The SCREECH survey acquired >3000 km of MCS, magnetic, gravity, and multibeam bathymetric data and 1000 km of wide-angle seismic reflection/refraction data off Newfoundland (Fig. F1). This was a two-ship program with MCS, magnetic, gravity and bathymetric data acquired aboard the Maurice Ewing (Cruise 00-07) and wide-angle seismic reflection/refraction data acquired by ocean-bottom seismometers/hydrophones deployed and retrieved from the Oceanus (Cruise 359-2). Coincident MCS reflection data and wide-angle seismic reflection/refraction data were collected along three primary transects (Fig. F1). MCS data were also collected on lines parallel and perpendicular to all transects, including the locations of both Leg 210 sites indicated by white stars in Figure F1. MCS data were recorded on the 6-km, 480-channel streamer of the Maurice Ewing; these data have a sampling interval of 4 ms, a shot-spacing of 50–62.5 m, a fold of 45–60, a recording length of ~16 s, and a common midpoint (CMP) spacing of 6.25 m. The tuned, 8540-in3, 20-gun array of the Maurice Ewing was the seismic source for both wide-angle and MCS seismic data. SCREECH Line 2, the central of the three primary transects, crosses Leg 210 Sites 1276 and 1277 (Figs. F1, F2), and the MCS data from this line are used for comparison with synthetic seismograms created from laboratory measurements. The profiles presented here were created by prestack time migration; a full description of processing is given by Shillington et al. (2004).
At Site 1276, 936.9 m of core were recovered between 800.0 and 1736.9 mbsf (Shipboard Scientific Party, 2004). Cored sediments range in age from early Oligocene to earliest Albian–latest Aptian. The sedimentary section was subdivided into five units based on variations in lithology and breaks in age as defined by biostratigraphy (Fig. F3). A brief description of each unit is given below; a complete description of the lithology and biostratigraphy can be found in Shipboard Scientific Party (2004).
Approximately 96 m of sediment lie between the two sills, including a ~17-m-thick interval from 1693 to 1710 mbsf that has very low velocities (~1.7 km/s) and densities (2.05 g/cm3). The explanation for the undercompaction of these sediments is still uncertain, but it is possible that the sills sealed off this interval and thus prevented normal compaction (Karner and Shillington, 2005; Shipboard Scientific Party, 2004).
As part of the standard shipboard analysis of all recovered cores, laboratory measurements of physical properties (e.g., density, compressional [P]-wave velocity, thermal conductivity, natural gamma radiation, etc.) were made on whole cores and/or selected samples. Of particular value to the present study are velocity and density (Fig. F3). Horizontal (x) and vertical (z) velocities were measured on representative sediment and rock samples every ~2 m throughout the recovered section (Shipboard Scientific Party, 2004). Cubes of rock ~8 cm3 in size were cut from the working half of the core, and P-wave velocity was measured in three directions using the P-wave velocity sensor 3 modified Hamilton Frame velocimeter. An acoustic signal of 500 kHz is transmitted and received between two transducers, passing though the sample, whose thickness is measured by a digital caliper. We chose to use the measurements of vertical velocity to create synthetic seismograms because they most closely approximate the path of seismic waves recorded in the seismic reflection data. This choice has implications for the time-depth relationship established below 800 m. The difference between vertical and horizontal velocity increased downhole from ~4%–5% at 800 mbsf to ~10% in the deepest sediments (Shipboard Scientific Party, 2004). Horizontal velocities are often faster in sediments because of grain orientation and cementation along near-horizontal bedding planes. The longest offset arrivals in the MCS reflection data will have a significant contribution from horizontal velocities. Therefore, the use of vertical velocities indicates that the depth-time relationship established below 800 mbsf from laboratory measurements will represent the slow end-member.
Two types of density measurements were taken on each core: (1) gamma ray attenuation (GRA) bulk density and (2) moisture and density (MAD). GRA data are evenly spaced measurements of density obtained over the full core before splitting. Although this procedure provides continuous, finely spaced measurements (~2.5 cm), the consolidated sediments and rocks retrieved using rotary core barrel drilling typically fracture when they are recovered, leading to breaks in the core, reduced core volume, and significant artificial variations in the density data. GRA density measurements are also too low because recovered cores do not completely fill the core liner (Shipboard Scientific Party, 2004). These artificial variations in density would cause significant noise when computing synthetic seismograms, and thus the GRA densities were not used. The MAD technique determined wet and dry bulk density, grain density, and porosity on discrete samples taken from every section of each core. Because the samples analyzed were small and hand-selected, they were not as compromised by the fractures that degrade the GRA density measurements, although they are affected by decompaction. Also, because of noncontinuous sampling and imperfect (85%) core recovery, they do not sample the details of density and velocity changes that would be measured by downhole logging. Densities were obtained from the same samples measured for velocity, so any sampling biases will be present in both data sets. Nonetheless, because these data contain fewer artificial variations because of core breaks or core diameter compared to GRA data, they are more suitable for our purposes.
The use of physical property data instead of traditional sonic and density logs requires several additional considerations to ensure that velocity and density measurements are used in a way that best represents the overall core properties:
These results show that care must be taken when interpolating physical property measurements between locations. For example, a measurement on a sample from a thin interval of a particular lithology could be extrapolated to exist over 2 m, or more in the case of gaps between cores. This could affect the reflection characteristics of a synthetic seismogram significantly, particularly if the measured velocity is much higher or lower than the average velocity trend. It may at least partially account for discrepancies between MCS data and synthetic seismograms created at sea during Leg 210. Those synthetic seismograms contained several bright reflections that could not be easily related to reflections in the MCS data (Shipboard Scientific Party, 2004). This was particularly true of synthetic reflections arising from within Unit 5, which is largely mudrock and corresponds to an interval of low-amplitude, discontinuous reflections in the MCS data (Shillington et al., 2004; Shipboard Scientific Party, 2004).
We tried three different approaches to processing physical property data in order to explore the effects of sampling bias on synthetic seismograms (Fig. F4):
The processing steps described above screen the data points in different ways to account for the presence of potentially anomalous samples and to test the effects of sample bias.
The final step was to create an "earth model" consisting of a series of layers, and it was ultimately used for the calculation of synthetic seismograms. These layers were interpreted by hand within the Nucleus software package to group together similar velocities and densities; 252 layers were included in this analysis. Given that there are 533 input data points, there is an average of <2 data points per layer. For each layer, P- and S-wave velocity, density, and P- and S-wave attenuation (QP and QS, respectively) must be given. The density and P-wave velocity assigned represent average values calculated between the interpreted top and base of the layer based on input data points. The result of this averaging step is that layers in the final earth model are assigned velocities that are neither as high nor as low as velocities in the original velocity function (Fig. F5). Both the input data points and the average velocities assigned to the layers in the earth model are shown in Figure F6. S-wave velocity is estimated from P-wave velocity. We use a QP of 1000 and a QS of 5000 for water, 200 and 100 for sediments, and 400 and 100 for sills (Fuchs and Muller, 1971; Minshull and Singh, 1993).
We generated synthetic seismograms for earth models derived from each of the three scenarios described above (Fig. F6) to examine the robustness of the reflections generated and to test their sensitivity to potential sampling biases in laboratory measurements. All three cases are discussed in "Results and Discussion."