Prospecting for Hydrothermal Vents Using Moored Current and Temperature Data: Axial Volcano on the Juan de Fuca Ridge, Northeast Pacific (2024)

1. Introduction

Axial Volcano, sitting astride the Juan de Fuca Ridge and above the Cobb Hotspot (Karsten and Delaney 1989; Desonie and Duncan 1990) at 45°58′N, 130°W, erupted lava onto the seafloor in January 1998 (Dziak and Fox 1999; Embley et al. 1999). Prior to the eruption, the caldera of the volcano had seen extensive surveys that located a number of focused- and diffuse-source hydrothermal venting sites (e.g., Baker et al. 1990; Embley et al. 1990). Following the eruption, water-column and seafloor surveys using Alvin and the Remotely Operated vehicle ROPOS (Embley et al. 1995; Embley and Baker 1999) found numerous, apparently new discharge sites and an abundance of near-bottom plumes (Embley et al. 1999). Comparison with pre-eruption water column and seafloor survey data suggested that the hydrothermal plumbing system had been substantially altered in the January 1998 event, leading to much increased heat release (Baker et al. 1999). In August 1998 we put a short-term mooring in place to sample near-bottom currents (u, υ) and temperature (T), with the intention of observing differences in water column signals of low- and high-rising plumes. Here we report those measurements and present an inverse calculation that is directed at locating, though imprecisely, yet undiscovered hydrothermal sources responsible for observed signals.

Hydrothermal source/plume terminology first needs clarification. Ocean ridge hydrothermal sources and the plumes they generate have generally been categorized in terms of diffuse- or focused (discrete)-source geometry, or the high and low temperatures of the vented fluids. Representative of focused (discrete) sources are black, white, or gray smokers—mineral chimney vents and their plumes that form when hot, chemical-laden hydrothermal fluid rises into cold, oxidizing ocean water. High-temperature (e.g., >200°C) and focused (discrete)-source labels are often used interchangeably or in tandem, presumably because of the preponderance of high temperatures measured in chimneyed sources. The terms diffuse-source or diffuse venting typically have come to mean source geometry consisting of one or more fissures, faults, or cracks that occur together with or in isolation from focused (discrete) hydrothermal sources. Diffuse and low-temperature labels are also often used synonymously, a connection presumably made because the temperature of hydrothermal fluids issuing from fissured geometry, for example, is commonly a few tens of degrees Celsius or less. Yet no condition prevents high temperature discharges from a fissure, for example, or low-temperature discharges from a small-diameter chimneylike structure.

Consequently, to say that plumes that rise several hundreds of meters above the seafloor stem from discrete, high temperature sources alone or that plumes that are trapped near the seafloor spring from diffuse, low temperature sources alone is too quick a judgment. Rise heights of plumes depend on source geometry, background stratification, source buoyancy flux, and cross-flow strength. Source buoyancy flux in turn depends on fluid flux as well as the fluid’s density anomaly, the latter calculated with respect to ambient water when salt and temperature are both considered. Sources with high-buoyancy flux, large volume fluxes of low-temperature water for example, can ascend relatively high into the water column. The plume from a high-temperature, heavily salt-laden fluid from a focused source, on the other hand, will likely be trapped near the seafloor, a distribution that might then lead to the plume being mistakenly labeled a diffuse-source plume. We think the terms high- and low-buoyancy flux to categorize sources and high- or low-rising plumes to indicate rise height are better descriptors than the descriptors diffuse/focused and/or high/low temperature. Consequently we will use these less conventional terms in what follows. High and low rising are, of course, relative; here, high rising will mean 100–200 m above the seafloor, while low rising will mean from the seafloor to 50 m above. A fair assumption is that high- and low-buoyancy flux sources lead, respectively, to high- and low-rising plumes.

2. Measurements

A single mooring consisting of six Miniature Autonomous Plume Recorders (MAPR: Baker and Milburn 1997) and two current meters (Fig. 1) was deployed at 45°55.9350′N and 129°59.0225′W on the eastern side of Axial caldera from 0200 UTC 8 August to 1700 UTC 12 August. Longer-term moorings were deployed at the end of the same cruise, but those data were not available for this analysis. Water depth at the site is 1515 m. MAPRs recorded T and pressure every 5 min at heights 10, 15, 30, 60, 100, and 150 meters above bottom (mab). Lower MAPRs were spaced to observe rise height and thickness of low-buoyancy source flux plumes. Aanderra RCM-7 Current Meters sampled current speed and direction at 31 and 151 mab. The lower RCM-7 recorded at 5-min intervals, while the upper recorded hourly. Ours is only the second set of current measurements on Axial Volcano (Cannon and Pashinski 1990) and the only set on the east side of the caldera yet to be reported. A mooring placed the previous year near our site and planned as a yearlong deployment has not been found. Since its deployment location subsequently became a site of new lava flow and since several acoustic and remotely operated vehicle searches have since failed to locate it, we presume the mooring was released during the January 1998 eruption and will never be recovered.

Observations from our mooring (Fig. 2) show large, rapid temporal variations in u, υ, and T, hereafter T_OBS. Currents were remarkably strong for a depth of ∼1500 m; mean speeds over 5 days of 6.9 and 7.4 cm s⁻¹ and maxima of 18.4 and 18.6 cm s⁻¹ were recorded at upper and lower sensors. Records on both meters show a strong semidiurnal signal and mean flow to the west-southwest (Fig. 2). Mihaly et al. (1998) have shown that near-bottom currents at nearby Endeavour Ridge have prominent spectral peaks at K₁, f, M₂, f–M₂, and M₄, as well as other tidal frequencies, where f is the inertial frequency and the f–M₂ frequency results from the f, M₂ nonlinear interaction. To examine component contributions, a sum of sinusoids at the five named frequencies and a mean were simultaneously least squares fit (Godin 1991) to observed u, υ (Fig. 2) in vector form. Time series are too short to resolve individual diurnal and semidiurnal tidal current constituents (Godin 1991);instead, composite diurnal (“K₁”) and semidiurnal (“M₂”) motions are represented by single frequencies;corresponding tidal amplitudes (Table 1) must be understood in that light. Weighting factors w_i = 1 + |(υ_i − 〈υ〉)/σ| were used to enhance the influence in fits of less frequently occurring, larger values. Here υ_i is a datum and σ is the variance of data about the record mean 〈υ〉. Correlation coefficients r between u, υ fits and data ranged from 0.68 to 0.94, with larger r values found for the stronger north–south flows. Fitted tidal–inertial amplitudes, in rotary vector form (Gonella 1972;Emery and Thomson 1998), are given in Table 1.

Mean velocities were 5.0 cm s⁻¹ at 241° and 3.3 cm s⁻¹ at 240° on lower and upper meters, respectively. The WSW direction of mean flow is similar to that of the large plume observed just after the January eruption (Baker et al. 1999). Inertial currents had amplitudes of 2–3 cm s⁻¹ during the time of observations (Table 1). Semidiurnal “M₂” amplitudes are as large as 8 cm s⁻¹ (Table 1). These are much larger than those typically measured above and slightly away from the ridge (Pashinski 1998). The aggregation of semidiurnal motions into a single “M₂” amplitude is contributory, but amplification effects of ridge/seamount topography must also be suspected (e.g., Allen and Thomson 1993; Beckman and Haidvogel 1997). Mihaly et al. (1998) show that tidal current amplitudes in this region and ridge setting have a nonstationary component, but that aspect of motion cannot be investigated with records as short as ours.

The six T_OBS time series have different temporal character over the 150 m of water column surveyed (Fig. 2). Note that T_OBS values have not been transformed to potential temperatures θ since T variability, not absolute magnitude, is of interest. By calculation and eye (Fig. 2), the three T_OBS records near the seafloor are well correlated. Above 30 m differences record-to-record increase with separation distance, trending toward negative correlation as separation distance reaches maximum. Each T_OBS record shows a substantial range of values; differences of extremes of T_OBS within each record extend from 0.13°C (at 10 mab) to 0.18°C (at 150 mab). Observed record means T_OBS increase monotonically with height from 2.44° to 2.62°C with dT_OBS/dz of 7.8 × 10⁻⁴ °C m⁻¹ at 12.5 m and 1.6 × 10⁻³ °C m⁻¹ at 125 m.

Scatterplots of |ΔT_OBS| cos(θ), where ΔT_OBS = (T_OBS − T_OBS) and θ is current direction, suggested that, for the lower-rising (30 mab) plume, the largest and smallest |ΔT_OBS| occurred when currents flowed from the NNE and SSE, respectively. For the higher-rising (150 mab) plume the situation is nearly reversed. Largest and smallest |ΔT_OBS| occurred when currents were from the SSE and NNE, respectively. Kinosh*ta et al. (1998) had observed an analogous occurrence of directionally variable ΔT on separate moorings in oscillatory flow near the Trans-Atlantic Geotraverse (TAG) hydrothermal mound. Instantaneous arrival direction may be a misleading indication of the direction from which heat originates, however, because ΔT measured at any one location represents an integrated history of all previous heat input into the water column, and that history may be quite complicated in regions where currents have high directional variability (Wetzler et al. 1999). Deconvolving that history is what the following is about.

3. Searching for sources—The inverse calculation

The range and time rate of change of ΔT_OBS is much too large to be ascribed to advection of lateral or even vertical gradients of background T. For example, CTD profiles to the side of Axial indicate that at 1500 m vertical gradients are on the order of 3 × 10⁻⁴ °C m⁻¹ or less. If T variability were caused by vertical motions without a hydrothermal heat component, vertical excursions of 400–600 m would be required to produce the observed range of ΔT_OBS. Instead, ΔT must represent recent hydrothermal discharge as distributed by water column advection and turbulent diffusion processes. That being the case, we have attempted the reconstruction of T time series through an inverse calculation (e.g., Wunsch 1996).

Plumes at each level studied (z_i = 30 and 150 mab) are taken to be vertically uniform and retain their initial thickness (Δz), accomplished analytically by setting K_z, the diffusion coefficient in the vertical direction, equal to zero. The reason for deprecating vertical diffusion is that over the time period corresponding to the average transit time of the plume from the most distant inverse source to the mooring, that is, 3.6 km/(5 cm s⁻¹) ≅ 20 h, the vertical diffusion length of a plume edge would be on the order of 4K_zt ≅ 5 m or less, if K_z has the canonical value 1 cm² s⁻¹. Even for most distant sources, this spreading distance is small compared to assumed plume thickness. The value of the horizontal diffusion coefficient K_H is more important. The central value for K_H about which sensitivity experiments were conducted was 0.1 m² s⁻¹. That value is based on dye-patch spreading-rate observations and analysis of Okubo (1971) for ocean diffusion over length scales of several hundreds of meters.

An advection–diffusion model of plumes is clearly a reductionist view. Individual hydrothermal plumes begin with turbulent convective ascent into a moving water column with much of the turbulence self-generated. Rise heights are known to depend on cross-flow strength, as well (e.g., Hanna et al. 1982). Laboratory and atmospheric observations show that under certain ratios of upward velocity and cross-flow, plumes can bifurcate horizontally, vertically, or in an orientation between (e.g., Ernst et al. 1994). Effects of turbulence on plume spreading is crudely parameterized by Laplacian diffusion. An advection–diffusion model lumps and smooths the effects of all these processes, with the result that one cannot expect in T_MOD the high-frequency content of the signal actually observed. More sophisticated models of plumes in cross flows that capture some of the turbulent and the bifurcating nature of plumes exist (e.g., Lavelle 1997), but those models are not configured for time-variable flow and are too numerically demanding at this time for an inverse calculation. Like most inverse computations, we start with a simpler model that can describe essential, but not all, features of the dataset.

Currents used in Eq. (2) were reconstructed from component amplitudes (Table 1). These series, ũ and υ̃, rather than the full u,υ series were used as a way of interpolating hourly sampled data at the upper current meter to 5 min and to extrapolate all current data backward in time 500 steps in order to create model plumes that would have fully developed over the region of interest by the time the mooring had begun sampling T.

To smooth T_OBS in a way consistent with the treatment of u and υ, T_OBS at 30 and 150 m were least squares fit with constants and with sinusoids at these seven frequencies: K₁, f, M₂, f–M₂, M₄, K₁–M₂, and f–K₁. The number of sinusoids was limited to seven after experimentation indicated limited improvement in the fits when additional higher and lower harmonic frequencies were included. The temperature series representing those best fits, T̃ (chain-dot line), are superimposed on T_OBS (dotted line) in Fig. 3. Here T_OBS and T̃ have correlation r values of 0.72 and 0.65 for upper and lower datasets, respectively. The fit to measured data at height 30 m is poorest over the interval 60–75 h (Fig. 3), when flow near the seafloor was also poorly represented with just a few periodic components. Reproducing the observed high-frequency variability of T_OBS, expected of plumes that originate as turbulent, convecting fluids, cannot be expected of tidal–inertial time series fits nor of time series arising from simple advection–diffusion models.

The T_MOD at the mooring from a uniformly spaced array (31 × 31) of potential sources (Fig. 4) was then least squares fit to T̃. The algorithm used (Hanson 1982) allows quick implementation of linear least squares with weights and constraints. A constrained least squares method is required because negative, and thus unphysical, source coefficients must be excluded. A source at the measurement location itself was disallowed. Fitting weights took the form previously described.

A series of inverse experiments was conducted to examine the sensitivity of results to differing values of diffusion coefficients, domain sizes, and masking thresholds, α (the ratio of the number of occurrences of the plume reaching the measurement location from any potential source to the total number of potential occurrences). Horizontal diffusion coefficients were varied from 1.0 to 0.01 m² s⁻¹. Source domains varied in five increments from 1200 m × 1200 m to 5400 m × 5400 m, assigning source dimensions (Δx and Δy) in each case to be one-half the grid cell size into which full source domains were divided. Masking threshold, α, as defined above, took values of 10% and 20%. We could find no appropriate statistic to summarize the general geographic similarity or inter-experiment scatter in the resultant source locations. In the following, we will present source location maps for the low- and high- buoyancy flux plumes using a reference set and two other sets of parameters at each level. The reference and other case settings are specified in the captions to Figs. 4 and 5.

4. Results

The series T_MOD from the inverse calculations for the reference set of parameters for upper (150 m) and lower (30 m) sensors are superimposed on T_OBS and T̃ in Fig. 3. Correlations of T_MOD with T̃ give r values of 0.8 and 0.72 for upper and lower records. Periodograms of the misfit, T_MOD − T̃, are spectrally white at low frequencies. Slopes of linear fits to periodogram amplitudes are, via Studen’s t-tests, zero at the 95% confidence level for spectral periods greater than 1.9 and 4.5 h for upper and lower records, respectively. The inverse model thus proves adequate in describing T̃ variations at those and longer periods. Of the T variance accounted for, ∼10 source locations account for 70%, and <30 source locations account for 90% at both 30 and 150 m levels for the reference set of parameters.

Inverse results for the lower and upper plumes are given in Figs. 4 and 5, respectively. For comparison, the location of vent sites that have already been observed by camera or submersible are marked in Figs. 4a and 5a. Intersecting lines mark the mooring site. Local bathymetry is superimposed. The heavy dashed curve partitions the region into allowed and disallowed source locations. Prevailing currents set plumes to the WSW, so the region westward and southward of the dashed line contain potential source locations with little chance to influence T_OBS. If locations that presented nonzero ΔT to sensors too infrequently (<α) were not masked off in the inverse calculation, potentially spurious sources might be indicated there based on little data.

Known sources of low-buoyancy flux within the calculation region (Fig. 4a) lie north and south of the mooring; not all provided a temperature signal to the sensors, however. Flow trajectories were such that low-buoyancy flux sources more than ∼0.5 km north of the mooring (Fig 4a) would have only slight influence on recorded T. Three known vents may have contributed to the higher-rising plume. Marker 33 Vent, discharging at a maximum of 78°C but with diffuse source geometry (Embley et al. 1999), lies just north of the mooring (Fig. 5a, diamond). Neither its buoyancy flux nor rise height is known, so the extent of its contribution to upper or lower plumes or both is arguable. Cloud vent, some 50 m east of Marker 33 (Fig. 5a, box), is more likely to contribute to the higher-rising plume. Cloud has focused source geometry with an orifice dimension of 0.5 m, a discharge velocity of approximately 30 cm s⁻¹, and a source temperature of approximately 22°C. Using a nominal value for deep-sea buoyancy frequency (N) of 10⁻³ s⁻¹ and the well-used expression for rise height, h = 3.75(B/N³)^1/4, for a point source plume with source buoyancy flux B rising into a quiescent environment, one can estimate an approximate rise height for Cloud’s plume of 120–130 m. Thus, even though Cloud is considered a low-temperature vent, its total buoyancy flux likely results in its plume rising well above the seafloor. Castle vent, lying SSE of the mooring at an approximate distance of 300 m, discharges water at 274°C (Embley et al. 1999). Its rate of discharge and temperature appears sufficiently large to force a plume 150 m above the seafloor, where it could then be recorded by our uppermost MAPR. This additional point about low- and high-buoyancy flux vent locations needs to be made. A comparison of Figs. 4a and 5a shows that high- and low-buoyancy flux vents may be collocated, but many low flux buoyancy vents have no corresponding high buoyancy flux vents nearby.

Locations of low buoyancy flux sources, indicated by the inverse calculation when parameters have reference values, are mapped over bathymetry in Fig. 4b. Size of the map symbols indicates rankings of sources by the amount of variance of T̃ that each source accounted for. Thus, larger symbols represent sites contributing most to fitted T̃ series (Fig. 3). In this and subsequent maps, sources shown account for more than 70% of the fitted variance. Sensitivity of results to parameter choices is shown, in part, by comparison of Figs. 4b with 4c and 4d. Figure 4c represents results when K_H is changed from 0.1 to 0.02 m² s⁻¹ and α is changed from 0.2 to 0.1; Fig. 5d has the same K_H and α as Fig. 4b, but the calculation area is enlarged 225%. The same variation of parameters holds for Figs. 5b–d. Note that the areas in the corresponding panels of Figs. 4 and 5 differ by 225%.

Domain sizes were chosen to make domain area (examined in graduated size steps) as small as possible, yet keep all but a few less important sources away from domain boundaries. Currents are more likely to be nearly isotropic and hom*ogenous (model requirement) the smaller the region, but source points on boundaries suggest that potentially important sources lie beyond; thus the region is not sufficiently large. The smallest domain sizes with which we experimented were clearly too small. Experiments also showed that the number of source points accounting for 90% of the variance varied with the size of the solution domain, with larger domains indicating more source points but also yielding better r values.

Inverse sources for the low-lying plume (Figs. 4b–d) lie to the NNE and immediately south of the mooring (Fig. 4, at the location of intersecting lines), though there is a suggestion of a source well to the east. The general pattern of inverse-derived source locations is generally consistent over the range of inverse experiments (Figs. 4b–d). Given the coarse assumptions of the model, data from a single mooring, and the length and density of observations it would be too much to expect more. Inverse-derived source locations for the upper plume (Figs. 5b–d) have a much different pattern than those of the low-lying plume. Those sources are primarily clustered just south of the mooring and along the NW–SE trend line some 2400 m to the NE of the mooring. Cluster locations are consistent over the range of those inverse experiments (Figs. 5b–d).

Comparisons of inverse results to the known location of vents suggest that the stated mooring location (ship’s location at the time of deployment) and the resting site of the mooring may be offset by 100–200 m. Figures 4b–d and 5b–d all suggest a source of venting immediately south of the mooring, but the nearest known vents that could have created the signals are the Marker 33 and Cloud vents lying to the NNE by 100–200 m. If the mooring location were relocated a few hundred meters to place Marker 33 and Cloud vents just to the south of the mooring, known (Figs. 4a and 5a) and inferred (Figs. 4b–d and 5b–d) source distributions would in be much better agreement. It is not hard to imagine that the deployment and the resting position of a mooring in 1500 m of water could be offset by just a few hundred meters. A postrecovery attempt to better locate the mooring site by observing the location of the mooring’s anchor by camera was not successful.

If the suggested difference in deployment and resting location of the mooring were in fact true, the low temperature vents 0.4–0.7 km to the south of the mooring (Fig. 4a) would have only marginal influence on the T signal at the sensors. This could explain their apparent absence in inverse results (e.g., Fig. 4b). The contribution of the high temperature Castle vent (Fig. 5a) to the upper-lying plume would also be marginal, as the inverse calculation results (Figs. 5b–d) are perhaps reporting.

Even if the model and the density of mooring observations were much improved, it seems clear that inverse procedures will never substitute for bottom camera and submersible surveys in pinpointing actual vent locations. On the other hand, inverse results can point to general locations where camera/submersible surveys should, but otherwise would not likely, be run. We believe that the second cluster of source points to the NNE (Figs. 5b–d) represents one such situation. The location, orientation, and contribution to variance reduction of that NW–SE-trending cluster of vents is consistent over all three calculations (Figs. 5b–d). The mooring data thus points to unexpected venting well away from new lava flows on the shoulder of the volcano in an area that is yet to be bottom surveyed.

A tow-yo CTD–nephelometer transect (T11) made on an east to west transect just to the south of the mooring position (Fig. 6a) shows the results of loading the water column from vent sources. Potential temperature anomaly Δθ, calculated in the manner of Baker (1994), and nephelometer voltages Δn, indicative of particle concentrations in the water column, were recorded along the transect during the same early August 1998 period. Figures 6b and 6d show Δθ and Δn distributions indicative of hydrothermal plumes (e.g., Baker 1994); bottom-trapped, yet high-rising plumes west of ∼129°59′W are surely the result of both high- and low-buoyancy flux discharges, source locations which have been identified by direct observations (Embley et al. 1999) and, less precisely, by our inverse calculation (Figs. 4b–d, 5b–d).

Collaborating evidence for high-buoyancy flux venting 2.4 km to the NNE of the mooring serendipitously comes from a second tow-yo CTD–nephelometer south to north transect made during the same cruise. Transect T20 passed directly through the cluster of predicted sources east of the summit (Fig. 6a). Data from T20 (Figs. 6c and 6e) show Δθ and Δn distributions with maxima of substantial intensity situated well above the seafloor. Both Δθ and Δn indicate that the plume is sufficiently thick to allow its detection on even the uppermost MAPR. T20 shows a strong plume to the north of 45°56.5′N, but very little plume presence to the south. Plume maxima actually occur near latitude 45°57′N where T20 crosses the NW–SE trend line of the predicted high-buoyancy flux sources.

If it were true that the T20 plume anomalies arise from sources on the shoulder of Axial, at or near locations suggested by the inverse calculation, one might expect, given prevailing WSW currents, higher plume intensities along T20 south of 45°57′N. Currents may not have been in the prevailing direction preceding and during the T20 survey, however. Between 0030 and 0130 UTC on 14 August, when the section between 45°57.55′ and 45°55.90′N of T20 was surveyed (Fig. 6a), a forward-in-time extrapolation of data from the current meters (Fig. 2), based on the tidal–inertial-mean fits previously described, indicates that currents were to the NNW for 3 h prior to and during the survey, turning to the WSW just afterward. Figure 2 shows that the NNW flow is recurrent at the mooring site. The predominance of the plume north of 45°57′N (Figs. 5c and 5e) is thus likely a consequence of the flow direction at the time of the survey. At other times the plume from those sources would have been advected generally toward the mooring location.

Plumes in cross flow typically have increasing thickness and decreasing concentrations downstream from their sources, just as depicted by the plumes in Figs. 6c and 6e. Locations of plume maxima often indicate a nearby source. The T20 data thus provides circ*mstantial evidence in support of the inverse prediction that hydrothermal sources on the volcano shoulder existed during early to mid-August 1998. We find it less compelling to believe that a plume with these features, that is, a 0.035°C potential temperature anomaly core and increasing thickness to the north in an apparent northerly flow, would represent vestiges of plumes transported to this locale from more distant sites.

Several correlative observations need to be mentioned with regard to predicted shoulder sources. Bottom-mounted hydrophones were placed by Sohn et al. (1999) at Axial Volcano within weeks of the 1998 eruption. In the subsequent 5½ months, two distinct sets of microseismic events were recorded. The first set consisted of events located around the rim of the caldera, and these occurred within the first several months. The second set, more interesting in the context of this paper, consisted of events that began in late March and persisted to late June 1998. These events were situated around a line segment that begins at the caldera and radiates onto the shoulder of the volcano in the ESE direction (Sohn et al. 1999). Those data show that the shoulder region of the volcano was seismically active at least up to six weeks before our cruise. Furthermore, NW–SE orientation of that line of events (Sohn et al. 1999) is similar to that of our predicted easterly vent sources, and both orientations are similar to the trend of the Cobb–Eickelberg seamount chain that intersects the Juan de Fuca Ridge at Axial Volcano (e.g., Karsten and Delaney 1989; Desonie and Duncan 1990). Perhaps the seismicity that Sohn et al. (1999) observed was indicative of magmatic activity, the heat from which was released as vented fluids on the volcano’s shoulder.

If the predicted shoulder sources were, in fact, active high temperature hydrothermal discharge sites in August 1998, newer observations at Axial suggest that the venting was transitory. The same eastern flank region of the volcano was reexamined by CTD–nephelometer survey in July 1999. Those data show little evidence of temperature or particulate anomalies of any kind, indicating first that the region is not continuously, if at all, bathed with plumes that originate in the caldera. More tellingly, it suggests that local venting on the NW–SE trend line of Figs. 5b–d ceased between the 1998 and 1999 surveys. If local flank venting were still active, the water column survey on the transect that retraced T20 should have detected a plume. While appreciable hydrothermal venting continues in the area of lava flow and other sites in the caldera, the 1999 water column survey also shows that plumes in those regions do not rise as high as they had in 1998. A regional decline in the supply of crustal heat is thus likely. A cessation of venting to the east of the caldera may have been a consequence of that decline, or perhaps those vents, if they did exist, were choked off by mineral deposits, as has been observed to happen in other hydrothermal vent fields (Hannington et al. 1995). A visual survey of the seafloor at the northern end of T20 (Fig. 5a) for vestigial evidence of venting has not yet been possible.

Inverse calculation results allow one other step to be taken, that is, a comparison of source heat flux from the two source types. Total heat flux, Q, from each source type is Σ Q_i, where Q_i are fitted coefficient values [Eq. (3)]. For the upper-plume and reference parameter values, coefficients accounting for 90% of the fitted variance (n = 29) yielded a total-heat flux value of Q^U₉₀ = 217 MW; when all nonzero coefficients are used, Q^U₁₀₀ = 358 MW (n = 87). Corresponding heat fluxes into the lower plume are Q^L₉₀ = 75 MW (n = 29) and Q^L₁₀₀ = 126 MW (n = 97).

Distributions of individual flux rates were exponential when ordered by the amount of fitted variance that each source accounted for. Heat flux to the upper plume from individual discharge areas (60 m × 60 m) (Fig. 5b) ranged from 23 MW to negligible amounts at 87 locations. The median nonzero flux value was 2.6 MW (720 W m⁻²). Heat flux to the lower plume from 97 source areas (40 m × 40 m) (Fig. 4b) ranged from 4.8 to 0.01 MW, with a median nonzero flux value of 1.1 MW (690 W m⁻²). Heat flux from individual high-buoyancy flux sources tends to be larger that heat flux from individual low-buoyancy flux sources. Rona and Trivett (1992) measured a total of 4.4 MW from seven vent sites at ASHES vent field at Axial Volcano in a 1986 survey. Our generally larger heat flux values may be the consequence of sampling at different locations on the volcano or related to the much larger fluxes induced by the 1998 Axial eruption, if only temporarily.

Heat flux normalized by the area shows that the upper plume received 35.9 MW km⁻², while the lower plume received 30.9 MW km⁻²; these are regionally averaged values, not source-area averaged. The lower plume is thus supplied by sources with just 86% of the heat flux per unit area supplied by sources to the higher plume. The ratio of heat in sources contributing to high- and low-rising plumes thus seems to be ∼1:1. Rona and Trivett (1992), Shultz et al. (1992), and Baker et al. (1993) all support the view that the heat supply to low-buoyancy flux plumes substantially exceeds that to high-buoyancy flux plumes. Rona and Trivett (1992), for example, found that the heat flux in low-buoyancy flux sources exceed that in high-buoyancy flux sources by a factor of 9 at the ASHES vent field. Our results challenge the assumed universality of previous results on the partitioning of heat between the source types. In some cases heat from sources with low-rising plumes may exceed by many times the heat from sources that lead to plumes hundreds of meters above the seafloor, but that need not always be the case, or may not be the case over areas larger than those previously sampled.

5. Conclusions

Relatively large mean and semidiurnal near-bottom currents can occur near the seafloor at the southeastern sector of Axial Volcano. Along with an abundance of hydrothermal sources, currents drive rapid T variations of nearly 0.2°C in the lower 150 m of the water column. Inverse methods applied to u, υ, and T can suggest, though with a single mooring only imprecisely, where hydrothermal venting sites are likely to be found. Results of inverse calculations for source locations and CTD–nephelometer tows together suggest that high-buoyancy flux venting occurred on the eastern shoulder of Axial during the summer of 1998. Venting on the eastern shoulder apparently stopped by July 1999 when a CTD–nephelometer survey over the same area failed to detect hydrothermal plumes. Nearer the mooring, the predicted high- and low-buoyancy flux source sites are largely similarly distributed to known vent sources. Our results challenge a common view that, by and large, heat flux from low-buoyancy flux sources greatly exceeds heat flux from high-buoyancy flux sources at ridge crests. Limitations of the present dataset and those imposed by assumptions underlying the inverse approach are many. Yet the inverse technique offers the promise, particularly if longer-term measurements from multiple arrays become available, of helping direct time-constrained seafloor and water column surveys searching for new hydrothermal sources. Moreover, we expect that in not too many years a data telemetering capability will be in place on Axial Volcano that will allow the near-real-time monitoring of many variables, including temperature and currents. Those data and an inverse calculation of the kind begun here might then be used to monitor post-eruptive thermal discharge sites in near-real time as well as, perhaps, provide a view of source locations of heat released during an eruptive event itself.