Ž .Journal of Volcanology and Geothermal Research 90 1999 263–280www.elsevier.comrlocatervolgeores

Lava tube morphology on Etna and evidence for lava flowemplacement mechanisms

Sonia Calvari a,b,), Harry Pinkerton b

a Istituto Internazionale di Vulcanologia, Piazza Roma 2, 95123 Catania, Italyb EnÕironmental Science Department, Lancaster UniÕersity, Lancaster LA1 4YQ, UK

Received 19 December 1998; accepted 29 January 1999

Abstract

Lava tubes play a pivotal role in the formation of many lava flow fields. A detailed examination of several compound‘a‘a lava flow fields on Etna confirmed that a complex network of tubes forms at successively higher levels within the flowfield, and that tubes generally advance by processes that include flow inflation and tube coalescence. Flow inflation iscommonly followed by the formation of major, first-order ephemeral vents which, in turn, form an arterial tube network.Tube coalescence occurs when lava breaks through the roof or wall of an older lava tube; this can result in the unexpectedappearance of vents several kilometers downstream. A close examination of underground features allowed us to distinguishbetween ephemeral vent formation and tube coalescence, both of which are responsible for abrupt changes in level or flowdirection of lava within tubes on Etna. Ephemeral vent formation on the surface is frequently recorded underground by amarked increase in size of the tube immediately upstream of these vents. When the lining of an inflated tube has collapsed,‘a‘a clinker is commonly seen in the roof and walls of the tube, and this is used to infer that inflation has taken place in thedistal part of an ‘a‘a lava flow. Tube coalescence is recognised either from the compound shape of tube sections, or frombreached levees, lava falls, inclined grooves or other structures on the walls and roof. Our observations confirm theimportance of lava tubes in the evolution of extensive pahoehoe and ‘a‘a flow fields on Etna. q 1999 Elsevier Science B.V.All rights reserved.

Keywords: lava flow; lava tube; tube coalescence; flow inflation

1. Introduction

There is a widespread misconception that lavatubes form only in pahoehoe flows and at very low

Žeffusion rates Greeley, 1971, 1987; Peterson and

) Corresponding author. Istituto Internazionale di Vulcanologia,Consiglio Nazionale delle Richerche, Piazza Roma 2, 95123C atania, Italy . Fax: q 0039-095-435801; E -m ail:sonia@iiv.ct.cnr.it

Swanson, 1974; Peterson et al., 1994; Hallworth et.al., 1987 . This is disputed by Calvari and Pinkerton

Ž .1998 who describe a complex system of lava tubeson the 1991–1993 ‘a‘a lava flow field on Etna wheretubes formed at effusion rates spanning more than

Ž .two orders of magnitude Calvari et al., 1994a . Thiseruption was not unique. Many descriptions of lavatubes on ‘a‘a flows on Etna have been published

Žduring the past three centuries Anonymous, 1669;Recupero, 1815; Lyell, 1858; Cucuzza Silvestri,

0377-0273r99r$ - see front matter q 1999 Elsevier Science B.V. All rights reserved.Ž .PII: S0377-0273 99 00024-4

( )S. CalÕari, H. PinkertonrJournal of Volcanology and Geothermal Research 90 1999 263–280264

. Ž .1977 . Greeley 1971 concluded that 18% of theflows on Etna have been emplaced at least partiallythrough lava tubes. However, more recent studieshave shown that this may be an underestimate.

Ž .Brunelli and Scammacca 1975 compiled a compre-hensive catalogue of all known lava caves in Sicily,including many in ‘a‘a flows. More recently, Pinker-

Ž . Ž .ton and Sparks 1976 ; Guest et al. 1980 ; FrazzettaŽ . Ž .and Romano 1984 and Calvari and Pinkerton 1998

have discussed the importance of lava tube develop-ment in ‘a‘a lava flows on Etna. Finally, many tubeshave been discovered and surveyed during the pastthree decades by the Gruppo Speleologico Etneo.Thus, lava tubes play an important role in the forma-tion of many lava flows on Mount Etna, and many ofthese lava flows are ‘a‘a.

While single ‘a‘a flow units on Etna are typically1 to 10 m thick, the resulting flow fields are com-

Žmonly 10 to 50 m thick Romano and Sturiale,. Ž .1982 , and at least one 1991–1993 attained a maxi-

Žmum thickness of 100 m Calvari et al., 1994a;.Stevens et al., 1997 . Occasionally, and for a short

time before a new vent opens from its margins, theflow front of a single inflated flow can attain a

Ž .thickness of 20–30 m Calvari and Pinkerton, 1998 .Tubes that form inside arterial lava flows during thefirst few weeks of an eruption can easily be relatedto their parent channels. During long-lasting erup-tions, however, later flows cover features related topreviously formed tubes, and they become very diffi-cult to detect. Only when the tube is obstructed, orpressure increases cause breakouts along the roof or

Žsides of the tube, does its path become visible Mat-.tox et al., 1993; Calvari and Pinkerton, 1998 . Such

breakouts can remain localised at changes in slope,producing tumuli. Post-eruption or post-drainage col-lapse of part of the roof can then form useful en-trances to the tube system.

Because of the compound nature of many flowfields on Etna, the flows in which many tubes formare covered by subsequent flows, and there may belittle relationship between the slope and direction ofsurface flows and the underlying tubes. This is inagreement with observations on pahoehoe flow fields

Ž .in Hawaii Peterson et al., 1994 .While our observations confirm the importance of

tube formation in both pahoehoe and ‘a‘a lava flowson Etna, our observations suggest that at least some

of the tubes that are considered to have formed onpahoehoe formed in large ‘a‘a lava flows. This com-plication arises because the type of lava that finallyforms on a flow field or along its margins may differfrom the type of lava within which the tube systemdeveloped. We know, for example, that pahoehoeand toothpaste morphologies on Etna are common onflows that develop at the margins of ‘a‘a lava flows

Žthrough third-order ephemeral vents Calvari and.Pinkerton, 1998 with very low discharge rates

Ž .Pinkerton and Sparks, 1976; Calvari et al., 1994a .These small flows form during drainage of primary‘a‘a lava, and are typical of the later stages of longduration eruptions on Etna. Thus flow fields thathave been active for years, and whose actual shapewas formerly controlled by the emplacement of large

Ž‘a‘a lava flows Wadge, 1978; Kilburn, 1989; Kil-.burn and Lopes, 1991 , may eventually have a sur-

face of mostly pahoehoe and toothpaste lava. Forexample the pahoehoe and toothpaste morphologiescommon in the proximal part of the 1991–1993 flowfield contrast with the predominantly ‘a‘a flows thatproduced extensive systems of lava tubes during

Žmost of this eruption Calvari and Pinkerton, 1998;.Calvari et al., 1994a . We conclude that some of the

lava tubes previously considered to form in pahoe-hoe lava flows on Etna and elsewhere may haveformed initially in ‘a‘a lavas which have subse-quently been covered by late-stage pahoehoe lava.This will be investigated further during future field-work on Etna.

In view of the importance of lava tubes during the1991–1993 and earlier eruptions on Mount Etna, wehave undertaken a systematic study of a number oflava tubes. These studies confirm the complexity ofunderground structures and lead to an improved un-derstanding of how tubes control the development ofmany long-lived flow fields on this volcano.

2. Lava tube morphology

During August and September 1997, September1998, and October 1998, we examined 16 lava tubes

Ž .on Etna Fig. 1 with the assistance of members ofGruppo Speleologico Etneo. Our observations con-firm that underground observations are an essentialcomponent of research into emplacement mecha-

( )S. CalÕari, H. PinkertonrJournal of Volcanology and Geothermal Research 90 1999 263–280 265

Fig. 1. Map of Mount Etna and location of the lava tubes described in the paper. 1sMicio Conti tube, prehistoric lava flows; 2sTreLivelli-KTM tube system, 1792–1793 eruption; 3sGrotta Cassone, 1792–1793 eruption; 4sLa Fenice tube, 1792–1793 eruption; 5sand 6sCutrona and Salto della Giumenta tubes, 1991–1993 eruption; 7sGallobianco tube, 1050–1100 eruption; 8s Intraleo tube, 1225eruption; 9s1985 proximal tube system.

nisms of lava flows on Etna. They have led to agreater appreciation of the complex processes thattake place during the emplacement of long-lived ‘a‘alava flows, and they help to confirm the importanceof many of the processes that we described duringour earlier work on the 1991–1993 flow field on

Ž .Etna Calvari and Pinkerton, 1998 . In the followingaccount, we describe structures within selected lavatubes on Etna, and we show how studies of theirinternal dimensions, morphology and field relation-

ships can provide important information on themechanisms of formation and propagation of tubesand their parent lava flows.

2.1. Longitudinal sections of laÕa tubes

The most important feature of many longitudinalprofiles of lava tubes is the alternation of very widechambers and narrow, low passages. We have ob-served this situation in many lava tubes on Etna and

( )S. CalÕari, H. PinkertonrJournal of Volcanology and Geothermal Research 90 1999 263–280266

note that this is a common feature in lava tubesŽelsewhere Ollier and Brown, 1965; Atkinson et al.,

.1975 . Often constrictions are so pronounced thatthey produce an insurmountable blockage within thetube, although the tube may continue downslope andperhaps be accessible through a different entrance.Our underground observations lead to the followingconclusions.

Large hemispherical chambers within a tube sys-tem commonly represent different stages of inflationof a flow front. The small, narrow passages that areencountered at the distal end of some of these cham-

Žbers are first-order ephemeral vents Calvari and.Pinkerton, 1998 through which the tube system

propagated downslope. When the inflationary stagelasted long enough to develop a stable crust, theresulting chamber was hemispherical. If inflationcontinued and propagated upslope before anephemeral vent formed, the chamber assumed a moreelongated shape. If drainage through a first-orderephemeral vent took place before the crust of theinflated flow became self-supporting, the roof col-lapsed and blocked further expansion of the tube.Occasionally, partial collapse formed a downwardbulge on the tube roof.

Longitudinal profiles of lava tubes can also becompound, where two or more tubes coalesce. Insuch cases, tube size typically decreases verticallytowards the upper part of the compound tube. Thiscan arise either from a decrease in effusion rate

Ž .during the eruption Wadge, 1981 , or because suc-Žcessive flows in any region tend to be smaller be-

.cause there are more of them , and hence havereduced discharge rates.

While many lava tubes have similar mean diame-ters for several hundred metres, some gradually in-crease in size with increasing distance whereas oth-ers decrease. There is no evidence to suggest thatgradual increases in mean diameter are a conse-quence of inflation. Instead they are inferred to arisefrom reduced flow velocities because of the com-bined effects of increased apparent viscosities andreduced topographic gradients. In order to maintainthe same discharge rate under those conditions, themean diameter of a tube will need to increase down-flow. However, this increase can be offset by the

Ž .tendency for flows and hence tubes to split withincreasing distance from source. Each splitting event

will reduce the mean diameter. It is therefore impor-tant to recognise that not all changes in mean diame-ter are induced by inflation and tube coalescence. Asan example of increasing mean diameter with in-creasing distance from the source, the 1792–1793Tre Livelli tube close to the eruptive fissure has amean diameter of 1.5 m and a mean slope of 208

Ž .Corsaro et al., 1990 . By contrast, the Grotta Cas-sone, a tube that formed in the same flow field at adistance of 2.5 km from the main vent, has a mean

Ždiameter of 3 m and a mean gradient of 58 Balsamo. Ž .et al., 1994 . If we assume 1 a constant discharge

Ž .rate between these tubes; and 2 that Jeffrey’s equa-tion can be used to relate flow velocities to tubedimensions and viscosity, we can readily show thath4 sin arh is constant, where h is the thickness of aflow, a is the gradient and h is the apparent viscos-ity. For the 1792–1793 tube system, under condi-tions of full tube flow, discharge rates in the proxi-mal region will be 11 m3rs for an apparent viscosityof 103 Pa s and a density of 2500 kgrm3. Tomaintain a similar discharge rate in the lower part ofthe tube system, the apparent viscosity will rise to4000 Pa s. Assuming similarities in rheological prop-erties with the 1983 Etna lava, this change can arisethrough a decrease in temperature from 10978 to

Ž .10858C Pinkerton and Norton, 1995 . While thisdecrease is considerably higher than suggested by

Žprevious workers Swanson, 1972; Calvari et al.,.1994a; Peterson et al., 1994; Keszthelyi, 1995 , it

does not take into account crystallisation during de-Ž .gassing Sparks and Pinkerton, 1978 .

2.2. Internal structures and other features in laÕatubes

During our surveys we noted the presence ofseveral features that help to unravel the flow andthermal history of lava tubes. These include single ormultiples layers of lava on the outer walls and roofof many lava tubes, different types of stalactites,lateral lava benches, peel-off and rolling-over struc-tures, longitudinal and transversal cracks, and differ-ent surface morphologies on the floor of tubes.

2.2.1. Lining on the walls and roofs of laÕa tubesAll lava tubes that we have studied are lined with

one or more layers of lava. These layers preserve a

( )S. CalÕari, H. PinkertonrJournal of Volcanology and Geothermal Research 90 1999 263–280 267

record of periods when lava partially or completelyfilled a lava tube. Sections through these layers arevisible only when part of it collapses from the roofor walls of a lava tube. Multiple concentric layershave been observed in several cases and they com-monly show an increase in thickness both towardsthe base of the tube and downward along the flowfield. These layers have thicknesses that vary from a

Ž .few millimetres Fig. 2 to 1 m, and we concludethat their thickness is a function of flow rheology,flow duration, discharge rate, and thermal history ofthe tube. Thinner layers form when a single pulse offluid lava flows through a tube and is rapidly drained.Thicker, generally multiple, layers of lava form whenseveral surges of relatively cool lava pour down atube. The lining of lava tubes is generally black ordark grey in colour, and occasionally red, indicatingdifferent states of oxidation.

Each layer corresponds to a cycle of infilling andemptying of the tube. This cyclic process can be

Ž . Ž .caused by 1 variations in vent discharge rate; 2partial or complete blockages upflow or downflow;Ž . Ž .3 tube coalescence; or 4 cycles of flow inflationat the margins or flow fronts. A revealing example of

the latter process can be seen in La Fenice tube,where four layers of comparable thickness in theupper tube are clearly mirrored by four connected

Ž .flows Fig. 3 .

2.2.2. Lateral benchesLateral benches are a very common feature of

Žlava tubes e.g., inside the Cutrona tube, the TreLivelli tube and the tube along the 1991 eruptive

.fissure . They indicate that a stable lava level wasmaintained for long enough to allow lateral solidifi-cation. Sometimes multiple benches are observedŽ .Wood, 1974 , and these record different stages ofstable magma supply rate. The inward accretion oflateral benches can produce a complex system ofstacked tubes that can resemble several generationsof tubes formed in successively higher flows. Thisoccurrence is common along tubes that formed ineruptive fissures, where tube width is generally lessthan 2 m and flow level decreases with time. Merg-ing of lateral benches to form ‘tubes within tubes’ ismore common when peeling-off and rolling-over

Ž .structures develop. Leotta 1994 considers these

Fig. 2. Multiple linings with individual mm-thick layers at the entrance of the Cassone tube, 1792–1793 eruption. Photo by Alfio Amantia,IIV-Catania.

( )S. CalÕari, H. PinkertonrJournal of Volcanology and Geothermal Research 90 1999 263–280268

Fig. 3. Multiple layers, each 10 to 15-cm thick, along La Fenicetube, 1792–1793 flow field. Note the section of the tube indicat-ing sealing of narrow channel and the tube roof section formed byjamming of XaXa lava blocks. The tube width is ;1 m.

structures to be very common in tubes formed insideŽ .eruptive fissures e.g., the 1792 fissure where, be-

cause of their width, peeling-off and rolling-overstructures can readily coalesce and effectively fuse toform an inner tube roof.

2.2.3. Longitudinal and transÕersal cracksLongitudinal cracks along the roof of lava tubes

are a common feature, and they form as a result ofroof deformation, either as a consequence of defor-mation under the weight of the roof, or due to thecombined weight of the roof and an overlying flow.Cracks can form in the middle of the roof, producinga downward directed bulge, or they can open at thesides of the collapsing central belt. Extreme deforma-

tion of longitudinal cracks can produce pillars, some-times elongated in the flow direction with ellipsoidalsection. They are made of ‘a‘a rubble or ropy blocksfrom the flow surface, coated by lava flowing intothe tube. Such pillars are common in the Micio Conti

Žprehistoric lava tube at the outskirts of Catania Fig..4 .

Transverse cracks are less common and wherepresent they are not accompanied by inward defor-mation of a tube’s roof, and appear to be related tothermal contraction. Transverse cracks are more sur-ficial structures than longitudinal ones: they developonly in the innermost lining of tubes, and are re-stricted to places where the ground slope increases.If a tube suddenly drains and cools, thermally-in-duced contraction cracks will develop preferentiallyon the steepest slopes. Steep gradients assist in thegeneration of locally high gravitational stresses andthis may explain their higher concentrations in theseregions. The number of transverse cracks per unitlength appear to be proportional to lining thicknessand local ground slope. Two examples are the north-ern branch of the Cutrona tube, which has transverse

Žcracks every 1–2 m on a slope of 258 Giudice and.Leotta, 1995 , and the upper part of the Tre Livelli

tube, which has cracks every 0.5 m on a slope of308.

2.2.4. GrooÕesHorizontal and inclined grooves and striae are

Ž .common features on the walls of lava tubes Fig. 5 .They form when cool ‘a‘a lava blocks on the surfaceof the flow move past the plastic walls of the tube,softened by the high temperature of the lava flowinginto it, and they record the flow direction. In someplaces, grooves and striae record sudden changes inflow direction. An excellent example can be seen inthe Intraleo tube where, on both sides of a lava tube,grooves change from horizontal to almost verticalover a distance of 10 m. Because this sudden changein direction commonly takes place at the point wherethe floor of an upper tube has collapsed into a lowertube, such sudden changes in groove direction areconsidered to be indirect evidence of tube coales-cence. Different groove directions are preserved onmultiple layers in the lining of lava tubes and chan-nels. Along the 1983 flow field close to Mt. Vetore,

( )S. CalÕari, H. PinkertonrJournal of Volcanology and Geothermal Research 90 1999 263–280 269

Fig. 4. Close to the entrance of the Micio Conti lava tube, formed in prehistoric pahoehoe lava flows. The wall partially separating twoŽ . Ž .parallel chambers may have formed either by a lateral coalescence of two parallel tubes; or b by downward collapse along axial cracks of

Ž .the still plastic roof of the tube; or c by multiple vent opening along a fingering pahoehoe flow front. Photo by Alfio Amantia,I.I.V.-C.N.R.

there is a lava fall along a channel where grooveshave different inclination in each layer. In this case,

they appear to record a progressive increase in slopeof the step below the lava fall due to mechanical

Fig. 5. Horizontal grooves above the roadside entrance to the Tre Livelli tube, 1792–1793 flow field.

( )S. CalÕari, H. PinkertonrJournal of Volcanology and Geothermal Research 90 1999 263–280270

erosion triggered by turbulence.

2.2.5. LaÕa stalactitesWhile some stalactites form from mineral precipi-

Žtates at the end of eruptions Giudice and Leotta,.1995 , the majority form in response to thermal and

mechanical processes during flow and subsequentdrainage. Stalactites on Etna have shapes and sizesthat differ significantly from some of their Hawaiiancounterparts. For example, the delicate, worm-likestructures found on the roofs of lava tubes on HawaiiŽ .Jaggar, 1931 have not been observed on Etna.

On Etna four kinds of stalactites can be distin-guished, and there are excellent examples on thewalls and roof of the Tre Livelli and Cassone tubes.Stalactites with very smooth surfaces form on ridges

Ž .that are elongated in the flow direction Fig. 6 .These stalactites are typically red in colour, and theyare considered to form by remelting by gases accu-

Žmulating below the roof Jaggar, 1931; Kauahikaua.et al., 1998 . On Etna they are typically up to a few

centimetres long and at most 2 cm wide at the base,

and they are conical in shape. Another type is rough,grey in colour and spiky, and it is usually found inconstrictions of the tube. This second group of sta-

Ž .lactites was also recognised by Jaggar 1931 . Theyare generally less than 0.5 cm wide, a few centime-tres long, and they are considered to have formedwhere lava completely filled a tube and then drained,either partially or completely. The resulting stalac-tites record the dribbling of lava from the roof. Thespiny nature of these stalactites is due to the pres-

Ž .ence of crystals mainly plagioclase and smallamount of interstitial glass. The third group of stalac-tites are morphologically similar, and they form whenpart of the roof or wall lining drops or rolls off,

Ž .leaving rough ‘pull-apart’ stalactites Fig. 7 . Thefinal type of stalactite, which is not very common, ischaracterised by bulbous shapes. Thin sections revealthat they are composed of multiple layers with anexternal boundary marked by a very thin film ofoxides. We interpreted these as stalactites that havebeen repeatedly coated by lava. Sometimes stalac-tites have an unusual shape due to welding, on thelower tip, of small and delicate ‘a‘a scoria. This is

Ž . Ž .Fig. 6. Smooth lava stalactites right formed by melting of ridges parallel to the flow direction on the roof of the 1792–1793 Cassonetube, close to the entrance.

( )S. CalÕari, H. PinkertonrJournal of Volcanology and Geothermal Research 90 1999 263–280 271

clear evidence that the tube was, at one time, filledwith ‘a‘a lava. Beautiful examples of these structuresare exposed along the upper part of the Tre Livellitube.

2.3. Master tubes and laÕa tube networks

Our observations confirm that many lava tubes onEtna form complex three-dimensional braided net-works. ‘Master tubes’ tend to form in zones of highground slope, in narrow valleys, and along eruptivefissures. The term ‘master tube’ refers to an arterialtube that delivers lava to the network of tube dis-tributaries throughout most of the flow field. Therelationship between the location of master tubes andground slope or surface morphology has been ob-served in other volcanoes, for example in IcelandŽ . Ž .Wood, 1971 , Mexico Keszthelyi and Pieri, 1993

Žand Australia Atkinson and Atkinson, 1995; Atkin-.son et al., 1975 . Where ground slope decreases,

Žtubes generally split into smaller branches Cavallaro.et al., 1985 . In this situation they are difficult to

detect, unless a higher gradient improves drainageŽ .Wood, 1975 .

Master tubes form in cooling-limited lava flowsŽ .Guest et al., 1987 when a steady supply allows theflow front to inflate, and where topography permits anumber of first-order ephemeral vents to open from

Žsuccessive front regions of connected flows Calvari.and Pinkerton, 1998 . The effect of these vents is to

drain the upslope portion of the early-formed tubesector.

If the inflated frontal zone has a self-supportingroof, a lava tube forms within this region. Two othermechanisms discussed by Calvari and PinkertonŽ . Ž .1998 occur a along narrow channels, where aslightly fluctuating effusion rate allows levee growth,´eventually leading to complete roofing-over of the

Ž .channel, and b along wide channels, where coolingŽrate and crustal growth play the main role Calvari

.and Pinkerton, 1998 . The latter two mechanismshave been found to be less efficient than the first.During the initial phase of the 1991–1993 eruption,at similar effusion rates, it took 15 days to form atube along a narrow channel, and 1 month along awide channel, whereas tubes developed in inflated

Žfrontal zones in 1 week Calvari and Pinkerton,.1998 .

Ž .Fig. 7. Pull-apart stalactites arrows formed in the 1792–1793 Cassone tube by detachment of an inner lining. The ruler is 25-cm long.Photo by Alfio Amantia, IIV-Catania.

( )S. CalÕari, H. PinkertonrJournal of Volcanology and Geothermal Research 90 1999 263–280272

2.4. Tube coalescence

Mature tube systems typically form complexthree-dimensional networks, and upper tubes canremain separated from lower, previously formed,tube systems. Our observations suggest that the sta-bility of the intervening crust depends on its tensilestrength, thickness, and on the pressure exerted bythe flow above it. Occasionally, an upper tube canbreak through the roof of a lower tube. This canresult in the reactivation of the lower tube systembecause the increased pressure at the snout allowsbreakouts at the front, roof or margins of the previ-

Žously inactive lower system Peterson and Swanson,.1974; Mattox et al., 1993 .

Tube coalescence generally occurs in a verticaldirection. Two classic examples are seen in the 1985

Ž .and 1792–1793 proximal lava tubes Fig. 1 , whererespectively four and three levels of tubes can beobserved. The Zafferana-Rifugio Sapienza road cutsthrough the 1792 Tre Livelli tube and reveals asection through the upper two tubes of the complexthree-level master tube. The uppermost tube did notmerge with the middle one because it was small, ithad not inflated, and it appears to have been em-

placed at a late stage of the eruption, giving suffi-cient time for the upper crust of the middle tube tobecome self-supporting. The upper tube has a stablebase 0.5 m thick, is about 1 m wide, and the finalflow that occupied the tube had a ropy surface.

Our observations indicate that vertical coales-cence generally takes place when the direct stressapplied by lava within a new upper tube causestensile failure of the roof of a lower tube. In Hawaii,tube coalescence was found to be more common in

Ž .near-vent areas Peterson and Swanson, 1974 , wherethe crust had insufficient time to cool, thicken, andbecome self-supporting before being covered by anew flow. We observed the same on Etna, wheretubes that formed at less than 2 km distance from theeruptive fissure produce complex three-dimensionalnetworks. Horizontal coalescence also takes place, asfor example in the Micio Conti and Gallobianco I

Ž . Žlava tubes on Etna Figs. 1, 4 and 8 Puglisi, 1981;.Fanciulli et al., 1989 .

Along the large 1991–1993 flow field, tube coa-Žlescence was common on the Cutrona tube Giudice

.and Leotta, 1995 at about 2 km from the main vent,and also on the Salto della Giumenta tube thatformed in Val Calanna, 5 km from the source

Ž .Fig. 8. Plan view of Gallobianco I lava tube after Fanciulli et al. 1989 . The kink close to Section 2 is due to lateral coalescence.

( )S. CalÕari, H. PinkertonrJournal of Volcanology and Geothermal Research 90 1999 263–280 273

Ž .Fig. 1 . The septum between superimposed tubes istypically in excess of 0.5 m on Etna. The minimumthickness required to prevent roof collapse is clearlya function of the relative size of superimposed tubesand the intervening crustal thickness. It is also afunction of distance from the source because of theincrease in thickness and strength of the upper crustduring flow.

2.5. A classification system for Etna laÕa tubes

Our underground observations suggest that, onEtna, lava tubes can be split into two main morpho-

Ž .logical types: simple and compound Fig. 9 . Asimple lava tube forms by roofing over of a single

Ž .lava channel Fig. 9a or by cooling of a stable crustŽ .around an inflated lava flow Fig. 9b . A simple lava

tube can be either symmetric or asymmetric. Thedegree of asymmetry is commonly related to theflow direction of the tube. If a tube is straight, it is

Ž .usually symmetric Fig. 9a–b . Asymmetry developsŽ .when a tube changes direction Fig. 9c . This asym-

metry is a consequence of greater deposition of lavaon the more slowly moving inner side of the bendcompared with the outer side. Asymmetry in simpletubes can also arise from differential loading of the

Ž .tube roof Fig. 9d , accumulation of lava flows onone side of the roof, opening of ephemeral vents onthe walls of the tube, or secondary collapse events.

Compound lava tubes form by coalescence ofadjacent tubes. In this way we apply to tubes the

Ž .same distinction suggested by Walker 1971 forlava flows. Compound tubes have symmetric trans-verse sections when tube coalescence is vertical

Ž . Ž . Ž . Ž .Fig. 9. Simple a to d and compound e–f transverse sections of lava tubes. a–b Simple and symmetric sections; c simple, asymmetricŽ . Ž .section due to bending of the tube; d simple asymmetric section due to roof loading; e symmetric, compound tube section with typical

Ž .key-shape, indicating coaxial capture; f asymmetric compound tube section.

( )S. CalÕari, H. PinkertonrJournal of Volcanology and Geothermal Research 90 1999 263–280274

Ž .Fig. 9e , and asymmetric sections when the uppertube was not directly above the lower one when the

Ž .intervening roof collapsed Fig. 9f . In addition tothese end-members, we have observed many exam-ples where multiple processes have operated, result-ing in symmetric simple or symmetric compoundlava tubes, and asymmetric simple or asymmetriccompound lava tubes. An examination of transversesections of primary lava tubes and of their modifica-tion with distance from the source is essential inunderstanding the mechanism of flow development.

Ž .For example, longitudinal downward bulges Fig. 6 ,often accompanied by medial cracks, are a character-istic feature of many lava tubes.

3. The role of mechanical and thermal erosion onEtna

Thermal and mechanical erosion are considered toplay an important role in the development of komati-

Žite lava flows Greeley and Hyde, 1972; Huppert andSparks, 1985; Huppert et al., 1984; Williams and

.Lesher, 1996 and in basaltic lavas and tubes inŽHawaii Peterson and Swanson, 1974; Kauahikaua et. Žal., 1998 and Etna Cumin, 1954; Greeley et al.,

.1998 . On Hawaii, combined rates of thermal andmechanical erosion between 5 and 10 cmrday have

Ž .been measured by Kauahikaua et al. 1998 .On Etna, apparent erosion has been reported for

only two tubes. The first was in a channel on ‘a‘alava that flowed over the December 15, 1991 flow

Žon the southern margin of Valle del Bove Calvari et.al., 1994a . This channel was located at the base of

Serra Pirciata, 1 km from the main vent. On January3, 1992 the lava stream was flowing inside a well-defined, 10-m wide channel with vertical walls and aslope of 118. Because the levees were 5 m high, and´the level of the flow inside the channel was 2.5 mbelow the channel rim, the flow thickness was esti-mated to be 2.5 m for a length of about 100 m, andthe measured mean flow velocity of 0.4 m sy1 re-

3 y1 Žsulted in a discharge rate of 8 to 10 m s Table 2.in Calvari et al., 1994a . Measurements were re-

peated at the same place on January 5 and 7, givingflow rates of 18.8 and 18.8 to 25 m3 sy1, respec-

Ž .tively Calvari et al., 1994a . On January 7, 1992 theupper level of lava in the lower half of the channelwas over 3 m below the channel rim. The upper

surface which, on January 3, was gently sloping, onJanuary 7 had a step with nearly vertical lava cas-cades that produced an increase in the surface flowvelocity upstream and may have induced local turbu-

Žlence at the base of the step Kauahikaua et al.,. Ž .1998 . Bonforte 1994 concluded that this increase

in flow depth resulted from combined thermal andmechanical erosion of the lava flowing on a substra-tum made of debris and reworked pyroclastics and

Žlavas from previous eruptive centres Calvari et al.,.1994b . However, our detailed underground observa-

tions support an alternative explanation.As noted above, our surveys in old lava tubes on

Etna have revealed that tube coalescence is verycommon. The abrupt changes in the direction ofgrooves on the walls of some lava tubes suggestsrapid changes in the flow direction of lava withinthese tubes. Good examples are the Cutrona tubeŽ .formed inside the June 2, 1992 lava flow , the

Ž .Intraleo tube 1225 eruption, Tanguy, 1981 , theŽGallobianco tube 1050–1100 eruption, Tanguy,

. Ž1981 , the Tre Livelli tube 1792–1793 eruption,.Corsaro et al., 1990 , and the Salto della Giumenta

Ž .tube formed inside the flow of December 24, 1991Ž .Fig. 1 . The Intraleo tube is particularly revealingbecause of the clear relationship between surfacemorphology of the lava flow field and internal struc-tures within the lava tube. Here, we observed a 10 mwide lava flow with a tube inside, whose roof andfloor had collapsed in the frontal region where itmerged with a previous, deeper lava tube. Grooveswere well developed on the walls of the lower halfof the upper tube. About 5 m from the tube junction,striae were almost horizontal, but their inclinationgradually increased downslope and became verticalwhere the upper tube broke through the roof of thelower tube. This tube coalescence occurred at thefront of the upper flow, which, at this point, had athickness of about 6 m. Because the roof of theupper flow also collapsed, it is probable that thisflow was still inflating when it merged with thelower tube. In this case, the upper crust had no timeto solidify and become self-supporting.

We believe that the mechanism we have justdescribed is similar to what took place on the Jan-uary 3, 1991. The January 3 lava flowed along thechannel of the December 15, 1991 flow. Stable crustalong the December 15 flow was first observed after

( )S. CalÕari, H. PinkertonrJournal of Volcanology and Geothermal Research 90 1999 263–280 275

a few days of emplacement of the flow. The crustwas only a few days old and probably less than 1 mthick when it collapsed below the weight of the newoverlapping flow.

Ž .Greeley et al. 1998 , in their very comprehensivereview of erosion within lava tubes, described onelocality which, they argue, supports mechanical ero-sion. Their locality lies within the proximal part ofthe Tre Livelli tube at a height of about 1700 m a.s.l.Ž .Fig. 1 . This tube formed along the 1792–1793 flow

Žfield between 1850 and 1450 m a.s.l. Corsaro et al.,.1990 . The eruptive fissure, which is still visible on

the ground, extends between 1850 and 1600 m a.s.l.,and the flow field reaches the lowest altitude of750 m a.s.l., close to the town of Zafferana EtneaŽ .Romano et al., 1979; Corsaro et al., 1990 . Were-examined the tube system in September 1997 andOctober 1998. The lining on the wall of the lowest ofthe three tubes comprising this tube system hascollapsed, exposing a layer of pyroclastics that liebeneath the 1792–1793 flow field. Greeley et al.Ž .1998 use this as evidence of mechanical erosion atthe base of the flow. However, because the collapseis very close to the eruptive fissure, it is possible thatthe pyroclastic layer has been exposed by the open-

ing of the fissure rather than by erosion of thesubstratum.

Ž .We confirm the conclusion of Greeley et al. 1998that there is no evidence of thermal erosion insidelava tubes or channels on Etna. If it occurs, it is onlyclose to the vent area, and it is much less extensive

Žthan in Hawaiian lava tubes Kauahikaua et al.,.1998 . However, we have unequivocal evidence of

tube coalescence. This was observed in virtually alltubes that we surveyed, and we argue that thismechanism plays a pivotal role in prolonging andlengthening lava tubes on Etna.

4. Relationship between surface features on lavaflows and lava tubes

There are a number of surface features on manylava flow fields that suggest the presence of lavatubes. The most common are skylights, tumuli, col-

Žlapses, breakouts and secondary vents Guest et al.,1980, 1984; Mattox et al., 1993; Calvari et al.,1994a; Calvari and Pinkerton, 1998; Kauahikaua et

.al., 1998 . However, in discussing the relationshipsbetween these features and tubes, it is important to

Fig. 10. Circular collapse zone along the southern margin of the 1991–1993 lava flow field, Mount Etna. The width of the collapse isapproximately 80 m, and it formed after the end of the eruption by drainage of a lava pond.

( )S. CalÕari, H. PinkertonrJournal of Volcanology and Geothermal Research 90 1999 263–280276

Fig. 11. Collapse depression formed by drainage of the flow front at the lower end of the 1792–1993 Cassone tube, a few metres downslopeŽfrom the point where the tube stops. The wall on the right represents the remains of the inflated flow front. Flows on the flat ground and in

. Ž .section behind Dr Calvari are secondary flows from third-order ephemeral vents Calvari and Pinkerton, 1998 .

Fig. 12. View from the NE of the southwestern wall of Valle del Bove, 2nd June 1992. The eruptive fissure of the 1991–1993 eruption isŽ .still degassing in the upper part. Note the alignment of skylights white patches on the black flow field . These record the path of the firstŽ .tube system which was active before the diversion of May 1992 Calvari and Pinkerton, 1998 .

( )S. CalÕari, H. PinkertonrJournal of Volcanology and Geothermal Research 90 1999 263–280 277

Ž . Ž .Fig. 13. Tumulus on the 1983 lava flow field. Note the small drained right and left and undrained middle tubes fed by the tumulus. Theruler on the undrained tube is 20 cm long.

clarify what constitutes a lava tube. If we define lavatubes as the cores of drained lava flows with acontinuous and stable crust, we find that many tube-like structures never drained, or drained only a smallproportion of their lava. This may be why WalkerŽ .1991 stated that, on Hawaii, tumuli are not relatedto lava tubes. Since these tumuli were observed onlow gradient zones of the flow field in lava flows inthe Cava Basalt, Mount St. Helens, WashingtonŽ . ŽGreeley and Hyde, 1972 and on Etna Calvari and

.Pinkerton, 1998 , it is highly probable that they didnot drain. However they may have had an importantrole in delivering lava underground to the distal parts

Ž .of the flow field see Kauahikaua et al., 1998 .Depressions and collapse structures are often as-

Ž .sociated with tubes Fig. 10 . Sometimes collapsedzones appear to be much larger than the size of the

Ž .tube Atkinson et al., 1975 , and this has beenexplained as a consequence of the drainage of a lava

Ž .pond Stephenson, 1996; Kauahikaua et al., 1998 .Large collapse areas may also form during the com-plete drainage of inflated flow fronts at the pointwhere an ephemeral vent opens. We have found twoexcellent examples of this mechanism. One is on the

1983 flow field of Etna close to Rifugio SapienzaŽ .1900 m a.s.l. ; the other is approximately 40 mdownslope from the entrance of the Grotta Cassone,close to the margin of the 1792–1793 flow fieldŽ .Fig. 11 .

Ž .Skylights Fig. 12 are extremely useful featuresbecause they allow direct observations and measure-

Ž .ments of lava flowing inside tubes. Tumuli Fig. 13usually form along main tubes where there is a

Ž .sudden break in slope see Greeley and Hyde, 1972 .Overpressure created by lava accumulating at thebase of a steep scarp, or local collapses inside tubesŽ .Calvari and Pinkerton, 1998 can give rise to contin-uous output of small flows that can accumulate to

Žform large deltaic lava fans Guest et al., 1984;.Calvari and Pinkerton, 1998 .

5. Discussion

In our study of the 1991–1993 lava flow field onŽ .Etna Calvari and Pinkerton, 1998 , we demonstrated

the importance of ephemeral vent formation andsuccessive budding of new flows from inflated flowfronts. Our recent detailed investigations of lava

( )S. CalÕari, H. PinkertonrJournal of Volcanology and Geothermal Research 90 1999 263–280278

tubes, including cross-sections and longitudinal pro-files, reveal that the processes of flow inflationrecorded on the surface can be preserved under-ground in the form of inflated parts of a tube. Wehave also located numerous ephemeral vents emerg-ing from the fronts of inflated tubes.

Our underground surveys also reveal informationon additional flow processes which are difficult tounderstand during eruptions. For example, we havedocumented several cases where the roof of a lavatube has collapsed, allowing lava from a surface flowto re-occupy a deeper level tube. Particularly strikingare examples where tube systems at different levels

Žintersect at right angles for example the proximal.1985 lava tube system . In some cases the evidence

is subtle, and requires detailed investigations ofchanges from simple tube cross-sections to bell-shaped sections. In other cases, grooves created bycrustal material moving against still plastic lava wallscan record sudden changes in flow direction fromhorizontal to inclined and often vertical flow. Thischange can be accompanied by an abrupt change inlevel of the floor of the lava tube. Additional evi-dence in support of this process includes abruptchanges in the number and thickness of layers thatline the walls along different parts of a tube system.

The sudden collapse of a tube roof and the subse-quent rapid flow of lava into an older tube systemhas clear implications for hazard assessment. Firstly,it means that stationary flow fronts may suddenly bere-activated if a drained tube is re-occupied by newlava. Secondly, during future flow diversion mea-sures, if a diverted lava flows on top of older lavastreams, and if roof collapse permits lava to re-oc-cupy an older tube system, the lava may be able tomigrate downflow considerably more rapidly than ifit continued to flow on the surface. In those circum-stances flow diversion may exacerbate the hazardsinstead of reducing them. It is therefore importantthat old tubes are identified and their roof thicknessdetermined before flow diversion measures are car-ried out. In addition, the probability of roof fracturemust be established prior to any diversion measure.

Considering a simplistic model for a flat, horizon-tal tube roof, we can calculate the minimum flowthickness required to fracture the roof of a tube from:

ssWar bt 2Ž .

where s is the tensile strength of the roof, W is theweight of the overlying lava flow over a length b, ais the width of the tube roof, and t is its thickness.This can readily be re-expressed in terms of thethickness of the upper flow, h, as follows:

hs s t 2 r r ga2Ž . Ž .where r and g are lava density and gravitationalacceleration, respectively. If we assume a value fortensile strength of 5 MPa, and a density of 2500kgrm3, then for a tube roof that is 3 m wide, theflow depth of the upper flow required to fracture thecrust ranges from 0.9 m for a crustal thickness of 0.2m, to 2.1 m for a crustal thickness of 0.3 m, 5.6 mfor a crustal thickness of 0.5 m, to 22.7 m for acrustal thickness of 1 m. These figures are in generalagreement with the dimensions of flows that causedroof collapse into underlying tube systems on Etna.

6. Conclusions

On Etna, and on many other volcanoes, lava tubesare recognised only when the roof of a tube col-lapses, or when new roads cut through tubes. Thethick, self-supporting crust of ‘a‘a lava flows makesmost tubes in ‘a‘a lava flows difficult to detect.Moreover, lava flows emplaced through tubes thatdid not drain are virtually impossible to recognise.For these reasons, and because of the large increase,since 1971, of the number of known lava tubes onEtna, we conclude that the figure of 18% of tube fed

Ž .flows on Etna Greeley, 1971 is underestimated. Wealso conclude that at least some of the lava tubes onEtna previously considered to have formed on pa-hoehoe flows may have formed on ‘a‘a flows whichwere subsequently covered by late stage pahoehoelava.

Observations of active ‘a‘a lava flows emplacedduring the 1991–1993 eruption suggested that infla-tion of the flow fronts of mature ‘a‘a lava flows was

Žan important aspect of tube formation Calvari and.Pinkerton, 1998 . This statement is confirmed by

features we observed underground. A striking featurein many etnean lava tubes is the association of largechambers with narrow passages. We interpret thesefeatures as the product of flow inflation in the frontalzone, followed by opening of first-order ephemeralvents on the snout region. This suggests that the

( )S. CalÕari, H. PinkertonrJournal of Volcanology and Geothermal Research 90 1999 263–280 279

production of multiple flows connected by secondaryvents is an essential mechanism of tube growth onEtna. This process can be observed on the surfacewhere new, long-lived vents open at the margins ofpreviously inactive ‘a‘a lava flows, and undergroundas multiple linings and lateral benches.

Tube coalescence is an important process duringthe emplacement of long-lived flows. This occurswhen an upper tube drains into a previous, lowertube because of roof collapse. The re-occupation of adeeper and possibly longer tube may cause a re-activation of distal parts of the flow field, and aconsequent increase in flow field length. The possi-bility of new flows breaking through the roofs ofpreviously inactive tubes has important conse-quences in hazard assessment. Our surveys indicatethat a crust thickness of 0.5 m is the minimumrequired for ‘a‘a lava flows, and we propose asimple formula to calculate roof stability for a typi-cal Etnean ‘a‘a lava flow.

In conclusion, our underground surveys help toŽ .confirm see Calvari and Pinkerton, 1998 that sub-

terranean effusive processes play an important rolein the development of ‘a‘a flow fields. They showhow tube inflation and coalescence can result inconsiderable lengthening of lava flow fields beyondthe distance that can be attained by channel-fed lavaflows. This work has clear implications for hazardassessments during future effusive eruptions on Etna.Finally, in view of the ease with which lava canbreak through the roofs of lava tubes, we considerthat the process of tube coalescence may be anintegral part of tube development, not only on Etna,but also on other basaltic volcanoes.

Acknowledgements

We are very grateful to members of GruppoSpeleologico Etneo for their assistance in reachingsome of the critical localities we describe above.Without their support and the photographic assis-tance of Alfio Amantia, IIV-Catania, this work wouldnot have been possible. We also wish to thank KathyCashman, John Guest, Angus Duncan and membersof Gruppo Speleologico Etneo for many stimulatingdiscussions on the formation of lava tubes. We alsothank Professors Lionel Wilson and Ron Greeley fortheir very useful comments on this paper. This study

was undertaken with the assistance of a grant fromthe commission of the European Communities underthe Fourth Framework Programme, Environment andClimate, Contract ENV4-CT97-0713.

References

Anonymous, 1669. An answer to some inquiries about the lateeruption of Mount Aetna on 1669. Communicated by someinquisitive English merchants now residing in Sicily. Philos.Trans. R. Soc. London 51, 1028–1034.

Atkinson, A., Atkinson, V., 1995. Undara Volcano and its lavatubes. Vernon and Anne Atkinson, Brisbane, Queensland, p.85.

Atkinson, A., Griffin, T.J., Stephenson, P.J., 1975. A major lavatube system from Undara Volcano, North Queensland. Bull.Volcanol. 39, 266–293.

Balsamo, A., Priolo, A., Priolo, G., Sanfilippo, G., Scammacca,B., 1994. Grotte vulcaniche di Sicilia, notizie catastali: quarto

Ž .contributo da Si CT 76 a Si CT 100 . Boll. Accad. GioeniaSci. Nat. Catania 27, 173–212.

Bonforte, A., 1994. Evoluzione del campo lavico dell’eruzioneetnea 1991–93. Universita degli studi di Catania, unpublisheddegree second thesis, 85 pp.

Brunelli, Scammacca, 1975. Grotte vulcaniche di Sicilia.Calvari, S., Pinkerton, H., 1998. Formation of lava tubes and

extensive flow field during the 1991–93 eruption of MountEtna. J. Geophys. Res. 103, 27291–27302.

Calvari, S., Coltelli, M., Neri, M., Pompilio, M., Scribano, V.,1994a. The 1991–93 Etna eruption: chronology and geologicalobservations. Acta Vulcanol. 4, 1–15.

Calvari, S., Groppelli, G., Pasquare, G., 1994b. Preliminary geo-´logical data on the south–western wall of the Valle del Bove,Mt. Etna, Italy. Acta Vulcanol. 5, 15–30.

Cavallaro, F., Puglisi, G., Tranchina, A., 1985. Morfologia epetrografia di colate con ‘lava tubes’ del basso versante sud-orientale dell’Etna. Boll. Accad. Gioenia Sci. Nat. Catania 18,305–324.

Corsaro, R., Giudice, G., Puglisi, G., 1990. Il sistema 3 livelli-ktm:studio comparato di una colata con gallerie di scorrimentolavico. Atti del 18 Convegno Regionale di Speleologia dellaSicilia, Ragusa, 66–76.

Cucuzza Silvestri, S., 1977. Le grotte vulcaniche dell’Etna e illoro studio. Atti del Seminario sulle Grotte Laviche, Catania,215–229.

Cumin, G., 1954. L’eruzione laterale etnea del novembre 1950–dicembre 1951. Bull. Volcanol. 15, 3–70.

Fanciulli, F., Licitra, G.M., Pandolfo, C., Puglisi, G., 1989.Contributo alla conoscenza di alcune grotte vulcaniche interritorio di Adrano. Centro Speleologico Etneo 1–11.

Frazzetta, G., Romano, R., 1984. The 1983 Etna eruption: eventchronology and morphological evolution of the lava flow.Bull. Volcanol. 47, 1079–1096.

Giudice, G., Leotta, A., 1995. Le alchimie di un vulcano: laGrotta Cutrona. Speleologia 33, 14–20.

( )S. CalÕari, H. PinkertonrJournal of Volcanology and Geothermal Research 90 1999 263–280280

Greeley, R., 1971. Observations of actively forming lava tubesand associated structures, Hawaii. Modern Geol. 2, 207–223.

Greeley, R., 1987. The role of lava tubes in Hawaiian volcanoes.U.S. Geol. Surv. Prof. Pap. 1350, 1589–1602.

Greeley, R., Hyde, A., 1972. Lava tubes in the Cava Basalt,Mount St. Helens, Washington. Bull. Geol. Soc. Am. 83,2397–2418.

Greeley, R., Fagents, S.A., Harris, R.S., Kadel, S.D., Williams,D.A., Guest, J.E., 1998. Erosion by flowing lava: field evi-dence. J. Geophys. Res. 103, 27325–27346.

Guest, J.E., Underwood, J.R., Greeley, R., 1980. Role of lavatubes in flows from the Observatory Vent, 1971 Eruption onMount Etna. Geol. Mag. 117, 601–606.

Guest, J.E., Wood, C., Greeley, R., 1984. Lava tubes, terraces andmegatumuli on the 1614-24 pahoehoe lava flow field, MountEtna, Sicily. Bull. Volcanol. 47, 635–648.

Guest, J.E., Kilburn, C.R.J., Pinkerton, H., Duncan, A.M., 1987.The evolution of lava flow-fields: observations of the 1981and 1983 eruptions of Mount Etna, Sicily. Bull. Volcanol. 49,527–540.

Hallworth, M.A., Huppert, H.E., Sparks, R.S.J., 1987. A labora-tory simulation of basaltic lava flows. Modern Geol. 11,93–107.

Huppert, H.E., Sparks, R.S.J., 1985. Komatiites I: eruption andflow. J. Petrol. 26, 694–725.

Huppert, H.E., Sparks, R.S.J., Turner, S.J., Arndt, N.T., 1984.Emplacement and cooling of komatiite lavas. Nature 309,19–22.

Jaggar, T.A., 1931. Lava stalactites, stalagmites, toes, and‘squeeze-ups’. The Volcano Letter 345, 1–3.

Kauahikaua, J., Cashman, K.V., Mattox, T.N., Heliker, C., Hon,K., Mangan, M., Thornber, C., 1998. Observations on basalticlava streams in tubes from Kilauea Volcano, Hawaii. J. Geo-phys. Res. 103, 27303–27324.

Keszthelyi, L., 1995. A preliminary thermal budget for lava tubeson the Earth and planets. J. Geophys. Res. 100, 20411–20420.

Keszthelyi, L.P., Pieri, D.C., 1993. Emplacement of the 75-km-long Carrizozo lava flow field, south–central New Mexico. J.Volcanol. Geotherm. Res. 59, 59–75.

Kilburn, C., 1989. Surfaces of ‘a‘a flow-fields on Mount Etna,Sicily: morphology, rheology, crystallization and scaling phe-

Ž .nomena. In: F.J. F.J. , Lava Flows and Domes. Springer-Verlag, pp. 129–156.

Kilburn, C.R.J., Lopes, R.M.C., 1991. General patterns of flowfield growth: ‘a‘a and blocky lavas. J. Geophys. Res. 96,19721–19732.

Leotta, A., 1994. Caratteristiche generali delle fratture eruttivesull’Etna. Boll. Accad. Gioenia Sci. Nat. Catania 27, 153–160.

Lyell, C., 1858. On the structure of lavas which have consolidatedon steep slopes; with remarks on the mode of origin of MountEtna, and on the theory of ‘craters of elevation’. Philos. Trans.R. Soc. London 148, 703–786.

Mattox, T.N., Heliker, C., Kauahikaua, J., Hon, K., 1993. Devel-opment of the 1990 Kalapana flow field, Kilauea Volcano,Hawaii. Bull. Volcanol. 55, 407–413.

Ollier, C.D., Brown, M.C., 1965. Lava caves of Victoria. Bull.Volcanol. 28, 215–229.

Peterson, D.W., Swanson, D.A., 1974. Observed formation of lavatubes. Studies in Speleology 2, 209–222.

Peterson, D.W., Holcomb, R.T., Tilling, R.I., Christiansen, R.L.,1994. Development of lava tubes in the light of observations atMauna Ulu, Kilauea Volcano, Hawaii. Bull. Volcanol. 56,343–360.

Pinkerton, H., Norton, G., 1995. Rheological properties of basalticlavas at sub-liquidus temperatures: laboratory and field mea-surements on lavas from Mount Etna. J. Volcanol. Geotherm.Res. 68, 307–323.

Pinkerton, H., Sparks, R.S.J., 1976. The 1975 subterminal lavas,Mount Etna: a case history of the formation of a compoundlava field. J. Volcanol. Geotherm. Res. 1, 167–182.

Puglisi, G., 1981. La grotta ‘Micio Conti’. Studio genetico edevolutivo di una grotta di scorrimento lavico. Speleoetna-Notiziario del Gruppo Grotte di Catania 3, 22–30.

Recupero, A., 1815. Storia naturale e generale dell’Etna. Catania,pp. 85–97.

Romano, R., Sturiale, C., 1982. The historical eruptions of Mt.Ž .Etna volcanological data . Mem. Soc. Geol. Ital. 23, 75–97.

Romano, R., Amore, C., Atzori, P., Carter, S.R., Cristofolini, R.,Di Geronimo, I., Di Grande, A., Duncan, A.M., Ferrara, V.,Ghisetti, F., Guest, J.E., Hammill, M., Lentini, F., Lo Giudice,E., Patane, G., Pezzino, A., Puglisi, D., Rasa, R., Schiliro, F.,´ ` `Sturiale, C., Torre, G., Vezzani, L., 1979. Geological Map ofMt. Etna. Progetto Finalizzato Geodinamica, C.N.R.-I.I.V.

Sparks, R.S.J., Pinkerton, H., 1978. Effect of degassing on rheol-ogy of basaltic lava. Nature 276, 385–386.

Stephenson, P.J., 1996. Ridges and Depressions in the UndaraLava Flow, North Queensland: Alternative Mechanisms fortheir Formation. Chapman Conference on Long Lava Flows,Townsville, Australia, 82–83.

Stevens, N.F., Murray, J.B., Wadge, G., 1997. The volume andshape of the 1991–1993 lava flow field at Mount Etna, Sicily.Bull. Volcanol. 59, in print.

Swanson, D.A., 1972. Magma supply rate at Kilauea Volcano,1952–1971. Science 175, 169–170.

Tanguy, J.C., 1981. Les eruptions historiques de l’Etna: chronolo-gie et localisation. Bull. Volcanol. 44, 585–640.

Wadge, G., 1978. Effusion rate and the shape of ‘a‘a flow-fieldson Mount Etna. Geology 6, 503–506.

Wadge, G., 1981. The variation of magma discharge duringbasaltic eruptions. J. Volcanol. Geotherm. Res. 11, 139–168.

Walker, G.P.L., 1971. Compound and simple lava flows and floodbasalts. Bull. Volcanol. 35, 579–590.

Walker, G.P.L., 1991. Structure, and origin by injection of lavaunder surface crust of tumuli, ‘lava rises’, ‘lava-rise pits’, and‘lava-inflation clefts’ in Hawaii. Bull. Volcanol. 53, 546–558.

Williams, D.A., Lesher, C.M., 1996. AnalyticalrNumerical Mod-eling of Long Komatiite Flows. Chapman Conference on LongLava Flows, Townsville, Australia, 96–97.

Wood, C., 1971. The nature and origin of Raufarholshellir. Trans.´Cave Research Group of Great Britain 13, 245–256.

Wood, C., 1974. The genesis and classification of lava tube caves.Trans. British Cave Research Assoc. 1, 15–28.

Wood, C., 1975. Factors contributing to the genesis of caves inlava. Seminario sulle Grotte Laviche, Catania, 101–113.