The Tarava Seamounts: a Newly Characterized Hotspot Chain on the South Paci¢C Superswell

Earth and Planetary Science Letters 207 (2003) 117^130 www.elsevier.com/locate/epsl

The Tarava Seamounts: a newly characterized hotspot chain on the South Paci¢c Superswell

Vale¤rie Clouard a;, Alain Bonneville b, Pierre-Yves Gillot c

a Jeune Equipe Terre-Oce¤an, Universite¤ de la Polyne¤sie franc°aise, P.O. Box 6570 Faaa, Tahiti, French Polynesia b Ge¤osciences Marines, CNRS, Institut de Physique du Globe, 4 place Jussieu, 75252 Paris Cedex 05, France c Laboratoire de Ge¤ochronologie (USP-IPGP), Sciences de la Terre, Ba“t. 501, Universite¤ Paris Sud, 91405 Orsay, France

Received 12 August 2002; received in revised form 26 November 2002; accepted 12 December 2002

Abstract

The Tarava Seamount chain, also known as the Savannah seamounts on some previous marine charts, is located 200 km south of the Society Islands and was surveyed in 1996. It is 700 km long and comprises 18 seamounts, two of which have been dredged and dated. This chain exhibits two main branches: a western branch parallel to Pacific plate motion before 43 Ma and an eastern one parallel to plate motion since 43 Ma. To first order, K^Ar ages obtained on two dredged samples (43.5 þ 0.6 Ma and 35.9 þ 0.5 Ma), morphology and alignments are compatible with a hotspot origin but a large discontinuity exists in the track at 43 Ma, where the chain splits into two distinct alignments. Several volcanic sources could be responsible for these features but a more convincing solution is proposed that involves the influence of the lithospheric stress field on the track of a single hotspot. Considering that the stress field can be separated into a field due to the deflection of the lithosphere under new volcanic load and a pre-existing regional field, we show that volcanoes appear within the width of the hotspot track only where the less compressive component of the horizontal stress field is minimal. This analysis highlights the key role of the lithosphere and of its pre-existing state of stress on the path of the hotspot tracks. ß 2003 Elsevier Science B.V. All rights reserved.

Keywords: South Paci¢c Superswell; Tarava Seamounts; hotspot; lithospheric stress ¢eld; de£ection

1. Introduction ascribed to deep hotspots, where the term hotspot refers to any magmatic source deeper than the Since Wilson [1], the origin of linear intraplate moving lithosphere. But as more data were col- volcanic chains with age progressions has been lected, di¡erences between deep hotspot theory (e.g., [2^4]) and observation were increasingly recognized [5]. Speci¢cally, on the Paci¢c plate, the * Corresponding author. Present address: Departamento de analysis of the relationship between oceanic pla- Geo¢sica, Universidad de Chile, Casilla 2777, Santiago, Chile. E-mail addresses: [email protected] (V. Clouard), teaus, linear volcanic chains and hotspots has [email protected] (A. Bonneville), shown that most of the recent intraplate chains [email protected] (P.-Y. Gillot). associated with present-day active hotspots corre-

EPSL 6530 10-2-03 118 V. Clouard et al. / Earth and Planetary Science Letters 207(2003) 117^130

16°S

18°S

20°S

22°S 156°W 154°W 152°W 150°W 148°W

depth (m) -5000 -4000 -3000 -2000 -1000 0

Fig. 1. Studied area. The Tarava Seamounts correspond to the seamounts inside the ellipse. They were surveyed in December 1996 by the French R/V l’Atalante. Major magnetic anomalies [15] are represented by white lines. The bathymetry in gray levels is derived from a 30Q grid (see text). spond to short-lived volcanism unrelated to any data collected on a linear volcanic chain south deep mantle plume [6]. In the case of the south- of the Society Islands (Fig. 1). In our discussion, ernmost part of the Austral alignment, because ‘hotspot’ will be used in the general sense of a the age^distance relation between seamounts and localized source of melt in the mantle, below the the hotspot fails, this short-lived volcanism has lithosphere and ¢xed with respect to it, whatever been related to stress in the lithosphere [7].At its depth of origin: just below the lithosphere, at large scale, Steinberger et al. [8] have shown a the transition zone or deeper in the mantle. correlation between hotspots and positive anoma- On a previous bathymetric map [9], the Tarava lies of scalar stress, and these authors hypothe- Seamounts were also known as the Savannah sea- sized that hotspot magma ascends only where mounts, but no o⁄cial name was assigned at that the lithosphere is under tension. To address the time. The name of the chain and of the seamounts question of the respective importance of litho- used in this paper were submitted to the Interna- spheric stress ¢eld and hotspot mechanism, we tional Hydrographic Organization. The complete present in this paper an analysis based on new topography of the chain has been revealed using

EPSL 6530 10-2-03 V. Clouard et al. / Earth and Planetary Science Letters 207(2003) 117^130 119 altimetry data constrained by echo-sounding ship on two seamounts have been selected for K^Ar data [10]. The Tarava Seamounts were completely dating: one at 2900 m depth on the eastern £ank charted during ZEPOLYF1 cruise with the of Punu Taipu and one at 1500 m depth on the French R/V l’Atalante [11]. This survey showed eastern £ank of Fafa Piti (Fig. 2). We used the that there are two di¡erent volcanic provinces, method developed by Cassignol and Gillot [18] the Tarava Seamounts themselves to the west and Gillot and Cornette [19]. In this method the and the Va’a Tau Piti Ridges to the east (Fig. precision and accuracy, especially when dating 1). Our study focuses on the Tarava Seamount pillow basalts, depend primarily on the sample chain. To describe the local and regional stress selection and preparation. From careful petro- ¢elds in the Tarava area, we used the main struc- graphic observation, we selected lavas which con- tural directions of both individual volcanoes and tain fresh and preserved microlitic plagioclases in the whole chain. The morphological parameters the groundmass from the inner part of the pillow. are measured from multibeam bathymetric data. Indeed, to reach a precise age value, it is necessary This analysis coupled with radiometric ages on to date a pure mineral phase which crystallized two seamounts leads us to propose a hypothesis while the lava solidi¢ed, avoiding both the early for the origin of the Tarava Seamount chain crystallized minerals responsible for 40Ar inheri- based on the role played by the lithosphere in tance and the deuteric mineral phases, such as controlling the emplacement of this chain. zeolites, clay minerals and oxides. For this, we applied a double separation by means of heavy liquids (methylene iodide diluted in acetone) and 2. Geological setting magnetic separators. Pure plagioclase microlites were extracted from a 30^60 Wm fraction obtained The Tarava Seamount chain is located in the by re-crushing an already selected fraction of 200^ south central part of the Paci¢c plate (Fig. 1) 400 Wm grains of pure groundmass (which elimi- between latitudes 16‡S and 20‡S and longitudes nates all pheno-and xenocrysts). The results of 156‡W and 150‡W. This region is approximately two independent replicate measurements on po- centered on the South Paci¢c Superswell area. tassium and radiogenic argon content are re- The South Paci¢c Superswell [12] is characterized ported in Table 1 where we used the International by a positive depth anomaly of the sea£oor sev- Conventional Constants [20] for the age calcula- eral thousand kilometers in extent with an unusu- tion. ally high concentration of volcanoes [13]. This Finally, we obtained an age of 43.5 þ 0.6 Ma shallow sea£oor is not simply explained by coa- for the Fafa Piti basalt and 36.1 þ 0.5 Ma for lescent swells produced by the di¡erent hotspots, the Punu Taipu basalt. but by a dynamic or thermal e¡ect of an anom- alous asthenosphere on the lithosphere [14]. The eastern part of the Tarava Seamounts lies on a 80 4. Flexural signature of the Tarava Seamounts and Myr old lithosphere [15] and the western part on the Va’a Tau Piti Ridges a lithosphere generated during the Cretaceous Magnetic Quiet Zone [15^17]. The Tarava Sea- The eastern extremity of the Tarava Seamount mounts are located on the Society swell, which chain is elongated by two deep volcanic ridges extends for 300 km from the axis of the Society called Va’a Tau Piti Ridges (Fig. 1), which exhibit chain [14], a recent volcanic alignment due to hot- an east^west trend. Their morphology is su⁄- spot volcanism. ciently distinct from that of the Tarava Seamount to suggest a di¡erent origin. As no basaltic samples have ever been dredged on these ridges, geo- 3. Radiometric ages physical proxies for age, based on the £exural signature, may provide a useful discrimination be- Samples of alkali basalt pillow lavas dredged tween Tarava and the Va’a Tau Piti volcanism.

EPSL 6530 10-2-03 120 V. Clouard et al. / Earth and Planetary Science Letters 207(2003) 117^130

155°W153°W 151°W Manuae 2000 Arere 17°S Orio'Mata Otu'eroa 18°S

4000 Honu 4000 Ua'ao Fafa Piti Otaha 4000 19°S (43.5 Ma) 2000

Punu Taipu (35.9 Ma) 20°S

Fig. 2. Bathymetric map of the Tarava Seamounts, derived from a 125 m multibeam bathymetric grid. Depth contours interval is 500 m. Names of the seamounts used in the text are reported with their radiometric age in parentheses.

To compute the £exural rigidity under these density, D the £exural rigidity of the lithosphere, two features, we use the three-dimensional ap- 9 is the biharmonic operator, q(x,y) is the vertical proach developed by Watts et al. [21]. We ¢rst load derived from the topographic grid, and g is solve the classical equation governing the defor- the acceleration due to gravity. Model parameters mation w (positive downward) of an elastic layer are shown in Table 2. D is related to Te, the under a volcanic load: elastic thickness of the lithosphere, by 4 3 D9 w þðb m b lÞgw ¼ qðx; yÞð1Þ 3 ET e D ¼ 2 ð2Þ where bl is the load density, bm is the mantle 12ð13X Þ

Table 1 K^Ar analyses of pure microlitic plagioclases sorted from the core of pillow basalts from Tarava Seamounts Sample K (replicate Mean K Radiogenic 40Ar Age for replicate Ar analyses þ determinations) analytical uncertainty (%) (%) (%) (atoms/g) (Ma) Fafa Piti 1.557 1.561 67.08 7.204U1013 43.68 þ 0.63 1.565 67.99 7.153U1013 43.38 þ 0.62 Punu Taipu 0.489 0.492 51.63 1.877U1013 36.22 þ 0.51 0.495 56.50 1.861U1013 35.94 þ 0.51

EPSL 6530 10-2-03 V. Clouard et al. / Earth and Planetary Science Letters 207(2003) 117^130 121

Table 2 Values of parameters and constants used in £exural and gravity modeling Name Symbol Value Unit Young’s modulus E 1.0U1011 Pa Poisson’s ratio X 0.25 Acceleration due to gravity g 9.81 m s32 Newton’s gravitational constant Q 6.67U10311 m3 kg31 s32 33 Load density bl 2600 kg m 33 Water density bw 1030 kg m 33 In¢ll density bf 2800 kg m 33 Crust density bc 2900 kg m 33 Mantle density bm 3350 kg m where E is Young’s modulus, and X the Poisson a 47 Myr old lithosphere. The low value of the ratio. Te depends mainly on the age of the litho- elastic thickness for Va’a Tau Piti Ridges suggests sphere at the time of loading [22]. We then com- a formation close to the former Paci¢c^Farallon pute a free air anomaly with a three-layer model: ridge. This observation is reinforced by the simi- the volcanic load based on gridded bathymetry, larity of their morphology to that of other Paci¢c the material ¢lling the moat de£ection and the volcanic ridges like the Musician [26] or the Foun- £exured crust. The gravity anomaly induced by dation [27] ridges. This type of volcanic alignment two density contrasts at depth zj can be expressed exhibits preferentially an intermediate orientation in the Fourier domain by the formula [23,24]: between the absolute and the relative plate motions [28,29]. That could explain the Va’a Tau Piti ! Xr 3 ! n31 M k M ð M k MÞ n n Ridges’ E^W trend, which is intermediate between G¼2ZQ vbe z0 Hðz 3z Þð3Þ n! 1 2 the N75‡ trend of the fracture zones and the n¼1 N160‡ trend of the absolute plate motion around where Q is the Newton’s gravitational constant, 70 Ma. and H and G the Fourier transforms of the topography and of the gravity anomaly respectively. The computed gravity anomaly for values of Te 5. Morphostructural analysis and regional stress ranging from 15 to 1 km is compared to the ob- ¢eld served one extracted from the gravity grid built up from shipboard and satellite-derived data [25] Bathymetric data were collected with the multi- in the Tarava and the Va’a Tau Piti area. We beam echo sounder SIMRAD EM-12D on board keep the value of Te for which the mis¢t between l’Atalante. This 13 kHz echo sounder is composed those two sets is minimal in a least squares sense. of two independent echo sounders of 81 beams on The best ¢tting models obtained are found with each side of the hull. For our interpretation, we Te values of 8 km for the Tarava Seamount chain use a 125 mU125 m topographic grid of the sea- and of 3 km for the Va’a Tau Piti Ridges. The £oor derived from these data. These data are also values obtained for Te are su⁄ciently di¡erent to merged with previous regional bathymetric syn- conclude that the Tarava Seamount chain was thesis [10] to produce a 30QU30Q grid for general formed on an older lithosphere than the Va’a view of the studied area. Tau Piti Ridges, and thus, that they are two dis- The Tarava Seamounts (Fig. 2) are composed tinct volcanic provinces, even though we did not of 18 seamounts characterized by three volcanic consider a possible overprinting volcanism. This is types: cone-shaped volcanoes, £at-topped guyots con¢rmed by the radiometric ages of the two and small ridges. The cone-shaped volcanoes are dated Tarava seamounts in addition to sea£oor six isolated seamounts lying on the sea£oor at ages (see above) that show that they formed on 4000^4500 m depth. Their £anks are covered by

EPSL 6530 10-2-03 122 V. Clouard et al. / Earth and Planetary Science Letters 207(2003) 117^130 small eruptive cones, and they never reached sea elongated in the direction of the maximum hori- level. The mean depth of their summit is 1000 m. zontal compression of the regional stress ¢eld [32]. Their average volume is 760 km3. Five guyots To test the isotropic distribution of the regional display characteristic £at-topped summits and stress ¢eld, we plot the orientations of all the rift dredges have con¢rmed the presence of limestone zones of the Tarava Seamounts versus their length caps [11,30]. Their mean summit depth is only 580 (Fig. 3a). Short rift zones present a distribution m, and their average volume reaches 3000 km3. roughly radial, and an asymptotic east^west direc- These guyots exhibit a stellate morphology indi- tion appears for rift zones expanding far enough cative of £ank rift zones. Volcanic £ank rift zones from the volcanoes. The same conclusion is extend from the surrounding sea£oor to the sum- reached when plotting direction versus length of mit. Rift zones on the Hawaiian Ridge are recog- asymmetrical edi¢ces (Fig. 3b) and the direction nized as zones of preferred magma intrusion and versus length of regional ridges, including the eruption [31]. It is thought [32] that their orienta- Va’a Tau Piti Ridges (Fig. 3c). The N90‡ trend tion is related both to the magmatic stress ¢eld must be normal to the direction of c3. When the and to the existing regional stress ¢eld: they Tarava volcanism was active, the existing east^ should form radially in the immediate area of west trend in volcanic features indicates that the the conduit, and then curve in the far ¢eld to non-hydrostatic component of c3 was tensile. asymptotically approach the regional stress ¢eld. This direction is present for the old Va’a Tau Rift zones are more likely to develop in a direc- Piti Ridges, and therefore was inherited from tion normal to the minimum horizontal compo- the young lithosphere, but it is not su⁄cient to nent, c3, of the regional stress, that is to say, conclude that the stress ¢eld is only tensile in

0°N a

90°W 90°E 30 20 10 10 20 30 Rift zone length (km)

0°N 0°N c b

90°W 90°E 90°W 90°E 60 4020020 40 60 200 15010050050 100 150 200 Elongated volcano length (km) Ridge length (km)

Fig. 3. Orientation versus length of rift zones in the study area: (a) of the Tarava Seamounts, (b) of the asymmetrical Tarava Seamounts, (c) of the ridges.

EPSL 6530 10-2-03 V. Clouard et al. / Earth and Planetary Science Letters 207(2003) 117^130 123 volcanic periods. On the Paci¢c plate, the Puka- for the formation of the Tarava Seamounts can be puka ridges, created between 0 and 26 Ma, have supported: been explained by a stretching model, but no sig- 1. There is no hotspot at all, rather di¡use melt ni¢cant extension has been observed in plate re- anomalies in the upper mantle, which are constructions [33]. Within the studied area, the guided by the lithospheric structure to erupt east^west direction also corresponds to the trend on the sea£oor. The Tarava volcanism is con- of Tahiti’s main rift zone [34]. Since Tahiti Island trolled by stress in the lithosphere rather than is less than 2 Myr old [35,36], the north^south by hotspots, and melt anomalies are channeled direction of the minimum horizontal stress must to the surface by pre-existing cracks in the lith- still be active. It is con¢rmed by a recent model osphere. This hypothesis was postulated for for magnitude and orientation of the lithospheric the southern Austral volcanism where tension stress where a large and tensile stress anomaly in the lithosphere was due to the load by a pre- centered on the Paci¢c Superswell area is evi- existing chain [7]. But in the case of the Tarava denced [8]. Steinberger et al.’s model [8] was volcanism, the seamounts began to erupt far deduced from global mantle circulation which from the pre-existing Va’a Tau Piti Ridges. evolves with convective time scales ( s 100 Myr These old ridges are not likely the cause for [37]). Thus it is likely that present-day north^ the western Tarava volcanism. Another litho- south tensile stress direction exists intrinsically spheric stretching model was proposed for the in the lithosphere for tens of millions of years, Pukapuka ridges [33], based on the association and in particular during the Tarava volcanism between these ridges and gravity lineations. period. A secondary direction, approximately But the Tarava Seamount morphology does N75‡, also appears in Fig. 2. This direction is not match the morphology of ridges like Puka- parallel to the regional fracture zones, which puka, and furthermore, we cannot prove that likely a¡ect the whole lithosphere. there is no hotspot age progression along the At the scale of the chain, the Tarava Seamounts Tarava Seamount chain as is the case for the are aligned along two main directions: the west- Pukapuka ridges. ern part of the chain trends N160‡ and the eastern 2. There was only one hotspot and its track on part N120‡. These two directions correspond the sea£oor was subject to signi¢cant perturba- roughly to the directions of the Paci¢c plate ab- tions. solute motion before and after 43 Ma. At ¢rst To test the latter hypothesis, let’s try to recon- order, it is in good agreement with the ages we struct the apparent path of the hotspot that could obtained for Fafa Piti (43.6 þ 0.5 Ma) and Punu have created the Tarava Seamount volcanism. To Taipu (35.9 þ 0.6 Ma), which belong respectively do that, we need to move the Paci¢c plate back in to the western and the eastern parts of the chain. time using stage poles. These stage poles represent In addition, recent geochemical analyses [38] show Paci¢c plate absolute motion obtained in the hot- that samples dredged on Fafa Piti and Punu Tai- spot reference frame. For the 7^100 Ma period, pu have compositions compatible with a hotspot we use the data set proposed by Wessel and origin. These observations suggest a hotspot ori- Kroenke [39] and for the 0^7 Ma period those gin for the Tarava Seamounts but with two linear of Yan and Kroenke [40]. We also use the ages chains of volcanoes and we have to test some of Fafa Piti and Punu Taipu to infer the present hypotheses for their origin. location of the Tarava hotspot, obtained by back- tracking from Fafa Piti to HS1 and from Punu Taipu to HS2. Then both tracks are computed for 6. Origin of the Tarava Seamount chain and 0 to 100 Ma and displayed on the Paci¢c plate discussion (Fig. 4). The Tarava hotspot track is comprised between these extreme tracks and its present-day To reconcile all these observations, constrained location is in an area close to HS1 and HS2. Fig. only by two radiometric ages, several hypotheses 4 shows that neither the Foundation nor the Pit-

EPSL 6530 10-2-03 124 V. Clouard et al. / Earth and Planetary Science Letters 207(2003) 117^130

180°W 165°W150°W135°W 120°W

0 0

15°S 15°S

30°S 30°S

180°W 165°W 150°W135°W120°W Fig. 4. Tracks of the hotspots that could have produced the Tarava Seamounts (see text). HS1 and HS2 could be the corresponding present-day locations of these two hotspots. The white numbers in italics correspond to the theoretical age in Ma along the tracks. Black crosses represent dated seamounts or islands, numbers in parentheses correspond to radiometric ages in Ma. Black disks correspond to the present-day locations of Pitcairn (PI) and Foundation (FO) hotspots. The black dashed line box corresponds to the area represented in Fig. 5. cairn hotspot could have produced the Tarava that the Tarava Seamounts have been generated Seamounts, even if we consider motion of these by one hotspot now extinct, active only during a hotspots. Pitcairn hotspot has produced the Pit- short period of time. According to our recon- cairn^Gambier^Mururoa alignment with an age struction (Fig. 5), and to the age of Punu Taipu, of 11 Ma for Mururoa [41]. In order to produce almost the youngest volcano of the chain, the the Tarava Seamounts, the Pitcairn hotspot must life-time of the Tarava volcanism was about 13 have moved 700 km between Punu Taipu and Myr. It is within the range of mean life-time Mururoa, in less than 25 Myr. This corresponds known for the other Superswell present-day hot- to an average rate of 3.5 cm/yr, much greater than spots [6]. any suggested migration rate for hotspots [42,43]. If we use a 100 km wide track as representative The same conclusion can be drawn for Founda- of the zone of in£uence of a given hotspot source tion hotspot. Hence, it seems unlikely that Foun- located between HS1 and HS2 (Fig. 5), we clearly dation or Pitcairn hotspots could have produced see that all the Tarava seamounts fall into the the Tarava Seamounts. track except the southwestern branch. But the To the north, there is no bathymetric evidence 43 Ma bend of the track, corresponding to the for any other volcanoes than the Tarava Sea- Hawaii^Emperor bend, is close to the north of mount chain along the tracks (Fig. 4). To the the studied area, and with the same age, we south, there is no volcano between Punu Taipu have Fafa Piti, 275 km southward. We have to and the present computed location area of the explain how volcanism could simultaneously be hotspot. Thus, it seems reasonable to propose active in two distant locations.

EPSL 6530 10-2-03 V. Clouard et al. / Earth and Planetary Science Letters 207(2003) 117^130 125

156°W 154°W 152°W 150°W

18°S 18°S

20°S 20°S

22°S 22°S 156°W 154°W 152°W 150°W

Fig. 5. Track of the Tarava hotspot reported as a transparent gray band, 100 km wide, on the sea£oor topography (30Q grid). White numbers correspond to the theoretical age in Ma along the tracks. White crosses indicate non-dated seamounts while black ones indicate dated seamounts, numbers in parentheses correspond to radiometric ages in Ma (Table 1, this study, and A. Hil- denbrand, personal communication).

Additionally, we looked at evolution of the and, without other stresses, volcanoes might stress ¢eld in the lithosphere associated with the therefore appear everywhere around previous edi- successive emplacement of volcanoes on the crust ¢ces, along the isoline de¢ned by in£ection points. as the plate was moving. The mis¢t between vol- In our case, we have to take into account the fact canism produced over a ¢xed hotspot and what that the lithosphere moves to the northwest of the we observe in the Tarava Seamount chain appears hotspot source, and that there is a north^south slightly before 43 Ma, when the Tarava volcanism direction of lithospheric tensile (or at least less began to form two volcanic lines. To understand compressive) stress. Then the probability for the state of stress in the lithosphere at that mo- new edi¢ces to appear is greater to the north or ment, we compute the de£ection produced only south of the last volcano because of residual by the northwestern part of the Tarava Sea- stress, and to the southeast because of the abso- mounts. This re£ects the e¡ective volcanic load lute plate motion. We constructed successive just before 43 Ma. The stress in the lithosphere sketches of loading at 46, 44, 43, 42, 41 and 40 associated with the £exural deformation changes Ma. Results are presented in Fig. 6, they clearly at the in£ection point of the de£ection curve [44] demonstrate that the observed emplacement of

EPSL 6530 10-2-03 126 V. Clouard et al. / Earth and Planetary Science Letters 207(2003) 117^130

Fig. 6. Scenario of lithospheric loading by successive emplacement of volcanoes along the Tarava hotspot track. The scenario has been divided into eight stages named by the seamount that loads the crust at the corresponding age. For all stages, the sea£oor topography used for loading the crust is represented by its present-day bathymetric contours (thin black lines), outside this region, the sea£oor depth is set at its regional value. The white lines correspond to isovalues of lithospheric de£ection under the volcanic loads. The thick black isoline joins all the in£ection points (see text) where new volcanic edi¢ces (contoured by a thick black line) are more likely to appear. The vertical black arrow corresponds to the loading volcano (stage name). Thin black ar- rows indicate the Paci¢c absolute plate motion at the stage time.

EPSL 6530 10-2-03 V. Clouard et al. / Earth and Planetary Science Letters 207(2003) 117^130 127

-4000 35°N -4 00 a 0

-4000 92 43 -40 00 42 39

-4000 39

30°N

-4 -4000 000

Hawaii-Emperor

170°E 175°E 180°

b 54

35°S

46 Louisville

40°S 175°W 170°W 165°W

Fig. 7. Part of the seamounts tracks for Hawaii^Emperor (a) and Louisville (b) corresponding to the change in Paci¢c plate motion around 43 Ma. Bathymetric contours and gray level intervals are 1000 m. Black crosses represent dated seamounts, numbers correspond to radiometric ages in Ma (from Clouard and Bonneville’s [6] compilation of Paci¢c ages).

EPSL 6530 10-2-03 128 V. Clouard et al. / Earth and Planetary Science Letters 207(2003) 117^130

Tarava volcanoes on the sea£oor is controlled by mounts between 44 and 41 Ma we found in the the successive loading by volcanic edi¢ces along Tarava Seamount chain. This hypothesis could the Tarava hotspot track. New edi¢ces appear have important consequences for the mechanical along the in£ection line, to the southeast because behavior of the Paci¢c plate at that time and of the plate motion, or to the south because of the needs further investigation that goes far beyond direction of lithospheric tensile stress. The vol- the scope of the present paper. canic activity began around 48 Ma and lasted normally until Honu seamount (predicted age = 43 Ma). This early stage was not accompanied 7. Conclusion by any £ood basalt episode, as predicted by the plume hypothesis [2,4]. A second stage starts The Tarava Seamount chain was created by a around 43 Ma, the time of the major change in newly recognized hotspot within the South Paci¢c Paci¢c plate motion. A dual line of volcanoes Superswell area, characterized among other things developed for some time initiated by the forma- by an unusually dense concentration of hotspots tion of new edi¢ces o¡ the hotspot track axis, to [13]. All other already known South-central Pacif- the east and to the south. The formation of dual ic hotspots are short-lived hotspots, not fed by volcanic lines when a major change in the tectonic deep mantle plume [6,46]. The activity of this hot- plate motion occurs has been modeled by Hiero- spot began around 50 Ma and vanished around nymus and Bercovici [45] and explained by the 35 Ma. This chain can be described by a western fact that the last volcano of the single previous and a eastern segment, which present respectively line is in an o¡-axis position with respect to the the orientation of the Paci¢c plate motion before new hotspot path and perturbs the dynamics of and after 43 Ma. The discontinuity between the the system. The Tarava dual lines propagated for western and eastern parts of the track can be ex- 2 or 3 million years until the activity stopped plained by the formation of a dual line of volca- along the southern line and continued only along noes at the time of the major change in the Paci¢c the eastern line. plate motion observed at 43 Ma on the Hawai- One has to note that two other linear volcanic ian^Emperor seamount hotspot track. The stress chains seem to have registered the so-called 43 Ma in the lithosphere induced by the £exural defor- change in the Paci¢c plate motion: the Hawaii^ mation under the load of volcanoes and inherited Emperor seamount chain and the Louisville sea- from the regional stress ¢eld channelled the vol- mount chain (Fig. 7). The Hawaiian ridge (Fig. canism and controlled the path of the hotspot 7a) displays a cluster of volcanoes between 42 and track. 39 Ma along two lines, like the ones observed on the Tarava Seamount chain. In this case as well, the western volcanic alignment is the one that Acknowledgements vanished after a few million years of activity. In the southern Paci¢c, the Louisville ridge (Fig. 7b) Most of the data come from the ZEPOLYF1 is also perturbed by the change in Paci¢c plate cruise, and we are grateful to the crew of R/V motion, though not exactly at 43 Ma. In fact, l’Atalante for the quality of these data. V.C. was this event seems to have occurred earlier and is supported from a grant of the ZEPOLYF pro- marked on the sea£oor by a cluster of seamounts gram funded by the French Ministry of Research formed between 53.5 and 44 Ma but without any and by the government of French Polynesia. The preferential orientation. The Tarava Seamount authors wish to thank Kelsey Jordahl, Keitapu chain has an intermediate location on the Paci¢c Maamaatuaiahutapu and Hilary Todd for their plate between these two chains and if we invoke a careful remarks during the preparation of this progressive delay in the propagation of the plate paper. Our manuscript has been greatly improved motion bending from south to north, we would by the comments of the reviewers, Daniel Asla- have a better ¢t with the ages of cluster sea- nian and Daniel Scheirer.[AC]

EPSL 6530 10-2-03 V. Clouard et al. / Earth and Planetary Science Letters 207(2003) 117^130 129

References [18] C. Cassignol, P.-Y. Gillot. Range and e¡ectiveness of un- spiked potassium-argon dating: experimental groundwork and applications, in: G.S. Odin (Ed.), Numerical Dating [1] J. Tuzo Wilson, A possible origin of the Hawaiian Is- in Stratigraphy, Wiley, Chichester, 1982, pp. 159^179. lands, Can. J. Phys. 41 (1963) 863^870. [19] P.-Y. Gillot, Y. Cornette, The Cassignol technique for [2] W.J. Morgan, Convection plumes in the lower mantle, potassium-argon dating, precision and accuracy: exam- Nature 230 (1971) 42^43. ples from the Late Pleistocene to recent volcanics from [3] D.A. Yuen, G. Schubert, Mantle plumes: A boundary Southern Italy, Chem. Geol. (Isotope Geosci. Sect.) 59 layer approach for Newtonian and non-Newtonian tem- (1986) 205^222. perature-dependent rheologies, J. Geophys. Res. 81 (1976) [20] R.H. Steiger, E. Jaeger, Subcomission on geochronology: 2499^2510. Convention on the use of decay constants in geo-and [4] M.A. Richards, R.A. Duncan, V.E. Courtillot, Flood ba- cosmochronology, Earth Planet. Sci. Lett. 36 (1977) salts and hot-spot tracks: Plume heads and tails, Science 359^362. 246 (1989) 103^107. [21] A.B. Watts, J.R. Cochran, G. Selzer, Gravity anomalies [5] E.A. Okal, R. Batiza, Hotspots: The ¢rst 25 years, in: and £exure of the lithosphere: A three-dimensional study B.H. Keating, P. Fryer, R. Batiza, G.W. Boehlert (Eds.), of the Great Meteor Seamount, northeast Atlantic, J. Geo- Seamounts, Islands, and Atolls, AGU Geophys. Monogr. phys. Res. 80 (1975) 1391^1398. 43 (1987) 1^11. [22] A.B. Watts, An analysis of isostasy in the world’s oceans, [6] V. Clouard, A. Bonneville, How many Paci¢c hotspots 1, Hawaiian-Emperor seamount chain, J. Geophys. Res. are fed by deep-mantle plumes?, Geology 29 (2001) 83 (1978) 5989^6004. 695^698. [23] R.L. Parker, The rapid calculation of potential anomalies, [7] M.K. McNutt, D.W. Caress, J. Reynolds, K.A. Jordahl, Geophys. J. R. Astron. Soc. 31 (1972) 447^455. R.A. Duncan, Failure of plume theory to explain mid- [24] R.L. Parker, Understanding inverse theory, Annu. Rev. plate volcanism in the Southern Austral Islands, Nature Earth Planet. Sci. 5 (1977) 35^64. 389 (1997) 479^482. [25] V. Clouard, A. Bonneville, H.G. Barsczus, Size and depth [8] B. Steinberger, H. Smelling, G. Marquart, Large-scale of frozen magma chambers under atolls and islands of lithospheric stress ¢eld and topography induced by global French Polynesia using detailed gravity studies, J. Geo- mantle circulation, Earth Planet. Sci. Lett. 186 (2001) 75^ phys. Res. 105 (2000) 8173^8192. 91. [26] J. Phipps Morgan, W. Weinrebe, C. Kopp, E.R. Flueh, I. [9] J. Mammerickx, S.M. Smith, I.L. Taylor, T.E. Chase. Ba- Grevemeyer, H. Borus, R. He¤kinian, C.E. Larsen, H. Lel- thymetry of the South Paci¢c. Scripps Institution of gemann, S.N. Lyons, W.J. Morgan, J. O’Connor, Y. Pan, Oceanography, La Jolla, CA, 1973. T. Stender, F. Wolter, Morphology and structure of two [10] L. Sichoix, A. Bonneville, Prediction of bathymetry in 400 km-long volcanic ridges in the Musicians seamount French Polynesia constrained by shipboard data, Geo- province, in: Eur. Geophys. Soc., number 74 in XXV phys. Res. Lett. 3 (1996) 2469^2472. General Assembly, Nice, France, 2000, p. 109. [11] ZEPOLYF1, Rapport de mission de la campagne ZEPO- [27] M. Maia, D. Ackerman, G.A. Dehghani, P. Gente, R. LYF1. Technical report, Universite¤ de la Polyne¤sie He¤kinian, D. Naar, J. O’Connor, K. Perrot, J. Phipps franc°aise, IFREMER, SHOM, Papeete, Tahiti, 1996. Morgan, G. Ramillien, S. Re¤villon, A. Sabetian, D. Sand- [12] M.K. McNutt, K. Fischer, The South Paci¢c Superswell, well, P. Sto¡ers, The Paci¢c-Antarctic Ridge-Foundation in: B.H. Keating, P. Fryer, R. Batiza, G.W. Boehlert hotspot interaction: A case study of a ridge approaching (Eds.), Seamounts, Islands, and Atolls, AGU Geophys. a hotspot, Mar. Geotechnol. 167 (2000) 61^84. Monogr. 43 (1987) 25^34. [28] W.J. Morgan, Rodriguez, Darwin, AmsterdamT A second [13] M.K. McNutt, Superswell, Rev. Geophys. 36 (1998) 211^ type of hotspot island, J. Geophys. Res. 83 (1978) 5355^ 244. 5360. [14] L. Sichoix, A. Bonneville, M.K. McNutt, The sea £oor [29] D. Epp, Possible perturbations of hotspot traces and im- swells and superswell in French Polynesia, J. Geophys. plications for the origin and structure of the Line Islands, Res. 103 (1998) 27123^27133. J. Geophys. Res. 89 (1984) 11273^11286. [15] M. Munschy, C. Antoine, G. Guille, H. Guillou, La [30] A. Bonneville, V. Clouard, S. Labrosse, Mission POLY- crou“te oce¤anique et les points chauds de la Polyne¤sie DRAG1. Rapports du programme ZEPOLYF, n2, Univ. francaise (Oce¤an Paci¢que central), Ge¤ol. Fr. 3 (1998) Fran. Paci¢que, Tahiti, Polyne¤sie Franc°aise, 1998, 20 pp. 5^13. [31] R.S. Fiske, E.D. Jackson, Orientation and growth of Ha- [16] E.M. Herron, Sea-£oor spreading and the Cenozoic his- waiian volcano rifts: The e¡ect of regional structure and tory of the east-central Paci¢c, Geol. Soc. Am. Bull. 83 gravitational stresses, Phil. Trans. R. Geol. Soc. London (1972) 1671^1692. 329 (1972) 299^326. [17] C.L. Mayes, L.A. Lawver, D.T. Sandwell, Tectonic histo- [32] K. Nakamura, Volcanoes as possible indicators of tecton- ry and new isochron chart of the South Paci¢c, J. Geo- ic stress orientation ^ Principle and proposal, J. Volcanol. phys. Res. 95 (1990) 8543^8567. Geotherm. Res. 2 (1977) 1^16.

EPSL 6530 10-2-03 130 V. Clouard et al. / Earth and Planetary Science Letters 207(2003) 117^130

[33] D.T. Sandwell, E.L. Winterer, J. Mammerickx, R.A. ing hotspots and re¢ning absolute plate motions, Nature Duncan, M.A. Lynch, D.A. Levitt, C.L. Johnson, Evi- 387 (1997) 365^369. dence for di¡use extension of the Paci¢c Plate from Pu- [40] C.Y. Yan, L.W. Kroenke, A plate tectonic reconstruction kapuka ridges and cross-grain gravity lineations, J. Geo- of the Southwest Paci¢c, 0^100 Ma, Proc. OPD Sci. Res. phys. Res. 100 (1995) 15087^15099. 130 (1993) 697^709. [34] P.-Y. Gillot, J. Talandier, H. Guillou, I. Leroy. Temporal [41] P.-Y. Gillot, Y. Cornette, G. Guille, Age (K-Ar) et con- evolution of the volcanism of Tahiti island: Geology and ditions d’e¤di¢cation du soubassement volcanique de l’atoll structural evolution, in: International Workshop on In- de Moruroa (Paci¢que Sud), C.R. Acad. Sci. Paris 314 traplate Volcanism: The Polynesian Plume Province, Pa- (1992) 393^399. peete, Tahiti, 1993, p. 29. [42] P. Molnar, J. Stock, Relative motions of hotspots in the [35] R.A. Duncan, I. McDougall, Linear volcanism in French Paci¢c, Atlantic and Indian ocean since late Cretaceous Polynesia, J. Volcanol. Geotherm. Res. 1 (1976) 197^227. time, Nature 327 (1987) 587^591. [36] C. Diraison, H. Bellon, C. Leotot, R. Brousse, H. Barsc- [43] B. Steinberger, R.J. O’Connell, Advection of plumes in zus, L’alignement de la Socie¤te¤ (Polyne¤sie franc°aise): vol- mantle £ow: Implications for hotspot motion, mantle vis- canologie, ge¤ochronologie, proposition d’un mode'le de cosity and plume distribution, Geophys. J. Int. 132 (1998) point chaud, Bull. Soc. Ge¤ol. Fr. 162 (1991) 479^496. 412^434. [37] D. Bercovici, Y. Ricard, M.A. Richards. The relation [44] U.S. ten Brink, Volcano spacing and plate rigidity, Geol- between mantle dynamics and plate tectonics: A primer, ogy 19 (1991) 397^400. in: M.A. Richards, R.G. Gordon, R.D. van der Hilst [45] C.F. Hieronymus, D. Bercovici, Discrete alternating is- (Eds.), The History and Dynamics of Global Plate Mo- lands formed by interaction of magma transport and lith- tion, AGU Geophys. Monogr. 121 (2000) 5^46. ospheric £exure, Nature 397 (1999) 604^607. [38] L. Dosso, A. Bonneville, The Cook^Austral volcanic [46] B. Steinberger, Plumes in a convective mantle: Model and chain: Chemical Geodynamics at work, in preparation. observations for individual hotspots, J. Geophys. Res. 105 [39] P. Wessel, L. Kroenke, A geometric technique for relocat- (2000) 11127^11152.

EPSL 6530 10-2-03