Previous   |   Next

doi:10.2204/iodp.proc.301.102.2005

Methods

Swath mapping

Prior to Expedition 301 site surveys, vessel-based swath coverage of this area was limited mainly to isolated tracks across the ridge flank (by ships heading to/from the active spreading center to the west) and to small areas containing particular features such as basement outcrops. The Ridge Multibeam Synthesis Project (RMBS) compiled data from numerous surveys, but Generic Mapping Tools grid files included only the near-ridge area
(ocean-ridge.ldeo.columbia.edu/general/html/home.html). Additional lines of swath data in the Expedition 301 field area were acquired during the 1996 Sonne survey (Rosenberger et al., 2000), but these data had never been merged with the RMBS synthesis.

Bathymetric data collected by the Sonne in 1996 and 2000 were acquired at typical survey speeds of 5.5–7.0 kt while shooting seismic profiles and 10–11 kt during transits between work areas. The hull-mounted Hydrosweep DS2 system produces usable data up to two times water depth with a horizontal resolution on the order of 90 m (at 2700 m depth and depending on ship's speed and other factors). To suppress refraction effects on the outer beams without knowing the local sound velocity profile, the Hydrosweep system uses a calibration mode to compare depth values of the central and outer beams in order to calculate a mean sound velocity. Using this configuration, residual depth errors are minimized to values <0.5% of the water depth, on the order of 5–15 m for typical water depths in this area (Grant and Schreiber, 1990).

Because the RetroFlux expedition mainly worked within a few small areas, with the ship held stationary during coring and heat flow operations, the Hydrosweep system on the Thomas G. Thompson was activated only during transits and for a few short, dedicated swath-map surveys run at 7–10 kt. This resulted in the collection of relatively little additional data. Operating parameters and data resolution for RetroFlux bathymetric data are very similar to those described above for the Sonne. Some additional data were collected in the Expedition 301 area during the EW0207 survey, but because these data generally covered areas that had already been surveyed, no attempt was made to include EW0207 multibeam data in the composite grid.

After individual file processing, described below, the data collected during recent Sonne cruises (1996, 2000) and the RetroFlux expedition were combined with raw multibeam files that had previously been used to assemble the RMBS synthesis and additional ridge-flank data archived by researchers with the National Oceanographic and Atmospheric Administration Vents program (A. Bobbitt and C. Fox, pers. comm., 2000–2001). Earlier generations of Seabeam and Hydrosweep data were usually navigated by Loran C and contained systematic positioning errors, but the quality of bathymetric data was often very high. The earlier data included multiple crossings of several interesting features and helped to define targets for mapping and other work during the RetroFlux and ImageFlux surveys.

Processing of bathymetric data began with the public-domain software package MBsystem (Caress and Chayes, 1996). Navigation was corrected, and hydrosweep data were ping-edited to flag the outer 5–10 beams for removal and eliminate a variety of artifacts, including abnormal depth values and gradients. Pronounced "rails" (depth-shifts parallel to the ship track) were present in much of the data, and in other places there was an abrupt step, curl-up, or curl-down, typically affecting the outer beams. The rail effect was reduced using a tool developed by H. v. Lom (University of Bremen), and other artifacts were removed by hand-editing individual beams. Some swaths also showed subtle roll bias, which was corrected before file merging. This careful data processing was necessary because most bathymetric structures in the study area (apart from seamounts) and data artifacts have similar bathymetric scales. For typical survey depths of 2200–3700 m, the resulting usable swath width was generally 4–6 km.

Edited data files from the numerous surveys were combined to form a single bathymetric grid. Sound velocity profiles created with the Levitus database (Caress and Chayes, 1996; Levitus, 1982) were assigned by area to assure consistency in absolute depths determined during different surveys. Data were gridded at various resolutions to evaluate resolution and data quality. We experimented with additional processing to remove artifacts, using band-pass and other filters, but because the bathymetric differences across the field area are generally so small, even modest additional filtering often resulted in a confounding loss of resolution of key features. For this reason, we elected to use a composite, 100 m grid file that includes minimal filtering beyond that associated with obvious beam artifacts. The composite bathymetric data set provides almost complete coverage of a 8000 km² region on the eastern flank of the JFR (Fig. F2). Redundancy is particularly high around First Ridge and Second Ridge, where there is also the densest coverage of seismic data, as described below.

Seismic data

Sonne

Nearly 500 seismic lines, comprising >1 TB of data, were collected as part of the ImageFlux survey. The MCS system operated by the University of Bremen was optimized for high resolution and imaging of sedimentary structures and layers at a fine scale. Three different seismic sources were run at the same time, with results recorded separately along each line: a small-chamber water gun (0.16 L; 200–1600 Hz; Sodera), a generator-injector (GI) gun (2 × 1.7 L; 30–200 Hz; Sodera), and a GI gun with reduced chamber volume (2 × 0.4 L; 100–500 Hz). Guns of larger chamber volume provide greater depth penetration, revealing the larger-scale structural framework within sediments and uppermost basement, whereas guns with smaller chamber volumes provide higher resolution, revealing finer details of the upper 200–300 m of sediment.

The three guns were triggered at time intervals of 10–13 s (depending on water depth), resulting in a shot distance of ~25–34 m (depending on ship speed) for the alternating mode operation of each gun type. Using an optimized trigger scheme, shot spacing is the same for each gun type and >50% greater than it would be for only a single gun. Recording time was 3 s for GI gun records (0.25 ms sample rate) and 1.5 s for water gun records (0.125 ms sample rate), which is sufficiently long to include the sediment/basement interface reflection. Using a 600 m long Syntron multichannel streamer (plus two 50 m stretch sections), with 48 recording groups separated by 12.5 m, a common midpoint (CMP) fold of 8–10 was achieved for CMP spacing of 10 m. This provides an excellent compromise between trace density, image quality, and noise reduction. Six remotely controlled birds kept the streamer at 3 m depth (±0.5 m), and magnetic compass readings allowed for the determination of the position of each streamer group relative to the ship's course, taking variations in streamer geometry into account. Geographic positions of each shot location were provided by closely sampled (1 s) differential Global Positioning System (GPS) recordings, and custom software calculated receiver positions and statics and carried out binning. Standard data processing included editing, bandpass filtering to eliminate low-frequency noise, correction for geometrical spreading, stacking, and time migration. These tasks were completed with the public domain package Seismic Un*x (Stockwell, 1997).

Parasound 4 kHz echo-sounding data were collected on the Sonne during 1996 and 2000 surveys at the same time that MCS data were acquired. The Parasound system is hull-mounted and compensates for heave, pitch, and roll. Footprint size is only 7% of water depth, diffraction hyperbolas are suppressed, and both lateral and vertical resolutions are significantly higher in comparison to conventional sediment profiling systems. Parasound data provide an extremely high resolution image of the shallowest tens of meters of sediment.

Two aspects of the ImageFlux seismic survey require additional discussion. In the First Ridge region (Figs. F1, F2), an area of ~6.4 km × 2.3 km was covered by parallel profiles separated by 25 m, including Leg 168 Sites 1030 and 1031 and Expedition 301 alternate Site FR-1. The goal of this exercise was to use closely spaced lines to extract detailed information about the three-dimensional geometry of structures in sediments and basement rocks. The closely spaced lines are oriented west-northwest, in the direction of maximum variation of basement structure. Traces of all two-dimensional (2-D) lines were binned and stacked based on a predefined grid consisting of cells that are 10 m wide (in the inline direction) and 25 m long (in the crossline direction). Thus, common-cell sorting was applied rather than CMP sorting, and cell coverage is between 5- and 10-fold. Trace interpolation could be avoided because of the small inline spacing. However, because of the limited crossline dimension of the grid, only a 2-D time migration was applied in the inline direction. Quality control for processing, and a link to the larger seismic grid on the eastern JFR flank, were provided by a number of long crossing lines.

In the Second Ridge area, another grid (~5.4 km × 3.6 km) was covered with more widely spaced (100 m) profiles. This area includes ODP Sites 1026 and 1027 and IODP Site U1301. These profiles were also oriented west-northwest, in the direction of primary structural dip, and were integrated in the larger seismic grid covering this area. Although the line spacing is not as close as that in the First Ridge area and, therefore, the data were not processed on a grid, the data provide useful insights concerning basement structure in the vicinity of ODP and IODP drill sites. Groundtruth for GI gun data collected during Sonne cruise SO149 was provided by modeling reflection patterns from Leg 168 core-scale density logs, as described by Zühlsdorff and Spiess (2001).

Maurice Ewing

MCS data were acquired during the EW0207 cruise across the Deep Ridge area, where two secondary drilling targets are located. These data were collected using a 6 km long, 480 channel Syntron digital streamer with receiver groups spaced at 12.5 m. Streamer depth and feathering were monitored with a mix of 13 depth-controlling and 11 compass-enhanced DigiCourse birds, plus a GPS receiver on the tail buoy. A 10 gun, 49.2 L air gun array was used as the source of acoustic energy with shot-by-distance at a 37.5 m spacing. Listening time was 10.24 s with a sampling rate of 2 ms. The recorded signal has a bandwidth ranging from ~2 to >100 Hz. The nominal CMP bin spacing is 6.25 m, and the CMP fold is 81.

The prestack processing strategy adopted for the EW0207 MCS data consisted of

• Standard straight-line CMP bin geometry

• F-K and bandpass (2, 7, 100, and 125 Hz) filtering to remove the low-frequency cable noise

• Amplitude correction for geometrical spreading

• Surface-consistent minimum phase predictive deconvolution to balance the spectrum and remove short period multiples

• Surface-consistent amplitude correction to correct for anomalous shot and receiver group amplitudes not related to wave propagation

• Trace editing

• Velocity analysis using the velocity spectrum method

• Normal moveout and dip moveout corrections to align signals for stacking

• CMP mute to remove overly stretched data

The prepared prestack data, with and without the automatic gain control, were then stacked. The poststack processing included seafloor mute, primary multiple mute to reduce migration noise, bandpass filtering (2, 7, 100, and 125 Hz), and time migration to collapse diffractions and position the recorded reflection events at their true subsurface locations.

Extracting an image of the crustal Layer 2A event from the EW0207 data requires a somewhat different processing scheme because this event is not a true reflection (Harding et al., 1993). The prestack data preparation is identical up to the velocity analysis, which is done on bandpass-filtered (2, 7, 40, and 60 Hz) constant-velocity stacks. When the normal moveout velocities that best flatten the retrograde branch of the Layer 2A refraction are chosen, the data traces with source-receiver offsets from 1500 to 4000 m are stacked. The stacked Layer 2A event is time-migrated and coherency-filtered. Surgical mute is used to extract the Layer 2A event, which is merged with the reflection section to form a final, composite seismic image.

Heat flow data

Acquisition

RetroFlux heat flow transects during the RetroFlux expedition were typically colocated along existing or planned seismic reflection profiles so that we could merge thermal data with sediment thickness and basement structure, as discussed later. Because the ImageFlux and RetroFlux expeditions were at sea at the same time, we did not always have seismic data in hand when heat flow transects were run. In some cases, seismic profiling followed collection of heat flow data; in a few cases, heat flow data were collected even though we knew that we would not have colocated seismic data to use for later analyses.

Multipenetration heat flow data were collected using an 11 thermistor violin-bow heat flow probe with in situ thermal conductivity capability (Davis et al., 1997a). Prior attempts to measure heat flow in the southeastern part of the survey area had been unsuccessful because the heat flow lance would not penetrate sandy or lithified sediments. For the RetroFlux program, we used a conventional weight stand and 3.5 m outrigger lance (sensor spacing = 30 cm), but also brought out a heavier weight stand to be used with a 2.0 m outrigger lance (sensor spacing = 15 cm). This modified system operated successfully in areas where previous attempts failed, but in some instances, the probe continued to settle episodically during measurement.

Heat flow transects typically consisted of 15–20 individual penetrations spaced 50–500 m apart, with data collected during a 12–24 h station. Temperatures of the 11 sediment thermistors, bottom seawater, and logger electronics, plus pressure, tilt, and reference resistance, were logged at 10 s intervals and stored in nonvolatile memory. A subset of the data was telemetered back to the ship in real time for monitoring of instrument performance. In situ thermal conductivity was determined during ~60% of the penetrations. For the other stations, we used local thermal conductivity versus depth functions determined from surrounding measurements, as discussed below. Additional thermal conductivity measurements were made with a needle-probe system on gravity and piston cores recovered during the survey.

Processing and uncertainties

Heat flow data were processed using methods based on those described by Villinger and Davis (1987) and Davis et al. (1997a). Modifications to the approach include

• Use of local thermal conductivity versus depth functions for stations that lacked in situ measurements;

• Iterative processing for these stations, with new conductivity values assigned for each estimate of penetration depth, followed by recalculation of equilibrium temperatures;

• Use of the "scatter" parameter (variance normalized by number of thermistors) (Villinger and Davis, 1987) to guide selection of the number of thermistors used in each heat flow determination; and

• Monte Carlo analysis of all penetrations (100–200 realizations in each ensemble), incorporating uncertainties in thermal conductivity, equilibrium temperatures, and the thickness of layers having different thermal conductivities.

Reported heat flow values are means from the Monte Carlo analyses.

Stein and Fisher (2001) describe the general processing scheme in detail, including estimation of uncertainties, which was implemented after the RetroFlux expedition using a graphically driven program that allowed viewing and editing of individual thermistor records. Interactive thermistor-by-thermistor processing was essential because many measurements included data that did not follow the standard pattern of heating and cooling after probe penetration. These were most common when making measurements in turbidites using the short, heavy probe, and resulted from a combination of probe motion and the extreme frictional heating and high thermal conductivity associated with sandy layers.

Additional uncertainties or systematic errors that were not estimated by Monte Carlo analysis include the effects of sedimentation and instrument tilt. Sedimentation corrections of 15%–20% may be appropriate for individual measurements made where sediments are thickest (e.g., Davis et al., 1999). We did not apply sedimentation corrections because this would make it difficult to compare data collected during different surveys unless a sediment thickness estimate accompanies each measurement, and this would not be possible for measurements that are not colocated with seismic data. In addition, we focused our studies on local variations in heat flow, associated mainly with basement relief, that are often larger than the maximum sedimentation correction. Accounting for the thermal influence of sedimentation may be important for assessing the extent to which heat flow values deviate from lithospheric conductive models, as discussed later.

The tilt sensor on one of the heat flow instruments used during the RetroFlux survey failed, but the mean tilt correction for measurements made with a working tilt sensor was <2%, suggesting that errors are likely to be small for penetrations where there are no tilt data. This is consistent with the very close spacing between most measurements (50–200 m) made during a single lowering of the probe from the ship, which helped to keep the probe hanging vertically immediately prior to penetration into the seafloor.

Estimation of temperatures at depth

Heat flow data were continued downward to estimated temperatures within the sediment and uppermost basement based on the interpretation that heat transport within the sediment is dominantly vertical and conductive (Davis et al., 1999). Data used for this analysis included seismic profiles, thermal conductivities, core-scale seismic velocities, and sediment thicknesses determined during Leg 168.

Two-way traveltime (TWT) was converted to subseafloor depth using velocity data collected from recovered sediments, adjusted to agree with depth to basement determined during drilling (Davis et al., 1999, 1997). The resulting range in apparent sediment seismic velocities is 1500–1700 m/s. Separate relations were developed for the western and eastern ends of the Leg 168 drilling transect (z in meters, t in seconds):

z = 817t + 98t² – 81t³

and

  z = 937t - 108t² + 187t³,

respectively.

A constant sediment velocity was assumed for sediment thicknesses greater than that encountered during drilling (~600 m). The greatest uncertainties in calculated depths to basement occur where picking the top of basement from the seismic data is difficult, as is common when the upper basement surface is irregular or steeply inclined or the seismic resolution is otherwise poor.

Uncertainties in estimated temperatures at depth arise from errors in individual heat flow determinations, TWT between the seafloor and uppermost basement picked from seismic profiles, cumulative thermal resistance versus traveltime relations, and the assumption that heat transport within the sediment section is one dimensional (1-D) and vertical (Davis et al., 1999). Heat flow uncertainties came from the Monte Carlo analysis discussed above. We picked upper and lower limits for the seafloor and basement reflectors along each profile, based on clarity in the processed seismic data, and used this range to estimate uncertainties associated with this part of the analysis. The thermal resistance versus traveltime relations are well calibrated in areas where we have drilling data, and we have never found thermal or geochemical evidence from nonconductive heat transport through sediments in this region except where sediments are extremely thin. Total uncertainties in temperatures estimated at the sediment/basement interface and in the location of isotherms at depth within the sediment section are likely to be ~15%–25%.

The thermal structure within oceanic basement is less well constrained. Vertical heat flow within upper basement could range from essentially fully conductive to fully advective. One of the goals of Expedition 301 and associated experiments is to determine the extent to which the upper 300–400 m of basement is isothermal, as a result of vigorous convection, or transports heat conductively. We will not know whether conditions in upper basement are vertically conductive or advective until after data are recovered from temperature sensors deployed at depth within long-term observatories (Fisher et al., 2005). For illustrative purposes, we assume that thermal conditions in uppermost basement are conductive, with an effective thermal conductivity of 1.6 W/m·K (e.g., Becker et al., 1983; Busch et al., 1992; Karato, 1983) and a P-wave velocity of 4500 m/s (e.g., Carlson, 1998; Jacobson, 1992; Rohr, 1994).

Selected heat flow transects are compared to 2-D conductive calculations of heat transport using a finite-element model (Zyvoloski et al., 1996). The grids for these models were created using digitized data from colocated seismic profiles. The goal in creating these models was to determine the extent to which local variability in seafloor heat flow might be attributed to conductive refraction associated with seafloor and basement relief and the contrast in basement and sediment thermal properties. Particularly in locations where there is significant basement relief, the 1-D assumptions associated with downward continuation of near-surface thermal data may be violated and conductive 2-D models provide a useful reference to which observations can be compared to assess the potential significance of hydrothermal processes.

Top of page   |   Previous   |   Next