The Geology of Tofua Island, Tonga!

GLENN R. BAUER2

ABSTRACT : Tofua Island is an oval, steep-sided composite volcano, 5 miles by 6miles in diameter, the summit of which has collapsed to form a caldera. Within thecaldera, Lofia cone is still active.

Four unit s have been mapped: (1) the Hamatua Formation, of precaldera age,includes basaltic andesites, pyroxene andesites, and pyroxene dacites; (2 ) the HokulaFroth Lava, a microvesiculated lava flow of andesite; ( 3) the Kolo Formation,composed of air-laid lapilli tuff-breccia, tuff, unconsolidated ash and cinder, smallbasaltic andesite lava flows, and one thick pyroxene andesite lava flow; and (4) theLofia Formation, composed of air-laid tuff, ash, basaltic andesite lava flows, andandesite lava flows. An erosional unconformity separates the Hamatua Formationfrom the Hokula Froth Lava, and another lies between the froth lava and the KoloFormation.

The rocks are typical orogenic andesites and related types, unusually high in CaO,which is reflected in the calcic nature of the plagioclase. The pyroxenes of thegroundmass are usually pigeonite and pigeonitic augite, whereas the phenocrystsare augite. Hypersthene is less common and occurs only as phenocrysts surroundedby reaction rims of pigeonite and/or pigeonitic augite .

Concentric normal faults associated with caldera collapse are common on thenorthern, eastern, and southern rims of the caldera. Some of the faults have servedas conduits for rising magma of the Kolo Formation; others , on which calderacollapse is continuing, do not exhibit .associared volcanism. Tensional cracks areabundant along the southern rim.

THE GEOLOGICAL INVESTIGATION of TofuaIsland is part of a larger study of the Tonga andthe Lau islands by the Hawaii Institute of Geo­physics. The investigation includes paleomag­netic and gravity studies of many of the islands,but only the areal geology and petrology ofTofua are dealt with in this pap er.

Field work on Tofua was performed duringthe first three weeks of June 1968. A base mapshowing the triangulation stations established bythe Tongan Government Survey in 1961 wasobtained from the Lands Department of Tonga,in Nuku'alofa. Although the triangulation sta­tions were only temporary , it was not difficult torelocate them, since all were at prominent land­marks or villages. Using them as control, atopographic map of Tofua (Fig. 2 A) with 200-

i Hawaii Insti tute of Geoph ysics Contr ibuti on N o.336. Investigation supported by Grant GA-1 473 fromthe N ational Science Foun dation. Manuscript receivedD ecember 18, 1969.

2 P.O. Box 3415, Honolulu, Hawaii 96801.

333

foot contour intervals was drawn up primarilyfor the gravimetric work, but it was also used asa base for the geologic mapping (Fig. 2 B) .Elevations were obtained by aneroid altimeter.Most of the geologic mapp ing was done con­currently with the topographic mapping.

Tofua's flanks are heavily wooded: rock out­crops are especially difficult to locate on thewindward, eastern, and southern slopes. Pyro­clastic deposits of the Kolo Formation cover thenorthwestern flank and part of the western flank,and the bedrock is concealed. Approximately aweek was spent working on the caldera rim andwithin the caldera, and two weeks were devotedto mapping the flanks of the cone.

PREVIOUS INVESTIGATIONS

Previous geolog ical studies in the TongaIslands (Lister, 1891 ; Hoffmeister, 1932) wereprimarily concerned with detailed descriptions

334

of the Tertiary limestones . Descriptions of theigneous rocks were brief.

In 1643 Abel Tasman became the first Euro­pean explorer to sight Tofua Island. He madenote of sailing near two high islands (Tofuaand Kao) and seemed surprised at their height,for he had just left the low limestone islandsthat lie east of Tofua. He sketched its pr ofile,and called the island "Amatofua" (Woide,1776) .

The volcanic nature of Tofua was first notedby Capt. James Cook in 1774 when he wrote of"To fooa" in his journal. Cook sailed a milenorth of the island and noted that it was ineruption. His crew reported that raindrops froma localized rainstorm, caused by the eruptioncloud, created a burnin g sensation in their eyes.Cook also wrote that the rocks along the north­west coast were cavernous and appeared to formcolumns, and that the foliage just below thesummit of the island on the northwest flank hadbeen recently burnt. He concluded that thesummit of Tofua contained a crater (Cook andForester, 1777, vol. 4, pp . 162-163) .

The first Caucasian to set foot on Tofua wasCapt. William Bligh, just after the 1789 mut inyaboard the "Bounty." He also described "Tofoa"as a volcanic mountain, and made reference todeep gull ies on the southwestern side (Dalyell,1812, p. 153).

Later, Lister, who visited the island in 1889,stated (1891 , p. 593) that "Tofua is a volcanoin a state of intermittent activity." He providedAlfred Harker, of Cambridge University, with acollection of rocks from Tonga, but rocks fr omTofua were not among them (Harker, 1891).

Marshall (1911, p. 22) made collections onthe island . He gave the first brief petrographicdescriptions , stating that the rocks were "augiteandesites ." Daly ( 1916, p. 340) repeatedMarshall 's statement.

No further work was done on Tofua until1959, when Richard (1962, p. 16) visited theisland and collected two samples which werechemically analyzed. The analyses indicated thatthe rocks are quartz basalts in Rittmann's(1952) classification. However, the MnO con­tent of the rocks as reported in the analyses isnearly 2 percent, which is much too high forrocks of this type. In the new analyses of Tofuarocks given in Table 1 of this paper, only normal

PACI FIC SCIENCE, Vol. 24, July 1970

amounts of MnO, in the vicinity of 0.16 per­cent, are shown.

GEOGRAP HICAL DESCR IPTI ON

The Tonga Archipelago lies between latitudes15° and 22°S and longitudes 173° and 177°W.Its northern end lies approximately 400 milessouth of Samoa, and it stretches on southwardfor approximately 160 miles. The northernmostislands are N iuafo 'ou and Niuatoputapu. Farthersouth these islands fall into three recognizedgroups, or island clusters: Vava'u, the mostnortherly; Ha'apai, about 70 miles to the south;and Tongatabu, 90 miles farth er south . Tofuais one of the westernmost islands of this Ha'apaiGroup (Fig. 1) .

Just west of the Tonga Trench a series ofcoralline islands lies in .a north-south alignm entalong the eastern side of the archipelago. Thevolcanic islands form a line paralleling thecoralline islands on the western side. Along thevolcanic line there are several active terrestrialand submarine volcanoes, Tofua being one of themore prominent of the terrestr ial ones. Accord­ing to Admiralty Chart 2421, the geographiccenter of Tofua is at 19°45'S and 175°05'W.Th e island is 17 miles northwest of KotuIsland , from which it is separated by the north­south-trending Tofua trough. The volcanicisland of Kao lies 4 miles to the north. It isjoined to Tofua by a submarine ridge. The"andesite line," marking the boundary betweencontinental- and oceanic-type volcanics, liesalong the Tonga Trench (Marshall, 1911 ; Mac­donald, 1949, p. 1590 ) , and the rocks of Tofuaare all volcanic and of continental orogenictypes.

GEOMOR PHOLOGY

Tofua Island is oval in pl an, and approxi­mately 5 miles by 6 miles across (Fig. 2 A ). Itis a single composite volcano, the summit ofwhich has collapsed to form a caldera. Thelower slopes rise at an angle of about 35 0 , butthe slope angle decreases near the summit. Whenviewed from the east, Tofua has an extremelyregular profile. Elevation of the caldera rimvaries less than 300 feet-from 1,400 to almost1,700 feet.

Geology of Tofua Island-i-Bxusn

V a va ' u I. ~ffty>"", 0 •

L ate I. 0 VAVA'U GROUP175· 174·

19·

N:> Met i s Shoa l

~I

HA 'APAI GROUPKa o I. C • \

0) c::JTofua I- '" Q L! L i f uka I.

t:>

Kotu I . !• •20· .. ,

.ceJ N o mu k a I •

..

0 10 20 30

M I LES

2 1·

~ TONGATABU GROUP

To ng o t abu I.

~'E u a I.

F IG. 1. Map of the main part of the Tonga Archipelago, showing the location of Tofua Island .

335

336

19·43' 5

19·46'5

Ao

~CALE IN MILES

PACIFIC SCIENCE, Vol. 24, July 1970

N

~

160 0'

.., --.---:

FIG. 2. Topographic map (A) and geologic map (B) of Tofua Island . Geology by Glenn R. Bauer, June1968.

Geology of Tofua Island-i-Baunn 337

19°43 ' 5

B

Homc tuc

o -Scale in MilesL ATE PL~TOCENE

--? -_________________R:s-:,N_T "

EROSION ALUNCONFORM ITY -HOKULA FROTH LAVA

EROS IONALUNCONFORMIT Y~~KOLO FORMATI ON

h - Basall ,pyroxene andesit e,pyroxene doci le lava flows

f l -Porphyri tic andes ite fr othlava f low

ka - Consol ida ted lopilh l uff­brecc ia

kb - Unconsolida ted tephra

kc - Andesile lava fl ow

10- Consolidate d l a p i lli ~ l uff,

luff, spotte r, unconsolidat edlopilli c inde r, inte rbe ddedbosalt fa andesite lava flows

lap - Ponded lava flow

Ib- Ash coneic - Large cndesitic Java f low

Faul l , defi ni tel y lo cated , showing dow n -th ro w n si de

Fau ll , appr oxima te ly locat ed

Buried fault

Geolc qrc contact def initely located

- - - - - - Geolog ic conta c t cppr oximutely loc ct ed

-....r Dip ond str ike of beds

• Villoge

15~63' Trinnqufc t ton sta l ion showing eleva t ion

338

Th e caldera is subcircular, with an averagediameter of 2. 5 miles. Half of its floor is oc­cupied by a crater lake of unknown depth , thesurface of which is 76 feet above sea level. Th ewalls of the caldera are extremely precipitous,but the southern, southwestern, and northernwalls are not as steep as the others. In someplaces the caldera floor is more than 1,400 feetbelow the rim, but the mean difference in eleva­tion is approximately 1,10 0 feet.

Two prom inent spatter-and-cinder cones oc­cupy the northern and eastern parts of thecaldera. Less prominent is an older, deeplyeroded ash cone near the western caldera wall.

Lofia cone, the northern cone, is still active,though currently it is in a fumarolic state. Itscrater is about 400 feet deep, with a flat floorformed by a ponded lava flow. The eastern coneis dormant and has three vents, two of whichare shown as craters on the topographic map .Its summit is 600 feet above the crater lake,though only 200 feet above the southern flankof Lofia cone.

The caldera is continu ing to collapse alongrelatively minor fault s on the northern, eastern,and southeastern sides. The fault displacementshave given th is part of the rim a hummockytopography, with small grabens and steps withescarpments averaging from 50 to 100 feet high .

Large stream valleys do not exist on Tofua:all the valleys are narrow and small, thoughsome are deep. They are mostly on the south­eastern to southwestern flanks-the windwardside. These steep-sided windw ard valleys extendno more than a quarter of a mile inland . Somevalleys near Hamatua village cut into rocks ofthe Hamatua Formation, but all valleys in thevicinity of Kolo and Manaka villages cut onlypostcaldera tuffs of the Kolo Formation. Earliervalleys, if they were present, have been buriedbeneath thick tuff deposits of the Kolo Forma­tion. Small gullies of intermittent streams alongthe northern coast cut only rocks of postcalderaage.

Intermitt ent streams have gullied the softtuff and ash fill in the northern and eastern partsof the caldera . The gullies are commonly 5 to10 feet deep, with very steep walls. More ex­tensive gullying has taken place on the southernand southwestern walls of the caldera, in rocksof precaldera age.

PACIFIC SCIENCE, Vol. 24 , July 1970

Th e wave-cut cliff that encircles most ofTofua averages 150 feet in height. The cliff isespecially precipitous where froth lava is ex­posed, the steepness being due to undercuttingin the lower clinker zone of the flow, with sub­sequent breaking of the overlying massive rockalong joints.

STRAT IGRAPHY AND LITHOLOGY

General statement

Four stratigraphic units have been mapp ed :(1) Hamatua Formation, (2) Hokula FrothLava, (3) Kolo Formation, and (4) Lofia For­mation . The sequence of the unit s was deter­mined by superposition. Some rocks of the KoloFormation are contemporaneous with some rocksof the Lofia Formation, but their spatial distri bu­tion is not the same and they have thereforebeen mapped separately in Figure 2 B.

Ninety-eight th in sections were examined indetail. Th e rocks were classified by color indexand the amount of free silica present as observedsilica minerals. However, the latt er method isnot completely dependable. Some of the moresilicic chemically analyzed samples do not con­tain any observable silica minerals . The excesssilica is probably contained largely in the glass.Rittmann's zonal method was used with four­axis universal stage to determine the composi­tion of zoned plagioclase ( Emmons, 1943,pp. 115-133 ) . Combined albite-Carlsbad twin­ning and maximum extinction angles were usedto determine the composition of unzoned pla­gioclase.

Hamatua Formation

The Hamatua Formation, which consists oflava flows erup ted before the collapse of thecaldera, is named for the village of Hamatua,on the southeastern edge of the island . Theformation crops out directly behind the villageas a thick hypersthene-bearing augite dacite lavaflow (chemical analysis 3, Table 1), forming anabrupt cliff 50 feet high. Below the village itforms rocky points. Although it is the mostvoluminous rock unit on the island, the Hama­tua Formation is exposed only in scatteredpatches on the flanks of Tofua cone, usually inthe form of rocky points along the shore, oralong stream beds. It is covered by residual

Geology of Tofua Island-BAu ER

soil on the southwestern flank, but is exposedin the sea cliff.

The H amatua Formation is best exposed inthe walls of the caldera, where it is more than1,600 feet thick. The average thickness of in­dividual flows is approximately 70 feet. Eachflow consists of a thick, dense central portion,with relatively thin clinker zones above andbelow it. Poorly formed vert ical joints aretypical in the central mass. N o columnar joint­ing was seen. The flows appear to be of theblock lava variety (Macdonald, 1967, p. 23) .T he clinker fragments are less spinose than thoseof typical aa flows. In most cases they gradeinto the dense central portion of the flow.

The Hamatua rocks are ligh t- to dark-g raybasaltic andesites, hypersthene-bearing augiteandesites, andesites, and hypersthene-bearingaugite dacites. Th e ligh ter rocks display con­spicuous flow structure, with altern ate ligh t anddark bands averaging about 2 em thick, whichare usually distorted due to deformation duringflowage.

In hand specimens the andesites and dacitesoften appear vitreous. Th in-section examinationshows that some are quite vitreous, though mostare composed of microlites of plagioclase feld­spar, clinopyroxene, magnetite, and silica min­erals, with only a minor amount of glass. Thetexture is usually hyalopilitic, but intergranularand intersertal textu res are also present. Thefeldspar laths sometimes exhibit subparallelflow structure. The basalt ic andesites usuallyhave intersertal to intergranular texture, thoughsome are hyalo-ophitic.

A 400-foot type section of the Hama tua For­mation was measured just below the rim in thecaldera wall between triangul ation stations B andB1 (Fig. 2 A) . The lower 350 feet consists ofthick flows of andesite containing phenocrystsof hypersthene and augite, and some nonpor­phyri tic flows. The bottom flow is nearly 80feet thick. At the top of the section is a basalticandesite flow containing relatively large crystalsof pigeonite.

A sample of hypersthene-bearing augite dacitewas collected behind Hamatua village at anelevation of 500 feet (analy sis 3, Table 1) .

All varieties of rocks in the Hamatua Forma­tion usually contain plagioclase phenocrysts andmicrophenocrysts , which commonly are zoned

339

from a calcic core to a more sodium-rich rim.The cores of the phenocrysts range from Anlloto An5 l ' Th e more sodic cores are found in thelate flows, most of which are basaltic andesite.Groundmass plagioclase is generally more sodicthan the phenocrysts, and sometimes is normallyzoned. Its average composition is about An4a•

When hypersthene is present it is alwayssurrounded by a rim of either quasi-uniaxialpigeonite or pigeonitic augite. Augite pheno­crysts usually do not show disequilibrium withthe surrounding forme r liquid, though in somethin sections they are slightly embayed. Ground­mass pyroxene ranges from entirely pigeonitein the more basic rocks to combinations ofpigeonite, pigeon itic augite, and augite in theandesites and dacites. Small magnetite crystalsare presen t in the groundmass, and some largegrains appear in glomerocrysts of plagioclase,augi te, hypersthene, and, in rare instances, pi­geonite. Interspersed throughout the ground­mass of the more silicic varieties are silicaminera ls which have been identified as quartzand tridymite. Neither alkali feldspar nor ap­atite was observed.

Small elongate angular inclusions of coarserrock are common in the more siliceous of theHamatua lavas. They may be the result of localretention of volatiles during cooling. Quartzand tridymite are more abundant in these in­clusions than in the rest of the rock.

Hoenla Froth Lava

Forming the sea cliff on the northern andwestern shores of Tofua is a rock which inhand specimen appears tuffaceous, but whichin outcrop shows evidence of emplacement byflowage. It resembles rocks which in other partsof the world have been called "froth lava"( Locardi and Mittempergher, 1965) . It is herenamed the Hokula Froth Lava, after the nearbyHokula village .

The Hokula Froth Lava directly overlies theHamatua Formation and is separated from it byan erosional unconformity. It is overla in byrocks of the Kolo Formation and is exposedonly in the sea cliff. Consequent ly, its extentcannot be determined. It probably came fromthe upper part of the mounta in, but it has notbeen found in the caldera wall.

T he type section of the Hokula Froth Lava

340

is approximately 0.1 mile west of Hokula villagewhere two separate flow units of froth lavaform a wave-cut cliff about 80 feet high . Thelower of these units has been oxidized to abright red color, whereas the upper is gray togreen. Structurally, both units are ident ical, eachconsisting of three parts : a massive central por­tion grading into clinker zones above and belowit. Usually, the lower clinker layer is thickerthan the upper one. The basal clinker of thelower flow is approximately 5 feet thick, con­sisting of cobble-size blocks that were slightlywelded together after the flow came to rest.The massive central portion of the flow hasan average thickness of 10 feet. In it there arelenticular masses that are lighter in color thanthe surrounding rock. The masses are composedof the same material as the enclosing rock, butare more vesicular. A thinner clinker layer over­lies the massive center with a gradational con­tact. Its nature is the same as that of the basalclinker. The upper flow unit repeats the se­quence described, the only difference being thatit is thinner, with a total thickness of onlyabout 15 feet.

Columnar jointing is poorly developed in thefroth lava, each of the columns being severalfeet across. The ocean waves erode the frothlava with ease, and the squarish caverns whichthey cut owe their shape to the joint pattern.

In thin section, the Hokula Froth Lava showsa great many microvesicles, which range inshape from round to angular. The rock is highlyvitreous , containing small microlites of plagio­clase and green pyroxene, and phenocrysts ofhypersthene, augite, and plagioclase in a glassmatrix. In the lower flow unit the glass is red;in the uppe r one it is greenish gray. Most ofthe glass is heavily clouded with dusty opaquematerial, but small amounts of clear glassare present in both units . The refractive indexof the clear glass ranges from 1.552 to 1.535( -+- .002). According to a graph published byGeorge (1924, p. 365) of refractive indexversus silica percentage, the froth lavas areandesitic, Tridymite has been identified in someof the samples, usually in the vesicles. Smalllithic inclusions are present.

Locardi and Mittempergher (1965, 1967)have described froth flows in central Italy. Theybelieve that the froth lava is the result of gas

PACIFIC SCIENCE, Vol. 24, July 1970

separation in a volatile-rich flow, and representsa transition between ordinary lava flows andignimbrites. If vesiculation is extreme, the rockdisintegrates and becomes an ash flow; but ifvesiculation does not proceed far enough tocause disintegration, the rock remains pumiceousfroth lava.

Lenticular pumice masses have been describedby Locardi and Mittempergher (1967, pp. 133­134) within the central portion of the Italianfroth lavas. They believe the elongated massesof pumice are caused by the pulling apart ofcertain bands where vesiculation was greatest.On Tofua Island the lenses of pumiceous ma­terial are found randomly scattered throughthe central portions of the flow units, but withtheir longest axes parallel to the top of theflow.

Inclusions in the Hokule Froth Lava

Angular cobble-size xenoliths are common inthe Hokula Froth Lava. Less abundant arerounded boulder-size xenoliths. Both types arefragments of rocks of the Hamatua Formation,or of other rocks presumably from the base­ment, brought to the surface by the risingmagma. The rocks range from fine to mediumgrain, the fine-grained varieties predominating.

The fine-grained xenoliths are texturally andmineralogically similar to the rocks of theHamatua Formation described above. Thecoarser grained rocks, however, appear to havecrystallized under deeper-seated conditions, andhave been identified as augite diorite micro­pegmatite. Intergrowth of quartz and plagio­clase feldspar is common and occurs as a matrixbetween the larger grains of normally zonedsubhedral plagioclase and greatly altered augite.Most of the augite has been altered to a low­birefringent green mineral or mineraloid . Onesample was stained for alkali feldspar, with anegative result. Except for the quartz-feldsparintergrowth, the texture between the largercrystals is hypidiomorphic granular.

Some of the finer grained xenoliths do notresemble the rocks of the Hamatua Formationin the nature of their pyroxenes . Frequentlythey contain phenocrysts of pigeonite, pigeoniticaugite with cores of pigeonite, augite, and hy­persthene with pigeonitic rims. The bulk ofthe groundmass pyroxene is pigeonite. These

Geology of Tofua Island-c-Bxtran

rocks commonly are coarser grained than theHamatua lavas, but are not typically plutonic.They may have been derived from shallow in­trusive bodies. Whether or not these rocks aredirectly related genetically to the Tofua volcanois not known.

Kolo Formation

The Kolo Formation consists of consolidatedlapilli tuff-breccia, with local deposits of tuffand interbedded rootless basaltic andesite lavaflows, unconsolidated ash and cinder, and onethick lava flow of porphyritic hypersthene-augiteandesite. The lapilli tuff-breccia is by far themost abundant member. It is best seen near thevillage of Kolo .

The lapilli tuff-breccia is an air-laid deposit,poorly sorted, with weakly developed bedding.It is composed of approximately 50 percentangular lapilli-size blocks and cinder, 40 per­cent ash, and 10 percent blocks larger thanlapilli . Locally, however, the proportions changeand zones of tuff become abundant. Clasts oflapill i size and larger appear to have been de­rived from the Hamatua Formation. The ashand cinder are commonly vitreous and representnew magmatic material. The refractive indexof the glass in them shows it to be basaltic.Within the lapilli tuff-breccia are small rootlessbasaltic andesite flows, which microscopicallyare similar to the andesitic lavas of the HamatuaFormation but relatively poor in phenocrysts. Aflow near Hokula has been identified as hypers­thene- and augite-bearing basaltic andesite . Thephenocrystic feldspar is bytownite, and thegroundmass feldspar is labradorite.

The consolidated lapilli tuff-breccia memberrests unconformably upon the Hamatua Forma­tion on the southern, eastern, and northernflanks of Tofua. It forms a blanket approx­imately 100 feet thick over the entire area,though its thickness increases where it fills val­leys previously cut into the Hamatua rocks. NearHamatua village it grades into consolidatedblack ash, which is as much as 200 feet thick,perhaps because it fills a valley.

Near Hokula village the lapilli tuff-breccia isseparated from the underlying Hokula FrothLava by an erosional unconformity and a layerof tuff. This unconformity can also be seen0.25 mile west of triangulation station G (Fig.

341

2A). There, a deep valley cut into the frothlava is filled with 150 feet of interbedded ashand lapilli tuff-breccia. Conformably overlyingthis sequence is a thick and extensive porphyritichypersthene-augite andesite block-lava flow(analysis 6, Table 1) . The flow, which is 70 feetthick, forms a broad plateau in the vicinity oftriangulation station G. In thin section thegroundmass of the flow is seen to be hyalopilitic,with microlites of labradorite, clinopyroxene,and magnetite. Hypersthene and augite phe­nocrysts, usually clumped together as glomer­ocrysts, appear ragged and have been embayedby the groundmass. Phenocrysts of pigeonitealso are present.

Source vents of the consolidated lapilli tuff­breccia dot the northeastern, eastern, and south­ern caldera rims. These vents are commonlyshallow, with gradually sloping walls. Theyresemble maars in other parts of the world . Afew small craters 30 feet across and 50 feetdeep are found on the northeastern rim. Con­sidering the number of blocks of rocks of theHamatua Formation in the tuff-breccia, and themaarlike appearance of the craters, it is probablethat the explosions that formed the tuff-brecciatook place very near the surface, and probablywere phreatic and phreatomagmatic.

The unconsolidated tephra of the Kolo For­mation is younger than the tuff-breccia. Uncon­solidated tephra overlies consolidated tuff-breccianear triangulation station B1 and on the south ­eastern rim of the caldera . The source of theunconsolidated tephra is a series of cindercones that erupted on the caldera rim. Neartriangulation station B1, recent faulting hastruncated the cinder cones, debris from whichnow forms talus resting against the calderawall. Most of the fragments in the tephra arepumiceous lapilli-size cinder, though some spin­dle bombs also are present. The refractive indexof the glass in the cinder is 1.559 (±.002),indicating that it is basaltic.

The relationship of the source vents of theKolo Formation on the caldera rim to the faultpattern is clear. Concentric faults formed duringcaldera collapse provided conduits for risingmagma . However, it is not known whether thethick andesite flow near triangulation station Goriginated at the caldera rim. Its source hasbeen buried beneath later deposits of the Iapilli

342

tuff-breccia and the pyroclastics of the LofiaFormation .

Lofia Formation

Rocks of the Lofia Formation are mainlyconfined within the caldera, though they arecontinuously exposed down to an elevation of950 feet on the northern flank of the volcano.Those outside the caldera constitute only a thinveneer, and were derived from the large spatte r­and-cind er cones in the caldera. For the mostpart they overlie the rocks of the Kolo Forma­tion, but on the caldera rim the contact betweenthe two formati ons is gradational, and in thatarea the rocks of the two formations wereerupted partly simultaneously. W ithin the cal­dera, the Lofia Formation overlies rocks of theHamatua Formation which have collapsed intothe caldera.

The name Lofia Formation is derived fromLofia cone, the still-active vent within thecaldera . Almost all of the members of theformation can be seen within Lofia cone itself.Pyroclastic rocks probably comprise 90 percentof the formation , and include lapilli tuff, tuff,spatter, and unconsolidated lapilli cinder. Theyare interbedded with small basaltic andesite andandesite lava flows. The older cone near thewestern caldera wall is composed mainly ofweathered ash. The lava flow pond ed on thefloor of Lofia crater has been mapped as aseparate unit (Fig. 2B ). Only one other lavaflow is large enough to be mapped separately.Th e flow is vesicular andesite .

The pyroclastic deposits of the Lofia Forma­tion are thin bedded and friable, with onlyrudimentary size sorting. Near the cinder coneslapilli ash predominates, whereas the walls ofthe caldera are mantled by ash with few lapilli,indicating that most of the larger fragments fellcloser to the vents. However, spatte r and "cow­dung" bombs averaging 50 ern in diameter areplastered against the eastern wall of the caldera,half a mile from their source. From their posi­tion it is clear that they were formed by adirected blast from Lofia cone.

The lapilli ash and tuff of Lofia cone containssome angular blocks. A block selected at ran­dom for thin sectioning proved to be micro­scopically very similar to a small porphyriticbasaltic andesite lava flow found interbedded

PACIFIC SCIENCE, Vol. 24, July 1970

with the tuff in the cone. Thin lava flows inter­calated in the spatter and tuff are exposed inthe walls of Lofia crater, and appear to be thesame as the block and the basaltic andesite flowon the flank of the cone. In th in section, theporphyritic basaltic andesite flow (analysis 7,Table 1) is seen to be crowded with phenocrystsof calcic plagioclase, augite , pigeon itic augite,pigeonite, augite with pigeonite cores, andolivine. In addi tion, rounded and embayedxenocrysts of augite are present. The olivinegrains do not have reaction rims, but are anhe­dral. Labradorite and pigeonite are present inthe groundmass, surrounded by dark glass. Inhand specimens small aggregates of olivine areseen.

The refract ive index of glass from samplesof cinder in the Lofia Formation ranged from1.559 to 1.563 ( ± .002), indicating silicapercentages of about 52 to 53, which are in therange of basaltic andesite . The glass is both redand green ish gray. The latter contains tinyneedles of green clinopyroxene and plagioclase.

All the lava flows of the Lofia Formationare within the caldera, and all are relativelythin when compared with those of the HamatuaFormation. Most of them appear to be rootless,but this is not certain since the large volume ofpyroclastics may cover their sources. Only onelava flow is seen to be continuous over any greatdistance. It is an andesit ic aa flow, approximately0.75 mile long and 0.25 mile wide, whichpoured out of the now-dormant south ern ventand terminated at the crater lake.

GEO LOG IC STRUCTURE

Fallits

The largest structu ral features on Tofua arefaults associated with the caldera collapse. Thesecan be grouped into three main classes: (1) themain faults bounding the sunken block of thecaldera, (2 ) faults parallel to the main caldera­boundary faults along which volcanic eruptionshave taken place, ( 3) faults along the rimof the caldera along which eruptions have nottaken place.

Th e maximum displacement of the caldera­bound ary faults is more than the 1,500-footheight of the caldera wall, since the base ofthe wall is buried by later volcanics. Even

Geology of Tofua Island-i-Bauna

though part of the caldera-boundary fault zoneis buried, its trace is visible as recent scarpletsin the volcanics of the Lofia Formation.

Although the caldera-boundary fault is shownin Figure 2A as a single continuous fault, inactuality it appears to be a zone of many closelyspaced fractures . Along the southern lakeshorethe caldera-boundary fault zone consists of twomajor faults, between which is a large down­dropped portion of the rim that has the formof a giant step, with its surface sloping west­ward approximately 9°. Cinder cones along thefault that forms the east side of the stephave erupted the unconsolidated tephra of theKolo Formation.

The faults of the second class are the mostnumerous of the three . They are concentric tothe caldera and are found on its northern,eastern, and southeastern rims outside the maincaldera-boundary faults . They seem to representan outward spreading of the zone of calderacollapse, but as yet displacements on them havebeen relatively small. They have acted as con­duits for rising magma; along the eastern rimof the caldera were the loci for the source ventsof the Kolo Formation. It is noteworthy thaton Tofua no eruptions have taken place onradial fissures.

Faults of the third class comprise the high­angle normal faults occurring on the easternrim of the caldera. Those associated with faultsof the second class usually are found closer tothe caldera . Commonly two or three parallelfaults have given rise to step faulting andgrabens . Vertical displacement on individualfaults is as much as 50 feet, but averages about25 feet. These faults represent rather superficialslumping along the main caldera-boundaryscarps.

A few faults within the caldera north andeast of Lofia cone, which have not been theloci of eruptions, are believed to be part ofthe caldera-boundary fault zone.

T ensional Cracks

Tensional cracks extend along the southeast­ern rim of the caldera, but are not found else­where. Many are a few inches wide, thoughsome are several feet wide and are open to anestimated depth of 60 feet. Little or no verticaldisplacement has occurred on them. The larger

343

cracks are concentric to the caldera, while thesmaller ones meander, following local jointpatterns. They are found with faults of thesecond and third classes, and it is inferred thatthe cracks may later develop into faults.

Pseudo-folds

Fold-like structures on Tofua were formedby mantling of preexisting topography by vol­canic rocks. Near triangulation station F, mantlebedding in the tuff-breccia of the Kolo Forma­tion along the caldera rim dips 20° in bothdirections from the rim and resembles a smallanticline. A similar structure is found wherepyroclastic rocks of the Lofia Formation reston the northern rim of the caldera. Structuresresembling plunging synclines are found alongthe western coast, where froth lava flows fol­lowed valleys cut into the precaldera surface.At the end of the froth lava eruption the fluidparts of the flows drained away, lowering thecenter of the flow but leaving chilled lavaadhering to the valley walls. The resulting lavasheet is U-shaped in cross section and resemblesa syncline. Similar "plastering" lava flows arefound on the flanks of Haleakala volcano inHawaii (Stearns and Macdonald, 1942, pp. 97-­98). No true folding was found on Tofua.

MINERALOGY

Plagioclase Feldspar

Plagioclase is by far the most abundant min­eral in the rocks of Tofua. Most of the rockscontain phenocrysts of plagioclase, and all haveplagioclase in the groundmass. Even vitreouscinder contains fine needles of plagioclase.

The plagioclase is usually subhedral. In somerocks some phenocrysts are euhedral, whereasother nearby phenocrysts of the same composi­tion are deeply corroded . Other rocks containphenocrysts of widely varying composition.Phenocrysts of plagioclase range in size fromabout 2.0 to 0.3 mm long; groundmass lathsrange from about 0.5 to 0.08 mm. The smallersizes are prevalent in the andesites. In theandesite and dacite of the Hamatua Formation,plagioclase phenocrysts often form glomerocrysts,either alone or with other minerals .

Whether as phenocrysts or in the groundmass,plagioclase laths usually show normal zoning,

344

although reverse zoning may also be present insome crystals. Normal oscillatory zoning is com­mon, and is probably due to a combination, assuggested by Vance ( 1962, p. 751 ) , of fallingtemperatures and fluctuation in the supersatura­tion and unders aturation of the melt in the con­stituents of the feldsp ar. Rounded cores are verycommon in the zoned phenocrysts. The com­position of the cores ranges from An90 to An40,whereas that of the rims ranges from An53 toAn32' Zoned groundmass plagioclase is moresodic, the cores ranging from An59 to An38' andthe rims from An40 to An35.

Some of the plag ioclase phenocrysts andlarger groundmass laths contain inclusions ofglass, magnetite dust, and microlites of clino­pyroxene. Similar inclusions have been reportedby Kuno (1950, p. 968), Swanson (1966, p.1303) , and Macdonald and Katsura (1965) .They are common in orogenic andesites andrelated rocks.

The forms of the inclusions are qu ite varied.In some cases they occur only as elongat ed blebsusually parallel to the (010) face in the core ofthe crystal. In the plagioclase of glomerocryststhe inclusions tend to form in zones of roundedblebs, giving the appearance of embayment andsubsequent growth of the feldspar grains. Insome examples blebs cut across oscillatoryzoning. Commonly the inclusions form at thejunction of the (001) plane with the ( 010)plane. When this occurs, the bleb may have atriangular cross section, suggesting that meltingof the crystal took place along lines of inherentweakness in the crystal structure. In some casesalmost all of the plagioclase crystal is crowdedwith dark blebs, but more commonly inclusionsare found in only about 25 percent of the lath .

Kuno (1950, pp . 967-968) explains the oc­currence of the glassy inclusions in the plagio­clase by a remelting of the host crystal along itscrystal faces when it is enclosed in a magma thatis in equilib rium with more calcic plagioclase.He considers the plagioclase gra ins that showthis phenomenon to be xenocrysts. Rapid crystalgrowth also can explain the trapping of blebsof liquid magma and mineral grains within theplagioclase.

Most of the feldspar phenocrysts in thebasaltic andesites of the Lofia Formation arecrowded with glassy inclusions, whereas the

PACIFIC SCIENCE, Vol. 24, July 1970

more acidic rocks of the earlier Hamatua Forma­tion contain only small amounts of plagioclasewith inclusions, and the coarse-grained xenolithsin the froth lava do not have any glassy blebs inthe feld spar. This suggests a quieter magmatichistory for the older rocks, and a more complexhistory, pr obably with a greater amount of con­tamination, for the more recent magmas.

Pyroxene

Pigeonite is the most abundant pyroxene inthe groundmass of the Tofua rocks. It may alsooccur as ph enocrysts in some basaltic andesitesand rarely in andesites. The optic plane of thepigeonite is parallel to (010) , indicating that itis a calcium-rich variety (Trager, 1959, p. 57) .Large pigeonite grain s usually are round ed andembayed, though large subhedral lath -shapedgrains do occur in a late basaltic andesite flow ofthe Hamatua Formation . Pigeonite is presentas cores surrounded by rims of either augite orpigeonitic augite in the basaltic andesite flow ofthe Lofia Formation.

Pigeonitic augite is a common constituent ofthe groundmass, where its shape is generallyanhedral. Phenocrysts of pigeon itic augite aretypical of the basaltic andesites, but are rare inthe other rocks. The phenocrysts sometimes havecores of pigeonite, and when this is the case the2Vz of the pigeonitic augit e tends to be small.On the other hand , if there is no core ofpigeonite, the 2V. of the pigeonitic augite ap­preaches that of true augite (4 0° ) .

Some reaction rims around hypersthene ap­pear to be pigeonitic augite, but the grains areso small that identification of the mineral is notcertain .

Augite is the commonest phenocryst in therocks of Tofua . It occurs also in the ground­mass, but it is not as plentiful there as arepigeonite and pigconitic augite. As a phenocryst,it is often a single crystal, but it is sometimesassociated with plagioclase, hypersthene, andmagnetite in glomerocrysts. Grain shape variesgreatl y from rock to rock, and sometimes withina single rock. The majority of aug ite phenocrystsare subhedral laths showing some resorpti on,but euhedral grains are not uncommon. Deeplyembayed and rounded xenocrysts of augite arepresent in the basaltic andesite flow of the LofiaFormation.

Geology of Tofua Island-s-Baunn

Augite phenocrysts range from 3.0 to 0.5 mmin length, whereas groundmass augite is com­parable in size to the pigeonite. In the augiteandesites of the Hamatua Formation the pheno­crysts are relatively uniform in size throughoutthe rock. In the more basic rocks there is greatervariation in size.

Some augites are slightly pleochroic, withX = green, Y = green, and Z = grayish green.However, pleochroism is not common; crystalsare usually gray green in thin section. The opticaxial angle usually is between 44° and 47°. Anaugite phenocryst with 2Vz = 60° was foundin the thick porphyritic augite-hypersthene an­desite flow in the Kolo Formation. Extinctionangles measured from flash figures show ZAc =44 °, and the refractive index Ny averages 1.692(± .002) . Hence the average weight percentcomposition of the augite is Mg47Fe2sCa25(Heinrich, 1965, p. 219). In some instancesaugite phenocrysts have overgrowths of augitewith higher birefringence in the same crystallo­graphic orientation, indicating an enrichment ofeither calcium or iron, or both, in the outer layer.

The augite sometimes contains small inclu­sions of other pyroxenes of lower birefringence,presumably pigeonitic (though no optic figureswere obtained) . Blebs of magnetite also aresometimes present, and both may occur in thesame crystal.

Hypersthene is the least abundant pyroxenein the Tofua rocks. It has been identified onlyas phenocrysts, either alone or in glomerocrystswith augiteand/ or pigeonite, plagioclase, andmagnetite. It always has a rim of either pigeon ­ite or pigeonitic augite. It is usually present inthe augite andesites and augite dacites of theHamatua Formation, but it has not been foundin the basaltic andesites.

Olivine

Olivine was found only in the basaltic an­desite flow of the Lofia Formation, where itforms both small dunite inclusions and largeindividual grains. The latter are somewhatrounded, but have no reaction rims.

The olivine has an optic axial angle close to90 °. Its refractive indices are N, = 1.655,N, = 1.673, and N, = 1.690 (-1-.002) . Itscomposition is about F09 0 (Heinrich, 1965,p. 145).

345

Silica Minerals

Quartz and tridymite occur together in themore silicic andesites. They are also present insome inclusions in the Hokula Froth Lava.They commonly line vesicles or occur as micro­veinlets showing micropegmatitic intergrowthwith plagioclase. This intergrowth is coarserthan the surrounding rock, and it is postulatedthat the coarser grain of the veinlets resultedfrom retention of volatiles locally after the endof the main period of crystallization of the rock.

Opaque M inerals

Magnetite commonly forms euhedral to sub­hedral grains averaging 0.05 mm across. Her­ringbone magnetite is sometimes found be­tween the feldspar grains of the groundmass ,but it is rare. In a pyroxene dacite flow largeoctahedra of magnetite are found in glomero­crysts of plagioclase, augite, and hypersthene.Magnetite dust is very abundant in blebs ofglass that are included in some plagioclase andaugite phenocrysts.

Ilmenite also is present in the Tofua rocks,but it is much less abundant than magnetite.

CHEMICAL ANALYSES

Seven rocks from Tofua were analyzed chem­ically. The analyses, together with their respec­tive CIPW norms and modal compositions ofthe rocks, are given in Table 1. The HokulaFroth Lava itself was not analyzed because ofits high degree of oxidation . Each of the modesis based on a count of 1,000 points in thin sec­tions of the analyzed samples. The compositionsof the rocks are seen to be similar to those ofrocks from other parts of the circum-Pacific re­gion (Kuno, 1950, p. 1004; 1967, pp. 647-648;Coats, 1967, pp . 694-717). Rocks of thecircum-Pacific orogenic belt commonly containhigher percentages of CaO than comparablerocks of most other parts of the world .

Kuno (1950, 1967) divided the andesitesand related rocks of the Izu-Hakone region ofJapan into two suites, the hypersthenic and thepigeonitic. The pigeonitic suite contains pigeon­ite in the groundmass, whereas the hypersthenicsuite is oversaturated with silica and containshypersthene in the groundmass . According toKuno (1950, 1959), the magma of the hy-

TABLE 1

CHEMICAL ANALYSES OF ROCKS FROM TOFUA ISLAND

(Shiro Imai, analyst)

SAMPLE NUMBER*

Si02Al20aFe203FeOMgOCaONa20K 20H 20+H20­MnOTi02

P20"Total

Total Feas FeO

Qoraban

h {eny fs

mti1ap

FeldsparClinopyroxeneOrthopyroxene

Silica mineralsOpaque mineralsOlivineGlassVesicles

56.0115.38

3.636.924.469.862.070.490.610.270.170.710.14

100 .75

10.19

14.22.8

17.331.1

7.03.82.97.45.85.31.40.3

Aba6

17.017.6

1 .013.6

42 .03.8

2

57.1314.59

2.588.434.229.172.230.530.620.210.180.740.15

100 .81

10.75

14.23.3

18.928.1

7.03.63.27.09.41.43.70.3

Ab 40

21.433.2

3.615.6

25.80.4

3

65 .8613.27

1.836.121.846.013.130.880.530.240.160.670.23

100 .77

7.77

27 .75.0

26 .219.73.41.02.53.66.52.61.20.7

Ab"7

42.430.0

0.86.19.2

9.81.6

1

55.9316.596.254.413.15

10.031.980.540.500.210.160.730.15

100 .63

10.03

CIPW Norms

i8 .63.3

16.834.8

5.94.31.13.60.99.01.40.3

Ab 3a

Modes

58.819.2

0 .48.68.2

4.8

5

56.0315 .83

3.477.323.619.952.290.520.480.150.160.770.11

100 .72

10.44

13.63.3

19.431.1

7.33.53.75.55.55.11.50.3

Ab 38

39.439.2

12.49.4

6

54.4216.564.655.324 .71

10.892.000.400.610.250.160.640.12

100 .76

9.50

13.02.2

16.835 .0

7.51.92.16.93.06.71.20.3

Ab32

12.26.82.6

13.4

28 .636.4

7

53.8014.47

2.717.077.34

11.921.480.330.510.290.170.490.09

100 .67

9.51

10.62.2

12.631.711.16.73.8

11.76.33.90.90.3

Ab 28

29 .829.8

5.49.0tr .9.0

14.0

• I. Aphanitic hypersthene-bearing augite andesite (labradorite dacite in Rittmann' s nomenclature) from a 30-foot lava flowof the Hamatua Formation, 300 feet below the caldera rim between triangulation stations Band Bl. (Field number Tofua33.)

2. Aphanitic pyroxene ande site (labradorite dacite in Rittmann's nomenclature) from a 70-foot-thick block lava flow of theHamatua Formation. 400 feet below the caldera rim between triangulation stations Band Bl. (Field number Tofua 32.)

3. Microporphyritic hypersthene-bearing augite dacite (dacite in Rittmann's nomenclature) from the Hamatua Formation at500 feet altitude behind Hamatua village. (Field number Tofua 57.)

4. Labradorite andesite (labradorite dacite in Rittmann's nomenclature), 0.1 mile northwest of Hokula village, as an inclu­sion in the Hokula Froth Lava. (Field number Tofua 68.)

5. Medium-grained labradorite augite diorite inclusion in the Hokula Froth Lava on the beach below Hokula village. (Fieldnumber Tofua 4.)

6. Porphyritic hypersthene-augite andesite (quartz-bearing labradorite andesite in Rittmann's nomenclature) from a70-foot lava flow of the Kolo Formation at tri angulation station G. (Field number Tofua 50.)

7. Porphyritic augite basaltic andesite (quartz basalt in Rittmann' s nomenclature) from a small lava flow on the southernflank of Lofia cone. (Field number Tofua 17.)

346

Geology of Tofua Island-BAUER 347

1.3./~c; 12

lJ... 4'

III I IDc 100~ 9

D 8..0 7..c 6Qlu~ 5Ql +

Q.. 4<,

12 -.II

10

0D 9

U8..

c7Ql

u~

6QlQ..

5

4

.3 -1--

0-0 -N 5~

+ "0t\I .3D .

Z2

cQl .1u~

Ql 0Q.. 50 52 54 56 58 6 0 62 64 66 68 70 72

Per c e nt Si 0 2

FIG. 3. Variation diagrams of the rocks of Tofua Island, compared with those of the pigeonitic and hyper­sthenic suites of the Izu-Hakone region in Japan . Solid dots, compositions of the Tofua lavas; crosses, averagesof basalts, basaltic andesites, andesites, and dacites of the pigeonitic suite of Izu-H akone; open circles, averagesof basaltic andesite, andesite, and andesite-dacite of the hypersthen ic sui te of Izu-Hakone, D ata on the Japa­nese rocks are from Kun o (1967, pp. 647- 648 ).

348

persthenic suite was contaminated by sialiccrustal material, whereas the magma of thepigeonitic suite was not. Later, Kuno (1 965,1967) , following Osborn ( 1962, pp. 211-226),suggested that the hypersthenic ( orogenic)suite could be derived from tholeiitic, h igh­alumina, and alkalic basalt magmas by differen­tiat ion unde r high partial pressure of oxygen,resulting in early crystallization of magnet ite.In Figure 3 the compositions of the rocks ofTofua are compared with the variation curvesfor the rocks of the pigeonitic and hypersthenicsuites of Izu-Hakone, and with the averagecompositions of the principal rock types ofthose suites. It will be seen ,that the CaO con­tent of the Tofua rocks is higher than the av­erage in either of the Izu-Hakone suites. Thehigh CaO content is reflected in the norm cal­culations in Table 1 in the calcic nature of theplagioclase. Every rock except that of column 3contains normative labradorite to bytownite.The high CaO content also explains the calcicnature of the plagioclase actually found in therocks, even in the most siliceous varieties. Thenorm calculations show a high wollastonitecontent, which is reflected in the mode by calcicclinopyroxenes. Augite and pigeonite are veryabundant, whereas noncalcic pyroxene (hyper­sthene) is rare.

The alkali :silica diagram at the bottom ofFigure 3 shows a nearly linear increase in thealkalies with increasing silica, but the total al­kali content is less than in the rocks of Izu­H akone. It is interesting to note the paralleltrends between the three suites. The K20 con­tent is low and is similar in the two regions.Dickinson ( 1968, pp. 2261-2267) notes thatK20 usually is low in andesites which areerupted near the trench side of the island arcs,which is the case with both the Tofua and theIzu-Hakone rocks. However, even the most sili­ceous Tofua sample does not display a largeincrease in the alkalies, as do the average pigeon­itic and hypersthen ic dacites of Izu-Hakone,

When total iron is plotted against Si02 , asin the upper diagram of Figure 3, the iron con­tent of the Tofua rocks falls in between thoseof the two Izu-Hakone suites when the Si02

content is less than 58 percent, though it isgreater than those of the Japanese rocks athighe r contents of Si02 • In the pigeonitic series

PACIFIC SCIENCE, Vol. 24, July 1970

of Izu-Hakone, iron enrichment takes place ata lower Si02 content than in the Tofua rocks.

N ormative orthoclase in the Tofua rocksranges from 2.22 percent in the basaltic andesiteto 5.00 in the dacite, but modal alkali feldsparwas not observed. N ormative orth oclase in theTofua rocks is comparab le in amount to that inthe Izu-Hakone rocks. Normative quartz tendsto be more abundant in the Tofua rocks than inthe Japanese rocks, but the average dacite of thehypersthen ic suite of Izu-Hakone has a norma­tive quartz value close to that of the augitedacite of the Hamatua Formation (Table 1, col­umn 3).

The alkali-lime index for the Tofua lavas is63.5, which falls with in the calcic division ofPeacock's ( 1931 , pp . 54-67 ) classification. TheTofua rocks belong to the calc-alkaline orogenicsuite.

Data on minor elements in lavas of the TongaIslands, including Tofua, will be published byHu bbard and Gast ( in preparation ) .

GEOLOGIC HI STORY

The early geologic history of the Tonga Ar­chipelago is unkno wn. The coralline islands ofTonga are primarily composed of Eocene lime­stone.

The eruption of the Hamatua lavas, build ingthe main cone of To fua Island, may have takenplace during the late part of the Pleistoceneepoch. No radioactive dates have as yet beenobtained for the Hamatua rocks, and the age ofthe cone can only be estimated from the amountof erosion it has undergone.

A period of quiescence followed the eruptionof the Hamatua lavas, during which moder­ately deep valleys were cut into the side of thecone. Eventually, the summit of the cone col­lapsed to form the caldera. It is not knownwhether the init ial caldera collapse took placebefore or after the eruption of the H okulaFroth Lava. Field and petrograph ic evidenceindicate that the froth lava contained a largeamount of volatiles. It is possible that duringthe quiescent period following the eruption ofthe Hamatua lava, volatiles were concentrated inthe upper part of the magma chamber, and thatit was the erupti on of the froth lava that ini­tiated the caldera collapse. An unknown amount

Geology of Tofua Island-c-Bxu s n

of the frot h lava flowed into the sea, and thevolume erupted may have been much greaterthan that now remaining above sea level.

Following its eruption, streams cut valleysinto the Hokula Froth Lava. Tensional cracksand normal faults associated with the contin­ually enlarging caldera provided conduits forrising magma which erupted at points alongthe caldera rim, building ash and cinder cones,mantling a large par t of the outer slope ofTofua with pyroclastics, and filling the valleysin the H okula Froth Lava. By the time of erup­tion of the Kolo Formation the composition ofthe magma had become more basic than that ofthe precaldera per iod, although occasional erup ­tions of andesite still occurred.

Within the caldera, rocks of the Lofia Forma­tion were erupted, par tly concurrently with erup ­tions of the Kolo Formation, but cont inuingafter the Kolo eruptions had ceased. Frequenteruptions of Lofia cone have taken place in his­toric time. The last was in 1959 (Richard, 1962,p. 16).

ORIGIN OF THE TOFUA ROCKS

For the most part the features of the Tofuarocks are typical of rocks of the circum-Pacificorogenic belt. Basaltic andesites predominate,with lesser amounts of andesite and basalt, anda little dacite. The problem of the origin of theTofua rocks is a part of tha t of the origin of thecircum-Pacific orogenic suite in general. BothMacdonald (1960) and Gorshkov (1962) havepointed out the isolation of such island arcs asthose of Tonga and the Mar ianas from the con­tinenta l platforms, and the improbability thatthe assimilation of sial has played any importantpart in their petrogenesis. The rocks of thesearcs must either have been derived directly fromthe mantle, or have resulted from partial remelt­ing of previously formed suboceanic basalts.The many ind ications of instability, reaction,and melting in the feldspar, pyroxene, andolivine phenocrysts in the Tofua rocks stronglysuggest the remelting of older basaltic rocks.

ACKNOWLEDGMENTS

Without the hospi tality and help of MakaIfavalu and his family, of Hokula, and that of

349

Tulanga Vaea, of Kolo, the field work on Tofuawould have been virtually impossibl e. Dr. GaryD . Stice gave aid and guidance in the field, andDr. G. A. Macdonald, Dr. R. Moberly, Jr., andDr. K. A. Pankiwskyj criticized this manuscript.The chemical analyses in Table 1 were made byShiro Imai, at the Japan Analyti cal ChemistryResearch Institute in Tokyo.

LITERATURE CITED

COATS, R. R. 1967 . Basaltic andesites . In :Basalts, The Poldervaart treatise on basalticrocks, vol. 2, pp . 689- 736. New York,Interscience Publishers.

COOK, JAMES, and GEORGE FORESTER. 1777.A voyage round the world performed in HisMajesty 's ships the Resolution and Adventure,in the years 1772, 1773, 1774, 1775, vol. 4.Dublin, 518 pp .

DALY, R. A., 1916, Petrography of the Pacificislands . Bulle tin of the Geological Society ofAmerica, vol. 27, pp . 325-344 .

DALYELL, JOHN. 1812 . Shipwrecks and disas- .ters at sea, vol. 3. Dublin, Ramsay and Co.515 pp .

DICKINSON, W. R. 1968. Circum-Pacific andes­ite types. Journal of Geop hysical Research,vol. 73, pp . 2261- 2269 .

EMMONS, R. C. 1943. The universal stage.Geological Society of America, Memoir 8,205 pp.

GEORGE, W. O. 1924. The relation of thephysical p roperties of natural glasses to theirchemical composition . Journal of Geology,vol. 32, pp . 353-372.

GORSHKOV, G. S. 1962. Petrochemical featuresof volcanism in relation to the types of theearth's crust. In : Crust of the Pacific Basin,pp. 110-115. Geophysical Monograph 6,American Geophysical Uni on, Washington,D.C.

HARKER, ALFRED. 1891. Notes on a collectionof rocks from the Tongan Islands. GeologicalMagazine, vol. 8, pp . 250- 258.

HEINRICH, E. W . 1965. Microscopic ident ifica­tion of minerals. New York , McGraw-Hill.414 pp.

HOFFMEISTER, J. E. 1932. Geology of Eua,Tonga. B. P. Bishop Museum Bulletin 96,93 pp.

350

HUBBARD, N. J ., and P. L. GAST. In prepara­tion. Some chemical evidence about the originof Tonga lavas.

KUNO, HISASHI. 1950. Petrology of Hakonevolcano and adjacent areas, Japan. Bulletin ofthe Geological Society of America, vol. 61,pp.957-1020.

- -- 1959. Origin of Cenozoic petrographicprovinces of Japan and surrounding areas.Bulletin volcanologique, vol. 20, pp. 37- 76.

- - - 1965. Origin of calc-alkaline rock series.Special Papers of the Geological Society ofAmerica, vol. 82, p. 116. (Abstract.)

- - - 1967. Differentiation of basalt magmas.In : Basalts, The Poldervaart treat ise on ba­saltic rocks, vol. 2, pp. 623-688. N ew York ,Interscience Publishers.

LISTER, J. J. 1891. Notes on the geology of theTongan Islands. Quarterly Journal of theGeological Society of London , vol. 47, pp .590- 617.

LOCARDI, E., and M. MITTEMPERGHER. 1965.Study of an uncommon lava sheet in theBolsena District (Central Italy) . Bulletin vol­canologique, vol. 28, pp . 75-84.

--- 1967. On the genesis of ignimbrites,how ignimbrites and other pyroclastic prod­ucts origin ate from a flowing melt. Bulletinvolcanologiq ue, vol. 31, pp. 131-152.

MACDONALD, G. A. 1949. Hawaii an petro­graphic province . Bulletin of the GeologicalSociety of America, vol. 60, pp . 1541-1 596 .

- - - 1960. Dissimilarity of continental andoceanic rock types. Journal of Petrology, vol.1, pp . 172-177.

- - - 1967. Forms and structures of extrusivebasaltic rocks, In : Basalts, The Polderv aarttreatise on rocks of basaltic composition,vol. 1, pp . 1-61. New York, IntersciencePublishers.

PACIFIC SCIENCE, Vol. 24, July 1970

MACDONALD, G. A., and TAKASHI KATSURA.1965. Eruption of Lassen Peak, CascadeRange, California, in 1915 : example of mixedmagmas. Bulletin of the Geological Societyof America, vol. 76, pp . 475-482 .

MARSHALL, PATRICK. 1911. Oceania . Hand­buch der regionalen Geologie, vol. 2, pp .1- 36. Heidelberg.

OSBORN, E. F. 1962. Reaction series for sub­alkaline igneous rocks based on different oxy­gen pressure conditions. American Mineralo­gist, vol. 47, pp . 211-226.

PEACOCK, M. A. 1931. Classification of igneousrocks. Journ al of Geology, vol. 39, pp . 54­67.

RICHARD, J. J. 1962. Catalogue of the activevolcanoes of the world includ ing solfatarafields, Part 13. Kermadec, Tonga, and Samoa.Int ernational Association of Volcanology,Napoli. 38 pp.

RITTMANN, ALFRED. 1952. Nomenclature ofvolcanic rocks. Bulletin volcanologique, vol.12, pp . 75- 102.

STEARNS, H. T., and G. A. MACDONALD. 1942.Geology and ground-water resources of theisland of Maui, Hawaii. Hawaii Division ofHydrography, .Bulletin 7, 344 pp .

SWANSON, D. A. 1966. Tie ton Volcano, a Mio­cene eruptive center in the southern CascadeMountains, W ashington. Bulletin of theGeological Society of America, vol. 77, pp .1293- 1314.

TROGER, W. E. 1959. Optische Bestimmungder gesteinsbildenden Mine rale. Stuttgart ,Schweizerbart' sche Verlag. 147 pp .

VANCE, J. A. 1962. Zoning in igneous plagio­clase: normal and oscillatory zoning. Ameri­can Journal of Science, vol. 260, pp . 746- 760.

WOlDE, CHARLES (translator) .1776. Abel Tas­man 's 1643 Journal, Microfilm copy 1205,University of H awaii Library.