Seismic stratigraphy depends on linking seismic reflections to lithostratigraphy and to physical properties of sediments and hard rocks by means of synthetic seismograms. In the absence of downhole logs, physical properties measurements on cores or core samples can be used for this purpose if recovery is sufficiently high. The high quality of the downhole logs at this site enables us to assess the quality of both densities and velocities determined from discrete samples through synthesizing seismograms. This is relevant for other sites that were not logged where we use index properties or MST data to tie the lithostratigraphy to MCS reflection data.
The sampling interval of the downhole velocity log at Site 1137 is 15 cm. In comparison, sampling intervals of index properties measurements may vary from 20 cm to >10 m where recovery is low or zero. Low core recovery results in phase shifts between velocity anomalies from logs and samples, which may amount to several meters (Fig. F85). Amplitude differences in velocities from downhole logs and discrete samples are primarily caused by different temperatures and pressures for the in situ and laboratory measurements, which affect material properties, as well as differences in measurement techniques. High-frequency positive velocity peaks from downhole logs are typically lower than corresponding velocities from samples. Alternatively, sample velocities also show large scatter when collected at small intervals, which partly results from measurement uncertainties. These uncertainties vary depending on what kind of sample is used (i.e., cube, minicore, half core with or without liner), how well contact between the sample and transducers can be established, and how much the core was disturbed by RCB drilling.
To quantitatively assess the correlation of velocities from samples and logs at this site, we computed the coherence between the two profiles for Site 1137. Coherence, a dimensionless parameter that is a function of wavelength (or frequency), is zero when two data series are completely out of phase (i.e., a phase difference of 90°) and is 1 when they are in phase. When coherence is larger than 0.5, the two time series are considered to be coherent.
We applied Welch's method of periodiogram averaging, as implemented in Matlab, using nonoverlapping windows 128 samples long, corresponding to 19.2 m at a sampling interval of 15 cm. In terms of spatial wavelength, this implies an assessment of the coherence at wavelengths between 30 cm and 48.4 m. The resulting estimate of the coherence function (Fig. F86) illustrates that a unit has to be at least ~5 m thick to be detected at the same location downhole both by logs and index properties measurements. The lack of coherence for thinner units mainly results from two factors: (1) "misplaced" sample-velocity measurements in terms of depth caused by discrepancies between curated and real depths from incomplete core recovery and (2) the inherent difference between in situ and laboratory measurements, including core disturbance. A third effect that may also play a role is velocity measurement uncertainties.
We experimented with different filters to remove noise from index properties data from various sites before computing coherence and found independently that a robust mode filter with a width of 3.5-5 m appears to give the most acceptable result for cleaning both velocity and density measurements from samples (see "Downhole Measurements"). Based on these results, we adapted the approach to apply a 5-m-long robust mode filter to all those index properties data used to synthesize seismograms for other sites. The filter was applied at 1-m increments. Both the velocity and the density log show a distinct inversion in Unit IV. It is caused by an altered crystal vitric tuff that separates the two lowermost basalt flows.
We used velocity and density data from logs (90-350 mbsf), filtered data from physical properties measurements (12-90 and >350 mbsf), and GRAPE measurements (0-12 mbsf) to synthesize seismograms for linking the cores and logs with MCS data. Velocity data were extrapolated from 0 to 18 mbsf, assuming constant velocity. We summed transit times to create a log of TWT vs. depth. We resampled this log linearly using a sampling interval of 0.1 ms, resulting in oversampling to avoid aliasing. We then resampled velocities and densities using the TWT array and obtained impedance from the product of velocity and density and computed reflection coefficients from impedance contrasts (Fig. F87).
We constructed synthetic seismograms with and without multiples and transmission losses as described in the "Seismic Stratigraphy" in the "Explanatory Notes" chapter. Comparison of the two synthetic seismograms (Fig. F87) shows that the phase of all large-amplitude peaks is nearly identical. However, lower-amplitude reflections within the sediments show more severe differences in phase, and the amplitude of the lowermost reflections within basement is diminished in the synthetic seismogram that includes transmission losses (Fig. F87). The attenuated reflections, including transmission losses, match the MCS data better than the synthetic trace without multiples and transmission losses (Figs. F87, F88). This is caused by the large impedance contrasts between massive basalts and overlying brecciated basalts, volcanics or sediments resulting in rapid loss of energy with increasing depth.
Superimposing the synthetic seismogram with (or without) multiples shows that it covers a range in TWT that is larger than in the MCS data. In other words, the seafloor and basement reflections cannot be matched at the same time, as the summed transit times have resulted in a total TWT that is larger than that measured from the MCS data at Site 1137. This implies that we have underestimated velocities overall. Core recovery should be biased toward more indurated sediments with higher velocities; therefore, the discrepancy is not likely caused by velocities based on discrete samples. Without check shots, it is not possible to investigate where velocities are incorrect. To tie our synthetic seismogram to the MCS data, we simply split the former into two segments, matched to the seafloor and to the basement, respectively.
Reflections M1 and M2 (early to middle Miocene) in the synthetic seismogram largely result from interbed multiples. However, they cannot be tied unequivocally to the MCS data, where one broad peak, rather than two distinct reflections, is present at the same TWT. We tentatively tie reflection M1 to a broad reflection in the MCS data. We cannot match reflection E (late Eocene to late Oligocene) to the MCS data, but it may correspond to a broad reflection at that depth. It is caused by a logged drop in density between 165 and 195 mbsf, and it also corresponds to an increase in shear wave velocity (Vs) and decrease in Poisson Ratio at this depth interval (see "Downhole Measurements").
Four large reflections mark the Cretaceous part of the section, which match the MCS data well: (C) the top of the glauconite bearing sandy packstone (Unit III, Campanian), (B3) the massive basalt at the top of acoustic basement (basement Unit 1), (B2) the top of the massive basalt at basement Subunit 7B, and (B1) the top of the massive basalt at basement Subunit 10B (i.e., the deepest unit drilled). However, the reflection caused by the massive basalts of basement Subunit 7B is slightly shallower than in the MCS data. This may be attributed to cumulative errors in the TWT computed from interval transit times, as no check shots are available for testing.
These results show that below acoustic basement, the only features resolved by the MCS data are massive basalts overlain by a low-velocity unit at least 15-20 m thick. For instance, we cannot resolve the boundary between basement Units 2 and 3, even though they are separated by a low-velocity brecciated basalt flow top ~10 m thick. At average velocities of ~5 km/s within basaltic basement, and a peak wavelet frequency of 40 Hz, the tuning thickness (one-quarter of the wavelet length) is ~30 m. This implies that a low-velocity unit separating two basalt units would have to be at least 30 m thick for both its top and its base to be imaged. The theoretical limit for detecting a high- or low-velocity layer within basement given a velocity of 5 km/s and a frequency of 40 Hz would be one-thirtieth of the wavelet length (Badley, 1985), or ~4 m. However, the integration of the MCS data with downhole logs and the lithostratigraphy demonstrates that in fact we cannot resolve any layers thinner than ~15 m at Site 1137.
All basement units imaged in the seismic reflection data at Site 1137 have an apparent dip to the east, and the glauconite-bearing sandy packstone overlying basement thickens to the east (Fig. F88). This indicates that the age of the oldest sediments probably increases toward the east. Basement Unit 7 is not laterally continuous for more than 1.5 km, and we can trace basement Unit 10 for a total of ~3 km in an east-west direction.