deep submarine pyroclastic eruptions: theory and predicted

Deep submarine pyroclastic eruptions:theory and predicted landforms and deposits

James W. Head III a;�, Lionel Wilson b

a Department of Geological Sciences, Brown University, Providence, RI 02912, USAb Environmental Science Department, Lancaster University, Lancaster LA1 4YQ, UK

Received 31 October 2001; received in revised form 19 August 2002; accepted 19 August 2002

Abstract

Submarine pyroclastic eruptions at depths greater than a few hundred meters are generally considered to be rareor absent because the pressure of the overlying water column is sufficient to suppress juvenile gas exsolution so thatmagmatic disruption and pyroclastic activity do not occur. Consideration of detailed models of the ascent anderuption of magma in a range of sea floor environments shows, however, that significant pyroclastic activity can occureven at depths in excess of 3000 m. In order to document and illustrate the full range of submarine eruption styles, wemodel several possible scenarios for the ascent and eruption of magma feeding submarine eruptions: (1) no gasexsolution; (2) gas exsolution but no magma disruption; (3) gas exsolution, magma disruption, and hawaiian-stylefountaining; (4) volatile content builds up in the magma reservoir leading to hawaiian eruptions resulting from foamcollapse; (5) magma volatile content insufficient to cause fragmentation normally but low rise speed results instrombolian activity; and (6) volatile content builds up in the top of a dike leading to vulcanian eruptions. We alsoexamine the role of bulk-interaction steam explosivity and contact-surface steam explosivity as processes contributingto volcaniclastic formation in these environments. We concur with most earlier workers that for magma compositionstypical of spreading centers and their vicinities, the most likely circumstance is the quiet effusion of magma withminor gas exsolution, and the production of somewhat vesicular pillow lavas or sheet flows, depending on effusionrate. The amounts by which magma would overshoot the vent in these types of eruptions would be insufficient tocause any magma disruption. The most likely mechanism of production of pyroclastic deposits in this environment isstrombolian activity, due to the localized concentration of volatiles in magma that has a low rise rate; magmatic gascollects by bubble coalescence, and ascends in large isolated bubbles which disrupt the magma surface in the vent,producing localized blocks, bombs, and pyroclastic deposits. Another possible mode of occurrence of pyroclasticdeposits results from vulcanian eruptions; these deposits, being characterized by the dominance of angular blocks ofcountry rocks deposited in the vicinity of a crater, should be easily distinguishable from strombolian and hawaiianeruptions. However, we stress that a special case of the hawaiian eruption style is likely to occur in the submarineenvironment if magmatic gas buildup occurs in a magma reservoir by the upward drift of gas bubbles. In this case, alayer of foam will build up at the top of the reservoir in a sufficient concentration to exceed the volatile contentnecessary for disruption and hawaiian-style activity; the deposits and landforms are predicted to be somewhatdifferent from those of a typical primary magmatic volatile-induced hawaiian eruption. Specifically, typical pyroclastsizes might be smaller; fountain heights may exceed those expected for the purely magmatic hawaiian case; cooling of

0377-0273 / 02 / $ ^ see front matter : 2002 Elsevier Science B.V. All rights reserved.PII: S 0 3 7 7 - 0 2 7 3 ( 0 2 ) 0 0 4 2 5 - 0

* Corresponding author..E-mail addresses: [email protected] (J.W. Head III), [email protected] (L. Wilson).

VOLGEO 2544 30-1-03

Journal of Volcanology and Geothermal Research 121 (2003) 155^193

R

Available online at www.sciencedirect.com

www.elsevier.com/locate/jvolgeores

mailto:[email protected]

mailto:[email protected]

http://www.elsevier.com/locate/jvolgeores

descending pyroclasts would be more efficient, leading to different types of proximal deposits; and runout distancesfor density flows would be greater, potentially leading to submarine pyroclastic deposits surrounding vents out todistances of tens of meters to a kilometer. In addition, flows emerging after the evacuation of the foam layer wouldtend to be very depleted in volatiles, and thus extremely poor in vesicles relative to typical flows associated withhawaiian-style eruptions in the primary magmatic gas case. We examine several cases of reported submarinevolcaniclastic deposits found at depths as great as V3000 m and conclude that submarine hawaiian and strombolianeruptions are much more common than previously suspected at mid-ocean ridges. Furthermore, the latter stages ofdevelopment of volcanic edifices (seamounts) formed in submarine environments are excellent candidates for a widerange of submarine pyroclastic activity due not just to the effects of decreasing water depth, but also to: (1) thepresence of a summit magma reservoir, which favors the buildup of magmatic foams (enhancing hawaiian-styleactivity) and episodic dike emplacement (which favors strombolian-style eruptions); and (2) the common occurrenceof alkalic basalts, the CO2 contents of which favor submarine explosive eruptions at depths greater than tholeiiticbasalts. These models and predictions can be tested with future sampling and analysis programs and we provide achecklist of key observations to help distinguish among the eruption styles.: 2002 Elsevier Science B.V. All rights reserved.

Keywords: submarine; basalt; eruptions; volcaniclastic; pyroclastic; hyaloclastite; hawaiian; strombolian

1. Introduction and background

Submarine volcanic eruptions occur at diver-gent plate boundaries (e.g. Buck et al., 1998;Macdonald, 1998; Per¢t and Chadwick, 1998;Head et al., 1996) and in intraplate areas, com-monly building seamounts (e.g. Keating et al.,1987; Wessell and Lyons, 1997; Schmidt andSchmincke, 2000). In addition to e¡usive £ows,submarine eruptions can produce pyroclastic de-posits (e.g. composed of ‘solid fragments ejectedfrom volcanoes’ ; Cashman et al., 2000, p. 421)and hyaloclastic deposits (e.g. consisting of ‘frag-ments of volcanic glass formed by non-explosiveshattering’ ; Batiza and White, 2000, p. 361). Ritt-mann (1960) coined the term hyaloclastite for ash-sized basaltic particles produced in situ by break-age of pillow rinds during submarine extrusions.As pointed out by Batiza and White (2000), theterm hyaloclastite is sometimes used broadly toinclude both explosively formed fragments aswell as those particles produced in situ by break-age of pillow rinds and this term is commonlyencountered in the literature where the di¡erencecannot be or has not been ascertained. We followthe usage described above and in particular usethe term pyroclastic to refer to solid fragmentsejected from vents where we think that this canbe determined, and hyaloclastite for ash-sized ba-saltic particles produced in situ by breakage of

pillow rinds during submarine extrusions whereit is possible to determine this. If both types ofdeposits might be involved, or in case of uncer-tainty, we use the term volcaniclastic in order tominimize potential confusion.A common assumption about submarine vol-

canic eruptions is that the pressure of the over-lying water column is su⁄cient to suppress juve-nile gas exsolution so that magmatic disruptionand pyroclastic activity does not occur, except atsu⁄ciently shallow depths (e.g. Batiza and White,2000). For example, one of the most distinctivestages recognized in the evolution of seamountsis that related to the summits of edi¢ces wheresu⁄ciently shallow water is reached that mag-matic volatiles can be readily exsolved and accu-mulated, leading to magma disruption on dis-charge and thus to pyroclastic deposits. Thisdepth is generally recognized to be about 200^1000 m and less, depending on magma composi-tion and volatile content (e.g. Kokelaar, 1986;Bonatti and Harrison, 1988; Gill et al., 1990;Oshima et al., 1991; Heikinian et al., 1991; Bi-nard et al., 1992; Wright, 1996, 1999; White,1996; Kano, 1998; Fiske et al., 1998, 2001;Hunns and McPhie, 1999) and is referred to asthe volatile fragmentation depth (Fisher andSchmincke, 1984).Observations of seamount summits at various

depths, however, have shown the presence of hy-

VOLGEO 2544 30-1-03

J.W. Head III, L. Wilson / Journal of Volcanology and Geothermal Research 121 (2003) 155^193156

aloclastic and pyroclastic deposits ranging fromscattered fragments deposited between pillowlavas to evidence for extensive deposits associatedwith eruptive vents. Lonsdale and Batiza (1980)reported the presence of extensive £ows of volca-niclastics on the summits of four seamounts 800^1200 m above the £anks of the EPR interpreted tohave formed in deep-water phreatomagmaticeruptions (see also Batiza et al., 1984). Smithand Batiza (1989) documented extensive volcani-clastic deposits at depths from 1240^2500 m onsix additional seamounts near the EPR and lo-cated several vent areas (see also Maicher et al.,2000; Maicher and White, 2001). These depositsoccur at a depth where pressure would seem topreclude the normal exsolution of magmatic vol-atiles to such a degree that they disrupt (e.g. Fish-er and Schmincke, 1984), unless the magma vola-tile contents are unusually high (Devine andSigurdsson, 1995; Dixon et al., 1997). There area number of other possible mechanisms for pro-ducing pyroclastics and hyaloclastites, however(Kokelaar, 1986).More recent exploration of the sea £oor has

revealed additional evidence for pyroclastic andhyaloclastic deposits at a range of depths and inseveral magmatic^tectonic environments. For ex-ample, Clague et al. (2002a) documented frag-mental volcaniclastic deposits associated with vol-atile-rich alkalic magmas in the North ArchVolcanic Field, HawaiPi, at depths of 4.3 km.Lava bubble-wall fragments (‘limu o Pele’) inter-preted to be formed by submarine hydrovolcanicexplosions have been found on the summit of Lo-Pihi Seamount at depths of 1.2 km (Clague et al.,2000). Similar ‘limu o Pele’ deposits have beenfound along the Gorda Ridge axis at 3.2 kmdepth, and interpreted to be evidence for strom-bolian activity (Clague et al., 2002c). Fouquet etal. (1998) found widespread volcaniclastic depos-its at the Mid-Atlantic Ridge (MAR) axis atdepths up to 1.7 km. Extensive recent explorationof the summit of LoPihi Seamount (Clague et al.,2002b) has shown evidence of phreatic, phreato-magmatic, hawaiian, and strombolian eruptionproducts and activity at 1.2 km, including an11-m thick section of layered volcaniclastic depos-its. The LoPihi summit deposits include ¢ne-

grained ash beds, spatter, bombs, ‘limu o Pele’,Pele’s hair, and cored bombs (Clague et al.,2002b).Thus, several major questions arise concerning

the formation of pyroclastic and hyaloclastic de-posits observed on the ocean £oor and on sea-mounts (Fig. 1). What processes do these depositsrepresent in terms of mode of eruption, mecha-nisms of fragmentation, style of emplacement ofdeposits, and formation of landforms? What isthe signi¢cance of these processes to the laterstages of evolution in the overall developmentand shallowing of seamounts? The purpose ofthis paper is to address several aspects of thesequestions. We present a treatment of the theoryof ascent and eruption of magma at a range ofdepths in the submarine environment, concentrat-ing on processes which can lead to explosive ac-tivity. We make inferences about the most likelymechanisms and their sequence and relative im-portance. Finally, we compare these to severalknown occurrences and make predictions aboutthe style of emplacement of deposits, and process-es of formation of landforms, that can be testedby further observations.

2. Theory of the ascent and eruption of magma inthe submarine environment

In the subaerial environment, magma ascendingto shallow depths at speeds up to several metersper second typically undergoes gas exsolution inthe upper several hundreds of meters of the dike,causing disruption and acceleration of magmathrough the conduit to produce lava fountaining(known as hawaiian-style activity) and a range ofpyroclastic deposits is produced (Head and Wil-son, 1987, 1989). In some cases, magma ascent isstalled or magma rise speeds relative to dike wallsare less than bubble rise speeds through the mag-matic liquid; in this case larger bubbles can over-take and coalesce with smaller ones to createsmall numbers of very large bubbles which burstat the surface of the magma column, causingstrombolian-style eruptive behavior and deposits.In the deep submarine environment typical of theMAR or East Paci¢c Rise, pressures are high

VOLGEO 2544 30-1-03

J.W. Head III, L. Wilson / Journal of Volcanology and Geothermal Research 121 (2003) 155^193 157

enough to greatly reduce or completely inhibit gasexsolution so that continuous magma disruptionand hawaiian-style pyroclastic deposits are nor-mally precluded (Head et al., 1996).However, there are circumstances which can

lead to magma disruption and explosive activityat great water depths. In the following sections weexplore these (Fig. 2), examining speci¢cally: (1)eruption conditions when no magmatic gas is ex-solved; (2) eruption conditions when magmascontain some gas bubbles but in a concentrationtoo low to cause magma disruption; (3) the (high)levels of magma volatile content needed to ensuremagma disruption and continuous explosive (ha-waiian-style) activity even at high water pressures;(4) the (even higher) amounts of magmatic gasneeded to allow gas exsolution and accumulationinto a foam layer at the top of a magma reservoirwhich is then also released in a steady explosivedischarge; and (5) the expected interactions be-

tween the overlying water and the explosivelyejected gas and pyroclasts in hawaiian-style erup-tions. Next, we (6) examine the circumstances inwhich slow magma rise speed can allow smallnumbers of gas bubbles to exsolve and then coa-lesce into a few large bubbles which cause inter-mittent explosive (strombolian) disruption of themagma surface when they reach the sea-£oorvent, and ¢nally (7) comment brie£y on the con-ditions in which gas accumulation at the top ofshallow dikes approaching close to the sea-£oorcould in principle lead to localised (vulcanian)explosions.

2.1. No gas exsolution

In this case (Fig. 2), there is no gas exsolutionat all by the time the magma has risen to the levelof the vent and decompressed to the ambientpressure level. This is unlikely unless magma

Fig. 1. General geologic setting, conditions, and environment of emplacement of magma and lavas in submarine seamounts. Wetake the speci¢c size, shape and depth for the edi¢ce Seamount 6, Eastern Paci¢c Ocean (data from Smith and Batiza, 1989), butthe example is similar to conditions and observations on several seamounts (e.g. Smith and Batiza, 1989). Also shown are thedepth ranges of various clast-forming processes on the sea £oor (after Kokelaar, 1986).

VOLGEO 2544 30-1-03


Fig. 2. Con¢guration of the ascent and eruption of magma in several possible submarine examples: (1) no gas exsolution; (2) gasexsolution but no magmatic disruption; (3) gas exsolution, magma disruption, and hawaiian-style fountaining; (4) volatile contentbuilds up in the magma reservoir leading to foam formation; (5) magma volatile content insu⁄cient to cause fragmentation nor-mally but low rise speed results in strombolian activity; (6) volatile content builds up in the top of a dike leading to vulcanianeruptions.

VOLGEO 2544 30-1-03


CO2 contents are extremely low, but if it is thecase the consequences are that there will be vir-tually no fountaining. This case was treated forthe subaerial environment by Head and Wilson(1987) for the PuPu POPo vent on Kilauea volca-no’s East Rift zone. Using the observed magmavolume £ux to estimate the gas-free magma risespeed driven by magma buoyancy (0.4 m/s), theycalculated that the gasless fountain height wouldbe less than 1 cm and that, in order to approachaverage observed fountain heights of 200 m bydischarge alone, the volume £ux required wouldbe 20 000 m3/s, more than 100 times the maximum£ux observed. Are there additional non-volatile ornon-buoyancy factors that could accelerate mag-ma rise to produce fountain heights of 10^20 m,causing ‘autodisruption’ of the lava as suggestedby Smith and Batiza (1989) to account for volca-niclastics on deep seamount summits? Analysis ofa wide range of dike widths and reservoir condi-tions under sea £oor environmental situations(Head et al., 1996) shows that even under themost extreme values of excess reservoir pressure,rise speeds will increase by less than a factor of 5,and fountains would not exceed a few tens ofcentimeters in height. Such a ‘dynamic lavamound’ fountaining process would not of itselflead to magma disruption, pyroclast formation,and dispersal.

2.2. Gas exsolution but no disruption

In this case (Fig. 2), some gas exsolves but thevolume fraction of gas in bubbles stays smallenough that magma disruption does not occuras long as the bubbles are uniformly distributedin the magma; a vesicular magma is erupted e¡u-sively from the vent. The critical gas volume frac-tion which must be exceeded to ensure magmadisruption is commonly assumed to be V75%(Sparks, 1978), but may range from 60 to 90%(Vergniolle and Jaupart, 1990) depending on themagma rheology and applied strain rate. Sam-pling of £ows and vent-related lavas will providequantitative information on the volatile contentand amount of vesicularity. The eruption speedin this case is greater than that when no gas ex-solves because there will be some expansion of the

gas between bubble nucleation and eruption, butis still likely to be less than 1 m/s (Head et al.,1996). Is contact-surface steam explosivity likelyto contribute to hyaloclastite formation in eitherof these ¢rst two cases? On the basis of experi-ments (Wohletz and McQueen, 1984) impulse val-ues of about 0.7 m/s are required to initiate fuel^coolant interactions leading to catastrophic dis-ruption and formation of abundant hyaloclastites.This value is in excess of the highest e¡usion ve-locities calculated for sheet £ows on the MAR(0.2^0.5 m/s) (Head et al., 1996), and so we donot expect magma disruption to occur.

2.3. Gas exsolution, disruption, hawaiian-stylefountaining

In this case (Fig. 2), enough gas exsolves thatthe volume fraction of gas in bubbles exceeds thecritical value cited in 2.2. Gas exsolution but nodisruption, at some level below the vent (or in theextreme case just at the vent) and magma is dis-rupted. A hawaiian-style lava fountain eruptionthen takes place, though its appearance and prod-ucts are expected to be di¡erent from those of asubaerial lava fountain eruption (Head and Wil-son, 1987, 1989) because of the immediate inter-action with the seawater (see 2.7. Volatile contentbuilds up in the top of dike leading to vulcanianeruptions).Magmas with volatile contents typical of those

commonly seen in subaerial eruptions would noterupt explosively at depths greater than 200^1000 m (e.g. Kokelaar, 1986). Therefore, the anal-ysis which follows is designed to establish whatmagma volatile contents would be needed in asteadily rising magma to ensure that a continuousstream of fragmented magma would be ejectedfrom the vent in an underwater hawaiian-styleor plinian-style jet. The requirement is thatthe exsolved magmatic volatile volume fractionreaches a critical value (which we assume to beV75% for consistency with earlier work) by thetime the magma has decompressed to the pressurelevel of the vent. Let this pressure be Pv, calcu-lated as a function of depth below sea level usingthe acceleration due to gravity, g, = 9.8 m s32 andassuming a cold seawater density of 1026 kg m33.

VOLGEO 2544 30-1-03


The partial volume fractions of the gas and liquidphases of the magma are given by the ratios oftheir masses and densities. Assume that in general¢nite amounts of both H2O (steam) and CO2 haveexsolved and let the volume fractions of thesegases be Vs and Vc, respectively; the magmaticliquid volume fraction is Vm. If the amounts ofthe gases which have exsolved are ns and nc, ineach case expressed as mass fractions of the totalmagma mass, we have:

V s ¼ ðnsQTÞ=ðmsPvÞ ð1Þ

V c ¼ ðncQTÞ=ðmcPvÞ ð2Þ

Vm ¼ ð13ns3ncÞ=bm ð3Þ

where Q is the universal gas constant (8314 kgkmol31), T is the magma temperature, and msand mc are the molecular weights of the steamand CO2, 18.02 and 44.00 kg kmol31, respec-tively. We have assumed (an adequate approxima-tion for the present purpose) that the gases obeythe perfect gas law and are in thermal equilibriumwith the magmatic liquid. We take as plausiblevalues for ma¢c magmas T=1255‡C=1528 Kand bm = 2700 kg/m3. If the gas is to occupy 75%of the total volume we require that (Vs+Vc) =3 Vm, and so:

ðns=msÞ þ ðnc=mcÞ ¼ ½ð13ns3ncÞPv�=ðQTbmÞ ð4Þ

Also, the de¢nition of the bulk density L of themagma is (1/L) = (Vs+Vc+Vm), and so in the gen-eral case we have:

ð1=L Þ ¼ ½ðQTÞ=Pv�½ðns=msÞ þ ðnc=mcÞ�þ

½ð13ns3ncÞ=bm� ð5Þ

In the special case of (Vs+Vc) = 3 Vm this be-comes

ð1=L Þ ¼ ½4ð13ns3ncÞ�=bm ð6Þ

Thus, for any chosen ocean depth, and hencepressure, we can ¢nd either the minimum value ofns needed to ensure explosive disruption of themagma given a choice of the value of nc, or theminimum value of nc required given a choice ofthe value of ns, together with the correspondingbulk density of the erupting fragmentated magma.

To illustrate the extreme ranges of conditionswe show in the ¢rst part of Table 1 the values ofns when nc = 0 and in the second part the values ofnc when ns = 0. Also shown are the correspondingbulk magma densities Ls and Lc given by Eq. 6and in addition the total amounts nst and nct ofthe volatile phases which would have to bepresent in the magma prior to any gas exsolution,allowing for the solubilitites nsd and ncd of thesephases in basaltic melts given by:

nsd ¼ 6:8U1038 P0:7v ð7Þ

(Dixon, 1997) and:

ncd ¼ 5:9U10312 Pv þ 5:0U1036 ð8Þ

(Harris, 1981; Dixon, 1997) where in each casethe solubility is expressed as a weight fractionand the pressure Pv is expressed in Pa. These val-ues of nst and nct represent minimum requirementsbecause they assume that no supersaturation ofvolatiles is needed to initiate gas bubble nucle-ation. However, signi¢cant supersaturations maybe required at the pressures in submarine magmareservoirs (Bottinga and Javoy, 1990).Vents at the summits of seamounts and mid-

ocean ridges occur at water depths from severalhundred meters to V3500 m. Table 1 shows thatfrom V0.8 to V5.3 wt% water or alternativelyV2^V12 wt% CO2 would need to be exsolvedto ensure continuous explosive activity at thesedepths (of course, if both volatiles are present,as is the case in practice, some critical combina-tion of the two is required). On the basis of com-parison of these values with typically ma¢c mag-ma volatile contents (e.g. Wallace and Anderson,2000), and with the most extreme volatile contentsfound in submarine basalts (V1.4 wt% CO2 and0.54 wt% H2O in a ‘popping rock’ from theMAR, Javoy and Pineau, 1991; Pineau and Ja-voy, 1994; 0.8^1.0 wt% CO2 for similar MAR‘popping rock’ samples, Gerlach, 1991; 0.68wt% H2O in a basalt possibly contaminated byseawater during a caldera collapse event at LoPihi,Kent et al., 1999) it is clear that the volatile con-tents required to produce hawaiian-style eruptionsare very unlikely to be common (see also Mac-donald, 1967; Fornari et al., 1988; Dixon, 1997;Dixon and Clague, 2001; Dixon and Stolper,

VOLGEO 2544 30-1-03


1995; Dixon et al., 1991, 1995). However, Dixonet al. (1997) documented alkali basaltic/nepheli-nitic lavas erupted north of HawaiPi which con-tained 1.9 wt% H2O and 5.4% CO2 ; Table 1shows that if these magmas had been erupted atdepths shallower than about 1250 m, they wouldcertainly have produced the submarine equiva-lents of hawaiian-style lava fountains even ifonly one or the other of the volatile componentswere present. Explosive activity at even greaterdepths is possible when both volatiles are takeninto account.We can demonstrate this, and also illustrate the

likely eruption speeds in submarine hawaiianeruptions, by examining the eruption of the veryvolatile-rich magma documented by Dixon et al.(1997) at a range of water depths. First we estab-lish the pressure at which such a magma wouldhave a total exsolved gas volume fraction equal to75%, thus causing magma fragmentation. This in-volves making an initial estimate of this pressure

and using Eqs. 7 and 8 to ¢nd the amounts ofCO2 and H2O dissolved in the magma; subtract-ing these values from the total volatile contents,5.4 wt% CO2 and 1.9 wt% H2O, gives the ex-solved amounts of gas, nc and ns, respectively.These values are used in Eqs. 1^3, with the cur-rent pressure estimate being substituted for Pv, to¢nd the partial volumes of CO2, H2O and mag-matic liquid; the total gas volume fraction is thenevaluated. If this is more or less than 75% thepressure is increased or decreased, respectively,and a new gas volume fraction is calculated.The process is repeated until enough values closeto 75% are available to make accurate interpola-tion of the pressure corresponding to exactly 75%possible. This is found to be Pf = 21.557 MPa, atwhich pressure nc = 5.3867 wt% CO2 (i.e. virtuallyall of the 5.4 wt% available) and ns = 0.9752 wt%H2O (about half of the 1.9 wt% available) haveexsolved. Eq. 6 shows that the bulk density Lm ofthe erupting gas^pyroclast mixture is 721 kg m33.

Table 1Values for a series of depths below the ocean surface(a) Gas phase is pure water

Depth P ns nst bs Ls

(m) (MPa) (wt%) (wt%) (kg m33) (kg m33)

500 5.13 0.802 1.139 7.27 680.31000 10.16 1.575 2.121 14.40 685.71500 15.18 2.337 3.061 21.54 691.22000 20.21 3.087 3.971 28.66 696.42500 25.24 3.825 4.858 35.80 701.93000 30.26 4.553 5.726 42.93 707.23500 35.29 5.359 6.665 50.06 713.2

(b) Gas phase is pure carbon dioxide

Depth P nc nct bc Lc

(m) (MPa) (wt%) (wt%) (kg m33) (kg m33)

500 5.13 1.935 1.939 17.76 688.41000 10.16 3.761 3.767 35.17 701.41500 15.18 5.520 5.529 52.58 714.42000 20.21 7.216 7.228 70.00 727.52500 25.24 8.852 8.867 87.41 740.63000 30.26 10.432 10.450 104.82 753.63500 35.29 12.147 12.168 122.23 758.1

Values include the total ambient pressure, P, the minimum total mass fraction, ns or nc, of the volatile (pure water or pure car-bon dioxide, respectively) that must be exsolved from a magma to allow explosive disruption of the magma to occur; the totalamount of gas dissolved in the magma, nst or nct, respectively, to permit this amount to exsolve; the density of the exsolved gasphase, bs or bc, respectively; and the bulk density, Ls or Lc, respectively, of the gas-pyroclast mixture that emerges through thevent.

VOLGEO 2544 30-1-03


The pressure of 21.56 MPa corresponds to anocean depth of 2134 m, and so any putative ex-plosive eruption must take place at a water depthshallower than this. We therefore select a series ofdepths dv below the ocean surface at which a pro-posed vent is located, calculate the ambient pres-sure Pv at the vent for each depth, and then ¢ndthe amount of energy, E, per unit mass of magmawhich is available to drive the eruption due tomagmatic gas expansion between the magma frag-mentation pressure Pf and the vent pressure Pv.This energy is given by (Wilson, 1980):

E ¼ QT ½ðnc=mcÞ þ ðns=msÞ�lnðPf=PvÞþ

½ð13nc3nsÞðPf3PvÞ=bm� ð9Þ

The available energy E is shared between thepotential energy needed to raise magma fromthe fragmentation level to the vent, the workdone against friction with the walls of the dikethrough which the magma rises, and the kineticenergy of the accelerating mixture of gas and py-roclasts. Wilson and Head (1981) show that thework done against friction after the magma frag-ments is very small, and so we only need to eval-uate the work done against gravity. If we assumethat the pressure gradient in the erupting magmais close to the lithostatic gradient in the surround-ing host rocks (arguments supporting this as-sumption are given by Wilson and Head, 1981),and take the host rock density to be bh = 2700 kgm33 (to allow for a small amount of vesicularityin the ocean £oor crust), then the vertical distancebetween the magma fragmentation depth and thesea-£oor vent is df where (Pf3Pv) = (bh g df ) andthe potential energy change per unit mass of mag-ma is (g df ) = [(Pf3Pv)/bh]. If the average speed ofgas and pyroclasts erupted through the vent is uvthe kinetic energy per unit mass is (0.5 u2v) and wehave:

0:5u2v ¼ QT ½ðnc=mcÞ þ ðns=msÞ�lnðPf=PvÞþ

ðPf3PvÞf½ð13nc3nsÞ=bm�3ð1=b hÞg ð10Þ

Table 2 shows the values of df , Lm and uv forvalues of the depth to the vent, dv, between zero(a subaerial eruption at sea-level) and the maxi-mum possible water depth to allow an explosive

eruption for this magma, 2134 m. Note the dra-matic increase in eruption speed as water depthbecomes less than about 1000 m; in very shallowwater or on land the eruption of a ma¢c magmawith the volatile content used in this illustrationwould lead to a very violent, possibly plinian-stylerather than hawaiian-style (Par¢tt and Wilson,1999), explosive eruption. Also note that all ofthe bulk magma densities Lm are less than thedensity of the seawater into which the magmaerupts, so that the jet of volcanic ejecta is initiallypositively buoyant in the seawater.As a second illustration of possible submarine

hawaiian-style activity we have repeated theabove analysis for the second most volatile richsubmarine basalt mentioned earlier, that de-scribed by Javoy and Pineau (1991) and Pineauand Javoy (1994) with volatile contents of V1.4wt% CO2 and 0.54 wt% H2O. Table 3 shows theresults ; the maximum depth for explosive activityis now very much less at V489 m and the max-imum eruption speed for eruptions in shallowwater is V250 m/s. Although less impressivethan the previous example, this magma too wouldproduce an extremely vigorous (V3000-m high)lava fountain if it erupted subaerially.

Table 2Submarine hawaiian-type lava fountain eruption conditionsfor a very volatile-rich magma containing 5.4 wt% CO2 and1.9 wt% H2O

dv df Lm uv/(m) (m) (kg m33) (m/s)

0 811 3.5 556250 716 92 340500 621 181 275750 526 269 2301000 431 355 1931250 338 440 1611500 241 522 1281750 147 601 942000 51 680 542100 13 711 272134 0 720 0

For various depths, dv, of the vent below the ocean surface,values are given for: the depth below the vent at which mag-ma is fragmentated, df ; the bulk density of the erupting gas^pyroclast mixture, Lm ; and the mean eruption speed of gasand pyroclasts, uv. The entry for dv = 0 corresponds to a sub-aerial eruption of this magma.

VOLGEO 2544 30-1-03


The above analysis has been repeated for awide range of permutations of initial magmaticCO2 and H2O contents. Fig. 3 shows the maxi-mum depth in the ocean at which a steady hawai-ian-style explosive eruption can take place as afunction of the H2O content for a series of ¢xedCO2 contents. For each CO2 content there is a

critical H2O content below which there is no fur-ther dependence on H2O; this is because in thesecases magma fragmentation takes place solely dueto the presence of CO2 bubbles before any H2Oexsolution has taken place.We discuss the subsequent interactions between

the products of hawaiian-style explosive eruptionsand the overlying water in 2.5. Submarine erup-tion plumes, after dealing in 2.4. Volatile contentbuilds up in the magma reservoir, with the otherpossible source of submarine hawaiian-style activ-ity, gas accumulation in a magma reservoir.

2.4. Volatile content builds up in the magmareservoir

It is theoretically possible to have circumstancesin which explosive activity can be generated by amagma whose primary volatile content is insu⁄-cient to allow magma fragmentation to occur bythe time the vent pressure is reached (Fig. 2). Therequirement is that the magma must reside in areservoir at a depth which is su⁄ciently shallowbelow the sea-£oor to allow some volatile (mainly

Fig. 3. The maximum depth in the ocean at which an eruption can be of the steady explosive hawaiian-style, shown as a functionof the initial magmatic H2O content for a series of initial magmatic CO2 contents. The depth ceases to depend on water contentwhen magma fragmentation takes place solely due the presence of CO2 bubbles before any H2O exsolution has taken place.

Table 3Submarine hawaiian-type lava fountain eruption conditionsfor a moderately volatile-rich magma containing 1.4 wt%CO2 and 0.54 wt% H2O

dv df Lm uv(m) (m) (kg m33) (m/s)

0 190 13 246100 152 149 147200 114 288 108300 76 427 78400 38 565 50489 0 686 0

For various depths, dv, of the vent below the ocean surface,values are given for: the depth below the vent at which mag-ma is fragmented, df ; the bulk density of the erupting gas-pyroclast mixture, Lm ; and the mean eruption speed of gasand pyroclasts, uv. The entry for dv = 0 corresponds to a sub-aerial eruption of this magma.

VOLGEO 2544 30-1-03


CO2) exsolution to occur over a ¢nite range ofdepths below the roof of the reservoir (Bottingaand Javoy, 1990). Gas bubbles form and driftupwards to the roof where they accumulate intoa foam layer in which the gas volume fraction isvery high ^ greater than 90% is possible before thefoam becomes unstable (Vergniolle and Jaupart,1990). If this magmatic foam is subsequentlyerupted, it will be able to simulate the hawaiian-style eruption of a very gas-rich magma. This pro-cess has been proposed for some subaerial hawai-ian eruptions (e.g. Vergniolle and Jaupart, 1990),though we note that it does not seem to be con-sistent with bubble growth rates in the ascendingmelts inferred from vesicle size distributions in theeruption products (Mangan et al., 1995). When allof the foam layer has been discharged, it may ormay not be possible for the gas-depleted magmato be erupted after it ^ this will depend on thedetails of the magma density and magma reser-voir pressure. However, it is not trivial to assumethat magmatic foams can be retained in reservoirsin this way (in subaerial or submarine environ-ments). When a foam layer exceeds a criticalthickness, which is at most a few tens of meters(Vergniolle and Jaupart, 1990), the gas bubbles atthe top of the foam will collapse and release gasinto a continuous pocket which can escape intoany cracks in the reservoir roof. Also, foam accu-mulation at the roof of a reservoir is itself a pro-cess encouraging the concentration of stress at theroof (Par¢tt et al., 1993) and may lead to eruptioninitiation before very much accumulation has oc-curred.If magmatic volatiles do exsolve and build up

at the roof of a magma reservoir in this way (Fig.2), it is necessary to ¢nd the minimum volatilecontent of the magma in the reservoir which willallow some gas exsolution to occur just below theroof. Consider a reservoir with its roof at a depthof 1000 m below the summit of a seamount whichhas its vent located at a water depth of 1500 m(Fig. 2). The pressure Pv at the vent is the sum ofthe overlying water weight plus the atmosphericpressure, (0.1 MPa+[9.8 m s32U (1500 mU1026kg m33 = ) 15.18 MPa. Using an edi¢ce bulk den-sity of be = 2700 kg m33 (to allow for a smallamount of vesicularity in the eruption products

from which it is constructed), the pressure, Pr,at the level of the reservoir roof (which is thelowest pressure likely to exist in the magma incontact with the reservoir roof, see Par¢tt et al.,1993) will be (Pv+[9.8 m s32U1000 mU2700 kgm33]) = 41.64 MPa. Table 1(a) shows that no H2Ocould exsolve at this pressure unless the watercontent of the magma was more than about 7.4wt%, and so we assume that only CO2 vapor ispresent. At a pressure of Pr = 41.64 MPa the den-sity bcr of CO2 with mc = 44.00 kg kmol31 at mag-matic temperature T=1528 K is bcr = [(mc P)/(QT)] =V144.22 kg/m3. We require that a foambe formed which contains a large enough massfraction of gas to allow magma fragmentationto occur when the foam decompresses to the pres-sure of a vent assumed to be at the 1500-m waterdepth of the summit of the seamount. We sawearlier that this corresponds to V5.52 wt%CO2. If the volume fraction of CO2 in the foamcorresponding to this minimum mass fraction isXgmin its mass fraction is (bcr Xgmin) whereas themass fraction of the liquid magma with densitybm =2700 kg/m3 is (bm {13Xgmin}). We thereforehave:

nc ¼ ðb crX gminÞ=ðb crX g þ bmf13X gmingÞ ð11Þ

and with nc = 5.52 wt%, i.e. 0.0552, we ¢ndXgmin = 0.5224. Thus, any foam that accumulatesto the extent that the gas occupies at least 52.24%of the volume will lead to a hawaiian-style erup-tion when it erupts at the sea £oor. Since foamswith thicknesses of a few tens of meters can re-main stable with gas volume fractions as high asat least V90% (Vergniolle and Jaupart, 1990),this is clearly a potential mechanism leading tosubmarine lava fountain eruptions at much great-er depths than those illustrated earlier. The likelymaximum volumes of magma that could beerupted in this way can be estimated by assumingthat a 30-m thick foam layer with 80% gas volumefraction occupied an area of 3 km3, approxi-mately that of the roof of the summit magmachamber of Kilauea Caldera, HawaiPi (e.g. Ryanet al., 1983). The corresponding dense rock equiv-alent volume is then 20U106 m3, approximatelyten times the amount released in a single eruptive

VOLGEO 2544 30-1-03


episode at Kilauea’s PuPu POPo vent (Wolfe et al.,1988).The total amount of gas expansion which can

occur to contribute signi¢cant energy to the erup-tion products is that which takes place betweenthe level at which the magma is fragmented by gasbubble expansion and the vent level. In the exam-ple above, the build up of a foam in the reservoirto a volume fraction of 52.24% would just allowmagma fragmentation at the vent but this wouldlead to essentially no acceleration due to gas ex-pansion, and would yield only a very small erup-tion velocity. Interestingly, this situation maxi-mizes the amount by which the gas bubbles inthe magma can expand: decompression from the242-MPa chamber roof pressure to theV15-MPavent pressure would lead to less than a 2-foldexpansion of recently nucleated V20-Wm diame-ter bubbles to at most V 40 Wm. If the foambuilds up to a greater volume fraction, magmadisruption occurs below the vent and althoughless pre-fragmentation bubble expansion occurs,more energy is available from subsequent gas ex-pansion. We can estimate likely magma eruptionspeeds as follows. Assume that the gas volumefraction Xg in the foam at the roof of the magmareservoir is some value greater than the minimumvalue of Xgmin = 0.5224 found above. The CO2 gasmass fraction ncr which corresponds to this vol-ume fraction Xg is found from the equivalent ofEq. 11:

ncr ¼ ðb crX gÞ=ðb crX g þ bmf13X ggÞ ð12Þ

Table 4 shows how ncr varies if we let Xg in-crease from the minimum of Xgmin = 0.5224 to val-ues as large as 0.9; gas mass fractions in excess of0.3, i.e. 30 wt%, are possible. Also shown in Table4 are the corresponding values of the bulk densityLm of the gas^pyroclast mixture which emergesthrough the vent when a foam with this large agas content erupts at the vent pressure Pv = 15.18MPa. The values are found from Eq. 5 by sub-stituting ncr for nc and setting ns = 0, a reasonableapproximation as Eq. 7 shows that no water islikely to be exsolved from magmas at this ventpressure unless they contain more than 0.7 wt%H2O. All of the densities are substantially lessthan that of the overlying seawater.

We anticipate that gas concentrations as largeas those found here will lead to very energeticsubmarine explosive eruptions. To estimate erup-tion speeds, we ¢rst need to know at what pres-sure Pd the magmatic foam will disrupt as it risesthrough the dike system from the level of thereservoir roof to the vent. We assume that, forfoams with Xg initially less than 0.75, this willoccur when the gas volume fraction has increasedto 75%, i.e. when [(ncrQT)/(mcPd)] = 3 [(13ncr)/2700], to give:

Pd ¼ ½2700ncrQT �=½3 mcð13ncrÞ� ð13Þ

We ¢nd the distance, dd, below the vent atwhich the pressure Pd is reached by assuming,as in 2.3. Gas exsolution, disruption, hawaiian-style fountaining, that the pressure gradient inthe erupting magma is close to the lithostatic gra-dient, so that (Pd3Pv) = be g dd. In the case offoams with Xg initially greater than 0.75, magmadisruption will presumably occur within the mag-ma reservoir roof as the foam collapses so that Pdwill be equal to Pr. We then ¢nd the mean erup-tion speed, uv, of gas and pyroclasts at the vent byequating the kinetic energy per unit mass, (0.5 u2v),to the energy due to gas expansion minus the

Table 4The variation of the mass fraction, ncr, of CO2 gas in afoam accumulating under the roof of a magma reservoir as afunction of the volume fraction, Xg, of gas in the foam

Xg ncr Pd dd Lm uv(wt%) (MPa) (m) (kg/m3) (m/s)

0.5244 5.52 15.18 0 714 small0.55 6.13 16.97 68 661 620.60 7.42 20.83 214 570 1150.65 9.02 25.76 400 487 1640.70 11.08 32.38 650 410 2170.75 13.81 41.64 1000 339 2790.80 17.60 41.64 1000 274 3150.85 23.24 41.64 1000 213 3620.90 32.47 41.64 1000 156 428

The reservoir is located at a depth of 1000 m below the sum-mit of a submarine volcano and the summit vent is at awater depth of 1500 m. Also shown are the pressure, Pd, atwhich the foam disrupts into pyroclasts as it rises through adike to the vent, the depth, dd, below the vent at which thisoccurs, the bulk density Lm of the erupting gas^pyroclastmixture, and the velocity, uv, with which the mixture eruptsinto the overlying water.

VOLGEO 2544 30-1-03


work done against gravity in rising the distancedd ; the equivalent of Eq. 10 in 2.3. Gas exsolu-tion, disruption, hawaiian-style fountaining be-comes:

0:5u2v ¼ ðQT=mcÞncrlnðPd=PvÞþðPd3PvÞf½ð13ncrÞ=bm�3ð1=b eÞg ð14Þ

Table 4 shows the values of Pd, dd and uv re-sulting from foam gas volume fractions betweenthe minimum of Xgmin = 0.5224 and a likely upperlimit of Xg = 0.9; clearly eruption speeds of hawai-ian-style lava fountains of more than 400 m/s arepossible, even for a vent 1500 m below the sur-face. If foam build-up occurred in the magmareservoir of a volcanic edi¢ce with its vent lyingat shallower water depths then, as shown by thetrends in Tables 2 and 3 for eruptions in which noconcentration of volatiles occurs in this way, evenhigher eruption speeds would be possible. Wenow explore the likely interaction of both kindsof submarine lava fountains with the overlyingwater column.

2.5. Submarine eruption plumes

We saw in 2.3. Gas exsolution, disruption, ha-waiian-style fountaining, and 2.4. Volatile contentbuilds up in the magma reservoir, that continuoussubmarine explosive activity, in which a steadystream of fragmented magma is ejected from thevent into the overlying seawater column, can oc-cur either when the magma is very rich in volatilesor when volatiles accumulate in a foam layer atthe roof of a magma reservoir. We now treat theconsequences of the interaction between a steadilyerupting gas-pyroclast jet and the overlying wateras follows. First, as illustrated in Fig. 4, assumethat a ¢ssure vent of width W and along-strikelength S is at a depth where the pressure is Pand is erupting a steady jet of disrupted magmaconsisting of clots of liquid with unvesiculateddensity bm entrained in a mixture of gaseouswater and carbon dioxide, all the componentsbeing at the same temperature T. In keepingwith the examples shown in Tables 2 and 3, thebulk density of the erupting mixture is Lm and the

mean velocity of the jet is uv. The mass £ux ofmagma emerging from the vent is M where:

M0 ¼ L muvWS ð15Þ

and the momentum £ux is Z0 where:

Z0 ¼ L mu2vWS: ð16Þ

The momentum £ux will be conserved as mix-ing with the surrounding water takes place, butthe total mass £ux will increase as ocean water isadded to the widening plume. We therefore needto evaluate the mass £ux immediately after theerupting jet has completed the ¢rst phase of itsinteraction with the surrounding seawater. We as-sume that the mixing takes place under conditionssimilar to those described by Prandtl (1949) forthe turbulent interaction of a jet of £uid entrain-ing a surrounding £uid having a similar density.This is not unreasonable given the densities foundabove for volcanic £uids just able to erupt explo-sively on the ocean £oor. The volcanic jet mixeswith the surrounding water in a ¢xed geometricpattern (Fig. 4) so that the inner edge of the zoneof mixing extends from the edge of the vent to thecenter-line of the jet in the time it takes the mix-ture to rise to a height hm =6 W, where W is thevent width. At the same time the outer edge of themixing zone extends away from the edge of thevent by an amount (hm/8) on either side of the jet,so that the full width of the widening jet just afterit is fully mixed with the surrounding water is[W+2 (hm/8)] = [W+2 (6W/8)] = (5/2) W. Giventhe range of dike widths, up to V2 m, estimatedfor submarine rise crest eruption (Head et al.,1996), we infer that hm is typically V12 m andthe width of the jet after mixing is V5 m. Weassume (and will justify retrospectively) that dur-ing this mixing process two things happen: ¢rst,enough heat is transferred to the entrained waterthat all of the magmatic water vapor condensesinto liquid water with density bw and, second,that all of the carbon dioxide dissolves into thewater. These assumptions can be used to calculatethe resulting bulk density of the mixture as fol-lows.Consider a reference volume V located immedi-

VOLGEO 2544 30-1-03


Fig. 4. Submarine hawaiian eruption model developed in this analysis. (A) Formation and emergence of gas/pyroclast jet and de-velopment of the zone of mixing. (B) Evolution and collapse of the zone of mixing to form £ows, welded agglutinates, and den-sity currents.

VOLGEO 2544 30-1-03


ately above the vent so that V is proportional tothe vent widthW. If the bulk density of the erupt-ing £uid is Lm then the total mass within the vol-ume V is (V Lm). The mass of volcanic watervapor must therefore be (ns V Lm) and this occu-pies a volume (ns V Lm/bs). Similarly, the mass ofCO2 is (nc V Lm) in a volume (nc V Lm/bc) and themass of rock is [(13ns3nc) V Lm] in a volume[(13ns3nc) V Lm/bm]. Now consider the condi-tions after mixing with the ocean water is com-plete and the edges of the jet have expanded tothe width (5/2) W. The reference volume has en-larged to (5/2) V, but within it the volume of rockis still [(13ns3nc) V Lm/bm]. Because the volcanicwater has condensed, its volume is now (ns V Lm/bw) and the volume of the dissolved carbon diox-ide is of course zero. So the volume of oceanwater that has been entrained must be {(5/2)V3[(13ns3nc) V Lm/bm]3(ns V Lm/bw)} and itsmass must therefore be [{(5/2)3[(13ns3nc) Lm/bm]3(ns Lm/bw)} {V bw}]. The total mass withinthe volume (5/2) V consists of this entrained watermass plus the original masses of rock, volcanicwater vapor and CO2, i.e. [(13ns3nc) V Lm], (nsV Lm) and (nc V Lm), respectively. The new bulkdensity, L, is therefore this total mass divided bythe new volume and, after a little simpli¢cation, isfound to be:

L ¼ bw þ ð2L m=5Þf13ns3½ð13ns3ncÞðbw=bmÞ�g

ð17Þ

Given this value for the bulk density of themixture of volcanic materials and seawater wecan evaluate the velocity of the mixture. The mo-mentum £ux from the vent, Z0, given by Eq. 16,must be equal (since momentum is conserved) tothe momentum £ux Z in the mixed, expandedplume, given by:

Z ¼ L u2ð5=2ÞWS ð18Þ

and hence:

u=uv ¼ ½ð0:4L mÞ=L �1=2 ð19Þ

The mean temperature of the mixture can befound by equating the enthalpies of the compo-nents before and after mixing has taken place andcan be represented by:

½ð13ns3ncÞVL m�crðT3T eÞ þ ðnsVL mÞhðT3T eÞþ

ðncVL mÞccðT3T eÞ ¼ ½fð5=2Þ3ðð13ns3ncÞL m=

bmÞ3ðnsL m=bwÞgfVbwg�hðT e3273:15Þ ð20Þ

where T is the eruption temperature of the mag-ma as before, Te is the ¢nal equilibrium temper-ature of the mixture and it has been assumed thatthe ocean water is just above the freezing point at273.15 K. cr and cc are the average speci¢c heatsat constant pressure over the V273^1528 K tem-perature range of interest of basaltic rock andcarbon dioxide,V1200 andV900 J kg31, respec-tively, and h(Tx3Ty) is the speci¢c enthalpychange of water between any temperatures Txand Ty, obtained from UKCPS (1970). Use ofthe enthalpy eliminates the need to keep trackof the temperature variation of the speci¢c heatof water and, when the pressure is less than theV22 MPa critical pressure, to include the latentheat.For each of the eruptions illustrated in 2.4. Vol-

atile content builds up in the magma reservoir,and 2.5. Submarine eruption plumes, Table 5shows the results of mixing with seawater. Table5(a) deals with the eruption of the very volatile-rich magma, Table 5(b) with the moderately vol-atile-rich magma, and Table 5(c) with the case offoam accumulation in a reservoir. For each waterdepth of the vent, the magma density and erup-tion speed in the vent from Tables 2^4 are re-peated for ease of comparison with the mixturedensity and upward speed after mixing has takenplace. A number of interesting features are seen inTable 5. First, the velocity, u, after mixing is typ-ically no more than one third of that before mix-ing. Second, whereas in all cases the bulk densityof the jet of disrupted magma emerging from thevent is at least four times smaller than the densityof the surrounding water before mixing takesplace, after mixing is complete the bulk densityof the jet is always greater than that of the sur-rounding water, typically by 10^20% (except forthe shallowest eruptions where the density di¡er-ence can become quite small). This ¢nding di¡ersfrom the qualitative expectation of Cashman andFiske (1991) that the mixtures should be less

VOLGEO 2544 30-1-03


dense than the surrounding ocean water: ourquantitative analysis shows that the in£uence ofthe loading of the entrained pyroclasts is greaterthan they anticipated. Third, the thermal calcula-tion shows that the temperature after mixing is

always much less than the boiling point at theambient pressure. Thus the earlier assertion thatall of the volcanic steam condenses during theinitial phase of mixing with the surrounding wateris always entirely justi¢ed. Finally, we give in Ta-

Table 5Illustrations of the consequences of steadily erupted gas^pyroclast jets interacting with seawater. Part (a) deals with the eruptionof the very volatile-rich magma shown in Table 2, part (b) with the moderately volatile-rich magma shown in Table 3, and part(c) with the case of foam accumulation in a reservoir shown in Table 4.(a) Very volatile-rich magma contains 5.4 wt% CO2 and 1.9 wt% H2O

Before mixing After mixing

dv Lm uv L u hf Te BP(m) (kg m33) (m/s) (kg m33) (m/s) (m) (‡C) (‡C)

0 3.5 556 n/a n/a n/a n/a (n/a)50 21 450 1031 41 84 5 (159)250 92 340 1049 64 206 13 (226)500 181 275 1072 71 260 26 (265)750 269 230 1094 72 265 39 (291)1000 355 193 1116 69 242 52 (312)1250 440 161 1137 63 201 63 (328)1500 522 128 1158 54 151 76 (343)1750 601 94 1179 43 94 87 (355)2000 680 54 1199 26 33 98 (366)2100 711 27 1206 13 9 104 (370)2134 720 small n/a small small 105 (372)

(b) Moderately volatile-rich magma contains 1.4 wt% CO2 and 0.54 wt% H2O

0 13 246 n/a n/a n/a n/a (n/a)50 80 177 1046 31.0 49.0 11 (159)100 149 147 1063 34.7 61.4 21 (184)200 288 108 1098 34.9 62.1 41 (212)300 427 78 1133 30.4 47.0 61 (236)400 565 50 1167 22.1 24.9 82 (252)489 686 small n/a small small 99 (263)

(c) Foam accumulates under the roof of a magma reservoir

Before mixing After mixing

Xgmin Lm uv L u hf Te BP(m) (kg m33) (m/s) (kg m33) (m/s) (m) (‡C) (‡C)

0.5244 714 small n/a small small n/a (n/a)0.55 661 60 1196 28 40 90 (343)0.60 570 106 1174 47 111 81 (343)0.65 487 166 1154 68 237 69 (343)0.70 410 220 1135 84 357 57 (343)0.75 339 284 1117 99 500 47 (343)0.80 274 320 1101 101 519 38 (343)0.85 213 368 1086 103 541 29 (343)0.90 156 435 1072 105 560 20 (343)

For each water depth of the vent, dv, in parts (a) and (b), and for each gas volume fraction in part (c), the magma density Lm

and eruption speed uv in the vent are repeated from Tables 2^4 for ease of comparison with the mixture density L and upwardspeed u after mixing has taken place. hf is the height to which the fountain rises over the vent and Te is the equilibrium tempera-ture after complete mixing between the erupted jet and the entrained seawater. The ¢nal column gives the water boiling point,BP, at the pressure implied by the depth dv for comparison with Te.

VOLGEO 2544 30-1-03


ble 5 the values of the maximum height hf , equalto [u2/(2 g)], to which the materials in the jet couldrise above the vent if there were no further sig-ni¢cant interaction with the surrounding sea-water. These are seen to range from tens of metersin the case of the moderately volatile-rich magma,Table 5(b), to hundreds of meters in the case ofthe very volatile-rich magma, Table 5(a), and thefoam, Table 5(c). In Table 5(c), where the vent isat 1500 m depth, the volcanic materials do notreach depths shallower than 950 m despite thehigh eruption speeds. However, in the case of avent at 50 m depth in Table 5(b), materials doalmost reach the surface (nominally failing to doso by 1 m); and in the case of a vent at 50 mdepth in the more volatile-rich case in Table 5(a),the 84-m rise distance implies that the mixture ofpyroclasts and entrained water should overshootthe ocean surface by V34 m.In fact, the motion of the pyroclasts and en-

trained seawater after mixing is complete justabove the vent is much more complex than im-plied by the calculation of hf described above. Theconditions are similar to those in subaerial erup-tions in which not enough atmosphere is en-trained into the jet of volcanic materials emergingfrom a vent to ensure convective rise of the result-ing plume and are consistent with the scenariodescribed by Kokelaar and Busby (1992) to ex-plain the deposits of a more silicic submarine ex-plosive eruption. In general the mixture collapsesback to the ocean £oor as a density current froma height of order hf at a speed similar to u, andthe entire structure resembles a vertical fountainover the vent. The internal motions of such asystem are complex, particularly because the jetrising from the vent is now entraining not sea-water but the descending part of the fountainwhich will be denser than seawater by an amountsimilar to that found in the above mixing calcu-lation. The treatment of subaerial volcanic foun-tains by Wilson and Heslop (1990) shows that thedynamic pressure exerted by the falling materialcan cause the pressure in the vent in such systemsto rise well above the ambient pressure, reducingthe velocity of the fountain at its base (and henceits rise height) because less gas expansion occurs.However, this phenomenon is very much less pro-

nounced in the submarine environment. The dy-namic pressure exerted by the falling material is[(1/2) L u2], and using the values in Table 5 this isonly a signi¢cant fraction of the ambient waterpressure at the vent depth for vents shallowerthan V50 m.More important is the uncertainty related to

the decoupling of the motions of the pyroclastsof various sizes from that of the entrained water(e.g. Cashman and Fiske, 1991; Kokelaar andBusby, 1992; Davis and Clague, 1998; White,2000). All of the clasts will have a ¢nite terminalvelocity in the water and will therefore lag behindthe upward water speed by this amount. The ter-minal velocities of 1-m, 10-cm, 1-cm and 1-mmsized particles will be of order 3, 1, 0.3 and0.1 m/s, respectively; these values assume clastdensities of V1500 kg m33 (implying signi¢cantvesicularity) and include due allowance for thebuoyancy of the clasts in the water (Cashmanand Fiske, 1991), giving them an e¡ective densityof V500 kg m33, but they will not be as much asa factor of 2 greater even if there is no vesicular-ity. Thus, only very large magma clots emergingthrough the vent will decouple signi¢cantly fromthe water motion near the vent in explosive sub-marine eruptions. Given the range of speeds aftermixing in Table 5(a)^(c), say 30^100 m/s, thetimescales [(2 u)/g] for clasts to pass completelythrough the fountain over the vent will be 3^10s, and the thicknesses of the chilled skins on anyclasts in direct contact with seawater will be up toV(U tf )1=2 =V3 mm, where U is the thermal dif-fusivity of rock, V1036 m2/s. Thus, in general(though see also the discussion below) cm-sizedand smaller pyroclasts will undergo signi¢cantheat loss, whereas larger magma clots will reachthe surface with a chilled skin surrounding alargely hot interior.The primary size frequency distribution of frag-

ments resulting from submarine explosive erup-tions is poorly known. Also poorly understoodare the auxilliary processes (discussed in a follow-ing section) that might operate on the pyroclastsin this environment, such as the in£uence of con-tact-surface steam explosivity on the pyroclasts inthe mixing zone (Kokelaar, 1986). For example,as more clasts come into contact with seawater in

VOLGEO 2544 30-1-03


the outer parts of the descending column (Fig. 4),water is vaporized, and the clasts cool and mayshatter, potentially leading to fuel^coolent inter-actions. Despite these uncertainties, the generalfate of pyroclasts in this environment is clear.The high density of the surrounding water col-umn, relative to the subaerial environment, meansthat particles are decelerated dramatically in thenarrow mixing zone and eruption jet dispersal isinhibited relative to the subaerial environment.Thus, most pyroclasts will begin to fall in theimmediate vicinity of the vent (within a few me-ters radius) due to the negative buoyancy of themixed zone.Particles may have several fates, depending on

their position in the jet and zone of mixing (Fig.4). In the inner part of the column, where coolingis least, clasts are hottest and clast number densityis highest, the descending clasts may reaccumulateand coalesce to form a lava £ow or a lava pondfeeding a £ow. Toward the outer part of the col-umn, clast number density is less, more coolinghas occurred, the mixing zone is negatively buoy-ant, and column collapse will result in densitycurrents descending down the margins of the col-umn, reaching the surface in the vicinity of thevent, and expanding out into the surroundingareas at initial speeds of order at least 30^100m/s (e.g. Table 5(b),(c)) as potentially erosive, py-roclast-charged density £ows. The runout distan-ces of such £ows will be controlled much more bymixing with the overlying water than by basalfriction. If the geometry of the mixing process isthe same as that in the vertical part of the plumewe might expect runout distances of order twicethe plume height, i.e. V1 km for foam drivenhawaiian plumes. In the outermost parts of thecollapsing column, extensive mixing with watermay elutriate many (relatively small) clast sizesinto the submarine equivalent of a subaerial co-ignimbrite eruption cloud (Fig. 4; Kokelaar andBusby, 1992). The relatively high e¡usion ratesassociated with hawaiian-type eruptions wouldlead to the prediction that any lava £ows associ-ated with these events would be characterized bylobate sheets, rather than pillows (e.g. Head et al.,1996; Gregg and Fink, 1995).In summary, submarine hawaiian eruptions will

produce low, narrow eruption jets, which willquickly collapse to produce proximal accumula-tions of hot coalescing pyroclasts within a fewmeters of the vent and cooler, less clast-rich den-sity £ows descending from the distal parts of thecolumn and spreading radially away from thevent for distances of the order of the height ofthe column (relative to the subaerial case, density£ows are subjected to additional drag at theirupper surfaces by the submarine medium). Land-forms anticipated from these eruptions (Fig. 4)might include cones surrounding the vent withrim crests within a few meters of the vent (muchcloser than in the subaerial case), possible lavaponds within the cone, and an apron of pyroclas-tic deposits surrounding the vent. The cone andthe £anking deposits should consist of interlayersof pyroclastic £ows and lava £ows with sheet £owmorphology, rather than pillow lava morphologydominating. At greater radial distances from thevent (Fig. 4), one would predict that £ows andlayers of agglutinated pyroclasts would dominateproximally in the cone, giving way to bedded py-roclastics and interlayered lava £ows distally.Explosive eruptions driven by foam layer col-

lapse should produce generally similar depositsand landforms to those from normal magmaticvolatile-induced hawaiian eruptions but withsome systematic di¡erences. Thus, typical pyro-clast sizes might be smaller (due to the greateramount of gas bubble expansion), and the pyro-clasts themselves may be dominated by shatteredbubble walls forming ‘limu o Pele’. Cooling ofdescending pyroclasts in the outer parts of thecollapsing fountains would be more e⁄cient be-cause fountain heights would exceed those ex-pected for purely magmatic cases, leading todi¡erent types of proximal deposits. Runoutdistances for density £ows would also be greater,potentially leading to surrounding pyroclastic de-posits in the tens to hundreds of meters range. Ingeneral, the volatile buildup at the top of the res-ervoir would leave a complementary volatile-de-pleted magma below. Thus, after the volatile-richlayer was discharged during the hawaiian-styleeruption event, it could be followed by a veryvolatile-depleted e¡usive phase, and vesicle-poorlavas overlying pyroclastic-rich cones might there-

VOLGEO 2544 30-1-03


fore be a distinctive signature of this eruptionsetting. However, Blake (1981) has shown thattypically less than 1% of the total volume of themelt in a reservoir leaves the reservoir in indi-vidual overpressurization-induced diking events.Thus, it is entirely possible that only the volatilerich foam layer might be erupted.

2.6. No normal fragmentation, but low rise speedcauses strombolian activity

Su⁄ciently volatile-poor magmas cannot eruptexplosively as long as the bubbles are uniformlydistributed in the magma; however, explosive ac-tivity may occur if the magma rise speed in thedike/conduit feeding the vent is so slow that thereis time for large, early-nucleated bubbles to over-take smaller, later-nucleated bubbles in a run-away bubble coalescence process (Figs. 2 and5A). The intermittent emergence of the resultinggiant bubbles through the lava at the vent drivesstrombolian explosions. Since this mode of erup-tion requires a relatively low magma rise speed inthe dike, it is most likely to be associated with lowe¡usion rate eruptions. If a lava lake is presentaround the vent (Fig. 5B), the intermittent largebubbles burst through its surface and throw upmagma clots whose size and shape (and launchspeed) are dictated by the radius of the bubble,its arrival speed at the surface (larger bubblestravel faster) and the degree of cooling and hencenon-Newtonian behavior of the lake surface (thehigher the eruption rate the more rapidly the lakeover£ows to feed £ows and the hotter its surfacestays, on average, leading to easier deformationand smaller pyroclast sizes). If no lava lake ispresent over the vent, the large bubbles risingup the dike simply blow o¡ the magma just aheadof them (Fig. 5A). In this case the magma is in-evitably hotter than if it had been sitting in a lavalake. The pyroclast sizes to be expected in thiscase are hard to predict. Because strombolian ac-tivity is most likely to occur in magmas with lowexsolved gas contents, this implies that the magmabetween the large bubbles will contain relativelysmall numbers of small gas bubbles and so willcertainly not have any strong tendency to disruptfurther into small fragments; in the extreme case

the largest magma clots ejected might have theirlongest axes comparable to the width of the dike.We can explore the ranges of eruption condi-

tions that give rise to strombolian activity in amagma that has too low a volatile content toallow hawaiian-style activity to occur. For anychosen magma viscosity and total volatile contentthere is a critical magma rise speed below whichthere will be extensive bubble coalescence andstrombolian activity will be unavoidable. A com-puter program to simulate coalescence numeri-cally was developed by Wilson and Head (1981)and Par¢tt and Wilson (1995). We have usedthis program to ¢nd the maximum magma risespeeds, us, allowing strombolian activity whenthe volatile involved is CO2 and the eruptivevent is at least 1500 m below sea level. Theserise speeds have implications for the widths ofthe dikes through which the eruptions are occur-ring. A magma with a given viscosity R will havea rise speed us which is a function of the dikewidth Ws and the pressure gradient dP/dz drivingthe motion (Wilson and Head, 1981) such that inlaminar £ow (which can be shown to be the rele-vant condition for all of the cases above by retro-spectively calculating the Reynolds number of themotion):

us ¼ ðW 2s dP=dzÞ=ð12R Þ ð21Þ

For submarine eruptions Head et al. (1996)show that a plausible pressure gradient is dP/dz = 2000 Pa/m, and with this value we ¢nd themaximum dike widths allowing strombolian activ-ity to occur shown in Table 6. The values, alwayssigni¢cantly less than 1 m, ensure that conductivecooling and solidi¢cation of the dike will be rela-tively rapid (Wilson and Head, 1988; Bruce andHuppert, 1989; Head et al., 1996) and that, in thesubmarine environment, strombolian eruptionevents will typically last on the order of hours,rather than days.Using these calculations, we can now estimate

the characteristics and evolution of the eruptionprocess and make predictions about the resultingdeposits and landforms. Initial stages of a strom-bolian eruption might be characterized by dikeemplacement and extrusion of lavas at very lowe¡usion rates, leading to production of short

VOLGEO 2544 30-1-03


Fig. 5. Submarine strombolian eruptions. (A) Con¢guration of a vent. (B) Con¢guration of a lava pond. (C) Examples of subae-rial eruptive products in the block to bomb size range (from Macdonald, 1967). Explanation: (A) bipolar fusiform bomb withlee side to top; (B) cross-section of bomb shown in (A); (C) unipolar fusiform bomb; (D) almond-shaped bomb; (E) cross-sec-tion of bomb shown in (D); (F) cross-section showing broad equatorial ¢n; (G) cylindrical ribbon bomb; (H) cross-section ofbomb shown in (G); (K) cow-dung bomb; (J) cross-section of bomb shown in (K). (D) Example of a possible fragment derivedfrom the top of a lava lake (from Kokelaar, 1986) and described in the text.

VOLGEO 2544 30-1-03


£ows and pillows rather than extensive lobatesheets (e.g. Head et al., 1996; Gregg and Fink,1995). As the rise rate stabilized, and prior tothe time that cooling closed o¡ ascent, strombo-lian activity would occur as the gas bubble risespeed exceeded the magma rise speed. Disruptionof magma would occur at the vent^water inter-face; the maximum bubble size for subaerial en-vironments is about 5^10 m (Wilson and Head,1981) and that of submarine environments underthese conditions is about 1.5^2 m, being mainlycontrolled by the higher ambient pressure. Bub-bles will rise more slowly (4^5 m/s) than in thesubaerial case, and smaller bubbles mean that

there will be lower internal excess pressures (1^2bar), and thus initial ejecta velocities will be small,but up to 20 m/s. As the bubbles burst (Fig. 5),fragments mingle with the surrounding water, andare cooled and drastically decelerated; fragmentsthat might travel to maximum distances of 50^60 m in the subaerial environment are restrictedto less than 10^20-m maximum distance in thesubmarine environment. Thus, water drag and de-celeration of pyroclastic fragments is a key factorin both strombolian and hawaiian-style submarineeruptions and the resulting landforms. The max-imum bubble sizes of 1^2 m mean that typicalfragment sizes will be much smaller than this.

Fig. 5 (Continued).

VOLGEO 2544 30-1-03


Expansion of the gas bubble just prior to burst-ing will tend to accelerate all particles to approx-imately the same initial velocity, but the greaterinertia of the larger particles will mean that theywill travel further, and one can anticipate blocksand bombs littering the surface beyond the edgeof ¢ner-grained deposits. Larger fragments (in theblock, 64^256-mm, and bomb, s 256-mm, range)will probably be derived from the plug of magmapushed in front of the rising gas bubble (Fig. 5A),or from the cooled upper surface of a lava pond(Fig. 5B). The full spectrum of morphologies ofbomb and block shapes might be anticipated (Fig.5D), and the details of shapes, sizes, abundancesand frequency distribution around speci¢c ventsare keys to the original eruption conditions. Forexample, a sample collected from a submarineeruption at Surtla (Fig. 5D; Kokelaar, 1986)could be similar to the types of features antici-pated to be derived from the top of a lavapond; the cooled upper thermal boundary layer(chilled at the top, vesicular below) might beripped apart by the expanding gas bubble (Fig.5B), folded over in £ight and, upon landing, stillbe hot enough to cause agglutination of lapillionto its lower surface.The pulses associated with the bursting of bub-

bles in strombolian events might provide the en-ergy necessary to initiate contact-surface steamexplosivity (e.g. Kokelaar, 1986, and see discus-

sion below). In this case, fragments might shatter,possibly repetitively, so that a series of shatteringevents might operate to reduce the average grainsize until mm-sized fragments were formed; atthat point, subsequent cooling would not generatesu⁄cient steam to continue shattering by this pro-cess. Even with the extensive particle shatteringthat might occur during the contact-surface steamexplosivity process, the volume density of par-ticles is still very likely to be insu⁄cient to makeextensive density £ows, and the particles wouldmore likely settle to the surface to form a thinlayer of cooled pyroclasts in the area immediatelysurrounding the vent (Fig. 5A,B).Ejecta built up around a submarine stromboli-

an vent would obviously be closer to the ventthan in the subaerial case. The resulting conewould be no more than 20^40 m in maximumwidth, and would have a rim crest whose positionwas dictated by a combination of fragment sizeand velocity, and thus accumulation rate. Rimcrests would probably be located at radii of lessthan about 10 m. Rim heights would depend oneruption duration, and on the basis of the simplecooling relations discussed above, might be lim-ited to few meters. Initial strombolian eventscould build a low ring of agglutinated pyroclastsaround the vent and thus provide the basis foraccumulating a small lava pond or lake. The pres-ence of a pond (Fig. 5B) could change the natureof the pyroclastic deposits ; bubbles bursting inthe cooler magma of the pond and disrupting asurface thermal boundary layer, the characteris-tics of which might vary as described above,would produce larger, cooler pyroclasts than inthe earlier stages, and this should be re£ected inthe cone deposits. Strombolian activity is unlikelyto be su⁄ciently vigorous and continuous undersubmarine conditions that it will lead to a distinc-tive mixing layer which undergoes collapse to pro-duce gravity-driven pyroclastic £ows. Lava £owsassociated with these types of cones, in additionto having a surface morphology consistent withvery low e¡usion rates (predominantly pillows),should also be relatively depleted in vesicles. Insummary, the types of deposits associated withstrombolian activity should on average be domi-nated by relatively large clasts deposited near the

Table 6Magma rise speed at great depth in the conduit system, us,and corresponding conduit width, Ws, which marks theboundary between strombolian activity and steady magmadischarge

Magma contains 0.4 wt%CO2

Magma contains 1.4 wt%CO2

R us Ws us Ws

(Pa s) (m/s) (m) (m/s) (m)

30 0.5 0.30 2.0 0.60100 0.2 0.35 0.8 0.69300 0.1 0.42 0.4 0.85

Values are given as a function of magma viscosity, R, in ba-salts containing 0.4 and 1.4 wt% CO2. At higher magma risespeeds in wider conduits bubble coalescence is negligible anda steady discharge of vesicular magma (or, if the water depthabove the vent is small enough, hawaiian-style explosive ac-tivity) will occur.

VOLGEO 2544 30-1-03


vent and often agglutinated and welded, coneswith rim crest radii measuring a few meters, alack of extensive pyroclastic £ows, and associatedlava £ows with predominantly pillow textures.The role of smaller-scale pyroclastic fragmentsgenerated by contact-surface steam explosivity isat present unknown, but could clearly alter thesize-frequency distribution of pyroclasts and re-sulting deposits.

2.7. Volatile content builds up in the top of dikeleading to vulcanian eruptions

In this case, which might occur following a pe-riod of strombolian or hawaiian activity, the topof the dike becomes sealed, and gas accumulatesin the upper part of the dike (Figs. 2 and 6) lead-ing to su⁄cient excess pressure that a vulcanian-type explosion takes place. In these events, theregion around the top of the dike (consisting ofcountry rock) will be broken into angular frag-ments, mixed with some juvenile material fromthe upper part of the dike, and ejected radiallyaway from the point of the explosion (e.g. Headand Wilson, 1979). In the subaerial environment,fragments can be accelerated and transported todistances of kilometers (Fagents and Wilson,1993), but in the submarine environment, thefragments will be rapidly decelerated by the sur-rounding seawater, drag forces being V103 timeslarger than subaerially, and will settle to the sur-face as a relatively chaotic mass of angularblocks, generally within just a few meters (lessthan 5 m) of the vent. There is no a priori reasonto expect that a vulcanian event will be followedby extensive extrusive volcanism (because the as-cent of material in the dike has already ceasedand additional cooling has taken place as the vol-atiles have built up) and thus these types of eventsmight be characterized by deposits of angularblocks of country rock without many subsequentlava £ow deposits. Should e¡usion take place, itwould more likely be at low rates and thus wouldbe characterized by short pillow-lava £ow domi-nated deposits, rather than extensive sheet £owdeposits. A potential major variation on thistheme is the possibility that the energy associatedwith the vulcanian explosion might provide the

threshold necessary to initiate contact-surfacesteam explosivity in the exposed magma in thetop of the dike. In this case, the proximal blockyejecta deposits might be veneered with hyaloclas-tic deposits. In addition, exposure of the top ofthe dike by the explosion might cause bulk inter-action steam explosivity as seawater surged intocontact with the exposed magma at the top of thedike. As soon as the water was converted tosteam, however, the expanding wave would pushwater back out of the vent, suppressing furthercontact and permitting additional cooling totake place. This process seems self-limiting, butperhaps a few surges of cooling and quenchingcould take place to produce local pyroclastic de-posits before equilibrium was reached.

3. Non-magmatic gas mechanisms for magmafragmentation and the production of hyaloclastites

The range of magma^water interactions thatmight occur in subaqueous and emergent basalticvolcanism has been described by Kokelaar (1986)(Fig. 7). This includes the range of conditions thatmight result in the explosive release of magmaticvolatiles that we have treated in detail above, aswell as a variety of other mechanisms which weconsider below and relate to magmatic gas releaseprocesses.

3.1. Cooling-contraction granulation

In this mechanism, which can occur at anydepth, magma comes into contact with waterand cools by conductive heat transfer at its sur-face, thus developing a temperature gradient be-tween the center and the surface of the fragmentor feature (Figs. 7 and 8). Rapidly following theextrusion of lava underwater, the outer boundarylayer becomes rigid and, when the interior cools,it will contract more that the outer layer canaccommodate, so that cracking or granulationresults. Among the deposits produced are domi-nantly sand- and granule-size grains which com-prise a mixture of glassy globules commonlyshowing evidence of shattering in situ, and highlyangular chunks and splinters of glass. These are

VOLGEO 2544 30-1-03


commonly observed at the margins of lava bodies,particularly pillows, which are susceptible to gran-ulation in situ.

3.2. Bulk interaction steam explosivity

In this mechanism, enclosure of water in mag-ma, or entrapment of water close to magma,causes the creation and explosive expansion ofsteam (Figs. 7 and 9). Signi¢cant fragmentationfrom this mechanism appears to be restricted todepths much shallower than that of the criticalpressure of water (V3 km in seawater). Severalscenarios can be visualized, including those wherewater is trapped adjacent to magma, such as whenlava £ows out onto a wet substrate and root-less explosions occur, the submarine equivalentof subaerial pseudocraters (Thorarinsson, 1953;Keszthelyi et al., 2000; Greeley and Fagents,2001). In other cases, magma might intrude awet slurry of water-saturated sediment as a dikeor a sill causing explosions and disruptions ofcountry rock and juvenile material (Zimanowskiet al., 1991; for analogous situations on Mars, see

Squyres et al., 1987; Wilson and Head, 1994;Head and Wilson, 2002). The main processesthat form clasts are the tearing apart of magmaaround explosively expanding steam and the shat-tering of rigid magma by associated pressurewaves (Kokelaar, 1986). Explosive expansion ofsteam depends on the ambient pressure and theamount and rate of heat exchange between themagma and water. The onset of fragmentationis likely to depend largely on factors that in£uencethe rate and amount of heat exchange, which in-clude the con¢guration and duration of entrap-ment or enclosure, and related convection pro-cesses. At depths below that of the criticalpressure, any transformation of liquid to super-critical vapor may involve only a several-fold vol-ume expansion, and pressure relief is less likely.At shallow depths, however, such an expansionmay go to several thousand times the originalvolume, and thus fragmentation is most likely toaccompany large volume expansions at depthsshallower than the critical depth of about 3 km.Deposits associated with these types of eventsshould be characterized by a combination of

Fig. 6. Con¢guration and geometry of a submarine vulcanian eruption.

VOLGEO 2544 30-1-03


blocky country rock ejecta and disrupted juvenilematerial forming volcaniclastics. If they form assubmarine pseudocraters they might be relativelyreadily recognized as a series of small craters sur-rounded by ramparts of blocky ejecta and ¢ner-grained pyroclastics. One might expect these fea-tures to be more abundant on sheet £ows than onpillow basalts on the basis of the ability of sheet

£ows to trap larger amounts of water over shorterperiods of time. Features formed from intrusiveevents such as dike injection would appear verysimilar to those from vulcanian events describedabove. Di¡erences might include variations in cra-ter geometry and country rock size-frequency dis-tribution based on the intrusion into a wet slurryin the bulk interaction steam explosivity example

Fig. 7. Submarine clast-forming processes and possible enhancement mechanisms. From Kokelaar (1986).

Fig. 8. Processes and environments of cooling-contraction granulation.

VOLGEO 2544 30-1-03


(e.g. shallower crater, smaller and less coherentfragments).

3.3. Contact-surface steam explosivity

This mechanism, also known as fuel^coolentinteraction, involves the explosive expansion andcollapse of steam formed at magma^water contactsurfaces (Figs. 1 and 7). It commonly requiresinitiation by vigorous impact between magmaand water; although no certain depth limit isknown, the likelihood of such explosivity de-creases rapidly with increasing depth. The initialsurface event is su⁄ciently vigorous to cause on-going melt fragmentation, mixing with water, andheat transfer so that steam explosivity is capableof sustaining the interaction until the entire meltis fragmented. In such a sustained interaction,fragmentation and heat transfer occur extremelyrapidly and they can be violently explosive; mostof the melt is ¢nely comminuted (a large propor-tion in the micron size range), and the explosion isextremely powerful (essentially a violent hydro-magmatic eruption) (Wohletz, 1983, 1986).Two alternative models seem to account best

for this process (Kokelaar, 1986; Wohletz, 1986).In one, the ‘spontaneous nucleation model’, su-

perheated water vaporizes instantaneously andproduces homogeneous boiling. In the ‘thermaldetonation model’, rapid vaporization occurs be-hind a propagating shock. Basically, when magmacomes into contact with water, a thin ¢lm isformed along the contact surface by coalescingsteam bubbles, and the ¢lm becomes unstable asit expands and collapses on a microsecond ormillisecond scale. If the steam ¢lm oscillates vig-orously, it may cause the magma to become ¢nelyfragmented and to mix turbulently with water.Explosions of this mixture can occur if eitherthe water is superheated to its spontaneous vapornucleation temperature (homogeneous boiling), orif the insulating steam ¢lm collapses or is dis-rupted by a pressure wave, causing further meltfragmentation, so that heat is rapidly exchangedand steam is instantly produced (thermal detona-tion). This interaction can be sustained since ineither case the explosion can cause further meltfragmentation and turbulent mixing with water(Kokelaar, 1986; Fig. 7). Kokelaar (1986) citesthe observations of decreasing amounts of clasticmaterial with increasing depth to propose thatcontact-surface steam explosivity is limited by in-creasing hydrostatic pressure. For example, forthermal detonation, vapor expansion of su⁄cient

Fig. 9. Processes and environments of bulk interaction steam explosivity.

VOLGEO 2544 30-1-03


violence is needed to cause continued fragmen-tation and heat exchange capable of propagat-ing the pressure wave and sustaining the interac-tion; increasing pressure with depth reduces theprobability of achieving such propagation. Evenat much lower pressures, however, experimentshave shown (Wohletz, 1986) that pressure has sta-bilized vapor ¢lms and limited reactions. Thus,although triggers appear to work, they need tobe increasingly vigorous at higher pressures andgreater water depths, so that expansion aids theprocess and so that it does not become self-limit-ing.In summary, increased con¢ning pressure sup-

presses the process of contact-surface explosivity(Figs. 1 and 7) by limiting initial fragmentationand heat exchange during ¢lm boiling, by pre-venting spontaneous nucleation vaporization,and by requiring increasingly vigorous triggersto initiate interaction (Kokelaar, 1986). In thegrowth and vertical evolution of seamounts(Fig. 1), the transition from lava £ows to volcani-clastic deposits does not necessarily require exso-lution of magmatic volatiles, but can be accom-plished in principle by contact-surface explosivityprocesses (Fig. 6). However, magmatic volatileexplosivity is likely to be one of the principlecauses of and initiators of contact-surface explo-sivity. Thus, great care should be taken to developcriteria to distinguish the characteristics of prod-ucts and landforms produced by magmatic vola-tile explosivity processes (Figs. 2, 4 and 5) fromthose related to contact-surface explosivity pro-cesses.

4. Summary of predictions of the style ofemplacement of deposits and processes offormation of landforms

On the basis of observations of deposits andlandforms (e.g. Schmidt and Schmincke, 2000;Batiza and White, 2000) on a variety of sea-mounts (e.g. Smith and Batiza, 1989) and sea£oor environments, and theoretical considerationof the ascent and eruption of basaltic magma (e.g.Head et al., 1996) under similar submarine con-ditions (Fig. 6), we now summarize a set of pre-

dictions (Table 7) that might be used to distin-guish submarine e¡usive and explosive depositsproduced under a variety of eruption conditions(Fig. 2). These predictions are designed to aid inthe observation and interpretation of deep-sealandforms and their deposits, and to re¢ne thetheoretical treatments of the behavior of eruptingmagma under submarine conditions.For magma compositions typical of spreading

centers and their vicinity, the most likely circum-stance in the depth ranges considered here is thequiet e¡usion of magma with minor gas exsolu-tion, and the production of somewhat vesicularpillow lavas or sheet £ows, depending on e¡usionrate (Head et al., 1996; Gregg and Fink, 1995).Eruption column heights would be measured incentimeters and would be insu⁄cient to causeany magma disruption. Magma disruption fromthe exsolution of purely primary magmatic vola-tiles to produce hawaiian-style continuous foun-taining (Figs. 2 and 4) would not be expected inthese environments because primary magma vol-atile contents are considerably less than the sev-eral wt% required to cause disruption. Such erup-tions might occur in some subduction zoneenvironments where unusually high primary mag-ma volatile contents are observed. Should such aneruption occur in the submarine environment atthese depths (Fig. 1), one would anticipate ahighly collimated gas^pyroclast column rising toa height of the order of a few meters, surroundedby a distinctive but narrow mixing zone whichwould collapse due to increased density to forman inner zone dominated by primary and rootless£ows and agglutinated hot pyroclasts, and an out-er zone or apron (meters to several tens of metersin extent) of pyroclastic density £ows (Fig. 2).Due to the relatively high e¡usion rate, associatedlava £ows would tend to be sheet £ows ratherthan pillow lavas.The most likely mode of occurrence for pyro-

clastic deposits in the submarine environment andwith ma¢c compositions is strombolian (Fig. 2),due to the arti¢cial local buildup of volatiles inmagma that has a low rise speed. In this case,magmatic gas collects by bubble coalescence, as-cends, and reaches su⁄cient concentration thatdisruption of the magma occurs, producing local-

VOLGEO 2544 30-1-03


ized blocks, bombs, and pyroclastic deposits(Fig. 5). Because of the very low magma risespeeds, associated lava £ows would be expectedto be dominated by pillow lavas rather than sheet£ows. These deposits can be distinguished fromhawaiian eruption products and landforms bythe abundance of blocks and bombs (Fig. 5C),the lack of extensive pyroclastic £ow deposits(eruptive pulses are episodic and volumetricallysmall), and di¡erences in the nature of associatedlava £ows (strombolian: pillow lavas, less vesicu-lar; hawaiian: sheet £ows, more vesicular).Another possible mode of occurrence of pyro-

clastic deposits in the submarine environment isthat resulting from vulcanian eruptions (Fig. 2).These deposits, being characterized by the domi-nance of angular blocks of country rocks depos-ited in the vicinity of a crater (Fig. 6), should beeasily distinguished from strombolian and hawai-

ian eruptions. Other modes of bulk-interactionsteam explosivity (Fig. 7) may have similar char-acteristics (Fig. 9, left). The production of pseu-docraters by the bulk-interaction steam ex-plosivity mechanism (Fig. 9, right) should bedistinguishable from strombolian eruptions byvariations in the shape and size-frequency distri-bution of ejecta and its distribution around thevent (compare Figs. 5 and 9, right).A special case of the hawaiian eruption style

may occur if magmatic gas buildup occurs in amagma reservoir (Fig. 2). In this case, a layer offoam may build up at the top of the reservoir insu⁄cient concentrations to reach and even exceedthe volatile contents necessary for disruption andhawaiian-style activity. In this case, the depositsand landforms are predicted to be somewhat dif-ferent from those of a typical primary-magmatic-volatile-induced hawaiian eruption (Fig. 4). Spe-

Table 7Predictions concerning characteristics and deposits for di¡erent eruption styles

Lava £ows Ventcharacteristics

Cones Pyroclastics deposits;agglutinates;blocks and bombs;hyaloclastite deposits

Density £ows

(1) No gas exsolution: No vesicles ? ^ ^ ^

(2) Gas exsolution, nomagmatic disruption:

Vesicular ? ^ ^ ^

(3) Hawaiian-stylefountaining: gasexsolution, magmadisruption:

Vesicular sheet£ows

Linear orcircular pit

Meters to severaltens of meters indiameter

Proximal weldeddeposits, distalfragmental andpartly agglutinated

Surround ventand cone to tensof meters’ radius

(4) Strombolian activity:magma volatile contentinsu⁄cient of causefragmentation; lowmagma rise speed causesarti¢cial volatileenhancement:

Less vesicular,tend towardshort pillow lavas

Crater Rim crest 6 10-mradius, agglutinatedpyroclasts

Abundant blocksand bombs, ¢nerfragments near vent

Density £ows rare

(5) Vulcanian eruption:volatile contentarti¢cially builds upin top of dike:

May be pondsin crater

Explosioncrater

Blocky cone Blocks of countryrock, minorhyaloclastites

^

(6) Enhanced hawaiian-style eruption: volatilecontent arti¢ciallybuilds up in magmareservoir forming foam:

Probably lessvesicular

Crater Cone broader andless steep

Deposits lessagglutinated, moredispersed. No blocksor bombs

Density £owsextensive, out tos 100-m radius

VOLGEO 2544 30-1-03


ci¢cally, typical pyroclast sizes might be smaller,pyroclasts might be dominated by broken bubblewalls (‘limu o Pele’), fountain heights may exceedthose described in Fig. 4, cooling of descendingpyroclasts would be more e⁄cient, leading to dif-ferent types of proximal deposits, and runout dis-tances for density £ows would be greater, poten-tially leading to surrounding pyroclastic depositsin the tens to hundreds of meters range. In addi-tion, £ows emerging after the consumption of thefoam layer would tend to be very depleted in vol-atiles, and thus extremely poor in vesicles relativeto typical £ows associated with hawaiian-styleeruptions in the primary magmatic gas case.Thus, su⁄cient di¡erences appear to exist to dis-tinguish between these two types of hawaiianeruption style (Table 5).

5. Relation of predictions to recent sea £oorobservations

Numerous sea £oor exploration e¡orts havedocumented unusual pyroclastic and hyaloclasticdeposits at a wide range of depths in the subma-rine environment. We now brie£y examine arange of these observations in order to assessthe applicability of the models developed in thiscontribution to the interpretation and observationof pyroclastic and hyaloclastic deposits.Fouquet et al. (1998) reported on the discovery

and documentation of extensive volcaniclastic de-posits along the MAR axis (southwest of theAzores) that they suggested could be formed bydeep-water explosive volcanic activity. They ex-plored a series of three progressively deeperMAR segments, ranging from depths of V400toV2000 m, each with di¡erent types of deposits.The shallowest segment (38‡20PN; V400 toV930 m water depth) is about 45 km long, lacksa deep axial rift valley, and contains a circular25-km diameter central volcano with a height ofV1200 m, which is bisected by a 2-km wide,500-m deep axial graben. Exposed in the grabenwalls and on the £oor is at least 400-m thicknessof layered volcaniclastic ejecta, with individuallayers ranging from a few mm to a few cm thick,consisting of sand and lapilli-sized clasts, and a

few m-thick poorly sorted lapilli layers. The inter-mediate depth segment (Menez Gwen; V700 toV1050 m water depth) is morphologically similarto the shallow one, with a central V16-km diam-eter volcano having a height of V700 m, bisectedby a 2-km wide, 300-m deep axial graben. A290-m thick volcaniclastic unit overlies a 60-mthick lava £ow section containing £ows up to3 m thick. The deepest segment (Lucky Strike;1570^V2000-m water depth) consists of a 15-kmwide rift valley containing a central 12U8-km vol-cano with a 1-km wide central caldera; threedominantly scoriaceous breccia summit cones sur-round a restricted area of layered volcaniclasticdeposits within the caldera. The central part ofthe caldera contains a lava lake with non-vesicu-lar lobate and sheet £ows. The volcaniclastic de-posits are well-layered, but much thinner (6 10 mthick) and less extensive than those on the twoshallower segments. The volcanic breccia depositstypical of the three summit cones (V1700 m) aredominated by massive fragmental units gradinglaterally into in situ breccias and coherent highlyvesicular scoriaceous lava £ows that locally formpillow lava.Sea bottom re£ectivity images enabled Fouquet

et al. (1998) to extrapolate their submersible diveobservations and showed that: (1) the volcaniclas-tic material is restricted to the center of the seg-ments, forming most of the surface of the volcaniccones, (2) the area of the surface covered by vol-canic deposits decreases with increasing depth(V67 to V15 to V2.8 km2 for the three seg-ments, respectively), (3) the deposits become thin-ner and less voluminous with depth, and (4) lava£ows are the only deposits at the southern andnorthern ends of the three segments, whereas thevolcaniclastics coexist with lava £ows in the cen-tral part of the segment (suggesting to Fouquet etal. (1998) a relation to ridge segmentation).The deposits themselves commonly consist of

well-strati¢ed ash layers, with normal and possi-bly reversed grading, very sharp bedding planesbetween lapilli and ¢ner ash units, worm burrowsat the tops of some beds, some intercalated mm-thick calcareous pelagic sediment, non-gradedlayers generally restricted to the very center ofthe segments, some layers with a matrix of pelagic

VOLGEO 2544 30-1-03


sediment, and decreasing total thickness as a func-tion of increasing water depth (s 400,V270 and6 10 m for the three segments). Thin sectionsshowed that the predominant clast type is scoria-ceous (30^50% vesicle volume) to pumiceous (50^80% vesicle volume) glass fragments, with highlyvesicular glass fragments showing often extremestretching, indicating acceleration and rapid de-formation prior to and during quenching. No evi-dence of welding was observed. More common inthe coarse-grained volcaniclastics are broken crys-tals and accidental lithic fragments. Geochemicalanalyses show that most of the glasses have awater concentration typical of N-MORB (6 0.4wt%).Fouquet et al. (1998) interpreted these charac-

teristics to mean that the deposits were largelyprimary volcaniclastics from episodic and perhapsshort-lived explosive eruptive events occurring re-peatedly over long periods of time, with the non-graded layers representing proximal deposits. Inaddition they cited the following characteristicsin support of their interpretation: (1) the layerednature is indicative of fallout deposits after mag-ma fragmentation caused by the combined e¡ectsof expanding magmatic volatiles as well as hydro-magmatic processes involving seawater, (2) thelack of welding suggests that if any hot eruptionfacies were generated, they were of limited extent,and (3) the abundance of lithic clasts and brokencrystals favors a magma^water explosive inter-action such as that seen in maars, tu¡ conesor Surtsey-type hydromagmatic volcanism. Theyconcluded that the bedded volcaniclastics resulted‘from submarine explosions involving a combina-tion of expansion of magmatic volatiles, bulk/sur-face steam explosivity of seawater and possiblythermal contraction fragmentation, which can oc-cur at any depth, perhaps aided by a rapid extru-sion rate’ (Fouquet et al., 1998). They concludedthat their evidence suggested that explosive erup-tions can occur and produce extensive depositsalong mid-ocean ridges at considerably greaterwater depths (up to V1700 m) than commonlybelieved. The preferential distribution of the de-posits at the medial topographic highs of the threeMAR segments suggested to them a preferentialconcentration of volatiles there.

The excellent descriptions and abundant evi-dence for submarine explosive volcanic eruptionsdocumented by Fouquet et al. (1998) permit us toapply the theoretical models and predictions de-scribed above to these three examples, in order todevelop further the interpretations of Fouquet etal. (1998), and to help distinguish among the rolesof magmatic gas expansion, bulk/surface steamexplosivity and thermal contraction fragmenta-tion. We ¢nd that the deposits and relationshipsdescribed at the two shallowest ridge segments aremost consistent with magmatic gas exsolution,disruption and hawaiian-style pyroclastic foun-taining. Evidence favoring hawaiian-style erup-tions includes: (1) relatively wide dispersal, largethickness and volume of deposits, (2) relativelysmall grain size, (3) lack of sorting, (4) lack ofwelding, (5) elongation and stretching of frag-ments, and (6) presence and nature of layering.We envision the following scenario for these

events (Fig. 4A,B): juvenile volatile expansionleads to a su⁄ciently high gas bubble content toproduce magmatic disruption and the rise of ha-waiian-style plumes. Fragmentation associatedwith the gas expansion and disruption producestephra composed of glassy vesicular bubble wallfragments, many of which have been stretchedand elongated. Rapid mixing with seawater causesimmediate quenching and the rising column be-comes denser than adjacent seawater and beginsto collapse. The outer margins of the collapsingcolumn form a dense turbulent slurry which de-scends the £anks as a density current; some sort-ing, grading, and erosion of underlying depositsoccurs. Multiple fountaining events cause succes-sive £ows and form additional layers ; periodsbetween such eruptions are characterized by cal-careous sediment accumulation and biogenic re-working in the upper pyroclastic layers. This sce-nario is consistent with the characteristics ofthe deposits : (1) non-graded layers generally re-stricted to the very center of the segments (prox-imal density current deposits), (2) well-strati¢edash layers, with normal and possibly reversedgrading; very sharp bedding planes between lapilliand ¢ner ash units (more distal density currentdeposits; the matrix of pelagic sediment seen insome layers could be eroded by the density cur-

VOLGEO 2544 30-1-03


rent from the underlying between-event calcare-ous sediment accumulation), and (3) worm bur-rows at the tops of some beds and some inter-calated mm-thick calcareous pelagic sediment(layers deposited between eruption events).In summary, hawaiian eruption style events ap-

pear to explain the vast majority of the observa-tions without requiring a major role for bulk/sur-face steam explosivity and thermal contractionfragmentation, although such processes will inevi-tably accompany submarine magmatic and explo-sive eruptions. If these deposits are produced byhawaiian eruptions, is enhanced primary mag-matic gas content or arti¢cial buildup in the sum-mit of the reservoir required (Fig. 2)? Volatilecontents cited by Fouquet et al. (1998) (typicalof N-MORB, 6 0.4 wt%) would require no en-hancements to cause disruptions at depths of lessthan V500 m. At depths typical of the deeperparts of the shallow segment and the intermediatedepth segment, some additional gas buildup isrequired for basalts of this composition. We favorthe arti¢cial buildup of gas in foams at the top ofthe magma reservoir for the shallow and inter-mediate depth cases for the following reasons:(1) the locations of these extensive deposits areat the center and topographic summit of a ridgesegment, where the central magma reservoir ispredicted to reside on the basis of the geometryof dike emplacement (e.g. see Head et al., 1996and references therein), and (2) the buildup offoam in the reservoir will lead to much more vig-orous pyroclastic fountains which we interpret tobe more consistent with the thickness, large lateralextent (out to 1 km) and volumes of the observeddeposits.Fouquet et al. (1998) note that the landforms

and deposits of the deepest segment (LuckyStrike) di¡er signi¢cantly from those of the twoshallower segments. These pyroclastic deposits arecharacterized by smaller volumes, less extensivedispersal, and coarser-grained clasts. The conesat the summit of the central edi¢ce are character-ized by volcanic breccia that they interpreted asbeing produced by autobrecciation, downslope re-sedimentation, and/or cooling-contraction granu-lation. We interpret these pyroclastic deposits tobe the distal products of strombolian eruptions

(Fig. 5A,B), with arti¢cial gas buildup beingcaused by relatively high rise rate of bubbles com-pared with magma, and the coalescence andgrowth of these bubbles leading to the disruptionof the magma surface and the ejection of pyro-clasts to the vicinity of the vent. Adjacent to thesedeposits within the caldera is a lava lake withnon-vesicular lobate and sheet £ows. The non-ve-sicularity of these deposits is consistent with theirdegassing during a period of strombolian activityand their subsequent eruption.In summary, we concur with the interpretation

of Fouquet et al. (1998) that these deposits are ofpyroclastic origin and we further suggest that ha-waiian-style eruptions dominate the two shallow-est MAR segments, whereas strombolian-styleeruptions dominate the deepest segment. We donot discount the possibility that hydrothermallycirculating water contributed to the explosivityof these eruptions, but stress that the small vesiclesizes and relatively uniform distributions ofvesicles in the bulk of the clasts imply that if ex-ternal water was incorporated into these magmas,it must have been absorbed by solution at thedepth of the magma reservoirs to ensure its uni-form distribution in the magma.Clague et al. (2002c) described submarine spat-

ter and volcanic bombs in alkalic basalts fromKauaPi and submarine spatter of tholeiitic compo-sition from Kilauea’s submarine east rift zoneeruption, and interpreted them to be indicativeof submarine strombolian eruptions. Abundantsamples of bubble-wall fragments (‘limu o Pele’)collected along the submarine rift zone of Kilaueaand the Gorda Ridge axis are interpreted asmildly explosive strombolian eruption events.From alkalic vents on the south side of KauaPi,

Clague et al. (2002c) described highly vesicularspatter, ribbon spatter, breadcrust bombs, a spin-dle bomb, and a large block of agglutinated spat-ter. These are precisely the kinds of features thatare predicted by the model calculations (Fig. 5A^D), and we concur with Clague et al. (2002c) intheir interpretation of these as representative ofstrombolian-style eruptions. Enhanced volatilecontent can be attributed to both the higher vol-atile abundance of alkalic basalts and the concen-tration of volatiles by bubble coalescence.

VOLGEO 2544 30-1-03


‘Limu o Pele’ were originally described by Honet al. (1988) in an environment where lava enteredthe ocean and produced bubbles in the lava dueto incorporation of seawater into the stream oflava and its expansion to steam. Following thisinterpretation, Clague et al. (2000) attributed‘limu o Pele’ found on the Puna Ridge at about2200 m depth, on the LoPihi Seamount at 1150^1950 m depth, and on Seamount 6 at 1600^2000 mdepth, to be due to the expansion of seawaterduring boiling following its contact with lava.However, the discovery of ‘limu o Pele’ from vol-canic cones in the North Arch lava ¢eld at 4160 mdepth, i.e. a depth signi¢cantly greater than thecritical point of seawater (Clague et al., 2002a),and from the Gorda Ridge (V2800 m depth;Clague et al., 2002c), led to the reinterpretationof these deposits as being derived from a sepa-rated magmatic gas phase during strombolianeruptions (Clague et al., 2002c). Clague et al.(2002c) further suggest that such strombolian ac-tivity may be very common along the entire mid-ocean ridge magmatic system, that seismic re£ec-tion pro¢les may in part be detecting the presenceof coalesced gas layers or volcanic foams in theupper ridge reservoirs that are feeding the strom-bolian eruptions, and that event plumes previ-ously interpreted to be hydrothermal dischargesmay also be composed of a signi¢cant componentof magmatic gas.We concur with the interpretation of Clague et

al. (2002c) that ‘limu o Pele’ could be produced bystrombolian activity. However, the presence of‘limu o Pele’ in itself does not necessarily distin-guish strombolian eruptions from hawaiian erup-tions. Thus, additional information is required todistinguish adequately between these two mecha-nisms, particularly since the buildup of foams atthe top of magma reservoirs can lead to hawaiianeruptions at signi¢cant depths. For example, therelative proportions of vesicles might be one in-dicator, with few to no vesicles favoring strombo-lian eruptions of more degassed magma, andmore abundant vesicles favoring hawaiian-styleeruptions. These types of observations, togetherwith the relative abundance of ‘limu o Pele’ versusspatter, might help to distinguish eruption types.Furthermore, the presence of magmatic foams

feeding hawaiian eruptions could make even moredramatic two of the factors cited by Clague et al.(2002c): (1) the possible presence of strong re£ec-tors at the top of magma reservoirs would besigni¢cantly enhanced due to magmatic foambuildup, and (2) event plumes consisting of a sig-ni¢cant magmatic gas component could be evenmore extensive if they were part of a hawaiian-style eruption that transferred the gas directly intothe ocean. The source of the event plume in asubmarine hawaiian eruption would not be thejet itself (because it collapses; Fig. 4B), but couldbe the more widespread sea £oor pyroclastic de-posit, producing an upwelling comparable to asubaerial co-ignimbrite plume.The summit of LoPihi Seamount lies at about

1200 m depth and exploration there has revealeda variety of pyroclastic deposits interpreted byClague et al. (2002b) to be evidence for phreatic,phreatomagmatic, hawaiian and strombolianeruptions. Extensive units including spatter,bombs and bubble wall fragments (‘limu o Pele’)are interpreted as evidence of stormbolian activ-ity. The presence of scoriaceous fragments andPele’s hair is cited as evidence of hawaiian erup-tive activity. The presence of cored bombs,coarse-grained basalt fragments, hydrothermallyaltered basalt and glass fragments, and hydrother-mal stockwork fragments are interpreted to rep-resent phreatic and possibly phreatomagmaticeruption styles, perhaps related to caldera or pitcrater formation. We concur with these interpre-tations but note that ‘limu o Pele’ is also verylikely to be associated with hawaiian-style subma-rine eruptions, as discussed above, and that vul-canian eruption products (Fig. 6) may be di⁄cultto distinguish from deposits of phreatic, phreato-magmatic and bulk interaction steam explosivity(Fig. 9) eruption styles.Particles attributed by Clague et al. (2002b) to

hawaiian eruptions are mostly alkalic basalts,while those attributed to other eruption typesare tholeiitic and transitional basalts. The volatilecontent of the LoPihi alkalic basalts (Clague et al.,2002c) (V1.5 wt% CO2) is insu⁄cient to causesimple hawaiian disruption (Fig. 2, example 3),according to the calculations we presented earlier(Table 3), and thus magmatic gas enhancement

VOLGEO 2544 30-1-03


through the development of magmatic foams atthe tops of subsurface reservoirs seems to be re-quired (Fig. 2, example 6). This situation seemsvery likely because of the high likelihood of thepresence of a large magma reservoir beneath theLoPihi summit.Deep sea drilling has also revealed that deep

submarine pyroclastic eruptions may be morecommon than previously thought. For example,hole 396B, which was drilled approximately 150km east of the MAR (22‡59.14PN, 43‡30.90PW),encountered (from top to bottom) 225 m of sedi-ment, 175 m of pillow basalts, and 90 m ofloose, sand-size, glassy basaltic debris (hyaloclas-tite) with some associated broken-pillow breccia(Schmincke et al., 1978). That the clastic unitwas primary and not produced during the drillingprocess was indicated by bedding in some of thevolcaniclastics, a high ratio of sideromelane tocrystalline basalt (up to 60%), and the high degreeof brecciation and vesiculation of overlying andunderlying pillow breccias. Furthermore, accord-ing to Schmincke et al. (1978) the near-synchro-neity of eruptions and the glass production pro-cess, and the lack of extensive transportation ofthe fragmental material from the site of breccia-tion, were suggested by: (1) the high ratio of glassto crystalline basalt, and (2) the relative chemicalhomogeneity of the glass in the bedded volcani-clastite. The units formed V13 Ma ago, and dueto the fact that eruptions occurred at depths esti-mated to be s 3000 m, Schmincke et al. (1978)concluded that the con¢ning pressures impliedby these depths precluded explosive volcanic ac-tivity due to primary magmatic volatile exsolutionand disruption, as well as sea-water vaporization.Instead, the major processes of formation of thesideromelane shards was attributed to ‘spalling ofpillow rinds to produce ‘‘spallation’’ shards [T]but implosion of pillows and granulation of lavamay also have occurred’ (Schmincke et al., 1978.p. 341). According to the authors, the process offragmentation of tachylite shards is unknown.Furthermore, Schmincke et al. (1978) note thatthe sideromelane in the deposit is extremely fresh,suggesting that the clastic unit was sealed o¡ frompercolating seawater since its formation about 13Ma ago.

On the basis of the theoretical treatments andpredictions described earlier in this paper, we be-lieve that the characteristics of the bedded clasticunit outlined by Schmincke et al. (1978) may bealternatively explained in the context of primarymagmatic volatile exsolution and disruption, suchas that characteristic of hawaiian and strombolianeruption styles. Given the plausibility of hawaiianeruptions even at these great depths due to thebuildup of magmatic gas foams in the top of areservoir (Fig. 2, example 6), the characteristics ofthe deposit might be explained as a hawaiian-styleeruption of such a magmatic foam for the follow-ing reasons.(1) Vesicle-rich overlying lava £ows suggest that

volatiles are exsolving during magma rise, storageand eruption in this environment.(2) The very close association of the £ows and

the bedded clastics suggests formation in the sameenvironment.(3) The absence of alteration (palagonitization)

suggests that the glass shards have not been incontact with seawater and thus that they under-went ‘rapid sedimentation’ preserving them in adry environment for 13 Ma; such an environmentmay have been the intermediate portions of themixing zone of a hawaiian eruption plume (Fig.4A,B) dominated by CO2 gas, where pyroclastic-glass and CO2-rich density £ows descended rap-idly from the plume and deposited material prox-imally and buried it rapidly before signi¢cant sea-water mixing and penetration had taken place.(4) The presence of indurated volcaniclastite

and the absence of crystalline basalt fragmentsin it suggest that this material may have beenthe result of emplacement in the warmer moreproximal parts of density £ows produced by thecollapsing column (Fig. 4B). These would bedominated by glass fragments and are more likelyto be welded.(5) The presence of loose and unindurated vol-

caniclastites and the common presence of crystal-line basalt fragments in the loose material sug-gests that these deposits may have been derivedfrom the outer part of the eruption plume mixingzone, where mixing with seawater is more thor-ough and collapse and resulting density currentsproduce unwelded deposits which incorporate the

VOLGEO 2544 30-1-03


substrate (the basalt fragments). The presence offoraminifera and bedding in these deposits furthersupports the emplacement in such an environ-ment.On the basis of these observations, and the

thickness of the deposits, we believe that the de-posits can be interpreted in the context of a ha-waiian-style eruption resulting from the release ofa magmatic foam built up at the MAR magmareservoir. While strombolian activity may havealso occurred, the volume of the deposits as wellas their apparent rapid and £ow-like emplace-ment, favors a major role for hawaiian eruptions.The thickness of the deposits suggests that thevent was in relatively close proximity and thatthere was more than one eruptive phase. Al-though spalling of pillow rinds and implosion ofpillows and granulation of lava may also haveoccurred (e.g. Schmincke et al., 1978), hawaiianeruptions seem to be a more likely dominant pro-cess.

6. Critical observations for interpretations andfurther assessment of models

Continued sea £oor exploration is essential inorder to test the models outlined here and to pro-vide new observations to re¢ne and modify them.Important observations that need to be made inorder to understand further the modes and stylesof submarine eruptions include the following:(1) Vent characteristics. What is the nature of

the vent? Does it appear constructional, or cre-ated from an explosion? Does it appear rootlessand superposed on a £ow or is it the source of£ows and other deposits? What is the width of thevent? What is the shape of the vent (elongate,circular, deep, shallow, etc.)?(2) Cones. What is the diameter of the cone, its

rim height, its shape, and the radius to the rimcrest? What are the surface deposits on the coneand how do they vary as a function of distancefrom the vent or rim crest? Is any stratigraphyexposed in the cone?(3) Blocks and bombs. Are these primary mag-

matic (e.g. cow-dung bombs) or angular blocks ofejected country rock? What is their size-frequency

distribution, and how does this change as a func-tion of distance from the vent? What is their mor-phology (see Fig. 5C)? What is the detailed struc-ture and morphology of bomb surfaces? What isthe radial range of the largest fragment and whatis its size and shape? Are blocks and bombswelded to the surface or do they have aggluti-nated undersides?(4) Volcaniclastic deposits. What is the distribu-

tion of these deposits relative to vents? What istheir thickness as a function of range? How doesthis vary as a function of azimuth? What is thenature of these deposits? What is the vesicle sizedistribution? Do they contain non-magmatic par-ticles (e.g. fragmental bedrock or incorporatedsediment)? What is the shape and relative abun-dance of di¡erent types of pyroclastic particles involcaniclastics? Is ‘limu o Pele’ present, in whatabundance, and what is its importance relativeto other pyroclastic fragments? Is Pele’s hairpresent? Is there evidence for graded bedding?What is the relationship of hyaloclastites to lava£ows? What are the key associations? Are blocksand bombs observed in association with proximaland distal pyroclastic deposits? Are ¢ner-grained(cm-sized) particles littered on top of pyroclasticdeposits? If so, what is their vesicularity?(5) Lava £ows. What is the form and surface

texture of lava £ows (e.g. pillows, sheet £ows,etc.)? Where do they occur in relation to otherdeposits in both time (superposed, covered) andspace (e.g. marginal to pyroclastic £ows, cappingthe vent around pyroclastic £ow deposits, etc.)?What are the general associations of each £owtype (e.g. pillow lavas and blocks and bombs,sheet £ows and agglutinated cone deposits, etc.)?What is the vesicularity of rocks associated witheach £ow type?

7. Summary and implications

The detailed models of the ascent and eruptionof magma in the submarine environment that aredeveloped here show that signi¢cant pyroclasticactivity can occur even at depths in excess of3000 m and that a wide range of pyroclastic de-posits should be anticipated. Mid-ocean ridge

VOLGEO 2544 30-1-03


magma reservoirs may be environments favoringthe buildup of magmatic foams and their rapidrelease, producing hawaiian-style eruptions andassociated pyroclastics. On the basis of the natureof ridge segmentation and magma reservoir for-mation and evolution, such foam buildups wouldbe favored at topographic highs in the middle ofridge segments. Furthermore, the episodic natureof dike emplacement at divergent plate bound-aries also produces conditions in which magmarise rates in dikes are often low, and bubble coa-lescence and magmatic gas rise rates are relativelyhigh, leading to magma disruption at the top ofthe dike and strombolian-style eruptive activity.The latter stages of development of volcanic edi-¢ces (seamounts) formed in submarine environ-ments are also excellent candidates for a widerange of submarine pyroclastic activity. This isdue not just to the e¡ects of decreasing waterdepth, but also to: (1) the presence of a summitmagma reservoir, which favors the buildup ofmagmatic foams (enhancing hawaiian-style activ-ity) and episodic dike emplacement (which favorsstrombolian-style eruptions), and (2) the commonoccurrence of alkalic basalts, the CO2 contents ofwhich favor submarine explosive eruptions atdepths greater than tholeiitic basalts. Pyroclasticdeposits resulting from vulcanian eruptions (char-acterized by the dominance of angular blocks ofcountry rocks deposited in the vicinity of a cra-ter), should be easily distinguishable from thosedue to strombolian and hawaiian eruptions, butwill be more di⁄cult to distinguish from productsof phreatic or phreatomagmatic eruptions styles.Bulk-interaction steam explosivity and contact-surface steam explosivity contribute to volcani-clastic (hyaloclastite) formation in these subma-rine environments.The presence of previous submarine explosive

eruptive activity can be detected through the anal-ysis of the resulting deposits, as outlined here. Inaddition, instrumentation of candidate sites ofsubmarine explosive activity might include oneor more of the following approaches.(1) Active seismic studies. Although magmatic

foam layers may be relatively thin (tens of me-ters), the location of volatile concentrations atthe top of magma reservoirs, both on mid-ocean

ridge summits and seamount summits, may bedetectable by seismic methods, as suggested byClague et al. (2002c). Furthermore, if volcaniclas-tic deposits are common in some parts of theupper crust, they may complicate the interpreta-tion of Layer 2a (e.g. Schmincke et al., 1978,where a 90-m thick layer is described from aDSDP drill hole); alternatively, high resolutionactive seismic techniques might be used to mapthe extent of such deposits.(2) Passive seismic studies. Submarine explosive

eruptions will be accompanied by seismic activity.Seismic activity associated with di¡erent explosiveeruption styles will di¡er from that typical of dikeemplacement (e.g. laterally migrating, episodicswarm activity; Fox et al., 1995) and is predictedto have the following characteristics.b for hawaiian eruptions: intermediate intensitysustained activity from a central source lastinghours to several days;b for strombolian eruptions: lower level intermit-tent activity from a central source or sources;b for vulcanian eruptions: high intensity spike-like activity from a central source.(3) Gas and thermal plume detection. A range of

gas release styles is predicted from submarine py-roclastic eruptions, and as pointed out by Clagueet al. (2002c), some of the event plumes that areassociated with seismic swarms (e.g. Embley andChadwick, 1994; Fox et al., 1995) could be re-lated to magmatic gas eruption as well as large-scale hydrothermal discharge (e.g. Baker, 1998).Our treatment shows that the most e⁄cient trans-fer of magmatic gas into seawater above dikes orvents occurs in strombolian eruptions, where largegas bubbles are segregated from the magma anderupted directly into the seawater. The H2O com-ponent of these bubbles will condense and theCO2 component will dissolve into the water toproduce narrow, CO2-rich, thermal plumes risingdirectly toward the surface. Magmatic gas re-leased during submarine hawaiian-style activityis predicted to be closely linked to hot pyroclaststhat mix with seawater at the margins of the erup-tive column and become dense enough to collapseand produce marginal density currents thatspread laterally for distances of up to a kilometer.Gas release from these types of events will thus

VOLGEO 2544 30-1-03


largely occur in the terminal phases of emplace-ment and will therefore be more dispersed in sea-water, more di¡use in area, and occur over a lon-ger time period than the strombolian case. Suchdi¡erences in predicted behavior suggest thatcareful monitering programs may be able to dis-tinguish hydrothermal discharges from both sub-marine explosive eruptions and the slow di¡usionof gas from intruded dike tops.Finally, the submarine environment on Earth is

similar in many ways to the surface environmenton Venus, where a thick CO2 atmosphere pro-duces surface pressures equivalent to depths ofabout 1000 m on the Earth’s sea £oor, su⁄cientto suppresses juvenile gas exsolution such thatmagmatic disruption and pyroclastic activity isnot favored under normal conditions (e.g. Headand Wilson, 1986, 1992). In previous treatments,strombolian activity seemed to be the most likelycandidate for explosive eruptions on Venus (Headand Wilson, 1986), but the results of the presentanalysis suggest that the buildup of magmaticfoams at the top of reservoirs on Venus (e.g. Pavriet al., 1992) may also favor the local occurrenceof hawaiian-style eruptions.

Acknowledgements

We gratefully acknowledge helpful discussionswith Rodey Batiza and David Clague. Thanks areextended to Leonid Dmitriev for productive dis-cussions and bringing to our attention the pres-ence of glassy sands and volcaniclastics recoveredfrom the MAR during the Glomar ChallengerDSDP drilling and dredging from the R/V Aka-demik Boris Petrov. Special thanks are extendedto Rodey Batiza for arranging for the participa-tion of J.W.H. in two cruises (R/V Atlantis II,October 1995; R/V KaPimikai-o-Kanaloa, October1998) and associated submersible dives to studyseamounts and submarine pyroclastic eruptions.Special thanks are extended to the pilots and sup-port crews of the submersibles Alvin and Pisces IVwith particular thanks to Pisces IV pilot TerryKerby for his interest and dedication to scienti¢cexploration. Thanks are further extended to Rich-ard Fiske and David Clague for detailed re-

views that resulted in a signi¢cant improvementof the manuscript. Financial support is gratefullyacknowledged from PPARC (Grant PPA/G/S/2000/00521 to L.W.) and the NASA PlanetaryGeology and Geophysics Program (NAG5-10690to J.W.H.).

References

Baker, E.T., 1998. Patterns of event and chronic hydrothermalventing following a magmatic intrusion: New perspectivesfrom the 1996 Gorda Ridge eruption. Deep-Sea Res. II 45,2599^2618.

Batiza, R., White, J.D.L., 2000. Submarine lavas and hyalo-clastite. In: Sigurdsson, H., Houghton, B., McNutt, S.R.,Rymer, H., Stix, J. (Eds.), Encyclopedia of Volcanoes. Aca-demic Press, San Diego, CA, pp. 361^381.

Batiza, R., Fornari, D.J., Vanko, D.A., Lonsdale, P., 1984.Craters, calderas and hyaloclastites on young Paci¢c sea-mounts. J. Geophys. Res. 89, 8371^8390.

Binard, N., Hekinian, R., Cheminee, J.L., Sto¡ers, P., 1992.Styles of eruptive activity on interplate volcanoes in theSociety and Austral hot spot regions: Bathymetry, petrolo-gy, and submersible observations. J. Geophys. Res. 97,13999^14015.

Blake, S., 1981. Volcanism and the dynamics of open magmachambers. Nature 289, 783^785.

Bonatti, E., Harrison, C.G.A., 1988. Eruption styles of basaltin oceanic spreading ridges and seamounts: E¡ect of magmatemperature and viscosity. J. Geophys. Res. 93, 2967^2980.

Bottinga, Y., Javoy, M., 1990. Mid-ocean ridge basalt degas-sing: Bubble nucleation. J. Geophys. Res. 95, 5125^5131.

Bruce, P.M., Huppert, H.E., 1989. Thermal control of basaltic¢ssure eruptions. Nature 342, 665^667.

Buck, W.R., Delaney, P.T., Karson, J.A., Lagabrielle, Y.(Eds.), 1998. Faulting and Magmatism at Mid-OceanRidges. Geophysical Monograph 106. American Geophysi-cal Union, Washington, DC.

Cashman, K.V., Fiske, R.S., 1991. Fallout of pyroclastic de-bris from submarine volcanic eruptions. Science 253, 275^280.

Cashman, K.V., Sturtevant, B., Papale, P., Navon, O., 2000.Magmatic fragmentation. In: Sigurdsson, H., Houghton, B.,McNutt, S.R., Rymer, H., Stix, J. (Eds.), Encyclopedia ofVolcanoes. Academic Press, San Diego, CA, pp. 421^430.

Clague, D.A., Davis, A.S., Bischo¡, J.L., Dixon, H.E., Geyer,R., 2000. Lava bubble-wall fragments formed by submarinehydrovolcanic explosions on Loihi Seamount and KilaueaVolcano. Bull. Volcanol. 61, 437^449.

Clague, D.A., Uto, K., Satake, K., Davis, A.S., 2002a. Erup-tion style and £ow emplacement in the submarine NorthArch Volcanic Field, Hawaii. In: Hawaiian Volcanoes:Deep Underwater Perspectives. Am. Geophysical Mono-graph 128, American Geophysical Union, Washington,DC, pp. 65^84.

VOLGEO 2544 30-1-03


Clague, D.A., Batiza, R., Head, J.W., Davis, A.S., 2002b.Phreatic, Phreatomagmatic, Hawaiian, and StrombolianEruption Products at the Summit of Loihi Seamount, Ha-waii. Am. Geophys. Union Spec. Publ. on Submarine Vol-canism, American Geophysical Union, Washington, DC, inpress.

Clague, D.A., Davis. A.S., Dixon, J.E., 2002c. Submarinestrombolian eruptions in Hawaii and along the GordaMid-ocean Ridge. In: Explosive Subaqueous Volcanism.Am. Geophysical Monograph, American GeophysicalUnion, Washington, DC, in press.

Davis, A.S., Clague, D.A., 1998. Changes in the hydrothermalsystem at Loihi Seamount after the formation of Pelee’s Pitin 1996. Geology 26, 399^402.

Devine, J.D., Sigurdsson, H., 1995. Petrology and eruptionstyles of Kick’em^Jenny submarine volcano, Lesser Antillesisland arc. J. Volcanol. Geotherm. Res. 69, 35^58.

Dixon, J.E., 1997. Degassing of alkalic basalts. Am. Mineral.82, 368^378.

Dixon, J.E., Clague, D.A., 2001. Volatiles in basaltic glassesfrom Loihi Seamount, Hawaii: Evidence for a relatively dryplume component. J. Petrol. 42, 627^654.

Dixon, J.E., Stolper, E.M., 1995. An experimental study ofwater and carbon dioxide solubilities in mid-ocean ridgebasaltic liquids, Part II. Applications to degassing. J. Petrol.36, 1633^1646.

Dixon, J.E., Clague, D.A., Stolper, E.M., 1991. Degassing his-tory of water, sulfur, and carbon in submarine lavas fromKilauea Volcano, Hawaii. J. Geol. 99, 371^394.

Dixon, J.E., Stolper, E.M., Holloway, J.R., 1995. An experi-mental study of water and carbon dioxide solubilities in midocean ridge basaltic liquids, 1. Calibration and solubilitymodels. J. Petrol. 36, 1607^1631.

Dixon, J.E., Clague, D.A., Wallace, P., Poreda, R., 1997. Vol-atiles in alkalic basalts from the North Arch volcanic ¢eld,Hawaii: Extensive degassing of deep submarine-erupted al-kalic series lavas. J. Petrol. 38, 911^939.

Embley, R.W., Chadwick, W.W., 1994. Volcanic and hydro-thermal processes associated with a recent phase of sea£oorspreading at the northern Cleft segment: Juan de FucaRidge. J. Geophys. Res. 99, 4741^4760.

Fagents, S.A., Wilson, L., 1993. Explosive volcanic eruptions ^VII. The ranges of pyroclasts ejected in transient volcanicexplosions. Geophys. J. Int. 113, 359^370.

Fisher, R.V., Schmincke, H.-U., 1984. Pyroclastic Rocks.Springer, Berlin.

Fiske, R.S., Cashman, K.V., Shibatas, A., Watanabe, K.,1998. Tephra dispersal from Myojinsho, Japan, during itsshallow submarine eruption of 1952^1953. Bull. Volcanol.59, 262^275.

Fiske, R.S., Naka, J., Iizasa, K., Yuasa, M., Klaus, A., 2001.Submarine silicic caldera at the front of the Izu^Bonin arc,Japan: Voluminous sea£oor eruptions of rhyolite pumice.Geol. Soc. Am. Bull. 113, 813^824.

Fornari, D.J., Per¢t, M.R., Allan, J.F., Batiza, R., 1988.Small-scale heterogeneities in depleted mantle source:Near-ridge seamount lava geochemistry and implications

for mid-ocean-ridge magmatic processes. Nature 331, 511^513.

Fouquet, Y., Eissen, J.-P., Ondreas, H., Barriga, F., Batiza,R., Danyuschevsky, L., 1998. Extensive volcaniclastic depos-its at the Mid-Atlantic Ridge axis: Results of deep-waterbasaltic explosive volcanic activity. Terra Nova 10, 280^286.

Fox, C.G., Radford, W.E., Dziak, R.P., Lau, T.-K., Matsu-moto, H., Schreiner, A.E., 1995. Acoustic detection of asea£oor spreading episode on the Juan de Fuca Ridge usingmilitary hydrophone arrays. Geophys. Res. Lett. 101, 281^295.

Gerlach, T.M., 1991. Midocean ridge popping rocks ^ Impli-cations for degassing at ridge crests ^ Comment. Earth Plan-et. Sci. Lett. 105, 566^567.

Gill, J., Torssander, P., Lapierre, H., Taylor, R., Kaiho, K.,Koyama, M., Kusakabe, M., Aitchison, J., Cisowski, S.,Dadey, K., Fujioka, K., Klaus, A., Lovell, M., Marsaglia,K., Pezard, P., Taylor, B., Tazaki, K., 1990. Explosive deepwater basalt in the Sumisu backarc rift. Science 248, 1214^1217.

Greeley, R., Fagents, S.A., 2001. Icelandic pseudocraters asanalogs to some volcanic cones on Mars. J. Geophys. Res.106, 20527^20546.

Gregg, T.K.P., Fink, J.H., 1995. Quanti¢cation of submarinelava-£ow morphology through analog experiments. Geology23, 73^76.

Harris, D.M., 1981. The concentration of CO2 in submarinetholeiitic basalts. J. Geol., 89, 689^701.

Head, J.W., Wilson, L., 1979. Alphonsus-type dark-halo cra-ters: Morphology, morphometry and eruption conditions.Proc. Lunar Planet. Sci. Conf. 10, 2861^2897.

Head, J.W., Wilson, L., 1986. Volcanic processes and land-forms on Venus: Theory, predictions, and observations.J. Geophys. Res. 91, 9407^9446.

Head, J.W., Wilson, L., 1987. Lava fountain heights at PuPuOPo, Kilauea, Hawaii: Indicators of amount and variationsof exsolved magma volatiles. J. Geophys. Res. 92, 13715^13719.

Head, J.W., Wilson, L., 1989. Basaltic explosive eruptions:In£uence of gas-release patterns and volume £uxes onnear-vent dynamic morphology, and the formation of sub-sequent pyroclastic deposits (cinder cones, spatter cones,rootless £ows, lava ponds, and lava £ows). J. Volcanol.Geotherm. Res. 37, 261^271.

Head, J.W., Wilson, L., 1992. Magma reservoirs and neutralbuoyancy zones on Venus: Implications for the formationand evolution of volcanic landforms. J. Geophys. Res. 97,3877^3903.

Head, J.W., Wilson, L., 2002. Mars: Geologic setting of mag-ma/H2O interactions. In: Smellie, J.E., Chapman, M.G.(Eds.), Volcano-Ice Interactions on Earth and Mars. Geo-logical Society of London Spec. Publ., London, in press.

Head, J.W., Wilson, L., Smith, D.K., 1996. Mid-ocean ridgeeruptive vent morphology and structure: Evidence for dikewidths, eruption rates, and evolution of eruptions and axialvolcanic ridges. J. Geophys. Res. 101, 28265^28280.

Heikinian, R., Bideau, D., Sto¡ers, P., Cheminee, J., Muhe,

VOLGEO 2544 30-1-03


R., Puteanus, D., Binard, N., 1991. Submarine intraplatevolcanism in the South Paci¢c: Geologic setting and petrol-ogy of the Society and Austral region. J. Geophys. Res. 96,2109^2138.

Hon, K., Heliker, C.C., Kjargaard, J.I., 1988. Limu o Pele: Anew kind of hydroclastic tephra from Kilauea Volcano, Ha-waiPi. Geol. Soc. Am. Abst. Prog. 20, A112^A113.

Hunns, S.R., McPhie, J., 1999. Pumiceous peperite in a sub-marine volcanic succession at Mount Chalmers, Queensland,Australia. J. Volcanol. Geotherm. Res. 88, 239^254.

Javoy, M., Pineau, F., 1991. The volatiles reecord of a ‘pop-ping’ rock from the Mid-Atlantic Ridge at 14‡N: Chemicaland isotopic composition of gas trapped in the vesicles.Earth Planet. Sci. Letts. 107, 598^611.

Kano, K., 1998. A shallow-marine alkali-basalt tu¡ cone in theMiddle Miocene Jinzai Formation, Izumo, SW Japan.J. Volcanol. Geotherm. Res. 87, 173^191.

Keating, B., Fryer, P., Batiza, R., Boehlert, W. (Editors), 1987.Seamounts, Islands and Atolls. Geophysical Monograph 43,Amer. Geophys. Union, Washington, DC.

Kent, A.J., Clague, D.A., Honda, M., Stolper, E.M., Hutch-eon, I.D., Norman, M.D., 1999. Widespread assimilation ofa seawater-derived component at Loihi Seamount, Hawaii.Geochim. Cosmochim. Acta 63, 2749^2761.

Keszthelyi, L.P., McEwen, A.S., Thordarson, T., 2000. Terres-trial analogs and thermal models for martian £ood lavas.J. Geophys. Res. 105, 15027^15049.

Kokelaar, P., 1986. Magma^water interactions in subaqueousand emergent basaltic volcanics. Bull. Volcanol. 48, 275^290.

Kokelaar, P., Busby, C., 1992. Subaqueous explosive eruptionand welding of pyroclastic deposits. Science 257, 196^201.

Lonsdale, P., Batiza, R., 1980. Hyaloclastite and lava £ows onyoung seamounts examined with a submersible. Geol. Soc.Am. Bull. 91, 545^554.

Macdonald, G.A., 1967. Basalts: The Poldervaart Treatise onRocks of Basaltic Composition. In: Hess, H.H., Polder-vaart, A. (Eds.), Wiley, New York, pp. 1^482.

Macdonald, K.C., 1998. Linkages between faulting, volcanism,hydrothermal activity and segmentation on fast spreadingcenters, In: Buck, W.R., Delaney, P.T., Karson, J.A., La-gabrielle, Y.(Eds.), Faulting and Magmatism at Mid-OceanRidges. Geophysical Monograph 106, American Geophysi-cal Union, Washington, DC, pp. 27^58.

Maicher, D., White, J.D.L., 2001. The formation of deep-sealimu o Pele. Bull. Volcanol. 63, 482^496.

Maicher, D., White, J.D.L., Batiza, R., 2000. Sheet hyaloclas-tite: Density current deposits of quench and bubble-burstfragments from thin glassy sheet lava £ows, Seamount 6,Eastern Paci¢c Ocean. Mar. Geol. 171, 75^94.

Mangan, M.T., Cashman, K.V., Newman, S., 1995. Vesicula-tion of basaltic magma during eruption. Geology 31, 157^160.

Oshima, S., Tsuchide, M., Kato, S., Watanabe, K., Kudo, K.,Ossaka, J., 1991. Birth of a submarine volcano Teisi Knoll.J. Phys. Earth 39, 1^19.

Par¢tt, E.A., Wilson, L., 1995. Explosive volcanic eruptions ^

IX. The transition between Hawaiian-style lava fountainingand Strombolian explosive activity. Geophys. J. Int. 121,226^232.

Par¢tt, E.A., Wilson, L., 1999. A Plinian treatment of falloutfrom Hawaiian lava fountains. J. Volcanol. Geotherm. Res.88, 67^75.

Par¢tt, E.A., Wilson, L., Head, J.W., 1993. Basaltic magmareservoirs: Factors controlling their rupture characteristicsand evolution. J. Volcanol. Geotherm. Res. 55, 1^14.

Pavri, B., Head, J.W., Klose, K.B., Wilson, L., 1992. Steep-sided domes on Venus: Characteristics, geologic setting, anderuption conditions from Magellan data. J. Geophys. Res.97, 13445^13478.

Per¢t, M.R., Chadwick, Jr., W.W., 1998. Magmatism at mid-ocean ridges: Constraints from volcanological and geochem-ical investigations. In: Buck, W.R. Delaney, P.T., Karson,J.A., Lagabrielle, Y. (Eds.), Faulting and Magmatism atMid-Ocean Ridges. Geophysical Monograph 106, AmericanGeophysical Union, Washington, DC, pp. 59^115.

Pineau, F., Javoy, M., 1994. Strong degassing at ridge crests:The behaviour of dissolved carbon and water in basaltglasses at 14‡N, Mid-Atlantic Ridge. Earth Planet. Sci.Letts. 123, 179^198.

Prandtl, L., 1949. The Essentials of Fluid Dynamics. Blackie,Edinburgh.

Rittmann, A., 1960. Vulkane und ihre Tatigkeit. Enke, Stutt-gart, 335 pp.

Ryan, M.P., Blevins, J.Y.K., Okamura, A.T., Koyanagi, R.Y.,1983. Magma reservoir subsidence mechanics: Theoreticalsummary and application to Kilauea Volcano, Hawaii.J. Geophys. Res. 88, 4147^4181.

Schmidt, R., Schmincke, H.-U., 2000. Seamounts and islandbuilding. In: Sigurdsson, H., Houghton, B., McNutt, S.R.,Rymer, H., Stix, J. (Eds.), Encyclopedia of Volcanoes. Aca-demic Press, San Diego, CA, pp. 383^402.

Schmincke, H.-U., Robinson, P.T., Ohnmacht, W., Flower, M.F.J., 1978. Basaltic hyaloclastites from hole 396B, DSDPLeg 46. In: Dmitriev, L., Heirtzler, J. et al. (Eds.), InitialReports of the Deep Sea Drilling Project 46, US Govern-ment Printing O⁄ce, Washington, DC, pp. 341^348.

Smith, T.L., Batiza, R., 1989. New ¢eld and laboratory evi-dence for the origin of hyaloclastite £ows on seamount sum-mits. Bull. Volcanol. 51, 96^114.

Sparks, R.S.J., 1978. The dynamics of bubble formation andgrowth in magmas: A review and analysis. J. Volcanol. Geo-therm. Res. 3, 1^37.

Squyres, S.W., Wilhelms, D.E., Moosman, A.C., 1987. Large-scale volcano-ground ice interaction on Mars. Icarus 70,385^408.

Thorarinsson, S., 1953. The crater groups in Iceland. Bull.Volcanol. 14, 3^44.

UKCPS (United Kingdom Committee on the Properties ofSteam), 1970. UK Steam Tables 1970. Edward Arnold, Lon-don.

Vergniolle, S., Jaupart, C., 1990. Dynamics of degassing atKilauea Volcano, Hawaii. J. Geophys. Res. 95, 2793^2809.

Wallace, P., Anderson, A.T., 2000. Volatiles in magmas. In:

VOLGEO 2544 30-1-03


Sigurdsson, H., Houghton, B., McNutt, S.R., Rymer, H.,Stix, J. (Eds.), Encyclopedia of Volcanoes. Academic Press,San Diego, CA, pp. 149^170.

Wessell, P., Lyons, S., 1997. Distribution of large paci¢c sea-mounts from Geosat/ERS-1: Implications for the history ofintraplate volcanism. J. Geophys. Res. 102, 22459^22475.

White, J.D.L., 1996. Pre-emergent construction of a lacustrinebasaltic volcano, Pahvant Butte, Utah (USA). Bull. Volca-nol. 58, 249^262.

White, J.D.L., 2000. Subaqueous eruption-fed density currentsand their deposits. Precambr. Res. 26, 87^109.

Wilson, L., 1980. Relationships between pressure, volatile con-tent and ejecta velocity in three types of volcanic explosion.J. Volcanol. Geotherm. Res. 8, 297^313.

Wilson, L., Head, J.W., 1981. Ascent and eruption of basalticmagma on the Earth and Moon. J. Geophys. Res. 86, 2971^3001.

Wilson, L., Head, J.W., 1988. Nature of local magma storagezones and geometry of conduit systems below basaltic erup-tion sites: PuPu POPo, Kilauea East Rift, Hawaii, example.J. Geophys. Res. 93, 14785^14792.

Wilson, L., Head, J.W., 1994. Mars: Review and analysis ofvolcanic eruption theory and relationships to observed land-forms. Rev. Geophys. 32, 221^263.

Wilson, L., Heslop, S.E., 1990. Clast sizes in terrestrial andmartian ignimbrite lag deposits. J. Geophys. Res. 95, 17309^17314.

Wohletz, K.H., 1983. Mechanisms of hydrovolcanic pyroclastformation: Grain-size, scanning electron microscopy, andexperimental results. J. Volcanol. Geotherm. Res. 17, 31^63.

Wohletz, K.H., 1986. Explosive magma^water interactions:Thermodynamics, explosion mechanisms, and ¢eld studies.Bull. Volcanol. 48, 245^264.

Wohletz, K.H., McQueen, R.G., 1984. Experimental studies ofhydromagmatic volcanism. In: Explosive Volcanism: Incep-tion, Evolution, and Hazards, Studies in Geophysics. Na-tional Academy Press, Washington, DC, pp. 158^169.

Wolfe, E.W., Neal, C.A., Banks, N.G., Duggan, T.J., 1988.Geologic observations and chronology of eruptive events.In: Wolfe, E.W. (Ed.), The Puu Oo Eruption of KilaueaVolcano, Hawaii: Episodes 1 through 20, January 3, 1983,through June 8, 1984. US Geological Survey ProfessionalPaper 1463, US Geological Survey, Reston, VA.

Wright, I.C., 1996. Volcaniclastic processes on modern subma-rine arc stratovolcanoes: Sidescan and photographic evi-dence from the Rumble IV and V volcanoes, southern Ker-madec Arc (SW Paci¢c). Mar. Geol. 136, 21^39.

Wright, I.C., 1999. Southern Kermadec submarine caldera arcvolcanoes (SW Paci¢c): Caldera formation by e¡usive andpyroclastic eruption. Mar. Geol. 161, 207^227.

Zimanowski, B., Frohlich, G., Lorenz, V., 1991. Quantitativeexperiments on phreatomagmatic explosions. J. Volcanol.Geotherm. Res. 48, 341^358.

VOLGEO 2544 30-1-03


deep submarine pyroclastic eruptions: theory and predicted

Documents