The Katmai Eruption of 1912: a Comparison with the Minoan Eruption of Santorini

There were no observers near the Novarupta eruption site, but the voluminous fall-out and ash-flow deposits are well preserved and well exposed. The vent area itself is also well preserved, almost uniquely so for an eruption so large, because compensatory caldera collapse took place 10 km away, atop Mount Katmai, reflecting a complex, hydraulically interconnected magma reservoir.

     1.     The pre-1912 edifice near Novarupta consisted of a cluster of overlapping andesite-dacite cones and domes (Fig. 1, 2). Among these, the Trident-Katmai subcluster exhibits ~ 12 vents only 1-10 km from Novarupta, represents ~ 50 km³ of pre-1912 eruption products, and covers an area of ~ 160 km², similar to the area of the Santorini islands and caldera. Also within 15 km of Novarupta are the Griggs, Mageik, and Martin stratocones, representing an additional 150 km² area and ~ 75 km³ of andesite-dacite eruption products, mostly lava flows. If the submarine Santorini edifice extends to Kolombos Bank, total areas are similar at ~ 300 km².

     2.     No major pyroclastic eruptions are known to have taken place in the  Katmai-Novarupta area prior to 1912, in marked contrast to the recurrently explosive pre-Minoan history of Santorini (Druitt et al. 1989). Andesite-dacite near-vent agglutinates and small-volume scoria-flow and lithic-pyroclastic-flow deposits are present on the Katmai volcanoes, but no pre-1912 plinian or pumiceous-pyroclastic-flow deposits have been identified.

     3.     No precursory pre-plinian ash layer has been found at Novarupta. Heiken and McCoy (1984) suggest that emplacement of a thin basal (phreatic) ash sequence prompted pre-climactic evacuation of Santorini's population. Areas near the Katmai volcanoes were uninhabited in 1912, but several days of premonitory seismicity prompted evacuation of a fishing encampment 30 km south-south-east (Griggs 1922).

     4.     The 1912 ejecta erupted from a new vent, which broke through Jurassic sedimentary rocks a few km north-west of the Trident-Katmai edifice (Fig. 2). In contrast to Novarupta's eccentricity, the Minoan plinian vent coincides with a long-lived conduit system nearly central to the island-edifice.

     5.     The 1912 pyroclastic eruption began with plinian ejection of ~ 10 km³ of rhyolite fall-out, continued with ~ 11 km³ of pumiceous pyroclastic flows (forming ignimbrite zoned from rhyolite to andesite), and concluded with two dacitic plinian episodes (Fig. 4) that contributed ~ 10 km³ of ejecta (Curtis 1968; Hildreth 1983; 1987). The sequence lasted ~ 60 hours, which included 2 significant breaks, each of 1-10 hours duration. The climactic Minoan eruption began with a few km³ of rhyodacitic plinian fall-out, followed by emplacement of numerous phreatomagmatic surge and fall units and wet pyroclastic flows, and concluded with several km³ of hot rhyodacitic pyroclastic flows. From analysis of eruption dynamics and mass eruption rates, Wilson (1980) estimated 20-40 hours for the plinian phase, ~ 1 hour for the phreatomagmatic phase, and 13-17 hours for the ignimbrite phase, giving a total of 34-58 hours for the entire Minoan eruptive sequence.

     6.     In contrast to Santorini, no significant phreatomagmatic episode took place at Novarupta. Sea-water access to the vent system probably accounts for the prominence of such activity during the Minoan eruption. The Novarupta vent was at an elevation of ~ 900 m, and the pyroclastic flows were emplaced entirely on land.

     7.     Pumice-rich debris flows were common during the eruptive episodes at both centres. At Novarupta, most consisted of plinian fall-out that was re-mobilized from snowclad slopes of surrounding mountains. At Santorini, most may have resulted from proximal accumulation of thick phreatomagmatic deposits re-mobilized down steep outer slopes of the edifice, in part during caldera subsidence.

     8.     The volume of 1912 ejecta is estimated to be 30 ± 5 km³, of which one-third is ignimbrite and about two-thirds plinian (in 3 discrete fall intervals) (Fierstein and Hildreth 1990). The Minoan fall-out volume was estimated to be ~ 28 km³ and may be significantly greater; the total ejecta volume may exceed 50 km³, of which only 5 km³ is on the islands (Sparks et al. 1984). Sparks and Huang (1980) argued that a large fraction of the downwind Minoan ash was coignimbrite elutriate, but at Katmai the chemical and petrographic zonation restricts such a component to < 15% of the regional fall-out, most of which is truly plinian (Fierstein and Hildreth 1990).

     9.     Column-height estimates based on volumetric discharge rates and lithic-clast isopleths are 30-35 km, 25-30 km, and 21-28 km for the three successive plinian events at Novarupta (Fierstein and Hildreth 1990)and 29 km (Wilson 1980) or 36 km (Sparks 1989) for the Minoan plinian event.

     10.     The Minoan ignimbrite is as thick as 60 m on Thera, but most of the pyroclastic-flow material was lost at sea. The 1912 ignimbrite, although deposited entirely on land (Fig. 3), is not yet deeply incised (Fig. 5) where its thick proximal parts are confined in the glacial Valley of Ten Thousand Smokes (VTTS); hence, neither its maximum thickness (100-200 m ?) nor its volume (11 ± 4 km³) is well established.

     11.     The Minoan ignimbrite is wholly nonwelded. Exposed distal and medial parts of the VTTS ignimbrite are also nonwelded, but andesite-bearing proximal parts exhibit incipient to partial welding (Fig. 5), both in surface exposure and at depth (as illustrated by blocks ejected from secondary phreatic craters).

     12.     The Minoan ignimbrite on Thera is reported to consist of > 40 thin flow units, mostly < 3 m thick and many thinner than 50 cm (Bond and Sparks 1976). The VTTS ignimbrite (where exposed to depths of 20-33 m on gorge walls) consists largely of one main flow unit, overlain locally by one or two areally restricted flow units, each only a few metres thick. Distally, however, the thick main unit self-segregates into dozens of secondary pseudo-flow-units (each only 3-50 cm thick), which apparenty developed in response to internal shear as the mass was slowing to a halt.

     13.     The 1912 ignimbrite contains only 1-6 wt % lithic fragments (although flow-veneer lags on near-vent ridgecrests have as much as 16 wt %), whereas the Minoan ignimbrite has 30-60 wt % and some preceding surges and debris flows have 20-30 wt % (Bond and Sparks 1976). The contrast is largely attributable to edifice disruption, phreatomagmatic explosions, and opening new ring vents during collapse of the Santorini caldera. At Novarupta there was little syn-eruptive collapse at the vent, the lesser fraction of lithics having simply been torn from the conduit and the slumping walls of a funnel-like orifice. Initial plinian deposits contain similar fractions of lithics (4-15 wt %) at both volcanoes. At Novarupta, both initial plinian and ignimbrite lithics are almost entirely the Jurassic sedimentary rocks exposed near the vent. At Santorini, progressive changes in lithic populations demonstrate shifting vent patterns during caldera subsidence (Heiken and McCoy 1984).

     14.     Caldera collapse in 1912 took place at the summit of Mount Katmai, 10 km east of the eruption site (Novarupta), providing an extraordinary example of hydraulic compensation in a complex magmatic plumbing system, which is presumed to underlie much of the Katmai-Trident-Novarupta multi-vent edifice (Curtis 1968; Hildreth 1983; 1987). In contrast, Santorini's composite caldera is nearly centred on the island-edifice, presumably directly above the site of the Minoan magma body. Because subsidence around Novarupta was very restricted, coarse and extremely thick fall-out deposits (Fig. 6) are preserved right into the vent area; on the Turtle (Fig. 1, 3) their thickness is thought to exceed 200 m.

     15.     The magma volume erupted in 1912 was in the range 11-16 km³. The Minoan magma volume is, variously estimated at 13-38 km³. In both cases, down wind fall-out volumes are moderately uncertain, as are the volumes of the unexposed ignimbrites, submarine at Santorini, non-incised in the VTTS. Pyle (1990) discusses revised volume estimates for the Minoan eruption.

     16.     Katmai caldera is only 4.2 km² in area at its floor and 7 km² at its rim; Santorini caldera is ~ 60 km², but Heiken and McCoy (1984) and Druitt et al. (1989) agree that only the northern 30-km² segment subsided during the Minoan eruption, the southern half being largely an older feature. Neglecting unknown elevations of collapsed summits, both Katmai and Santorini calderas are 600-700 m deep. Accepting a 30-km³ Minoan subsidence, the collapse volume would be ~ 20 km³, as compared to the estimated 13-38 km³ of Minoan magma ejected. Because Katmai's caldera volume is only ~ 5 km³, the 1912 magma system was apparently undercompensated, even if one adds a possible ~ 1 km³ of subsidence around Novarupta.

     17.     Minoan ejecta are predominantly rhyodacite pumice ( ~ 71% SiO₂ normalized to 100% dry), although streaks, blebs, and lumps of olivine-bearing andesitic pumice (58-59% SiO₂) are common (Bond and Sparks 1976). Novarupta ejecta are about half high-silica rhyolite pumice (77% SiO₂) and half andesite-dacite scoria and pumice zoned from 58 to 66% SiO₂. Crystal-rich pumice, probably cumulate in origin, was ejected in small amounts during both eruptions. The rocks of both systems are typical medium-K arc suites, the Alaskan one slightly more calcic.

     18.     Minoan rhyodacite contains 8-12 wt % phenocrysts of plagioclase, hypersthene, Fe-Ti oxides, and augite. Novarupta rhyolite has only ~ 1 wt % of plagioclase, quartz, hypersthene and Fe-Ti oxides, whereas the 1912 intermediate magmas have > 30 wt % plagioclase, augite, hypersthene, Fe-Ti oxides, and minor olivine. Rare hornblende occurs in Minoan andesite scoriae and in the crystal-rich cumulate pumice; trace apatite occurs in both suites.

     19.     Post-climactic lavas include a rhyolite dome (77% SiO₂) at Novarupta (Fig. 2, 3) and a dacite dome (66% SiO₂; Fenner 1930) on the floor of Katmai caldera (both probably emplaced in 1912), as well as 0.4 km³ of 60-64%-SiO₂ lavas erupted in 1953-63 from a new vent (Fig.2 only 4 km south of Novarupta. Central to the Santorini (composite) caldera, the Kameni islands represent 2-3 km³ of 2-pyroxene dacite lava (~ 64-68% SiO₂) that has accumulated sporadically in at least 11 eruptive episodes, from 197 BC to the present century. Neither system is dead yet, and both probably retain significant reservoirs of silicic magma.

Addendum

Evacuation of Akrotiri is unlikely to have been prompted either by a single large earthquake or by minor phreatic ashfalls. People might, on the other hand, be sufficiently terrorized by sustained volcanic tremor (or by a series of shocks lasting for several days or weeks) to abandon their homes, property and accustomed routines. In 1912, a week of precursory seismicity eventually led to evacuation of Katmai village (30 km south-south-east of Novarupta) two days before the eruption began. Residents of Savonoski (30 km north), however, did not evacuate until the eruption had already started.

The Kameni islands were constructed upon a caldera floor that lies on average > 300 m below sea-level; they now reach a maximum elevation of + 124 m, having accumulated incrementally during a long series of episodic post-Minoan eruptions, predominantly lava flows. The earliest eruption reported historically (in 197 BC) forms a subaerial part of present-day Palaea Kameni, suggesting that much of the submarine Kameni edifice had already been constructed. At Katmai, two lava domes were emplaced immediately after the 1912 explosive sequence, and 40-50 years later several andesitic lava flows vented only 4 km from Novarupta. At Santorini, comparably early post-caldera lavas are likely to have been extruded on the caldera floor. An extended series of unreported intra-caldera eruptions between ~ 1628 BC and 197 BC seems required in order to have built the Kameni edifice up to sea-level by 197 BC. Such activity may well have influenced or delayed early recolonization of Thera.

New geologic data discussed at the Congress make it clear that the Santorini caldera is structurally composite. The large presumptive volumes of ejecta released during the LP-1, LP-2, Cape Riva and Minoan eruptions virtually require (at least) four subsidence events. The 1912 Katmai eruption shows that subsidence area and location need have no simple relationship to vent locations. A plinian outburst from a centrally located vent near present-day Nea Kameni might typically evolve into a ring-vent ignimbrite eruption, with accompanying subsidence taking place in either the northern or southern caldera segments or alternatively throughout the whole caldera. Subsidence is most likely to occur above all or part of a shallow magma reservoir, and geometric details are likely to be influenced by older caldera faults and regional structures. A plinian vent, on the other hand, might initially open anywhere along a ring-fracture system, or along a regional structure like the 'Kameni Line', or even well off the edifice, as at Novarupta.

----------------------------------------------

------------------------------------------------