How Many Pacific Hotspots Are Fed by Deep-Mantle Plumes?

How many Paciﬁc hotspots are fed by deep-mantle plumes?

ValeÂrie Clouard* Jeune Equipe Terre-Oceán, Universite´ de la Polyne´sie Française, B.P. 6570, Faaa, French Polynesia Alain Bonneville* UMR Geósciences Marines, Institut de Physique du Globe, 4 place Jussieu, 75005 Paris, France

ABSTRACT The Paci®c plate is usually considered to host 14 hotspots, but most of this volcanism does not seem to originate from deep-mantle plumes. To reach this conclusion, we tried to establish how many of the seamount alignments on the Paci®c plate correspond to classic hotspots, i.e., long-lived hotspots linked to oceanic basaltic plateaus. We retraced the tracks of the 14 Paci®c hotspots by using (1) the absolute stage poles representing the Paci®c plate absolute motion since 145 Ma, (2) an updated compilation of radiometric ages of seamounts and oceanic plateaus totaling 266 seamounts or islands, (3) the detailed bathymetry of the Paci®c Ocean, and (4) the present locations of the hotspots. This analysis allowed us to correlate only three hotspots with the beginning of their tracks possibly corresponding in space and time to an oceanic plateau: Easter to the eastern Mid-Paci®c Mountains, Louisville to the Ontong Java plateau, and, with less con®dence, Marquesas to Hess Rise and Shatsky Ridge. In addition, the Hawaii hotspot has produced long-lived volcanism. These four are the only classic hotspots on the Paci®c plate. However, seven hotspots present short tracks (Ͻ35 m.y.) that cannot be traced to an oceanic plateau and thus are not related to any deep-mantle phenomena: Foundation, Macdonald, Pitcairn, Rarotonga, Rurutu, Samoa, and Society. The two northeast Paci®c hotspots, Kodiak-Bowie and Cobb, and the Caroline hotspot are unclassi®able because of close proximity to a subduction zone where the prior history of volcanism has been lost.

Keywords: Paci®c plate, mantle plume, hotspot track, oceanic plateau.

INTRODUCTION trainment of the surrounding mantle; the head compilation of Smith and Sandwell (1994) Along with the fracture zones, transform can expand to more than 1000 km in diameter that allows detection of any seamount with a faults, and swells, voluminous oceanic pla- after impingement against the lithosphere. diameter greater than 15 km. Then the stron- teaus and linear volcanic chains are the main Then, over a short period of time, the plume gest constraint to link a seamount to a pre- morphological features of interior of the Pa- head produces extensive volcanism, and the sumed hotspot is its age. We have built our ci®c plate (Figs. 1 and 2). Oceanic plateaus plume tail subsequently continues to produce seamount age compilation (Clouard and Bon- and oceanic ¯ood basalts are thick layers of linear volcanic chains (McDougall and Dun- neville, 2000) from numerous papers, and the basalt generated near or at oceanic spreading can, 1980) for millions of years as the plate corresponding database, including all the ref- centers, usually triple junctions, but not due to drifts over the ®xed plume tail. erences, is available on line at http:// normal sea¯oor spreading (Cof®n and Eld- On the Paci®c plate, between 30ЊS and www.upf.pf/geos/data/ages࿞paci®c.pdf. Paleo- holm, 1994). They were formed by extensive 50ЊN, 10 oceanic plateaus have been recog- magnetic ages and foraminifera dates have not and voluminous volcanism over a short period nized so far, and 14 hotspots are usually iden- been considered. Only radiometric ages were of time (Richards et al., 1989). Their chemical ti®ed (Fig. 1). Our goal was to check how thus used, and Ar/Ar ages were preferred over characteristics are those of enriched mid- many of these hotspots are linked to old ¯ood- K/Ar ages, when both exist at the same place. oceanic ridge basalt, and their isotopic analy- basalt features or long-lived trails and thus are When several ages separated by less than 2 sis shows oceanic-island af®nities (Mahoney, plume-fed classic hotspots. For this, we have m.y. exist for the same seamount or island, we 1987). Both oceanic plateaus and intraplate reconstructed the different hotspot tracks by assigned the average age to the feature. The volcanic chains have been related to hotspot using four updated data sources: a detailed data set as ®nalized is composed of 266 ages theory (Morgan, 1971; Richards et al., 1989). bathymetric map of the Paci®c Ocean, the on seamounts, islands, and plateaus (Fig. 1); Morgan (1971) ®rst thought these ¯ood ba- ages of seamounts and oceanic plateaus, the the oldest feature is 138 Ma. Despite this ap- salts to be the result of the arrival of a mantle- known active hotspot locations, and the stage parently large-sounding number, age data are plume head under the lithosphere. Several ex- poles for plate motion in the hotspot reference actually very sparse over the Paci®c plate and perimental or numerical models have frame. obviously concentrated on a few seamount developed this idea (e.g., Grif®ths and Camp- alignments. The present locations of the 14 bell, 1990). These authors assumed that the DATA AND METHOD hotspots usually considered on the Paci®c source of the plume is at the thermal, possibly Linear volcanic chains, parallel to the ab- plate are reported in Table 1 with the age of chemical, boundary layer, the ``D'' layer just solute motion of the plate when they were cre- their most recent volcanic activity. Hawaii, above the core-mantle boundary. As the ated, are supposed to be the trace of a hotspot Society, Macdonald, and Pitcairn are active, plume rises, it develops a large head by en- on the sea¯oor. To identify such seamount and there are in situ observations of their vol- *E-mail addresses: ClouardÐ[email protected]; alignments, we ®rst needed a complete knowl- canism. Historic underwater activity has hap- BonnevilleÐ[email protected]. edge of the sea¯oor topography. We used the pened in Ta'u, Samoan Islands. All the other

᭧ 2001 Geological Society of America. For permission to copy, contact Copyright Clearance Center at www.copyright.com or (978) 750-8400. Geology; August 2001; v. 29; no. 8; p. 695±698; 2 ®gures; 1 table. 695 Easter Easter hotspot is one of the most puzzling hotspots of the Paci®c plate because of its un- certain location. Its name comes from its ini- tial assumed location beneath Easter Island (Morgan, 1972). Present volcanism appears in the island's vicinity on both sides of the East Paci®c Rise, simultaneously and in several locations. Bonatti et al. (1977) ®rst suggested that rather than numerous hotspots, there was a mantle hotline, corresponding to mantle con- vecting rolls. However, O'Connor et al. (1995), through the use of new Ar/Ar ages, supported the idea of a hotspot rather than a hotline on the eastern side of the East Paci®c Rise. To account for the bathymetry from the Tuamotu Islands to the Easter microplate, we located the Easter hotspot to the west of the East Paci®c Rise, in agreement with Okal and Cazenave (1985). The construction of the track shows that it could have generated the Tuamotu Islands and part of the Line Islands, south of the equator. Epp (1984) pointed out that these chains could be the result of a hotspot interacting with a ridge, west of the East Paci®c Rise. An 8 Ma seamount is a strong constraint, as well as Cretaceous ages on the western Tuamotu Islands. The origin of the track ®ts with the eastern part of the Mid- Paci®c Mountains, which are composed of guyots overlying a broad plateau; Morgan (1972) supposed that they were the continuation of the Line-Tuamotu alignment. A con- Figure 1. Locations and ages of seamounts and islands on bathymetric map (equal-area nection with the western Easter hotspot leads Hammer projection) of Paci®c plate (Smith and Sandwell, 1994). Names of 14 hotspots usu- to the same conclusion. According to our re- ally considered on Paci®c plate are in white. Present hotspot locations are indicated by white disks; black star inside disk indicates recent observed activity. Black and white num- construction, the Tuamotu Plateau is not the bers are ages (in Ma) of volcanic features considered in this study. Flood-basalt plateau end of the track. We thus agree with the in- names are given in black italics in Figure 2. terpretation of Ito et al. (1995), who suggested that this plateau is not a typical oceanic plateau. hotspotsÐCaroline, Cobb, Kodiak-Bowie, usually cannot be so linked. This is the case Easter, Louisville, Foundation, Rarotonga, Ru- for Louisville, Emperor, Kodiak-Bowie, Cobb, rutu, and MarquesasÐhave shown activity and Caroline hotspots. Because of the unusual Marquesas within the past million years. We inferred their geometry of the Tonga Trench, the inferred The Marquesas hotspot created the Marque- zero-age location by classical backtracking extrapolation of the Louisville hotspot track sas Islands between 1.3 and 5.8 Ma. A radio- from the last dated volcanic event. can be studied before 100 Ma on the Paci®c metric age of 0.5 Ma was determined for the To reconstruct the apparent path followed plate. The track of the Hawaiian hotspot, if it south of the archipelago and provides a strong by a hotspot on the sea¯oor, we have moved is assumed to be a continuation of the Em- constraint on the most recent hotspot location. the hotspot's present location back in time by peror Seamount chain, disappears in the Aleu- The predicted trend of the track disagrees with using the set of stage poles proposed by Wes- tian subduction zone, but because this hotspot the observed volcanic alignment direction. sel and Kroenke (1997). This data set repre- is used as a reference for the determination of However, if we consider a control by en ech- sents the most updated synthesis for the ab- absolute plate motion by virtue of its contin- elon fractures oblique to the regional trend, solute Paci®c plate motion since 145 Ma. uous volcanism since 65 Ma, it was consid- there is no longer any contradiction. However, However, for the last interval (0±3 Ma), we ered as a classic hotspot in the present study. the Line Islands between 5Њ and 15ЊN exhibit have chosen to use the pole previously pro- Thus, only the northeast Paci®c hotspots and dates that could agree with generation by the posed by Yan and Kroenke (1993). Caroline hotspot were not classi®ed. For each Marquesas hotspot. The link between the of the 10 remaining hotspots, the track was northern end of the Marquesas chain and the HOTSPOT TRACKS compared to the sea¯oor topography, and the southern end of the Line Islands chain could With the chosen set of stage poles in the predicted ages along the track were compared therefore correspond to the Marquesas-Line hotspot frame of reference, the backtracks to the actual radiometric ages of the sea- swell described by Crough and Jarrard (1981). were determined over the past 145 m.y. for mounts. When the shift between predicted and From north of the Hawaiian chain to north of the 14 hotspots (Fig. 2). Our aim was to link radiometric ages is less than 5 m.y. for a fea- Hess Ridge, the track ®ts with the Liliuokalani hotspots with oceanic plateaus, but hotspots ture older than 20 Ma, we concluded that the Ridge, a clear but nondated linear chain. We whose tracks disappear at a subduction zone hotspot generated the feature. thus assumed that this hotspot produced the

696 GEOLOGY, August 2001 Louisville trail could reappear on the other side of the Indo-Australian plate, if the Lou- isville hotspot is that old. There is a seamount chain north of the Tonga Trench, but no ages are known. At 125 Ma, the track is close to the southeast edge of the Ontong Java Plateau. Isotopic evidence suggests a plume-initiation model to account for the Ontong Java Plateau formation (Mahoney and Spencer, 1991), and the duration of its volcanism is supposed to be 3 m.y., between 124 and 121 Ma (Tarduno et al., 1991). However, as Mahoney and Spen- cer (1991) pointed out, both temporal varia- tions in the isotopic compositions of the lavas and a southward motion of the Louisville hotspot are needed to connect Louisville with the Ontong Java Plateau. With this restriction, the Louisville hotspot and the Ontong Java Pla- teau could be linked, although a connection with the Eltanin fracture zone cannot be com- pletely put aside.

Other Hotspots All other hotspot tracks considered in our study vanish after a short period of activity. For the Society hotspot, there is no evidence of any volcanic activity prior to that of the Society Islands. Maupiti is the oldest island (4.3 Ma), and there is no young oceanic pla- Figure 2. Hotspot tracks on Paci®c plate. Tracks are represented by continuous gray lines teau in the vicinity. The Macdonald hotspot and have been computed for 0±145 Ma. Dashed lines indicate that no corresponding topo- can be backtracked to 19 Ma on the western graphic features have been found on sea¯oor and thus track segment is questionable. Lines side of the Cook-Austral alignment. The so- are not drawn before estimated beginning of hotspot life. Changes in gray along track (leg- end in upper inset) correspond to successive stage poles used for Paci®c plate motion. called Rarotonga hotspot seems to have pro- Symbols represent dated seamounts and islands. duced only the Rarotonga island, and the Sa- moa hotspot trail seems to have vanished after Hess Ridge ca. 100 Ma and the Shatsky Ridge erence frame from 66 to 12.5 Ma. Conse- the Samoan Islands were formed, prior to 14 between 145 and 125 Ma. From the estimation quently, the track exactly ®ts the hotspot trail Ma. Turner and Jarrard (1982) postulated that of its eruptive rate and magnetic anomalies, from 12.5 Ma to the Kermadec Trench. How- the Rurutu hotspot agrees with ages and lo- Sager and Han (1993) inferred a mantle-plume ever, the exact location of the active hotspot, cations of some of the Austral Islands. Its pre- origin for Shatsky Ridge. The Marquesas hot- if any, is still a matter of debate (Wessel and sent location has been determined by a 0.2 Ma spot could have generated these plateaus. Kroenke, 1997; GeÂli et al., 1998). With the age on a seamount south of Rurutu Island pole we used for the recent period, the hotspot (Bonneville et al., 2000). Its track is clear over Louisville is located near a 0.5 Ma seamount, which is the past 8 m.y. and reaches the East Mariana The Louisville chain is, with the Hawaiian- in good agreement with the interpretation of basin, which is covered by a 400-m-thick ba- Emperor chain, the volcanic alignment most GeÂli et al. (1998). The track that was formed salt layer. However, because there is just one often used to determine the Paci®c plate ab- before 66 Ma has disappeared into the Ker- dated volcanic feature between 8 and 117 Ma solute-motion stage poles in the hotspot ref- madec Trench. If the Paci®c plate is rigid, the along the predicted track of the Rurutu hotspot, we considered it as a short-lived hotspot. TABLE 1. PACIFIC PLATE HOTSPOTS Pitcairn hotspot is an active hotspot respon- sible for the Pitcairn-Gambier-Moruroa align- Name Recent Long. Lat. Beginning of the activity or activity (Ma) (ЊE) (Њ) associated oceanic plateau if any ment, but it is dif®cult to extrapolate to older ages. Because paleopositions of Mid-Paci®c Caroline 1.2 164.0 5.0 N Track Ͼ30 Ma subducted Cobb 0.12 230.0 46.0 S Track Ͼ30 Ma subducted Mountain seamounts are close to the Pitcairn Easter 8.0 245.0 26.5 S East Mid-Paci®c Mountains at 140 Ma hotspot, they have been previously linked to- Foundation Ͻ1 245.0 36.0 S 30 Ma at least gether (Winterer et al., 1993). However, our Hawaii 0 204.7 18.9 N Track Ͼ75 Ma subducted Kodiak-Bowie 0.02 225.0 53.0 N Track Ͼ20 Ma subducted reconstruction indicates a gap of more than 90 Louisville 0.5 or 12.5 221.0 50.8 S East Ontong Java Plateau at 125 Ma m.y. We thus considered Pitcairn as a short- Marquesas 0.5 223.0 11.2 S Hess Rise at 110 Ma, Shatsky Ridge at 145 Ma lived hotspot. The Foundation hotspot is a Macdonald 0 219.7 29.0 S 19 Ma near-ridge volcanic source of unknown loca- Pitcairn 0 230.7 25.3 S Maximum 10 Ma tion. From Foundation seamounts, dated be- Rurutu 0.2 208.7 23.5 S East Mariana basin at 117 Ma? Samoa 0 190.5 14.3 S 14 Ma tween 5 and 21 Ma, its track can be followed Society 0 211.5 18.2 S 4.5 Ma north of the Austral Islands along the 30 Ma Rarotonga 1.1 201.0 21.5 S 1.1 Ma Ngatemato chain. The beginning of the track

GEOLOGY, August 2001 697 could be ®tted with the Magellan Rise, which spite their apparent ®xedness, it seems unlike- tong Java oceanic plateaus: Earth and Plane- is the smallest Paci®c plateau, formed ca. 135 ly that they were related to any deep-mantle tary Science Letters, v. 104, p. 196±210. McDougall, I., and Duncan, R.A., 1980, Linear vol- Ma (Tamaki and Larson, 1988). However, phenomena. Therefore, deep-mantle plumes canic chains, recording plate motions?: Tec- more data from between 30 Ma and 135 Ma are unnecessary to explain most, if any, vol- tonophysics, v. 63, p. 275±295. are needed to con®rm this hypothesis. canic chains on the Paci®c plate, as proposed McNutt, M.K., Caress, D.W., Reynolds, J., Jordahl, All the tracks of these atypical short-lived with other arguments by Anderson (1998) in K.A., and Duncan, R.A., 1997, Failure of plume theory to explain midplate volcanism in hotspots pass through French Polynesia, a global study. the Southern Austral Islands: Nature, v. 389, which also corresponds to the South Paci®c p. 479±482. superswell. Those hotspots induced a complex ACKNOWLEDGMENTS Morgan, W.J., 1971, Convection plumes in the low- overprinting volcanism and numerous trends We sincerely thank Marcia McNutt and Don An- er mantle: Nature, v. 230, p. 42±43. Morgan, W.J., 1972, Plate motions and deep mantle of seamount and island chains since 40 Ma. derson for their careful review. This paper bene®ted from the constructive remarks of Anne Davaille, convection, in Shagam, R., et al., eds, Studies However, these trends agree with the absolute Paul Wessel, Norman Sleep, and Richard Von Her- in Earth and space sciences (Hess Volume): motions of the Paci®c plate, and these volca- zen. V. Clouard was supported by a grant from the Geological Society of America Memoir 132, nic sources, like plume-fed hotspots, seem ZEPOLYF program, funded by the French State and p. 7±22. the French Polynesian government. O'Connor, J.M., Stoffers, P., and McWilliams, also to be ®xed relative to plate motions, at M.O., 1995, Time-space mapping of Easter least for the time scale considered. Except for chain volcanism: Earth and Planetary Science REFERENCES CITED Letters, v. 136, p. 197±212. the Foundation hotspot, all of them are located Anderson, D.L., 1998, The scales of mantle con- far from the East Paci®c Rise, and all char- Okal, E.A., and Cazenave, J., 1985, A model for the vection: Tectonophysics, v. 284, p. 1±17. plate tectonic evolution of the east-central Pa- acterized by short-lived volcanism. Bonatti, E., Harrison, C.G.A., Fisher, D.E., Hon- ci®c based on Seasat investigations: Earth and Another example of short-lived intraplate norez, J., Schilling, J.G., Stipp, J.J., and Zen- Planetary Science Letters, v. 72, p. 99±116. tilli, M., 1977, Easter volcanic chain (south Richards, M.A., Duncan, R.A., and Courtillot, V.E., volcanism could be found in the northwestern east Paci®c): A mantle hot line: Journal of Paci®c, where there are 120±80 Ma seamounts 1989, Flood basalts and hot-spot tracks: Plume Geophysical Research, v. 82, p. 2457±2478. heads and tails: Science, v. 246, p. 103±107. on older crust. From gravity and bathymetric Bonneville, A., Le SuaveÂ, R., Audin, L., Clouard, Sager, W.W., and Han, H.-C., 1993, Rapid formation data, Winterer et al. (1993) proposed that V., Dosso, L., Gillot, P.-Y., Hildenbrand, A., of the Shatsky Rise oceanic plateau inferred these seamounts were formed on an earlier su- Janney, P., and Jordhal, K., 2000, Tino Mana, from its magnetic anomaly: Nature, v. 364, the missing hot spot found in the Austral Is- perswell, the Darwin Rise. The paleolocation p. 610±613. lands: Eos (Transactions, American Geophys- Smith, W.H.F., and Sandwell, D.T., 1994, Bathy- of the Darwin Rise ®ts well with the present ical Union), v. 81, p. F1368. metric prediction from dense satellite altim- location of the South Paci®c superswell, and Clouard, V., and Bonneville, A., 2000, Ages of sea- etry and sparse shipboard bathymetry: Jour- northern Paci®c seamounts are often thought mounts, islands and plateaus on the Paci®c nal of Geophysical Research, v. 99, plate: Tahiti, University of French Polynesia, to have originated from French Polynesian p. 21 803±21 824. 42 p. Tamaki, K., and Larson, R.L., 1988, The Mesozoic hotspots. Our reconstruction shows a mini- Cof®n, M.F., and Eldholm, O., 1994, Large igneous tectonic history of the Magellan microplate in mum 80 m.y. volcanic gap in the hotspot provinces: Crustal structure, dimensions, and the western central Paci®c: Journal of Geo- tracks. Hence we suppose that these sea- external consequences: Reviews of Geophys- physical Research, v. 98, p. 2857±2874. ics, v. 32, p. 1±36. mounts are not related to any present hotspots, Tarduno, J.A., Sliter, W.V., Kroenke, L., Leckie, M., Crough, S.T., and Jarrard, R.D., 1981, The Marque- Mayer, H., Mahoney, J.J., Musgrave, R., Sto- but are merely due to past, short-lived sas-Line swell: Journal of Geophysical Re- rey, M., and Winterer, E.L., 1991, Rapid for- hotspots. search, v. 86, p. 11 763±11 771. mation of Ontong Java Plateau by Aptian Dickinson, W.R., 1998, Geomorphology and geodyn- mantle plume volcanism: Science, v. 254, amics of the Cook±Austral Island±Seamount p. 399±403. CONCLUSIONS chain in the South Paci®c Ocean: Implications Turner, D.L., and Jarrard, R.D., 1982, K/Ar dating If we restrict the term plume-fed hotspots for hotspots and plumes: International Geology of the Cook Austral island chain: A test of the to long-lived hotspots and those that generated Review, v. 40, p. 1039±1075. hot spot hypothesis: Journal of Volcanology and Geothermal Research, v. 12, p. 187±220. ¯ood-basalt plateaus, then with the very con- Epp, D., 1984, Possible perturbations to hotspot traces and implications for the origin and Wessel, P., and Kroenke, L., 1997, A geometric servative hypotheses we used, only three or, structure of the Line Islands: Journal of Geo- technique for relocating hotspots and re®ning with less con®dence, four of the traditionally physical Research, v. 89, p. 11 273±11 286. absolute plate motions: Nature, v. 387, admitted Paci®c hotspots belong to this cate- GeÂli, L., Aslanian, D., Olivet, J.-L., Vlastelic, I., p. 365±369. Winterer, E.L., Natland, J.H., van Waasbergen, R.J., gory. Hawaii presents evidence of long-lived Dosso, L., Guillou, H., and Bougault, H., 1998, Location of Louisville hotspot and ori- Duncan, R.A., McNutt, M.K., Wolfe, C.J., Sil- volcanism. The Easter hotspot could be linked gin of Hollister Ridge: Geophysical con- va, I.P., Sager, W.W., and Sliter, W.V., 1993, to the east part of the Mid-Paci®c Mountains, straints: Earth and Planetary Science Letters, Cretaceous guyots in the northwest Paci®c: An and the Louisville hotspot could be linked to v. 164, p. 31±40. overview of their geology and geophysics, in Pringle, M.S., et al., eds., The Mesozoic Pa- the Ontong Java Plateau. The Marquesas hot- Grif®ths, R.W., and Campbell, I.H., 1990, Stirring and structure in mantle starting plumes: Earth ci®c: Geology, tectonics, and volcanism: spot is the most questionable, and could have and Planetary Science Letters, v. 99, p. 66±78. American Geophysical Union Geophysical produced Hess Ridge at 110 Ma and Shatsky Ito, G., McNutt, M., and Gibson, R.L., 1995, Crust- Monograph 77, p. 307±334. Ridge at 140 Ma. Seven other hotspots simply al structure of the Tuamotu Plateau, 15ЊS: Im- Yan, C.Y., and Kroenke, L.W., 1993, A plate tectonic reconstruction of the Southwest Paci®c, do not show evidence of a continuous volca- plications for its origin: Journal of Geophysi- cal Research, v. 100, p. 8097±8114. 0±100 Ma, in Berger, W.H., et al., Proceedings nic trend between basaltic plateaus and the Mahoney, J.J., 1987, An isotopic survey of Pa- of the Ocean Drilling Program, Scienti®c re- sults, Volume 130: College Station, Texas, present hotspot locations. ci®c oceanic plateaus: Implications for their Ocean Drilling Program, p. 697±709. As pointed out by McNutt et al. (1997) and nature and origin, in Keating, B., et al., eds., Dickinson (1998), it seems reasonable to think Seamounts, islands, and atolls: American Manuscript received November 20, 2000 Geophysical Union Geophysical Mono- that most of the volcanic alignments belong- Revised manuscript received March 26, 2001 graph 43, p. 207±220. Manuscript accepted April 10, 2001 ing to the South Paci®c superswell region are Mahoney, J.J., and Spencer, K.J., 1991, Isotopic ev- in fact diffuse and secondary volcanism. De- idence for the origin of the Manihiki and On- Printed in USA

698 GEOLOGY, August 2001