Effect of the Gala¤Pagos Hotspot on Sea£Oor Volcanism Along the Gala¤Pagos Spreading Center (90.9^97.6‡W)

Available online at www.sciencedirect.com

Earth and Planetary Science Letters 217 (2004) 331^347 www.elsevier.com/locate/epsl

E¡ect of the Gala¤pagos hotspot on sea£oor volcanism along the Gala¤pagos Spreading Center (90.9^97.6‡W)

Mark D. Behn a;Ã, John M. Sinton b, Robert S. Detrick c a Department of Terrestrial Magnetism, Carnegie Institution of Washington, 5241 Broad Branch Road NW, Washington, DC 20015, USA b School of Ocean and Earth Science and Technology, University of Hawaii, Honolulu, HI 96822, USA c Department of Geology and Geophysics, Woods Hole Oceanographic Institution, Woods Hole, MA 02543, USA

Received 23 June 2003; received in revised form 13 October 2003; accepted 14 October 2003

Abstract

Studies along the Mid-Atlantic Ridge (MAR) and East Pacific Rise (EPR) show that seamount abundance is a strong function of spreading rate. At the MAR, small axial seamounts are a dominant morphologic feature of the inner valley floor, while at the EPR seamounts are rarely observed within the neovolcanic zone. The Gala¤pagos Spreading Center (GSC) provides an excellent location to test the influence of hotspot-related magma supply variations on seamount formation at a relatively constant spreading rate. In this study, we use multibeam bathymetry data to examine the distribution of axial seamounts with distance from the hotspot along the GSC. A numerical algorithm is developed to identify isolated volcanic edifices by searching bathymetry for closed, concentric contours protruding above the surrounding seafloor. Seamount populations are fit with a maximum likelihood model to estimate the total number of seamounts per unit area and the characteristic seamount height. The number of seamounts in the axial zone decreases significantly as the Gala¤pagos hotspot is approached, varying from MAR-like abundances west of 95.5‡W to EPR-like abundances east of 92.7‡W. The along-axis variation in seamount density corresponds to a change in lava flow morphology from dominantly pillow flows in the west to a combination of pillows and sheet/lobate flows in the east. These changes in volcanic style suggest a transition from point-source to fissure-fed eruptions as magma supply increases. We find no evidence for along-axis variations in lava viscosity and, therefore, interpret the differences in seamount abundance and flow morphology to indicate an increase in effusion rates approaching the hotspot. Comparing seamount abundance with axial morphology, crustal thickness, and the presence and depth of an axial magma chamber (AMC), we find that the transition from point-source to fissure-fed eruptions is most sensitive to the presence of a steady-state AMC. In summary, the correlation between seamount abundance, lava morphology, and magma supply at the GSC suggests that effusion rate is directly related to the presence of a steady-state crustal magma chamber. ß 2003 Elsevier B.V. All rights reserved.

Keywords: Gala¤pagos Rift; volcanism; seamount; hotspots; plume^ridge interaction; mid-ocean ridges

1. Introduction * Corresponding author. Tel.: +1-202-478-8837; Fax: +1-202-478-8821. E-mail address: [email protected] (M.D. Behn). The style of sea£oor volcanism at mid-ocean

EPSL 6909 15-12-03 Cyaan Magenta Geel Zwart 332 M.D. Behn et al. / Earth and Planetary Science Letters 217 (2004) 331^347 ridges is strongly dependent on spreading rate (e.g., [1,2]). At the fast-spreading East Paci¢c Rise (EPR), eruptions typically occur within a narrow neovolcanic zone [3^5] centered over a shallow steady-state crustal magma chamber [6^ 8]. Volcanic activity tends be continuous along- axis, with eruptions occurring frequently on time scales of 1^50 yr [2,3]. In contrast, the slow- spreading Mid-Atlantic Ridge (MAR) is characterized by a wide, discontinuous neovolcanic zone Fig. 1. Bathymetry of the Gala¤pagos hotspot region with [9^12] and the absence of a steady-state magma depths shallower than 2750 m shaded. Location of G- chamber (e.g., [13]). Eruptions at the MAR occur PRIME study area is outlined. Vectors show absolute plate less frequently than those at the EPR, with a typ- motion of the Cocos and Nazca plates [70]. White star illus- ical eruptive interval of 1000^10 000 yr [2,14,15]. trates the present location of the Gala¤pagos hotspot [71]. Lava £ow morphology is also a function of spreading rate, with fast-spreading ridges domi- along the Gala¤pagos Spreading Center (GSC) nated by sheet and lobate £ows and slow-spread- (Fig. 1). The GSC is an intermediate spreading ing ridges dominated by pillow lavas (e.g., rate ridge, with a full opening rate that increases [2,16,17]). In general, sheet and lobate £ows result slightly from 4.6 cm/yr at 98‡W to 5.6 cm/yr at from high e¡usion rate, ¢ssure-fed eruptions, 91‡W [27]. Variations in axial topography [28], while pillow lavas are the product of lower e¡u- mantle bouguer gravity anomalies [29], crustal sion rate, point-source eruptions [18]. This change thickness [30,31], and the chemical and isotopic in volcanic style is re£ected in the morphology of composition of dredged basalts [30,32^36] indi- the neovolcanic zone at fast- and slow-spreading cate that elevated magma supply associated with ridges. For example, at the MAR, the neovolcanic the Gala¤pagos hotspot in£uences the GSC over a zone is dominated by hummocks, hummocky region extending V1300 km along-axis. In addi- ridges, and small, nearly circular volcanoes (sea- tion, constraints on the presence and depth of an mounts) composed primarily of pillow lavas [19^ axial magma chamber (AMC) from multichannel 21]. At many MAR segments, these features coa- seismic re£ection data [30] make the GSC an ideal lesce to form narrow axial volcanic ridges aligned location to study the in£uence of magma supply parallel to the spreading axis [22,23]. In contrast, on constructional volcanism in the absence of var- at the EPR where sheet and lobate £ows prevail, iations in spreading rate. the neovolcanic zone is observed to have smooth- In this paper, we use multibeam bathymetry er along- and across-axis relief than the MAR and data to quantify the distribution of axial sea- is typically devoid of seamounts (e.g., [24]). mounts along the GSC as a function of distance These variations in volcanic style with spread- from the Gala¤pagos hotspot. A numerical algo- ing rate illustrate the in£uence of magma supply rithm is developed to identify isolated volcanic on crustal construction. However, spreading rate edi¢ces, and the characteristic height and density also in£uences lithospheric thickness [25], mantle of the seamount populations are calculated from upwelling patterns [26], and the partitioning be- a maximum likelihood model. We show that axial tween magmatic and amagmatic extension [1]. seamount abundance decreases dramatically as Thus, it is di⁄cult to isolate the e¡ects of magma the hotspot is approached, suggesting a transition supply on constructional volcanism based solely from point-source to ¢ssure-fed volcanism as on observations from di¡erent spreading environ- magma supply increases. These variations are ments. compared to changes in axial morphology, mag- To address this issue, we investigated the in£u- ma chemistry and viscosity, and the depth and ence of the Gala¤pagos hotspot on seamount for- presence of an AMC. In particular, we ¢nd that mation and the style of constructional volcanism seamount formation is most sensitive to the ab-

EPSL 6909 15-12-03 Cyaan Magenta Geel Zwart M.D. Behn et al. / Earth and Planetary Science Letters 217 (2004) 331^347 333 sence of a shallow AMC, suggesting that the style of constructional volcanism is linked to the amount of magma present in the crust at the time of eruption.

2. Data and methods

As part of the Gala¤pagos Plume^Ridge Inter- action Multidisciplinary Experiment (G-PRIME) in April^May 2000, multibeam bathymetry data were collected along the GSC between 90.5 and 98‡W [30]. These data were integrated with previous multibeam surveys [28,37] to provide the highest resolution bathymetry currently available for the neovolcanic zone of the western GSC. The integrated bathymetry data were used to identify axial seamounts and estimate variations in seamount abundance along the GSC. Because we Fig. 2. Example of seamounts identi¢ed assuming K 9 2. were interested in seamounts associated with crus- Contour interval is 10 m. Thick black lines illustrate best-¢t- tal construction at the ridge axis, we limited our ting ellipse to the basal contour of each seamount. investigation to sea£oor younger than V500 ka. Wessel and Lyons [38] identi¢ed large o¡-axis mounts. We ¢nd that values of Kmax = 2^3 are seamounts in the Paci¢c basin using an algorithm suitable for seamount identi¢cation in most that searched for local maxima in satellite gravity spreading environments and we show the sensitiv- data. This approach is well suited for identifying ity of our results to a chosen value of Kmax. Fig. 2 seamounts in situations where seamount height is illustrates the application of our algorithm over a signi¢cantly greater than the local topographic small region of the GSC bathymetry data. variations. However, tectonic faulting at mid- Seamount population parameters were calcu- ocean ridges often results in 2^3 km of topo- lated using the maximum likelihood model of graphic relief within 10 km of the ridge axis, while Smith and Jordon [40]. This technique assumes axial seamounts tend to have heights ranging seamount height follows an exponential distribu- from 50 to 500 m [20]. Therefore, to e⁄ciently tion in which the expected number of seamounts identify seamounts on rough axial topography, with a height h v H is given by X(H)=Xoexp we developed a numerical algorithm that searches (3LH), where Xo is the average number of sea- bathymetry for closed, concentric contours pro- mounts per unit area, and L31 is de¢ned as the truding v 50 m above the surrounding sea£oor. characteristic seamount height of the population. A contour interval of 10 m was chosen for sea- Model parameters were calculated from the data mount identi¢cation, however we found that our by sorting the identi¢ed seamounts into 10-m results are relatively insensitive to contour inter- height bins, starting with a minimum height of vals between 10 and 25 m. 50 m. To eliminate elongated topographic highs associated with either faulting or axial volcanic ridges, 2.1. Veri¢cation of seamount identi¢cation we applied a maximum aspect ratio condition to algorithm our identi¢cation algorithm. The aspect ratio, K, of each topographic high was estimated by ¢tting Before estimating the seamount population pa- an ellipse to the basal contour [39] and only those rameters for the GSC, we tested our identi¢cation features with K 9 Kmax were identi¢ed as sea- algorithm using bathymetry data from the north-

EPSL 6909 15-12-03 Cyaan Magenta Geel Zwart 334 M.D. Behn et al. / Earth and Planetary Science Letters 217 (2004) 331^347 ern MAR (24^30‡N) [41] and the southern EPR Table 1 (15^19‡S) [42]. At the MAR, we identi¢ed 455 Calculated seamount populations for the MAR and EPR K X L31 h seamounts with h 50 m and K 2, resulting max # Seamounts o (H/R) v 9 3 3 33 (10 3 km 2) (m) in the population parameters Xo = 144 þ 7U10 km32 and L31 = 40 þ 2 m. By comparison, in MAR (24^30‡N, sea£oor area = 11 000 km2) an earlier study in this region Smith and Cann 2 455 144 þ 7 40 þ 2 0.15 þ 0.05 [20] manually identi¢ed 481 seamounts with the 2.5 544 171 þ 7 40 þ 2 0.14 þ 0.05 X 3 570 177 þ 7 41 þ 2 0.14 þ 0.05 same height and shape criteria and found o = EPR (15^19‡S, sea£oor area = 8500 km2) 3 3 3 195 þ 9U10 3 km 2 and L 1 = 58 þ 2 m for the 2 32 53 þ 9 15 þ 3 0.06 þ 0.05 population. We attribute the smaller population 2.5 40 66 þ 10 15 þ 2 0.06 þ 0.02 density calculated using our identi¢cation algo- 3 43 78 þ 12 14 þ 2 0.06 þ 0.02 rithm to di¡erences in the spatial coverage and resolution of the bathymetric data used in the analyses, as well as the more rigorous de¢nition dependent on the data resolution, contour algo- of K required by explicitly ¢tting an ellipse to the rithm, and identi¢cation criteria. However, the basal contour of each seamount. Speci¢cally, our results shown in Table 1 clearly indicate that numerical approach eliminates the potential for our identi¢cation method can provide quantita- arti¢cial rounding, which may occur when man- tive estimates of seamount populations in diverse ually ¢tting an ellipse to the basal contour of a spreading environments. seamount. Our approach is also advantageous because it reduces the potential for biases in seamount identi¢cation between di¡erent investiga- 3. Characteristics of the GSC seamount population tors. The smaller characteristic height predicted for the MAR population is caused by calculating Seamounts were identi¢ed in the axial zone of seamount height from the peak to the basal closed the GSC between 97.6 and 90.9‡W. The axial contour. This e¡ect is most pronounced for sea- morphology of the GSC is broadly divided into mounts formed on a slope, where only the portion three distinct domains re£ecting the increasing in- of the seamount above its uppermost edge will be £uence of the Gala¤pagos hotspot to the east included in the calculated height. [28,30]. West of the 95.5‡W propagating rift, the At the southern EPR, we calculated seamount ridge axis is characterized by a well-developed rift 33 32 population parameters Xo =53þ9U10 km valley similar to that observed at the slow-spread- and L31 = 15 þ 3 m for young crust located within ing MAR (Fig. 3a). We ¢nd that the axial valley V10 km of the ridge axis. The calculated value of £oor in this region is dominated by the presence 33 32 Xo is similar to the value of 31 þ 1U10 km of small seamounts, suggesting that west of reported by White et al. [43] for isolated sea- 95.5‡W point-source eruptions are the primary mounts on 0^1 Ma crust between 15.3 and 20‡S style of constructional volcanism. This observa- on the EPR. We also ¢nd that the average sea- tion is consistent with a previous study by Klein- mount height to radius ratio, h , is signi¢cantly rock and Brooks [44], who found the sea£oor smaller for the EPR seamounts (h = 0.06) than surrounding the 95.5‡W propagator to be popu- for those at the MAR (h = 0.15). Table 1 summa- lated by numerous small volcanic ‘knobs’ or sea- rizes the population parameters calculated for mounts. In many locations, groups of seamounts both the MAR and EPR, and illustrates the sen- are observed in chains aligned sub-parallel to the sitivity of these parameters to the maximum as- ridge axis. The number of seamounts is generally pect ratio allowed for the basal contour. Our re- found to be greatest on the rift valley £oor. This sults are consistent with earlier observations of a observation is consistent with a model in which spreading rate dependence on seamount abun- most seamounts form within the rift valley and dance (e.g., [20]). Finally, we note that seamount are subsequently destroyed by faulting as they identi¢cation is an inherently subjective process, are transported o¡-axis [45].

EPSL 6909 15-12-03 Cyaan Magenta Geel Zwart M.D. Behn et al. / Earth and Planetary Science Letters 217 (2004) 331^347 335

Fig. 3. Bathymetry illustrating the axial morphology of the (a) rift valley, (b) transitional, and (c) axial high domains. Contour interval is 25 m. Seamounts with K 9 2 are indicated by the best-¢tting ellipse to the basal contour. The axial valley £oor of the rift valley domain is dominated by the presence of small seamounts. In contrast, the neovolcanic zone of the axial high domain is characterized by smooth topography and few seamounts.

EPSL 6909 15-12-03 Cyaan Magenta Geel Zwart 336 M.D. Behn et al. / Earth and Planetary Science Letters 217 (2004) 331^347

Between 95.5 and 92.7‡W the ridge axis enters a per unit area with K 9 2 decreases from transitional regime in which the well-de¢ned rift 234 þ 14U1033 km32 in the rift valley domain valley is replaced by low-relief abyssal hills (Fig. to 54 þ 10U1033 km32 in the axial high domain. 3b). Seamount formation in this region is less The characteristic seamount height also decreases prevalent than in the rift valley domain, with sea- from west to east, with L31 = 37 þ 2 m in the rift mount abundance decreasing from west to east. valley domain and L31 = 28 þ 5 m in the axial high Closest to the hotspot where magma supply is domain. The sensitivity of the population param- highest (92.7^90.5‡W), the ridge axis is de¢ned eters to Kmax is shown in Table 2. Although the by a prominent EPR-like axial high (Fig. 3c). choice of Kmax a¡ects the absolute seamount den- The neovolcanic zone associated with the axial sities, it does not in£uence the along-axis gradient high is characterized by smooth topography and in seamount abundance. few axial seamounts, suggesting that low-relief ¢s- sure-fed eruptions are the dominant form of sea- 3.1. Comparison to fast- and slow-spreading ridges £oor volcanism. Seamount population parameters were calcu- As discussed earlier, axial seamount abundance lated for each of the three morphologic domains tends to increase with decreasing spreading rate (Fig. 4). We ¢nd that the number of seamounts (Fig. 5). However, while fast-spreading ridges typ-

Fig. 4. (a) Bathymetry along the GSC with small circles showing the location of each seamount with K 9 2. Boxes illustrate regions in which seamounts were identi¢ed. Locations of Fig. 3a, b, and c are indicated. (b^d) Number of seamounts versus seamount height in the rift valley, transitional, and axial high domains, respectively. Blue diamonds are seamount counts in 10-m height bins, red circles are the cumulative counts. Blue line illustrates the maximum likelihood ¢t to the binned data. Note the decrease in seamount density with decreasing distance to the Gala¤pagos hotspot.

EPSL 6909 15-12-03 Cyaan Magenta Geel Zwart M.D. Behn et al. / Earth and Planetary Science Letters 217 (2004) 331^347 337

Table 2 zone of the Reykjanes Ridge is elevated relative Calculated seamount populations for the GSC to non-hotspot a¡ected regions of the MAR. How- K X L31 h max # Seamounts o (H/R) ever, the Reykjanes Ridge does not exhibit an 33 32 (10 km ) (m) along-axis gradient toward the hotspot. This may Rift valley (98^95.5‡W, sea£oor area = 4400 km2) be due to the slow-spreading rate of the Reykjanes 2 270 234 þ 14 37 þ 2 0.16 þ 0.06 Ridge (1 cm/yr half-rate [27]), suggesting that the 2.5 329 284 þ 16 38 þ 2 0.15 þ 0.05 transition from point-source to ¢ssure-fed erup- 3 364 323 þ 17 37 þ 2 0.15 þ 0.05 Transitional (95.5^92.7‡W, sea£oor area = 6200 km2) tions is sensitive to the integrated e¡ects of spread- 2 124 102 þ 9 31 þ 3 0.14 þ 0.05 ing rate and plume in£uence on crustal magma 2.5 158 126 þ 10 31 þ 2 0.13 þ 0.04 supply. We will discuss possible explanations for 3 199 172 þ 12 30 þ 2 0.12 þ 0.04 the di¡erences between the Reykjanes Ridge and 2 Axial high (92.7^90.5‡W, sea£oor area = 3500 km ) the GSC in more detail in Section 5.3. 2 33 54 þ 10 28 þ 5 0.14 þ 0.08 2.5 41 78 þ 12 26 þ 4 0.13 þ 0.07 3 43 82 þ 12 26 þ 4 0.13 þ 0.07 3.2. Seamount formation versus magma supply

To investigate the relationship between sea£oor volcanism and magma supply we compared the ically have uniformly low seamount abundances, variations in seamount abundance to crustal slow-spreading ridges are characterized by greater thickness and the presence and depth of a variability. The along-axis variation in seamount steady-state magma lens (Fig. 6). Crustal thick- abundance at the GSC is similar in magnitude to ness along the GSC has been determined from a the di¡erences observed between the MAR and combination of multichannel seismic data and EPR (Fig. 5). Moreover, the change in seamount three wide-angle seismic refraction experiments abundance along the GSC is also larger than the observed variability at the MAR. Because spreading rate varies by only V0.5 cm/yr half-rate along the GSC, we attribute the primary cause of these di¡erences to the increasing in£uence of the Ga- la¤pagos hotspot to the east. In particular, we believe that elevated magma supply associated with the hotspot induces a transition from dominantly point-source volcanism in the west to ¢ssure-fed eruptions in the east. Kleinrock and Brooks [44] reported seamount abundances near the 95.5‡W propagating rift that are signi¢cantly higher than our results for the entire rift valley domain. We hypothesize that because the rift is propagating into old crust, local magma supply will be low relative to the ridge axis directly to the east and west of the propagator. Thus, the elevated seamount abundance surrounding the propagating rift is consistent with Fig. 5. Seamount density, Xo, as a function of spreading rate. the along-axis trend of increasing seamount abun- Gray symbols indicate Xo calculated for K 9 2 in this study. dance with decreasing magma supply at the GSC. White symbols represent results of previous studies, A88 [72], It is interesting that along-axis variations in F87 [24], J00 [45], KB94 [44], MS95 [46], SC92 [20], and W98 [43]. Dashed line indicates the general increase in sea- seamount abundance and volcanic style are not mount density with decreasing spreading rate. The variation observed at the Reykjanes Ridge south of Iceland in seamount abundance at the GSC is similar in magnitude [46]. Rather, seamount abundance in the axial to the di¡erences observed between the MAR and EPR.

EPSL 6909 15-12-03 Cyaan Magenta Geel Zwart 338 M.D. Behn et al. / Earth and Planetary Science Letters 217 (2004) 331^347

Fig. 6. Relationship between seamount formation and along-axis variations in magma supply. (a) Histogram of the number of seamounts normalized to the total bathymetry coverage in each of the 12PU12P regions shown in Fig. 4a. (b) Histogram of the mean pillow fraction in dredge samples (see Table 3). (c) Variations in Moho and AMC depth along the GSC. Black dots illustrate picks from along-axis multichannel seismic re£ection pro¢les, gray circles show picks from across-axis pro¢les [30,47]. White squares indicate crustal thickness estimated from Gala-01, -02, and -03 seismic refraction experiments [31]. (d) Axial depth along the GSC. The most rapid increase in seamount abundance occurs immediately to the west of the disappearance of the AMC.

[30,31]. These data show a progressive increase in responds closely with the development of the axial crustal thickness from V5.6 km at 97‡W to V7.5 high. km at 92‡W, with the most rapid increase occur- In contrast to the ‘threshold’ e¡ect observed in ring between 94 and 91.5‡W (Fig. 6c). In addition, AMC depth and axial morphology the number of multichannel seismic data collected near the ridge axial seamounts varies gradually along the GSC axis reveal the presence of an AMC east of (Fig. 6). In particular, little variation in seamount V94‡W [47]. Between 94 and 92.7‡W the AMC density is observed east of 94‡W where seamount is relatively deep (2.5^4 km), while east of 92.7‡W abundance is uniformly low. The most rapid it shallows to depths of 1.5^2.3 km (Fig. 6c). The change in seamount density occurs between 94 transition in AMC depth occurs abruptly and cor- and 96‡W, immediately to the west of the appear-

EPSL 6909 15-12-03 Cyaan Magenta Geel Zwart M.D. Behn et al. / Earth and Planetary Science Letters 217 (2004) 331^347 339 ance of the AMC and west of the largest gradient in crustal thickness. These results suggest that seamount formation is more closely related to the presence of a steady-state magma lens than to either magma chamber depth or crustal thickness. A similar relationship is observed between lava £ow morphology and magma supply at the segment scale along the EPR. White et al. [48] found that ¢elds of pillow mounds form near overlapping spreading centers on the southern EPR (16^ 19‡S), but not in mid-segment regions. The formation of these pillow ¢elds was attributed to lower e¡usion rate eruptions corresponding to the decrease in magma supply near segment ends. Comparing the location of the pillow ¢elds to the portions of the ridge axis where an AMC re£ector is observed [8], we ¢nd that while the mid-segment regions typically show a strong re- £ector, the overlapping spreading centers do not. Thus, a similar relationship between lava £ow morphology and the presence of a steady-state magma chamber is observed on the segment scale of the EPR and along the GSC.

3.3. Seamount height versus AMC depth

Assuming that magma rises buoyantly to the summit of a seamount, Smith and Cann [20] pro- posed that seamount height should be directly re- Fig. 7. (a) Seamount height versus AMC depth. Model curves show height predicted by a buoyancy model with lated to the depth of the magma source. To test h =[(b rock3bmagma)/(bmagma3bwater)]z, where bmagma is the this hypothesis, we compared seamount height to magma density, bwater is the density of water, b rock is the AMC depth for seamounts formed on crust where mean density of the overlying rock, and z is depth of the an AMC was imaged. Assuming that AMC depth magma source [20]. Rock density is calculated assuming porosity, x, decays exponentially with depth following corresponds to depth of the magma source, we 3V z x(z)=xoe . The grain density of unfractured rock is as- ¢nd that a buoyancy model does not ¢t the data 3 3 sumed to be 2950 kg/m . bmagma = 2740 kg/m , bwater = 1000 3 (Fig. 7). Rather, most seamounts are found to kg/m , and xo = 0.3 are used for all calculations. (b) Sea- have a smaller height than would be predicted mount volume versus AMC depth. Model curves assume by buoyancy. buoyancy and a seamount height to radius ratio of 0.15. There are several potential explanations for the Note that the buoyancy model tends to overpredict both the height and volume of the observed seamounts. tendency of the buoyancy model to overestimate seamount height. Assuming seamounts are mono- genetic features (e.g., [9]), it is possible that for a typical seamount-building event the total volume erupted volume could be large, but only a small of magma erupted is not su⁄cient for seamounts portion of this volume will contribute to sea- to reach their equilibrium height. Alternatively, if mount formation with the rest supplying earlier- seamounts form in the ¢nal stages of an eruptive stage sheet and lobate £ows. Finally, tectonic ex- sequence [49] their size may not re£ect the total tensional stresses will tend to decrease the e¡ec- volume of the event. In this scenario the total tive overburden pressure on a magma chamber,

EPSL 6909 15-12-03 Cyaan Magenta Geel Zwart 340 M.D. Behn et al. / Earth and Planetary Science Letters 217 (2004) 331^347

Table 3 Description of lava £ow morphologies for G-PRIME dredges Stn Latitude Longitude Site Flow morphology 2 1‡35.3PN 90‡49.0PW volcano in transform pillow 6 1‡54.1PN 91‡10.8PW ridge axial high pillow 7 1‡55.2PN 91‡16.4PW ridge axis pillow 9 1‡56.0PN 91‡19.3PW ridge axial high sheet/lobate 10 1‡57.6PN 91‡21.7PW large £at-topped seamount sheet/lobate 11 1‡58.3PN 91‡24.1PW small ridge on axis sheet/lobate 13 2‡00.3PN 91‡33.5PW southern limb of OSC sheet/lobate 15 2‡02.7PN 91‡36.3PW northern limb of OSC pillow 17 2‡04.2PN 91‡48.0PW small axial seamount sheet/lobate 19 2‡06.6PN 91‡57.3PW large axial seamount sheet/lobate 20 2‡06.9PN 92‡00.5PW northern limb of OSC pillow/lobate 23 2‡06.5PN 92‡13.3PW narrow ridge tip pillow 24 2‡07.5PN 92‡14.6PW northern limb of OSC pillow 25 2‡08.7PN 92‡19.3PW axial graben sheet/lobate 29 2‡12.1PN 92‡37.1PW narrow axial high pillow 32 2‡16.1PN 92‡52.9PW volcanic mound on axis pillow/lobate 33 2‡16.9PN 92‡58.4PW axial graben pillow 38 2‡21.4PN 93‡16.1PW northern side of axial deep pillow 39 2‡20.5PN 93‡13.1PW axial graben pillow 40 2‡19.5PN 93‡09.5PW small axial seamount sheet/lobate 41 2‡18.8PN 93‡05.6PW northern side of axial deep pillow/lobate 43 2‡24.7PN 93‡15.3PW volcanic mound on axis sheet/lobate 45 2‡26.6PN 93‡21.1PW seamount in axial graben pillow 47 2‡27.8PN 93‡33.7PW seamount in axial graben pillow 48 2‡29.3PN 93‡39.3PW small high in axial graben pillow 49 2‡30.0PN 93‡52.1PW small ridge in axial high pillow 50 2‡30.0PN 93‡46.5PW £ank of axial seamount pillow 53 2‡30.5PN 94‡07.6PW axial deep pillow 56 2‡33.2PN 94‡14.3PW ridge on axial high pillow 58 2‡31.6PN 94‡20.9PW axial seamount pillow 59 2‡32.5PN 94‡25.9PW axial seamount pillow 60 2‡34.3PN 94‡35.9PW local high in axial graben pillow 62 2‡34.9PN 94‡39.6PW volcanic mound in axial graben pillow 64 2‡36.0PN 94‡49.1PW axial rise structure pillow 66 2‡36.9PN 94‡58.6PW local high in axial graben pillow 67 2‡37.6PN 95‡01.9PW large axial seamount pillow 68 2‡37.0PN 95‡08.4PW rise in axis pillow 70 2‡37.8PN 95‡18.9PW narrow axial ridge pillow/lobate 71 2‡25.3PN 95‡36.1PW ridge in north graben pillow 73 2‡17.9PN 95‡42.0PW mound in axial graben pillow 79 2‡05.9PN 96‡43.4PW seamount in axial graben pillow 80 2‡06.8PN 96‡37.7PW seamount in axial valley pillow 81 2‡07.1PN 96‡41.7PW axial seamount pillow 82 2‡07.2PN 96‡46.5PW axial seamount pillow/lobate 83 2‡08.1PN 96‡49.0PW small axial seamount sheet/lobate 84 2‡08.2PN 96‡52.5PW elongated axial seamount pillow 87 2‡08.2PN 97‡05.9PW £ank of axial seamount pillow 88 2‡08.4PN 97‡11.5PW axial seamount pillow 91 2‡08.5PN 97‡36.2PW seamount on axial volcanic ridge sheet/lobate 92 2‡11.5PN 97‡46.9PW irregular axial seamount pillow/lobate

EPSL 6909 15-12-03 Cyaan Magenta Geel Zwart M.D. Behn et al. / Earth and Planetary Science Letters 217 (2004) 331^347 341 thus reducing seamount height relative to an iso- sea£oor observations are required to verify that a static model. direct relationship exists between £ow morphology and the style of constructional volcanism at the GSC. 4. Lava £owmorphology

Observations at the MAR suggest that most 5. Discussion seamounts are constructed primarily of pillow lavas. In contrast, at the EPR where few sea- We interpret the variations in seamount abun- mounts are observed, sheet and lobate £ows dance and lava £ow morphology at the GSC to tend to dominate [2,17]. To determine whether re£ect a change in the primary style of sea£oor the along-axis variation in seamount abundance volcanism from point-source to ¢ssure-fed erup- at the GSC is correlated with changes in lava tions. Analogue models have shown that lava £ow £ow morphology, we examined samples recovered morphology is strongly dependent on lava viscos- from 50 dredges during the G-PRIME cruise. Un- ity and e¡usion rate (e.g., [50^52]). These labora- fortunately, it is often di⁄cult to distinguish be- tory simulations predict that pillow lavas associ- tween £ow morphologies from dredge samples ated with seamount-building eruptions are alone. However, because direct sea£oor observa- characterized by high viscosities and low e¡usion tions have not been made along the western GSC, rates. In contrast, low-relief sheet and lobate £ows these samples provide the only constraints on £ow associated with ¢ssure-fed eruptions are predicted morphology currently available. To minimize the to have lower viscosities and higher e¡usion rates. uncertainty involved in classifying £ow morphol- Intriguingly, the viscosity of mid-ocean ridge ba- ogy from the dredge samples, we assign only a saltic magmas is typically assumed to be relatively pillow or sheet/lobate classi¢cation to each sam- constant (V100 PaWs) and independent of spread- ple. Pillows were distinguished by their massive ing rate and £ow morphology. This observation character, rounded outer surface, and clear pat- implies that e¡usion rate may be the primary con- tern of radial fractures. In contrast, sheet/lobate trol on the style of sea£oor volcanism. However, £ows had less rounded surfaces, lacked radial few studies have systematically investigated the fractures, and often had glassy rims on two sur- relationship between lava viscosity and £ow mor- faces. In certain cases, both pillow and sheet/lo- phology over a wide range of spreading environ- bate £ows were found in a single dredge haul. ments. Thus, to determine whether along-axis var- Flow morphology classi¢cations and a brief de- iations in lava viscosity contribute to the decrease scription of the dredge sites are presented in Table in seamount abundance toward the Gala¤pagos 3 (further information including digital images of hotspot, we calculated viscosities for a suite of the dredge samples are available at http://obslab. lavas dredged during the G-PRIME cruise. whoi.edu/ew0004/index.html). We ¢nd that west of V93.5‡W pillows are the primary £ow mor- 5.1. Variations in lava viscosity phology, with a few sheet/lobate £ows observed at the extreme western end of the study area. In Melt viscosity is a combined function of major contrast, east of V93.5‡W £ow morphologies element composition, temperature, and H2O con- tend to be more equally divided between pillows tent [53,54]. The GSC is characterized by system- and sheet/lobates. These results suggest that the atic variations in basalt chemistry re£ecting in- along-axis variation in seamount abundance is creasing degrees of fractional crystallization to correlated with a change in lava £ow morphology. the east [34,55]. This crystallization trend is Of course, dredge samples containing pillows do clearly observed in the along-axis variation in not necessarily re£ect the presence of a seamount, average Mg# (molar MgO/(MgO+FeO)) for the just as there are many pillows observed along the GSC lavas (Fig. 8a). We used the MELTS algo- EPR that do not form pillow mounds [48]. Future rithm [56] to calculate melt viscosity for 33 dredge

EPSL 6909 15-12-03 Cyaan Magenta Geel Zwart 342 M.D. Behn et al. / Earth and Planetary Science Letters 217 (2004) 331^347 samples for which major element and H2O data tions in viscosity were controlling the increase in are available [55]. Magma pressure was estimated seamount abundance to the west. from the depth of the dredge hole and a quartz^ The bulk viscosity of a lava is also sensitive to fayalite^magnetite (32) bu¡er was assumed for its crystal content [57]. To quantify this e¡ect, we oxygen fugacity. The MELTS algorithm is advan- used an image analysis technique to estimate crys- tageous because it estimates liquidus temperature tallinity for each dredge sample from scanned thin from the composition of the melt, and then incor- sections. A color value was assigned to each pixel porates this temperature into the calculated vis- of the scanned image. The proportion of pixels cosity. In general, we ¢nd that the variations in corresponding to plagioclase and olivine, the lava chemistry result in higher liquidus tempera- dominant phenocrysts in the GSC lavas, was tures and lower melt viscosities with increasing then used to estimate the crystallinity of the sam- distance from the hotspot (Fig. 8). This trend is ple. Typically multiple regions of a single thin the opposite of what would be expected if varia- section were analyzed independently to minimize

Fig. 8. Along-axis variations in magma properties. (a) Mg# (100U[molar MgO/(MgO+FeO)]) for dredged lavas [55] (b) Liquidus temperature calculated from the MELTS algorithm. Calculations assume hydrostatic pressure and a QFM-2 bu¡er for oxygen fugacity. (c) Dissolved H2O content of lavas [55]. (d) Crystallinity measured by image analysis of scanned thin sections. Small symbols represent samples for which H2O data are not available. (e) Bulk lava viscosity incorporating the e¡ects of major element chemistry, liquidus temperature, H2O content, and crystallinity. White symbols indicate dredges on seamounts, black symbols illustrate all other dredges. The absence of an along-axis gradient in the bulk viscosity indicates that viscosity variations alone are not responsible for the decrease in seamount abundance approaching the hotspot.

EPSL 6909 15-12-03 Cyaan Magenta Geel Zwart M.D. Behn et al. / Earth and Planetary Science Letters 217 (2004) 331^347 343 the e¡ects of porosity and to insure that our esti- eruptions. In reality, di¡erences in the rheologic mates were consistent for the entire sample. Based properties of wax and basaltic magmas could in- on these analyses, we believe the reported crystal- troduce signi¢cant error into these estimations of linities are accurate to þ 1%. e¡usion rate and the time of seamount formation. East of 93‡W we ¢nd crystallinity to be uni- The predicted along-axis variations in e¡usion formly low, with values 9 5% (Fig. 8d). However, rate at the GSC may be associated with system- west of 93‡W crystallinity becomes more variable atic changes in the geometry of the eruptive con- ranging from 1 to 17%. Integrating crystallinity duit or the pressure gradient within the conduit. and melt viscosity, the bulk viscosity of each sam- The mean £ow rate through a circular pipe is 32:5 2 ple was calculated using Rbulk = Rmelt(1^1.7P) de¢ned as u =(R /8W)(dp/dx), where R is the [58], where Rmelt is the melt viscosity and P is pipe radius, W is viscosity, and dp/dx is the pres- the crystal fraction. The higher crystallinities sure gradient. Thus, e¡usion rate is most sensitive west of 93‡W tend to o¡set the decrease in melt to conduit geometry, with wider pipes resulting in viscosity away from the hotspot. The result is rel- higher e¡usion rates due to decreased shear along atively uniform bulk viscosities of 10^100 PaWs for the conduit walls. Similarly, a 1-m wide dike will all GSC lavas (Fig. 8e). These results are in good be characterized by a greater mean £ow rate than agreement with laboratory measurements of vis- a 1-m diameter pipe, suggesting that ¢ssure-fed cosity for basaltic melts near the liquidus [59]. eruptions will have larger e¡usion rates than Furthermore, there is no clear correlation between point-source eruptions. the bulk viscosity of lavas dredged from sea- Along-axis changes in e¡usion rate could also mounts and those dredged from other locations. be associated with variations in the pressure gra- Based on these observations, it appears that dient within the conduit. The excess driving pres- along-axis variations in lava viscosity are not re- sure necessary for an eruption has been shown to sponsible for the decrease in seamount abundance increase with the depth of the magma reservoir approaching the Gala¤pagos hotspot. and the strength of the overlying crust [62].This will result in elevated pressure gradients to the 5.2. Variations in lava e¡usion rate west, where the magma reservoirs are likely deep- est. However, after the initial eruption the pres- The absence of systematic variations in lava sure gradient in the conduit will be controlled viscosity suggests that the decrease in seamount only by the relative densities of the magma and abundance along the GSC is controlled primarily the overlying rock. The presence of gas bubbles by changes in e¡usion rate. Although other fac- has a strong e¡ect on magma density. Bottinga tors, such as slope [60], sea£oor roughness [50], and Javoy [63] showed that above 2 kbar (V7 and cooling rate [61] can a¡ect lava £ow mor- km) the vesicularity of basaltic magmas increases phology, it is unlikely that any of these parame- rapidly with decreasing depth due to the exsolu- ters vary su⁄ciently along-axis to explain the ob- tion of CO2. This implies that magmas residing in served variations in seamount abundance. Based shallower reservoirs will tend to have lower den- on our calculated viscosities, we use the experi- sities than those at greater depths. Therefore, the mental results of Gregg and Fink [52] to estimate appearance and shoaling of the AMC may gener- e¡usion rates for eruptions along the GSC. For a ate elevated magma pressure gradients to the east lava viscosity of 100 PaWs, the transition between along the GSC. pillow and lobate £ows occurs at an e¡usion rate Finally, tectonic stresses produced by extension of V0.5^1 m3/s for point-source eruptions. As- will tend to decrease the pressure gradient in the suming seamounts are constructed of pillow lavas, conduit and reduce e¡usion rate. Because tectonic the minimum formation time for a small sea- stresses are likely largest in the rift valley domain mount of volume 50U106 m3 would be V2 yr. where the lithosphere is thickest, this mechanism Of course, this calculation assumes that the labo- is consistent with a decrease in e¡usion rate to the ratory results can be scaled directly to submarine west. However, because seamount abundance ap-

EPSL 6909 15-12-03 Cyaan Magenta Geel Zwart 344 M.D. Behn et al. / Earth and Planetary Science Letters 217 (2004) 331^347 pears to be closely linked to the presence of an dicts the progression of lava £ow morphologies AMC, it is unlikely that long-wavelength varia- that have been observed during a single event at tions in extensional stresses are the sole explana- the GSC (86‡W) [49]. tion for the changes in volcanic style. This model is attractive because it links seamount abundance and lava £ow morphology 5.3. A model for seamount formation with the presence of a steady-state magma chamber, providing an explanation for the di¡erences The correlation between seamount abundance, in volcanic style observed along the GSC as well lava £ow morphology, and magma supply at the as those between fast- and slow-spreading ridges. GSC indicates that e¡usion rate may be directly Furthermore, it is consistent with the segment related to the presence of a steady-state crustal scale variations in lava morphology observed magma chamber. To understand the observed re- along the southern EPR [48]. One region that is lationship between magma supply and construc- not simply explained by this model is the Reyk- tional volcanism we examine variations in the janes Ridge, where wide-angle seismic data [68] characteristics of individual eruptions as a func- and controlled source electromagnetic soundings tion of spreading rate. Although few high-quality [69] indicate the presence of a crustal magma data are available for submarine eruptions, both chamber near 57.75‡N. This location corresponds total erupted volume and ¢ssure length appear to to the region in which Magde and Smith [46] be correlated with spreading rate. Speci¢cally, identi¢ed numerous axial seamounts. One possi- Sinton et al. [64] found that the length of eruptive ble explanation for the dominance of point-source ¢ssures increases with spreading rate, but the vol- volcanism is that the en-echelon structure of the ume of magma erupted during a single event de- Reykjanes Ridge inhibits the formation of a con- creases with increasing spreading rate. The most tinuous AMC. This could result in small, isolated rapid change in these parameters occurs below a magma bodies that are more likely to generate half-rate of V3 cm/yr, approximately the rate seamount-building eruptions. However, until corresponding to the disappearance of a steady- more detailed constraints on the presence and state magma chamber [65]. This implies that the depth of an AMC along the Reykjanes Ridge variations in ¢ssure length and erupted volume are available, this model cannot be tested. may also be sensitive to the presence of a steady-state magma chamber. Based on these observations, we propose a model in which ridges 6. Conclusions characterized by a steady-state magma chamber undergo short, high e¡usion rate eruptions. In In this study, we have developed a robust meth- this scenario, a small volume of lava is trans- od to identify isolated volcanic edi¢ces by search- ported quickly to the surface along a long ¢ssure, ing bathymetry for closed, concentric contours resulting in the formation of low-relief sheet and protruding above the surrounding sea£oor. Using lobate £ows. In contrast, the larger eruptive vol- this technique, we ¢nd that the number of sea- umes characteristic of ridges lacking a steady-state mounts in the axial zone of the GSC decreases AMC cannot be erupted over a short enough time dramatically as the Gala¤pagos hotspot is ap- period to thermally maintain a long ¢ssure. We proached from the west. These variations are sim- hypothesize that these ¢ssures collapse to several ilar in magnitude to the di¡erences in seamount point sources [66], which continue to erupt at a abundance observed between fast- and slow- low e¡usion rate over a longer period of time. spreading mid-ocean ridges. Corresponding to Such a model is consistent with the observation the along-axis variation in seamount density is a that basaltic ¢ssure eruptions often initiate with change in lava £ow morphology from dominantly lava supplied along a long ¢ssure and subse- pillow £ows in the west to a combination of pil- quently converge to one or more vents within lows and sheet/lobates in the east. These changes hours to days (e.g., [67]). Moreover, it also pre- in volcanic style suggest a transition from point-

EPSL 6909 15-12-03 Cyaan Magenta Geel Zwart M.D. Behn et al. / Earth and Planetary Science Letters 217 (2004) 331^347 345 source to ¢ssure-fed eruptions as magma supply References increases. Lava £ow morphology has been shown to be strongly dependent on both lava viscosity and the [1] K. Macdonald, Mid-ocean ridges: Fine scale tectonic, e¡usion rate of an eruption. However, because volcanic and hydrothermal processes within the plate boundary zone, Annu. Rev. Earth Planet. Sci. 10 (1982) dredged lavas show no evidence for along-axis 155^190. variations in viscosity at the GSC, we interpret [2] M. Per¢t, W. Chadwick Jr., Magmatism at mid-ocean the di¡erences in £ow morphology and seamount ridges: Constraints from volcanological and geochemical abundance to re£ect an increase in e¡usion rates investigations, in: W. Buck, P. Delaney, J. Karson, Y. nearer the hotspot. Comparing seamount abun- Lagabrielle (Eds.), Faulting and Magmatism at Mid- Ocean Ridges, Geophys. Monogr. 106, AGU, Washing- dance with axial morphology, crustal thickness, ton, DC, 1998, pp. 59^115. and the presence and depth of an AMC, we ¢nd [3] P. Lonsdale, Structural geomorphology of a fast-spread- that the transition from point-source to ¢ssure-fed ing rise crest: The East Paci¢c Rise near 3‡25PS, Mar. eruptions is most sensitive to the presence of a Geophys. Res. 3 (1977) 251^293. steady-state AMC. Therefore, it is likely that the [4] K. Macdonald, P. Fox, The axial summit graben and cross-sectional shape of the East Paci¢c Rise as indicators e¡usion rate of a typical sea£oor eruption is di- of axial magma chambers and recent volcanic eruptions, rectly related to the presence of a steady-state Earth Planet. Sci. Lett. 88 (1988) 119^131. crustal magma chamber. This model provides [5] E. Hooft, H. Schouten, R. Detrick, Constraining crustal a viable explanation for the di¡erences in vol- emplacement processes from the variation in seismic layer canic style observed at (1) fast- and slow-spread- 2A thickness at the East Paci¢c Rise, Earth Planet. Sci. Lett. 142 (1996) 289^309. ing ridges, (2) the segment scale on the EPR, [6] R. Detrick, P. Buhl, E. Vera, J. Mutter, J. Orcutt, J. and (3) hotspot-in£uenced ridges, such as the Madsen, T. Brocher, Multi-channel seismic imaging of a GSC. crustal magma chamber along the East Paci¢c Rise, Na- ture 326 (1987) 35^41. [7] J. Sinton, R. Detrick, Mid-ocean ridge magma chambers, J. Geophys. Res. 97 (1992) 197^216. Acknowledgements [8] R. Detrick, A. Harding, G. Kent, J. Orcutt, J. Mutter, P. Buhl, Seismic structure of the southern East Paci¢c Rise, We thank the shipboard scienti¢c party and Science 259 (1993) 499^503. crew of R/V Maurice Ewing Leg 00-04 for their [9] R. Ballard, T. van Andel, Morphology and tectonics of help in data acquisition and the Ecuadorian gov- the inner rift valley at 36‡50PN on the Mid-Atlantic Ridge, Geol. Soc. Am. Bull. 88 (1977) 507^530. ernment and the Parque Nacional Gala¤pagos for [10] K. Macdonald, The crest of the Mid-Atlantic Ridge: permission to work in their waters. Sara Wilson Models for crustal generation processes and tectonics, collected the crystallinity data present in this in: P. Vogt, B. Tucholke (Eds.), The Geology of North study and Bu¡y Cushman helped with the geo- America, The Western North Atlantic Region, v. M, chemical analyses. Garrett Ito, Juan Pablo Ca- GSA, 1986, pp. 51^68. [11] J. Karson, G. Thompson, S. Humphris, J. Edmond, W. nales, Debbie Smith, Wenlu Zhu, and Selwyn Bryan, J. Brown, A. Winters, R. Pockalny, J. Casey, A. Sacks provided insightful comments on various Campbell, G. Klinkhammer, M. Palmer, R. Kinzler, M. aspects of this study. We also thank Scott White Sulanowska, Along-axis variations in sea£oor spreading and Je¤ro“me Dyment for their constructive re- in the MARK area, Nature 328 (1987) 681^685. views. The GMT software package (Wessel and [12] D. Smith, J. Cann, Building the crust at the Mid-Atlantic Ridge, Nature 365 (1993) 707^715. Smith, 1998) was used extensively in the prepara- [13] R. Detrick, J. Mutter, P. Buhl, I. Kim, No evidence from tion of this paper. Support was provided by a multichannel re£ection data for a crustal magma chamber CIW/DTM Postdoctoral Fellowship to M.D.B., in the MARK area on the Mid-Atlantic Ridge, Nature NSF Grant OCE-9818632 to the University of 347 (1990) 61^64. Hawaii, and NSF Grant OCE-9819117 to the [14] W. Bryan, J. Moore, Compositional variations of young basalts in the Mid-Atlantic ridge rift valley near lat. Woods Hole Oceanographic Institution. SOEST 36‡49PN, Geol. Soc. Am. Bull. 88 (1977) 556^570. contribution #6294, WHOI contribution #11042. [15] T. Atwater, Constraints from the Famous area concerning [VC] the structure of the oceanic section, in: M. Talwani, C.

EPSL 6909 15-12-03 Cyaan Magenta Geel Zwart 346 M.D. Behn et al. / Earth and Planetary Science Letters 217 (2004) 331^347

Harrison, D. Hayes (Eds.), Deep Drilling Results in the as inferred from gravity analyses, J. Geophys. Res. 100 Atlantic Ocean, Vol. 2, AGU, Washington, DC, 1979, pp. (1995) 3733^3745. 33^42. [30] R. Detrick, J. Sinton, G. Ito, J. Canales, M. Behn, T. [16] E. Bonatti, C. Harrison, Eruptive styles of basalt in oce- Blacic, B. Cushman, J. Dixon, D. Graham, J. Mahoney, anic spreading ridges and seamounts: E¡ect of magma Correlated geophysical, geochemical and volcanological temperature and viscosity, J. Geophys. Res. 93 (1988) manifestations of plume^ridge interaction along the Gala¤- 2967^2980. pagos Spreading Center, Geochem. Geophys. Geosyst. 3 [17] M. Kennish, R. Lutz, Morphology and distribution of (2002) 10.1029/2002GC000350. lava £ows on mid-ocean ridges: A review, Earth Sci. [31] J. Canales, G. Ito, R. Detrick, J. Sinton, Crustal thickness Rev. 43 (1998) 63^90. along the Gala¤pagos Spreading Center and the compen- [18] R. Gri⁄ths, J. Fink, The morphology of lava £ows in sation of the Gala¤pagos hotspot swell, Earth Planet. Sci. planetary environments: Predictions from analog experi- Lett. 203 (2002) 311^327. ments, J. Geophys. Res. 97 (1992) 19739^19748. [32] J.-G. Schilling, R. Anderson, P. Vogt, Rare earth, Fe and [19] D. Smith, J. Cann, Hundreds of small volcanoes on the Ti variations along the Gala¤pagos Spreading Center and median valley £oor of the Mid-Atlantic Ridge at 24‡^ their relationship to the Gala¤pagos Mantle Plume, Nature 30‡N, Nature 348 (1990) 152^155. 261 (1976) 108^113. [20] D. Smith, J. Cann, The role of seamount volcanism in [33] M. Fisk, A. Bence, J.-G. Schilling, Major element chem- crustal construction at the Mid-Atlantic Ridge (24‡^ istry of Gala¤pagos Rift zone magmas and their pheno- 30‡N), J. Geophys. Res. 97 (1992) 1645^1658. crysts, Earth Planet. Sci. Lett. 61 (1982) 171^189. [21] W. Bryan, S. Humphris, G. Thompson, J. Casey, [34] J.-G. Schilling, R. Kingsley, J. Devine, Gala¤pagos hot Comparative volcanology of small axial eruptive centers spot-spreading center system, 1, Spatial petrological and in the MARK area, J. Geophys. Res. 99 (1994) 2973^ geochemical variations (83‡W^101‡W), J. Geophys. Res. 2984. 87 (1982) 5593^5610. [22] D. Smith, J. Cann, M. Dougherty, J. Lin, S. Spencer, C. [35] S. Verma, J.-G. Schilling, Gala¤pagos hot spot-spreading MacLeod, J. Keeton, E. McAllister, B. Brooks, R. Pas- center system, 2, 87Sr/86Sr and large ion lithophile element coe, W. Robertson, Mid-Atlantic Ridge volcanism from variations (83‡W^101‡W), J. Geophys. Res. 87 (1982) deep-towed side-scan sonar images, 25‡^29‡N, J. Volcan. 10838^10856. Geotherm. Res. 67 (1995) 233^262. [36] S. Verma, J.-G. Shilling, D. Waggoner, Neodymium iso- [23] J. Head, L. Wilson, D. Smith, Mid-ocean ridge eruptive topic evidence for Gala¤pagos hot spot-spreading centre vent morphology and substructure: Evidence for dike evolution, Nature 306 (1983) 654^657. widths, eruption rates, and evolution of eruptions and [37] R. Hey, J. Sinton, M. Kleinrock, R. Yonover, K. Mac- axial volcanic ridges, J. Geophys. Res. 101 (1996) donald, S. Miller, R. Searle, D. Christie, M. Atwater, N. 28265^28280. Sleep, H. Johnson, C. Neal, ALVIN investigation of an [24] D. Fornari, R. Batiza, M. Luckman, Seamount abundan- active propagating rift system, Mar. Geophys. Res. 14 ces and distribution near the East Paci¢c Rise 0‡^24‡N (1992) 207^226. based on Seabeam data, in: B. Keating, P. Fryer, R. [38] P. Wessel, S. Lyons, Distribution of large Paci¢c sea- Batiza, G. Boehlert (Eds.), Seamounts, Islands, and mounts from Geosat/ERS-1: Implications for the history Atolls, Geophys. Monogr. 43, AGU, Washington, DC, of intraplate volcanism, J. Geophys. Res. 102 (1997) 1987, pp. 13^21. 22459^22475. [25] P. Huang, S. Solomon, Centroid depths of mid-ocean [39] A. Fitzgibbon, P. Maurizio, R. Fisher, Direct least ridge earthquakes: Dependence of spreading rate, J. Geo- squares ¢tting of ellipses, in: Proc. 13th Int. Conf. on phys. Res. 93 (1988) 13445^13477. Pattern Recognition, Vienna, Austria, 1996, pp. 253^257. [26] J. Phipps Morgan, D. Forsyth, Three-dimensional £ow [40] D. Smith, T. Jordan, Seamount statistics in the Paci¢c and temperature perturbations due to a transform o¡set: Ocean, J. Geophys. Res. 93 (1988) 2899^2918. E¡ects on oceanic crust and upper mantle structure, [41] G. Purdy, J.-C. Sempe¤re¤, H. Schouten, D. Dubois, R. J. Geophys. Res. 93 (1988) 2955^2966. Goldsmith, Bathymetry of the Mid-Atlantic Ridge, 24‡^ [27] C. DeMets, R. Gordon, D. Argus, S. Stein, E¡ect of 31‡N: A map series, Mar. Geophys. Res. 12 (1990) 247^ recent revisions to the geomagnetic reversal time scale 252. on estimates of current plate motions, Geophys. Res. [42] D. Scheirer, K. Macdonald, D. Forsyth, S. Miller, D. Lett. 21 (1994) 2191^2194. Wright, M.-H. Cormier, C. Weiland, A map series of [28] J. Canales, J.J. Dan‹obeitia, R. Detrick, E. Hooft, R. Bar- the southern East Paci¢c Rise and its £anks, 15‡S to tolome¤, D. Naar, Variations in axial morphology along 19‡S, Mar. Geophys. Res. 18 (1995) 1^12. the Gala¤pagos Spreading Center and the in£uence of the [43] S. White, K. Macdonald, D. Scheirer, M.-H. Cormier, Gala¤pagos hot spot, J. Geophys. Res. 102 (1997) 27341^ Distribution of isolated volcanoes on the £anks of the 27354. East Paci¢c Rise, 15.3‡^20‡S, J. Geophys. Res. 103 [29] G. Ito, J. Lin, Mantle temperature anomalies along the (1998) 30371^30384. present and paleoaxes of the Gala¤pagos Spreading Center [44] M. Kleinrock, B. Brooks, Construction and destruction of

EPSL 6909 15-12-03 Cyaan Magenta Geel Zwart M.D. Behn et al. / Earth and Planetary Science Letters 217 (2004) 331^347 347

volcanic knobs at the Cocos-Nazca spreading system near [58] B. Marsh, On the crystallinity, probability of occurrence, 95‡W, Geophys. Res. Lett. 21 (1994) 2307^2310. and rheology of lava and magma, Contrib. Mineral. Pe- [45] G. Jaroslow, D. Smith, B. Tucholke, Record of seamount trol. 78 (1981) 85^98. production and o¡-axis evolution in the western North [59] H. Shaw, Rheology of basalt in the melting range, J. Pe- Atlantic Ocean, 25‡25P^27‡10PN, J. Geophys. Res. 105 trol. 10 (1969) 510^535. (2000) 2721^2736. [60] W. Zhu, D. Smith, L. Monte¤si, E¡ects of regional slope [46] L. Magde, D. Smith, Seamount volcanism at the Reyk- on viscous £ows: A preliminary study of lava terrace janes Ridge: Relationship to the Iceland hotspot, emplacement at submarine volcanic rift zones, J. Volcan. J. Geophys. Res. 100 (1995) 8449^8468. Geotherm. Res. 119 (2002) 145^159. [47] T. Blacic, G. Ito, R. Detrick, J. Canales, J. Sinton, Shal- [61] R. Gri⁄ths, J. Fink, E¡ects of surface cooling on the low crustal structure along the western Gala¤pagos Spread- spreading of lava £ows and domes, J. Fluid Mech. 252 ing Center, 91.3 to 95.2‡W: Correlations between axial (1993) 667^702. magma lens, layer 2A, and topographic characteristics, [62] A. Rubin, Tensile fracture of rock at high con¢ning pres- Eos Trans. AGU 83 (2002) T72C-01. sure: Implications for dike propagation, J. Geophys. Res. [48] S. White, K. Macdonald, R. Haymon, Volcanic mound 98 (1993) 15919^15935. ¢elds on the East Paci¢c Rise, 16‡^19‡S: Low e¡usion [63] Y. Bottinga, M. Javoy, MORB degassing: Bubble growth rate eruptions at overlapping spreading centers for the and ascent, Chem. Geol. 81 (1990) 255^270. past 1 Myr, J. Geophys. Res. 107 (2002) 10.1029/ [64] J. Sinton, E. Bergmanis, K. Rubin, R. Batiza, T. Gregg, 2001JB000483. K. Gronvold, K. Macdonald, S. White, Volcanic erup- [49] R. Ballard, R. Holcomb, T. van Andel, The Galapagos tions on mid-ocean ridges: New evidence from the super- Rift at 86‡W, 3, Sheet £ows, collapse pits and lava lakes fast-spreading East Paci¢c Rise 17‡^19‡S, J. Geophys. of the rift valley, J. Geophys. Res. 84 (1979) 5407^5422. Res. 107 (2002) 10.1029/2000JB000090. [50] J. Fink, R. Gri⁄ths, A laboratory analog study of the [65] J. Phipps Morgan, Y. Chen, The genesis of oceanic crust: surface morphology of lava £ows extruded from point Magma injection, hydrothermal circulation, and crustal and line sources, J. Volcan. Geotherm. Res. 54 (1992) £ow, J. Geophys. Res. 98 (1993) 6283^6297. 19^32. [66] P. Bruce, H. Huppert, Thermal control of basaltic ¢ssure [51] R. Gri⁄ths, J. Fink, Solidi¢cation and morphology of eruptions, Nature 342 (1989) 665^667. submarine lavas: A dependence on extrusion rate, J. Geo- [67] G. Wadge, The variation of magma discharge during ba- phys. Res. 97 (1992) 19729^19737. saltic eruptions, J. Volcan. Geotherm. Res. 11 (1981) 139^ [52] T. Gregg, J. Fink, Quanti¢cation of submarine lava-£ow 168. morphology through analog experiments, Geology 23 [68] D. Navin, C. Peirce, M. Sinha, The RAMESSES experi- (1995) 73^76. ment ^ II. Evidence for accumulated melt beneath a slow [53] Y. Bottinga, D. Weill, The viscosity of magmatic silicate spreading ridge from wide-angle refraction and multi- liquids: A model for calculation, Am. J. Sci. 272 (1972) channel re£ection seismic pro¢les, Geophys. J. Int. 135 438^475. (1998) 746^772. [54] H. Shaw, Viscosities of magmatic silicate liquids: An em- [69] L. MacGregor, S. Constable, M. Sinha, The RAMESSES pirical method of prediction, Am. J. Sci. 272 (1972) 870^ experiment ^ III. Controlled-source electromagnetic 893. sounding of the Reykjanes Ridge at 57‡45PN, Geophys. [55] B. Cushman, J. Sinton, G. Ito, J. Dixon, Hydrous melting J. Int. 135 (1998) 773^789. creates lower mean F at the Gala¤pagos Spreading Center [70] A. Gripp, R. Gordon, Current plate velocities relative to ^ modeling results from a major and volatile element the hotspots incorporating NUVEL-1 global plate motion study, Eos Trans. AGU 83 (2002) T72C-02. model, Geophys. Res. Lett. 17 (1990) 1109^1112. [56] M. Ghiorso, R. Sack, Chemical mass transfer in mag- [71] M. Kurz, D. Geist, Dynamics of the Gala¤pagos hotspot matic processes. IV. A revised and internally consistent from helium isotope geochemistry, Geochim. Cosmochim. thermodynamic model for the interpolation and extrapo- Acta 63 (1999) 4139^4156. lation of liquid^solid equilibria in magmatic systems at [72] G. Abers, B. Parsons, J. Weissel, Seamount abundances elevated temperatures and pressures, Contrib. Mineral. and distributions in the southeast Paci¢c, Earth Planet. Petrol. 119 (1995) 197^212. Sci. Lett. 87 (1988) 137^151. [57] R. Roscoe, Suspensions, in: J. Hermans (Ed.), Flow Properties of Disperse Systems, North-Holland, Amster- dam, 1953, pp. 1^38.

EPSL 6909 15-12-03 Cyaan Magenta Geel Zwart