of hybrid andesites (formed by the mixing of olivine basalt and dacite in approximately equal proportions) in the Cape Riva and late Therasia products. Pressurisation of the upper crustal plumbing system by sustained, high-flux injection of dacite and basalt may have triggered the transition from prolonged, largely effusive activity to explosive eruption and caldera collapse.
The South Aegean Volcanic Arc overlies a slowly subducting, cool slab of oceanic-to-transitional crust, and hosts the hazardous Christiana–Santorini–Kolumbo volcanic field. In order to investigate ...the primitive melts feeding the volcanic field, we present major and trace element analyses of 130 olivine-hosted melt inclusions from Santorini, integrated with previously published H
2
O and CO
2
data. Following post-entrapment corrections, we identify four endmember primitive melt types preserved in Fo ≥ 80 olivines, ranging from low-K island-arc basalts with La/Yb ~ 1.5 and 1.5–3.0 wt% H
2
O to andesites with La/Yb ~ 6–10 and 3.0–3.5 wt% H
2
O. They are consistent with melting at 1.3 to 2.3 GPa and 1350–1440 °C of variably depleted peridotitic mantle fluxed by slab-derived melts and fluids. The chemical signatures of sediment melts dominate, while those of fluids derived from the ocean crust are low compared to global datasets. This is consistent with thick sediment accumulations observed in the Hellenic trench, and with low calculated fluid fluxes from the downgoing slab. The low H
2
O contents estimated for the primary melts (0.8–1.8 wt%) may imply a component of decompression melting beneath the arc. Coupled with a well-constrained chronostratigraphic context, the melt inclusion archive provides a time series of mantle-derived input into the silicic crustal magmatic system over the last 530 ka. Primitive melts with La/Yb ≤ 5 have been erupted encased in olivines over the last 530 ky, without any evident time variation. Melt inclusions with La/Yb > 5 have, on the other hand, been restricted to two periods: (1) prior to the onset of major explosive volcanism at ~ 360 ka, and (2) the products of the 3.6 ka Late-Bronze-Age eruption and the 22-to-3.6 ka inter-Plinian period immediately preceding it. The observations may be explained by time-varying differential extraction of melts from deep storage zones in the mantle or lower crust, related to lithospheric rifting and caldera collapse events. Temporal variations in the supplies of slab-derived melts and fluids may also play a role.
X-ray computed microtomography (μCT) was carried out on four pyroclasts from the 1997 Vulcanian explosions of Soufrière Hills Volcano, Montserrat. Three-dimensional data from multiple image stacks ...with different spatial resolutions (0.37, 4–8, and 17.4 μm px
−1
) were combined to generate size distributions of vesicles, inter-vesicle throats, crystals, and Fe–Ti oxides over a 3.4–860-μm size range, and to compare the results with those obtained by 2D image analysis on the same samples. Qualitative textural observations are in good agreement with those made in 2D, but μCT provides better resolution of textural features and spatial relationships. Calculation of size distributions requires automated decoalescence of the connected vesicle network. Problems related to this process, in part due to the high porosity of pumice, result in potential artefacts in the calculated size distributions, which are discussed in detail. The main modes of the 3D vesicle volume distributions are systematically shifted to larger sizes compared with those of the 2D distributions. Sample total vesicularities obtained in 3D are within 13 vol.% of those found in 2D, and within 10 vol.% of those measured by He-pycnometry. Total number densities of vesicles and Fe–Ti oxides from the two methods are consistent only to the first order, 3D values ranging from 37% to 309% of those in 2D. Vesicle coalescence, investigated by examining inter-vesicle throat size distributions, occurred in all pyroclasts between neighbouring vesicles of many sizes. The larger the vesicle, the more connected it is.
Caldera-forming eruptions of island volcanoes generate tsunamis by the interaction of different eruptive phenomena with the sea. Such tsunamis are a major hazard, but forward models of their impacts ...are limited by poor understanding of source mechanisms. The caldera-forming eruption of Santorini in the Late Bronze Age is known to have been tsunamigenic, and caldera collapse has been proposed as a mechanism. Here, we present bathymetric and seismic evidence showing that the caldera was not open to the sea during the main phase of the eruption, but was flooded once the eruption had finished. Inflow of water and associated landsliding cut a deep, 2.0-2.5 km
, submarine channel, thus filling the caldera in less than a couple of days. If, as at most such volcanoes, caldera collapse occurred syn-eruptively, then it cannot have generated tsunamis. Entry of pyroclastic flows into the sea, combined with slumping of submarine pyroclastic accumulations, were the main mechanisms of tsunami production.