The creation of ocean crust by lava eruptions is a fundamental Earth process, involving immediate and immense transfers of heat and chemicals from crust to ocean. This transfer creates event plumes (“megaplumes”), massive ellipsoidal eddies with distinctive and consistent chemical signatures. Here we report the discovery of unique event plumes associated with a 2008 eruption on the Northeast Lau Spreading Center. Instead of a large plume hundreds of meters thick, we detected at least eight individual plumes, each ∼50 m thick and apparently only 1–3 km in diameter, yet still rising 200–1000 m above the eruption site. Low and uniform 3He/heat (0.041 × 10−17 mol/J) and dissolved Mn/heat (0.04 nmol/J) ratios in water samples were diagnostic of event plumes. High H2 concentrations (up to 9123 nM) and basalt shards confirmed extensive interactions between molten lava and event plume source fluids. Remote vehicle observations in 2009 mapped a new, small (1.5–5.8 × 106 m3) lava flow. Our results suggest that event plumes are more variable, and thus perhaps more common, than previously recognized. Small event plumes may be preferentially associated with small or sheet-flow eruptions, and massive event plumes with slowly extruding pillow mounds 25–75 m thick. Despite this correlation, and high H2 concentrations, existing theory and seafloor observations argue that cooling lava cannot transfer heat fast enough to create the buoyancy flux required for event plumes. The creation of event plumes under a broad range of eruption conditions provides new constraints for any theory of their formation.
Intrusions and eruptions of magma build the ocean crust, yet the physical and chemical interactions that occur between crust and ocean during these events remain poorly understood. The serendipitous discovery of an event plume, or “megaplume,” on the Juan de Fuca Ridge in 1986 was the first indication that eruptions could be accompanied by brief and massive releases of hot, chemical-rich fluids [Baker et al., 1987]. The 1986 event plume, a oblate ellipsoid 700 m thick and 20 km in horizontal diameter (∼150 km3) with a total hydrothermal heat anomaly of ∼1017 J [Baker et al., 1989], was apparently accompanied by an eruption of ∼0.05 km3 of lava [Chadwick and Embley, 1994; Chadwick et al., 1991]. Since then, three additional ridge crest eruption events on the Juan de Fuca and Gorda ridges have produced six separate event plumes (“Confirmed” in Table 1) with volumes ranging from ∼15 to 128 km3 [Baker, 1998]. In addition, five plumes (“Possible” in Table 1) with some characteristics similar to event plumes have been observed at ridges worldwide, with spreading rates from 11 (Gakkel Ridge) to 140 (Manus Basin) mm/yr (Figure 1). None of the plumes outside the NE Pacific, however, were mapped in three dimensions or could be linked to a specific seafloor eruption.
Table 1. Physical Characteristics of Event Plumes and Associated Eruptions
The most fundamental characteristics of all event plumes to date are their large size and chemical uniformity. Confirmed diameters range from 5 to 20 km (and perhaps up to 70 km [Murton et al., 2006]), and thicknesses from 0.5 to 1.2 km. Concentrations (in terms of the species/heat ratio) of diagnostic magmatic tracers such as 3He and dissolved Mn (DMn) are nearly uniform among all event plumes and lower than the same ratios found in chronic (non-event) plumes and high-temperature seafloor discharge [Lupton et al., 1989, 1999a, 2000; Massoth et al., 1995, 1998]. Thick, high-rising, and broadly dispersed plumes can also occur within very deep axial valleys, such as the Gakkel Ridge, where vertical density gradients are low and plumes become trapped by unbroken valley walls [e.g., Edmonds et al., 2003]. However, these plumes have neither the three-dimensional symmetry nor chemical characteristics of event plumes.
Here we report observations of hydrothermal discharge associated with a seafloor eruption in November 2008 on the Northeast Lau Spreading Center (NELSC), a back-arc ridge behind the Tonga arc (Figure 2). In May 2009 a response cruise sponsored by NOAA VENTS and NSF [Resing and Embley, 2009] made repeat water column measurements at the presumed eruption site, and used the remotely operated vehicle (ROV) JASON to confirm that lava was erupted in 2008, collect rock samples, and sample vent fluids. The eruption may have been one of a series of small fissure eruptions at this site that began as early as mid-2008 [Rubin et al., 2009]. Associated with the eruption was a unique suite of small plumes matching the distinctive chemical composition of previous event plumes, but with volumes 10–1000 times smaller. This relatively small event extends the correlation between event plume and lava heat content across two orders of magnitude. The discovery of extraordinarily high concentrations of H2 gas and abundant glass shards in the plumes confirms the interaction of event plume fluids and molten lava [Kelley et al., 1998]. Despite these observations, a detailed analysis of lava cooling rate using existing theoretical calculations and field observations shows that heat transfer from cooling lava rapid enough to create the buoyancy flux needed to generate event plumes (hours, not days) is highly improbable. Our observations provide new constraints for theories of why seafloor eruptions are associated with brief and massive discharges of unique hydrothermal fluids.
2. Geological Setting
The NELSC is one of several spreading ridges accommodating back-arc extension behind the Tonga arc [Hawkins, 1995; Pelletier et al., 1998; Zellmer and Taylor, 2001] (Figure 2). In the northern Lau Basin, the ridges forming the Mangatolo Triple Junction and the Fonualei Ridge Spreading Center have opening rates of 85 to 94 mm/yr [Zellmer and Taylor, 2001], implying a similar rate for the NELSC. The southernmost segment of the NELSC is a knife-edge ridge ∼15 km long with a bathymetric high in the middle and prominent cones (Maka and Tafu) at either end (Figure 2). In 2008 Nautilus Minerals Inc. made the first visual confirmation of high-temperature venting (∼315°C) on the summit of Maka, the southern of these two cones (J. J. Lowe, Nautilus Minerals Inc., personal communication, 2008).
Our main survey tool was a Sea Bird 911plus Conductivity-Temperature-Depth (CTD) package, with light backscattering and oxidation-reduction potential (ORP, sometimes referred to as Eh) sensors. A total of 15 vertical casts and tow-yos were conducted on and around the southern segment of the NELSC during 20–27 November 2008 (Figure 2). The voltage output of the light-backscattering sensors is equivalent to nephelometric turbidity units (NTU) [American Public Health Association, 1985]; ΔNTU is the value in ambient nonplume water. ORP is very sensitive to short-lived reduced chemicals in hydrothermal plumes, such as Fe+2 and HS- [Walker et al., 2007]. The ORP sensor measures the electrode potential (E (mV)) between seawater and a reference solution. Absolute values of E can vary because of instrumental drift and hysteresis (the response is instantaneous but recovery time can last several to tens of minutes), especially in concentrated plumes, so we rely on the time derivative, dE/dt (mV/s) [Nakamura et al., 2000] to identify the precise location of anomalies. Because E declines when it encounters reduced substances, the anomalies are negative. Miniature Autonomous Plume Recorders (MAPRs) [e.g., Baker et al., 2010], with identical scattering and ORP sensors, supplemented the CTD on some tows.
We calculate the hydrothermal temperature anomaly (Δθ) in a neutrally buoyant plume from the expression
where θ is potential temperature, σθ is potential density, and m0, m1, and m2 are constants in a linear or polynomial regression between θ and σθ in hydrothermally unaffected water around the plume. Because we will compare species/heat ratios in plumes in the Lau Basin with those in the northeast Pacific, we use the technique of McDougall  and McDuff  to adjust for the effect of hydrography and vent fluid characteristics on the observed Δθ. Those authors show that the plume heat flux at the equilibrium level, ΔθQ, where Q is the fluid volume flux, equals the vent heat flux, ΔθvQv, when multiplied by two correction terms, one related to ocean stratification and one to the vent source θv and salinity (Sv). Thus
where R = [(α/β)(dθ/dS)], Rv = [(α/β)(dθv/dSv)], dθ/dS is the vertical θ-S gradient of the ambient water column through which the plume rises, dθv/dSv is the ratio of the differences in θ and S between seawater and vent fluid, α is the coefficient of thermal expansion, and β is the coefficient of haline contraction (see the Notation list at the end of the paper). To ensure that the plume and vent heat fluxes agree, the observed Δθ must be multiplied by a correction factor, ([1/(1 − R)][1 − (R/Rv)])−1.
Because we have no information on vent source S during the creation of any event plume, we assume, for consistency with other event plumes, that Sv = S and set Rv = ∞. For the northeast Pacific, R = −0.97, giving a correction factor of 0.51 and indicating that the observed Δθ values must be multiplied by a correction factor of 2. For the Lau Basin, R = −2.4, the correction term is 0.29, and the observed Δθ values must be multiplied by a correction factor of 3.4. Lavelle et al.  calculated similar correction factors using a more generalized approach that extended the corrections to nonlinear hydrographic profiles and event plumes. They also showed that the corrections are independent of the entrainment coefficient of the ascending plume.
Although we have no information to make corrections for Sv, it is instructive to understand the effect such changes would have on the Δθ calculations (Figure 3). In the northeast Pacific, the effect is not drastic. For vent fluid temperatures of 350° and 700°C a significant change in the correction factor (>2 times increase) occurs only for 350°C fluids with salinities fresher than ∼5 psu. At the NELSC, a steeper dθ/dS gradient in the local water column leads to much greater variability. At 350°C, Rv is either greater than or less than R depending on dSv. Around the dSv value of −24.64 (i.e., a vent fluid S of ∼10 psu), the correction factor takes on very high positive or negative values, leading to high variability in the corrected Δθ. At 700°C, the correction factor stays monotonic but increases steadily as Sv decreases. The high sensitivity of the correction factor to low vent fluid salinity provides a strong constraint on possible values of Sv for the source fluids of the NELSC plumes.
Water samples were collected on every CTD cast using PVC sampling bottles. Samples for helium isotopes were sealed into copper tubing using a special hydraulic crimping device [Young and Lupton, 1983]. Helium isotope ratios were measured at the NOAA/PMEL laboratory in Newport, Oregon, using a dual collector mass spectrometer designed specifically for helium measurements [Lupton, 1990]. We determined the regional 3He concentration profile from background stations occupied away from the eruption site during the cruise and then subtracted a depth-appropriate background value from all plume samples. Immediately after sample collection, dissolved H2 was measured shipboard using a standard headspace equilibration technique as described by Kelley et al. . Dissolved Mn was determined with a precision of ∼1 nM by the direct injection method as described by Resing et al. .
We can evaluate the accuracy of corrected species/heat ratios in plumes by comparing them to values measured on samples collected from vent orifices located immediately below the plume samples. Regressing Δθ against 3He for eight plume samples <300 m above the summit of Maka (cast V08C-27) gave a 3He/heat ratio of 3.3 × 10−17 mol/J (r2 = 0.87, ± 0.5 × 10−17 mol/J at the 95% C.I.), in reasonable agreement with the value of 5 × 10−17 mol/J from high-temperature (315°C) fluids on Maka measured on samples collected in 2009. This source-plume agreement demonstrates that we can confidently distinguish the order-of-magnitude differences in 3He/heat ratios found between chronic (non-eruption) plumes and those generated by eruption events (as discussed below).
4. Plume Observations and Comparison to Previous Event Plumes
4.1. Hydrothermal Setting
The first exploration for hydrothermal activity on the NELSC was a series of widely spaced light-scattering profiles collected in 2004, of which only three were over the southern segment [German et al., 2006] (Figure 4). The profiles over Maka and the segment center show a broad plume layer between 1200 and 1500 m (a weak plume between 1250 and 1350 m was also present on a profile ∼6 km southwest of Maka). No anomaly was seen over Tafu, the northern cone. In 2006, Kim et al.  conducted a tow (CTD 12) and cast (CTD 11) along the segment, again finding a thick plume over Maka. No shallow plume was seen between the cones as in 2004, but weak, near-bottom, light-transmission anomalies occurred near the segment center at ∼15.395°S (Figure 4). Kim et al.  also detected a thin plume apparently originating from the summit of Tafu, and dredged a high-temperature chimney at ∼1900 m on the northwest flank of Maka, coincident with a water column anomaly interpreted as a buoyant plume.
4.2. Physical Characteristics
Previous event plumes have differed from chronic plumes in their symmetry and size, indicative of their origin by brief and massive injections of hot fluids. The shape of confirmed event plumes approximates an oblate ellipsoid, with diameters ranging from 5 to 20 km and thicknesses from 500 to 1200 m (Table 1). Event plumes detected over the NELSC were unique in both their small size and their abundance.
In 2008, high-rising plumes were initially detected on tow-yo T08C-07 (20 November), our first operation at the NELSC (Figure 5). On the final upcast, a series of thin plumes with high ΔNTU and Δθ values was detected between 600 and 1200 m, as much as 1000 m above the adjacent ridge crest axis. Extreme dE/dt anomalies, <−6 mV/s, confirmed that the plumes were relatively young (also confirmed by extraordinarily high H2 concentrations; see section 4.3.2). Our next tow along the ridge, T08C-09 (21 November), found another remarkable series of thin and intense plumes between 1600 and 900 m (Figure 6). We interpret this distribution as a record of sequential and discrete fluid discharges creating plumes of diminishing buoyancy flux (i.e., diminishing rise height) injected from the ridge crest into a southward current. (Note the southward advection of the plume from Maka in Figure 5.) Maximum dE/dt anomalies were much smaller than on T08C-07, implying somewhat older plumes. Plumes shallower than ∼1500 m had a consistent thickness of ∼40–70 m. Maximum rise height and Δθ values were comparable to the original 1986 “megaplume,” EP86 [Baker et al., 1987], although for that event a single plume encompassed the entire water column from 1200 to 2000 m.
This plume distribution lasted only a few days. By the time of the final CTD profiles of the cruise (V08C-18 (24 November), V08C-27 (27 November), T08C-18 (27 November)), we detected no plumes shallower than those from the Maka summit (∼1300 m minimum depth), and even the Maka plumes were absent from profiles over the eruption site. This rapid advection of the event plumes away from the eruption site also implies that the plumes we observed on 20 and 21 November were no older than approximately one week.
Using the conservative assumptions that the plumes were symmetrical and that T08C-09 passed through the center of each (its path followed the axis of the ridge crest), the total volume of all the event plume lenses seen on that tow is ∼1 km3, ∼15 times less than EP93A, the smallest event plume previously recorded [Baker et al., 1995]. (For operational reasons, no cross-axis tows were conducted to confirm this assumption.) Such thin plumes are not without precedent during eruption events, however. A CTD cast at the CoAxial eruption site, serendipitously conducted while the seismic activity was still underway, found a single 50 m thick plume layer ∼600 m above the seafloor [Baker et al., 1995]. A vertical profile through EP93A, about 10 d later, also showed pronounced layering throughout its 600 m thickness. Thus smaller eruptions associated with smaller event plumes (e.g., at the NELSC) may release insufficient energy to uniformly mix the water column over vertical length scales of hundreds of meters. The small, layered plumes we observed in 2008 also suggest that event plume structure grades from a large, highly symmetrical vortex (i.e., a “megaplume”) created during a large eruption, such as the 1986 or 1987 Juan de Fuca Ridge events, to an array of thin, layered “blobs” as created during the NELSC eruption. These blobs could be symptomatic of pulsed eruption activity.
Because our physical and chemical plume measurements strongly suggested a very recent eruption, perhaps only days or hours old, we had the opportunity to characterize the plumes within ∼50 m of the seafloor at a remarkably early stage after an eruption. In 2008 our sole clue to the eruption location was the intersection of the seafloor and the plumes seen on tow T08C-09 (Figure 6). A Δθ of 0.7°C recorded just above the seafloor at 15.389°S during T08C-09 was convincing evidence of the source location, so we conducted two additional tows along the ridge crest that would pass through that area 18 h (T08C-10) and 7 days (T08C-18) after T08C-09 (Figure 7). During each tow, the CTD altitude was kept at <50 m (and usually <30 m) through the target area. A ROV transect conducted in July 2008 along this part of the ridge crest had found only the low-temperature Nautilus vent (J. J. Lowe, Nautilus Minerals Inc., personal communication, 2008).
Temperature and dE/dt anomalies along those three tows suggested the presence of numerous fluid sources concentrated along the shallowest portion of the ridge crest (<∼1650 m), between 15.381° and 15.392°S (Figure 7). Δθ values >0.2°C and dE/dt anomalies <−0.5 mV/s were common. The tows also identified an apparent low-temperature vent site with a strong dE/dt response south of 15.40°S, deeper than ∼1800 m, that apparently corresponds to the deep plume seen between 15.4° and 15.41°S on T08C-09 (Figure 6).
A repeat CTD tow conducted in May 2009 found no evidence of continuing discharge from the shallowest portion of the ridge crest. However, detailed mapping of the eruption area by JASON did confirm the presence of fresh, unsedimented lava, called the Puipui flow, extending about 1.7 km along the ridge crest, with a cross-axis width of 250–400 m [Rubin et al., 2009]. Over most of its length, the flow comprises highly vesicular (up to ∼50% near the eruptive vents) basalts in the form of high-effusion-rate sheets and lobate flows, with pillow lavas mostly limited to flow margins [Clague et al., 2010]. JASON also observed weak diffuse venting (<20°C) at two locations on the lava flow, including the Nautilus site discovered in 2008 (Figure 7).
The high Δθ anomalies seen both in the event plumes and just above the seafloor over the lava flow provide an important constraint on the salinity of the source fluids. For the local hydrography, the Δθ correction factor becomes unreasonably high or even unstable as high-temperature source fluids become fresher (Figure 3). Thus it is probable that the source fluids for both sets of plumes could not have been of unusually low salinity. This is a significant restriction, since immediate post-eruption fluids from high-temperature vents can have Sv-S values <−30 [Von Damm et al., 1995; Lilley et al., 2003], which would result in strongly negative Δθ anomalies at the NELSC.
4.3. Chemical Characteristics
4.3.1. 3He and DMn
The most diagnostic characteristics of event plume water chemistry are the uniquely low and consistent ratios of 3He and DMn to hydrothermal heat (Figures 8, 9, and 10 and Table 2) [Lupton et al., 1999a, 2000; Massoth et al., 1995, 1998]. Historical means ( ±1σ) are 0.18 ± 0.022 × 10−17 mol/J for 3He/heat and 0.059 ± 0.014 nmol/J for DMn/heat. (Note that both values are lower than previously reported because of the hydrographic temperature corrections to plume Δθ, as discussed in section 3. Some DMn values reported by Massoth et al. [1995, 1998] are actually total dissolvable Mn, which can be a few percentage points higher than true DMn.) Because both 3He and hydrothermal heat are conservative tracers, we focus on their ratio to place event plumes in the context of magmatic activity.
Table 2. Chemical Characteristics of Event Plumes and Vent Fluids
Lupton et al. [1999a, 2000] noted that 3He/heat ratios fall into two groups: low ratios (mostly <1 × 10−17 mol/J) in event plumes and high-temperature fluids from magmatically quiescent locations, and high ratios (mostly >1 × 10−17 mol/J) from eruption-associated chronic plumes, especially plumes sampled very close to newly erupted lava flows. The two groups could thus be considered end-members that, when combined, are consistent with the theoretical 3He/heat ratio within the upper mantle, ∼2 × 10−17 mol/J [Lupton et al., 1989].
A compilation of time series measurements in chronic plumes following an eruption shows that those 3He/heat ratios, unlike in event plumes, vary not only temporally but also among eruptions (Figure 11). 3He/heat ratios in post-eruption chronic plumes can exceed the magmatically quiescent value typical of each eruption site by a factor of two or three. Lilley et al.  observed a similar trend in post-eruption time series measurements of vent fluids following the 1991 eruption near 9.9°N on the East Pacific Rise and the 1999 magmatic event at the Endeavor segment of the Juan de Fuca Ridge. 3He/heat ratios in post-eruption chronic plumes can remain high for at least a year. High-resolution sampling suggests that the highest values often occur some months after the eruption, perhaps when the extruded and/or intruded magma has cooled enough that cracks allow seawater access to 3He and other volatiles in the basalt [e.g., Baker et al., 1995; Lupton et al., 1999b]. The high variability of post-eruption chronic plumes indicates that their source fluids have an origin separate from those of event plumes.
Ratios found in the 2008 event plumes and associated discharge had many similarities to previous eruptions, but with some distinct differences that make the 2008 Lau plumes unique (Figures 9, 10, and 11 and Table 2). For 3He/heat, event plume samples were the lowest ever measured, 0.041 × 10−17 mol/J. As at all other eruption sites, the post-eruption plumes over the lava and <∼100 m above the seafloor (i.e., 1500–1650 m) had ratios (0.1 to 0.3 × 10−17 mol/J) consistently higher than the event plumes; these ratios showed no temporal trend on five casts collected over a week's time. However, these plumes were unique among other eruptions in that their 3He/heat ratios were 10–20 times lower than historical chronic plume and vent samples associated with recent magmatic activity on other mid-ocean ridges. And unlike other eruptions, these post-eruption plumes had 3He/heat ratios much lower than the non-eruption-associated discharge sources on the NELSC. Plumes sampled in 2008 from ridge sources deeper than the eruption or on the adjoining NELSC segments (1706–1888 m) were 10 times higher, comparable to chronic plumes just after eruptions on other mid-ocean ridges (Figure 11). Ratios from the Nautilus and Maka vents (sampled in 2009) are even higher, at 1.6 to 5 × 10−17 mol/J.
Unlike 3He/heat, DMn/heat ratios in the 2008 event plumes were within the range of past eruption plumes (Figure 10 and Table 2). Post-eruption and non-eruption plumes had similar DMn/heat ratios, both falling in the middle of the broad range of DMn/heat ratios from high-temperature vents at magmatically quiescent sites (Figure 10). Modest increases in chronic plume DMn/heat have sometimes accompanied past eruptions [Baker et al., 1995; Massoth et al., 1994], but even these increases are small relative to geographic variability.
Concentrations of H2 have rarely been measured in event plumes. Small concentrations of H2 in event plumes were first seen in EP96A and B over the Gorda Ridge (up to 47 nM in EP96A) [Kelley et al., 1998]. These observations were noteworthy because evidence suggests that H2 in plumes results from water/rock reactions, particularly at high temperatures, wherein H2 is produced from the reduction of water by reduced iron compounds in the rock [Sansone et al., 1991; Seyfried and Ding, 1995; Lilley et al., 2003; McCollom and Seewald, 2007]. In slow spreading environments, H2 can be produced by the serpentinization of mantle rocks [Charlou et al., 2002; Proskurowski et al., 2006]. The 2008 event plumes were unique in having consistently high (up to 9123 nM) H2 concentrations [Baumberger et al., 2009]. Unlike 3He and DMn, H2 concentrations are highest in the event plumes and lowest in the chronic plumes (Figure 8). All shallow plume samples had high but variable H2/heat ratios (1.6 to 6.0 nmol/J). This variability presumably arose from some combination of differing extents of lava-fluid interaction during the formation of individual plume bursts, and progressive H2 loss by microbial oxidation. The few measurements available for microbial oxidation rates in hydrothermal plumes indicate H2 half-lives ranging from a few hours to a few days [Kadko et al., 1990; Lilley et al., 1995; McLaughlin-West et al., 1999]. The fact that samples from adjacent event plumes in 2008 had roughly similar H2/heat ratios, and that these ratios did not decrease with distance from the eruption site (i.e., time since discharge) (Figure 6), suggests that variability in source fluid H2 was more important than oxidation through time in creating the event plume variability in H2/heat ratios.
We suggest that the 100-fold difference in maximum H2 concentrations (and in H2/heat ratio) between EP96A [Kelley et al., 1998] and EP08 is dominantly a function of eruption style, not H2 oxidation, because both event plumes were of roughly similar ages. Based on the seismic record [Fox and Dziak, 1998], EP96A was most likely ∼1 day old when discovered, and certainly no more than 10 days old. This young age is consistent with an age estimate of <3 days based on the ratio of dissolved to particulate Fe in the plume [Massoth et al., 1998]. Clague et al.  have now documented that an almost unfailing characteristic of submarine eruptions, sampled throughout the Pacific at depths from 1400 to 3800 m, is the production of glassy pyroclastic fragments that are remnants of bubbles of magmatic gas. Such fragments were abundant on and around the Puipui flow on the NELSC [Clague et al., 2010], and imply extensive opportunities for lava-seawater interaction during the eruption. Of the dozens of eruption sites that Clague et al.  sampled in the Pacific, the 1996 Gorda Ridge site alone was without pyroclastic debris.
All other NELSC plume samples, whether from just above the lava field or from venting much deeper than the eruption, had uniformly low H2/heat ratios (<0.41 nmol/J, Figure 8), as also found in chronic venting from the Gorda Ridge eruption [Kelley et al., 1998]. Fluids supplying these plumes had much less, if any, contact with molten lava.
We can assess the scale of H2 water-rock interaction by estimating the volume of rock needed to supply the observed H2. In the 2008 event plumes, H2 and Δθ are linearly correlated (r2 = 0.68), so from the total hydrothermal heat inventory in the plumes (1.2 × 1015 J) we estimate the total H2 at 4.4 × 106 M. To produce this quantity of H2 an equivalent molar amount of water is needed, which corresponds to 7.9 × 104 kg, or 79 m3, of H2O. We assume that 83% of Fe in the basalts is FeO, which is the median Fe2+/(FeTOT) value of six Lau Basin basalts having similar major element compositions and a Fe2+/(FeTOT) range of 70 to 90% [Nilsson and Peach, 1993]. A similar molar quantity of FeO, or 3.2 × 105 kg, must react with the H2O. Analysis of the Puipui basalts shows a FeO abundance of 9.1%, thus requiring 3.5 × 106 kg, or ∼1200 m3, of basalt. The area of the flow, based on ∼14 h of ROV mapping, is 0.6–0.9 km2. These are likely minimum values, as the off-axis extent of the flow was investigated in only one region. Estimates of the flow volume from JASON observations [Rubin et al., 2009] and bathymetry differencing between 2006 and 2009 (W. W. Chadwick, Jr., personal communication, 2011) yield a flow volume of ∼1.5 to 5.8 × 106 m3. Only 0.08–0.02% of the lava had to react with seawater to produce the H2 inventory in the event plumes (assuming the entire lava volume was emplaced at one time). Because this small volume is equivalent to an average thickness of ∼1–2 mm over the flow area, it is consistent with H2 forming primarily during the fraction of a second required for the surface (i.e., the outermost 1–2 mm) of molten lava to attain near-ambient temperature [Gregg and Fornari, 1998].
Another indication that the event plume source fluids experienced lava-seawater interaction was the discovery of abundant glassy shards in the plumes (Figure 6). Such fragments are commonly found in association with submarine eruptions [Maicher and White, 2001; Eissen et al., 2003; Clague et al., 2009]. The largest shards sampled in 2008 had Stokesian fall velocities on the order of 100 m/d and were found as high as 600 m above the depth of the eruption zone. The presence of such particles in thin (∼60 m) plumes is consistent with the presumption that these plumes were no more than a few days old when observed.
5. The Role of Lava in Event Plume Formation
The 2008 NELSC eruption extends the correlation between event plume heat anomaly and lava heat content across a factor of ∼100 in both variables (Figure 12). The event plume heat as a percentage of total available lava heat varies widely, from 4 to 160%. This range could vary because there is no causal relationship between the two, because of the uncertainties in accurately determining the volumes of event plumes and lava flows, or because the percentage of lava heat transferred to event plumes has varied among sites. If event plume heat derives from lava cooling, we would expect both a correlation between the two heat inventories and the lava heat inventory to always exceed the event plume inventory. Uncertainties arise in these calculations, of course, because underestimation of event plume heat is always possible (e.g., undetected event plumes) and the lava heat determination depends on an accurate knowledge of the lava volume. Nevertheless, the fact that the 1993 CoAxial event plumes held >100% of the calculated lava heat (assuming the lava volume has not been underestimated by more than ∼50%) is a challenge for the lava-cooling hypothesis.
If event plumes do form solely by the cooling of erupted lava [Palmer and Ernst, 1998, 2000], most of the available lava heat (Figure 12) must be transferred fast enough to satisfy the short time constraints of event plume formation. Models of event plume creation [Lavelle, 1995] and measurements of the precipitation rate of dissolved Fe in event plumes [Massoth et al., 1995, 1998] provide convincing evidence that event plume formation time is hours rather than many days. Our observations of fast-settling basalt shards, high H2 concentrations, high dE/dt anomalies, and rapid advection of the observed plumes out of the eruption area also require a short formation interval.
Although there are no direct measurements of the in situ cooling rate of solidifying submarine lavas, theoretical calculations and field observations provide reliable estimates of typical cooling rates. Gregg and Fornari  calculated the growth rate of the solid crust (defined as the depth of the glass transition temperature (732°C)) of a submarine lava flow and found the rate to be proportional to the square root of time (also derived by Griffiths and Fink ). Ideally cooling lava crust reaches a 0.015 m thickness in ∼40 s, 0.1 m in ∼30 min, and 1 m in ∼48 h. Cooling to 732°C releases about half the original lava heat. The total heat lost from a thickness of cooling crust is somewhat greater, however, because the solidified crust continues to cool as the 732°C isotherm progresses inward.
The validity of this theoretical calculation can be tested against a real lava flow using data from a high-precision pressure sensor serendipitously encased within the 1998 sheet flow eruption at Axial volcano [Fox et al., 2001]. Pressure changes showed that initial inflation (during lava extrusion) lasted 72 min, followed by 81 min of steady deflation and 44 min of a second minor inflation/deflation event. Thus for at least 3 h and 17 min after the eruption began, the thin sheet flow remained fluid enough to inflate 3.5 m and deflate 2.5 m. Chadwick  modeled the drainback process by measuring the thickness of drainback cavity lids (∼0.1 m) and crustal shelves (∼0.015 m) on lava pillars within the cavities. These thicknesses and their formation time (∼1 h and 24 s, respectively), calculated from the pressure sensor record, agree with the theoretical growth rate of lava crust calculated by Gregg and Fornari . (The measured growth time of the cavity lid (1 h) is somewhat longer than the theoretical value (0.5 h) because of additional heat supplied by the flowing lava and uncertainties in applying theory to actual rock.) Moreover, the fact that most of the emplaced lava volume remained fluid, and thus hotter than ∼1000°C, for at least 3 h after the eruption began restricts the lava heat loss to only ∼10–25% of its total content (depending on the extent of crystallization in the lava).
An additional proposed cooling mechanism is lava fountaining during volatile-rich eruptions of the Strombolian or Hawaiian style [e.g., Head and Wilson, 2003]. Despite the near-ubiquitous association of quenched pyroclastic debris and eruptions, however, pyroclastic fragments apparently account for only a negligible mass fraction of the erupted lava even when they appear “abundant” on the seafloor. The mass fraction of pyroclasts has been quantified at only one site, “NESCA” in the Gorda Ridge Escanaba Trough, where it totaled <0.02% of the total lava mass [Clague et al., 2009]. Thus their rapid quenching can transfer only a negligible fraction of the lava heat to seawater.
These results indicate that even for thin sheet flows (as at the NELSC) near-complete cooling in <∼1 day is improbable. The case for rapid heat transfer at other sites is even more problematical. Eruptions at Cleft, CoAxial, and Gorda Ridge all produced thick (25–75 m) constructions of pillow mounds that likely required several days to several weeks to complete [Chadwick and Embley, 1994; Gregg and Fink, 1995; Chadwick et al., 1998; Rubin et al., 1998]. At the Gorda Ridge site, the pillow mound is unusually thick (up to 75 m) and steep in some places, thin (∼10 m) elsewhere, and evidently emplaced slowly over several weeks [Chadwick et al., 1998; Rubin et al., 1998]. This complicated mound topography bears little resemblance to the ideal layer-by-layer, emplacement-then-cooling construction required by the model of Palmer and Ernst [1998, 2000]. Clague et al.  note that the absence of pyroclastic debris at the Gorda Ridge site (about the same depth, ∼3200 m, as the NESCA site) is another indication of a slow extrusion rate eruption. (The absence of pyroclastic debris here also weakens the proposal of Clague et al.  that gas bubbles are the primary source of magmatic gases (e.g., 3He) found in all event plumes.)
Conductive heat flow observations provide a longer view of pillow mound cooling rates. Measurements at the summit of the thickest portion of the Gorda Ridge mound eight months after the eruption found heat flow (up to 6950 mW/m2) at least 100 times higher than on “old” crust just adjacent to other lava flows [Johnson and Hutnak, 1997]. Combining the Gorda Ridge results with similar measurements at the CoAxial eruption mound, Johnson and Hutnak  suggest that the cooling half-life of those flows was ∼2 yr, far too long to power event plumes.
The Gorda Ridge flow may be an extreme example of a slowly cooling lava eruption, but this site nevertheless produced two event plumes. The first (EP96A) was discovered 11 days after regional seismicity began and only 1 day after the onset of a 10-day long seismic event centered at the location of the lava mound and EP96A [Fox and Dziak, 1998]. Thus, even given a convincing relationship between erupted lava heat and event plume heat, any theory that postulates lava cooling as the direct source of event plume buoyancy flux must account for rapid heat transfer from flows ranging from thin, rapidly erupted sheets to thick, slowly extruded pillow mounds.
The discovery of an eruption event on the NELSC in 2008 provided the first new clues in the puzzle of event plume formation in more than a decade. We now know that event plumes with a chemically unique and uniform composition occur in volumes spanning three orders of magnitude, that event plume source fluids interact enough with molten lava to generate very high concentrations of H2, and that the anomalous heat content in event plumes is roughly equivalent (∼70–95% on average) to the heat content of the erupted lavas in examples spanning two orders of magnitude. These results seem to point to the generation of event plumes by a rapid transfer of heat and chemicals from lava to seawater, but there are serious physical constraints to this hypothesis. Theoretical calculations and field observations of cooling lava flows, both thin sheet flows and thick pillow mounds, all conclude that erupted lava cools far more slowly than required to create the intense buoyancy flux needed to lift voluminous plumes up to a kilometer above the seafloor [e.g., Lavelle, 1995]. If lava cooling does generate event plumes, then we must be quite ignorant about how lava actually cools on the deep seafloor.
Heat source is one key to understanding event plumes; the other is fluid chemistry (see Lupton et al. [1999a, 2000] and Palmer and Ernst [1998, 2000] for a thorough discussion). Although this paper does not focus on the origin of event plume fluid chemistry, the 2008 event did provide some new chemical constraints. First, unlike every other sampled event, the near-bottom chronic plumes associated with the erupted lava had 3He/heat ratios much lower than nearby plumes not affected by the eruption. Those ratios were also much lower than historical post-eruption plumes. It is possible that the low NELSC 3He/heat ratios represented an evolving transition from event plume discharge to typical post-eruption discharge, but if so, that transition must have been much slower than observed following other events. Second, the extraordinarily high H2 concentrations in the event plumes apparently demand significant lava-fluid interaction regardless of the source of the fluids. And third, the combination of event plume Δθ anomalies comparable to prior event plumes and a relatively high dθ/dS gradient in the Lau Basin virtually eliminates the possibility that the source of event plumes are low-salinity fluids such as sampled immediately post-eruption at other locations [Von Damm et al., 1995; Lilley et al., 2003].
Relative measure of light backscattering above ambient (dimensionless).
Relative measure of oxidation-reduction potential (mV).
Potential temperature of seawater or vent fluid (°C).
Hydrothermal temperature anomaly (°C).
Potential density ((kg/m3)-1000).
Salinity of seawater or vent fluid (psu).
Volume flux of plume or vent discharge (cm3/s4).
Coefficient of thermal expansion (1/°C).
Coefficient of haline contraction (1/psu).
This research was sponsored by the NOAA VENTS Program, the NOAA Office of Ocean Exploration and Research, and NSF Marine Geology and Geophysics. D. A. Clague, G. J. Massoth, and J. W. Lavelle provided valuable discussion and comments for improving the manuscript. Reviewers T. Gregg, J.-L. Charlou, and an Associate Editor also offered detailed and constructive suggestions. This paper is PMEL contribution 3682.