We used densities and compressional wave velocities from index properties and MST measurements (see "Physical Properties") to synthesize seismograms at Site 1135, as no downhole logs were collected. Because of sediment disturbance caused by RCB coring and the inherent difficulties in determining velocity, a thorough inspection of the data is necessary. In addition, the sampling rate with depth fluctuates as a function of core recovery. Therefore, smoothing and resampling of the data (Fig. F15A, F15B) are necessary before using them for seismogram construction. Densities determined from core samples and GRAPE, and velocities from discrete core samples are scattered. In particular, the GRAPE densities cannot be used in their raw form.
Compressional wave velocities and densities from core samples are more densely spaced in the upper half of this hole compared to the lower section, especially below ~315 mbsf, where recovery was poor (Fig. F85, in the "Site 1137" chapter). A detailed comparison of velocities from samples and logs was possible at Site 1137 (see "Downhole Measurements" in the "Site 1137" chapter). Based on coherence analysis of the velocity data at Site 1137 (Fig. F90 in the "Site 1137" chapter), we filtered the sample velocities from Site 1135 using a 5-m running window, at a step (i.e., sampling rate) of 1 m. We used a robust mode filter (i.e., a maximum likelihood probability estimator that calculates the mode within a 5-m-long data window). In addition, we computed the median of the filtered data set and replaced outliers whose values exceed 2.5 times the L1 scale with the median, as described in "Downhole Measurements" in the "Site 1137" chapter.
The smoothed velocity profile (Fig. F15) shows velocity inversions at 305, 335, 430, and 490 mbsf. Because of poor core recovery below 315 mbsf, the exact shape or width of such anomalies should not be overinterpreted.
Comparison of filtered GRAPE, and discrete sample densities shows that above ~315 mbsf the GRAPE measurements yield higher values than those from discrete samples. The average difference is of the order of 0.05-0.1 g/cm3 (Fig. F15). This is unexpected, as densities from discrete samples are usually equal to or higher than GRAPE density data. This is because GRAPE density data show lower than expected values wherever the liner is not entirely filled with core (see "Physical Properties," ). The density data in the sedimentary section from Site 1137, where, in addition, densities from downhole logs are available, do not exhibit this effect. Rather, densities from discrete samples, GRAPE, and logs agree quite well for soft sediments at Site 1137 (Fig. F90 in the "Site 1137" chapter).
Because the discrete samples are taken soon after the core is split and covered, it is highly unlikely that lower than expected bulk densities result from evaporation of pore water, especially as the observed difference would equate a pore-water loss of 5%-10%. A calibration error of the pycnometer is also quite unlikely, as the sample densities are anomalously low only above ~310 m (see "Physical Properties"). The copious small dropstones in this core may account for the difference. Their anomalously high densities are reflected in the GRAPE measurements, but not necessarily in the sample densities, as the dropstones are not abundant enough to find their way into the one or two small samples per core taken for density measurements.
As a consequence, we regard the GRAPE densities as more reliable than the densities from discrete samples above ~315 mbsf. Conversely, below this depth, where consolidated sediments were sampled, densities from discrete sample measurements are equal to or higher than GRAPE densities, as here the liner is not entirely filled with core. We created a composite density profile, using GRAPE data above 315 mbsf and data from discrete samples below this depth. We filtered density and velocity data identically (Fig. F15).
We synthesized seismograms for Site 1135 (see "Seismic Stratigraphy" the "Explanatory Notes" chapter). We resampled densities and velocities every 0.5 ms as a function of TWT and created profiles for impedance, reflection coefficients, and a seismic trace. The seismic trace is based on convolution with a Ricker wavelet with a peak frequency of 40 Hz. As no density data below 450 mbsf are available for this site, the lowermost 70 m of this site was not included in the synthetic seismogram. Reflection coefficients with and without multiples and transmission losses are fairly similar at this site. No major transmission losses occur because basement is not included in the model and because the section is fairly thin, compared with other sites such as Sites 1137 and 1138, where transmission losses are more prominent. Whether or not interbed multiples create distinct reflections depends on whether they constructively or destructively interfere, respectively. This is not the case at this site, and the two synthetic seismograms show only minor differences (Fig. F16).
Tying MCS data to a synthetic seismic trace based on the physical properties of core samples is difficult because core recovery is typically biased toward more indurated sediments. Thus, TWTs calculated from interval transit times based on velocities from such incomplete recovery may be too low for parts of the profile. Also, no density or velocity data are available above 20 mbsf. We used the uppermost measured densities and velocities for extrapolation to the seabed, assuming constant values. This likely results in overestimating both densities and velocities for this part of the section. Determining velocities on sediments disturbed by RCB coring also results in underestimated velocities. In the absence of logs and check shots, we cannot expect MCS reflections at the same TWTs as in the synthetic seismogram. Despite all these shortcomings, Figure F17 illustrates that the overall match is better than expected, and most major reflections in the data can be tied to the synthetic seismogram and, therefore, to the lithostratigraphy (Fig. F16).
The thick Eocene part of this section includes five regionally continuous reflections (labeled E1-E5) that match impedance contrasts caused by slight changes in density and velocity in Subunit IIA (Fig. F16). The cause of these changes in terms of lithology cannot be determined without a more detailed analysis of the sediments. They are underlain by two late Paleocene-middle Eocene reflections (P1 and P2), which we tentatively tie to the lower and middle part of Subunit IIB.
Three Maastrichtian-late Paleocene reflections (labeled M1-M3) correspond to Subunit IIIA (M1-M3). Reflection M3 is located just above the K/T boundary, and below the boundary between Subunits IIB and IIIA. Here, a major increase in velocity marks the transition between white nannofossil ooze and white nannofossil chalk. M1 is the most prominent reflection above basement in this section and coincides with a velocity inversion, followed by a major increase in velocity below, roughly marking the boundary between Subunits IIIA (white nannofossil chalk) and IIIB (white to light gray calcareous chalk) within the Maastrichtian (Fig. F16).
A reflection (labeled CM) within Subunit IIIB of Campanian-Maastrichtian age is also related to a velocity inversion. Subunit IIIC, tentatively dated Turonian-Campanian, includes three high-amplitude reflections (labeled TC1-TC3) in the synthetic seismogram. Their exact match to the MCS data is not obvious, as cumulative errors in computed TWT, caused by poor core recovery, result in substantial phase changes between observed and modeled reflections in the lower part of the section. Reflection TC3 roughly corresponds to the boundary between Subunits IIIB and IIIC. We have tentatively tied TC1, caused by a velocity inversion, to a prominent reflection in the seismic data at ~2.6 s TWT (Fig. F17).
Some of the reflections have no clear relationship to lithologic boundaries. Numerous chert layers are present in the lithologic column (Fig. F16), however, these layers generally do not appear to correlate with seismic reflections in either the synthetic seismogram or the MCS data. In part, this may be attributable to poor core recovery, which renders exact locations of chert layers uncertain. Also, a thin (<<1 m) chert layer is not thick enough to cause a reflection. Given an average velocity of ~2.1-2.4 km/s in the Cretaceous sedimentary section, and a source with a peak frequency of 40 Hz, the tuning frequency (one-quarter of the wavelet length) is 13-15 m, and the theoretical resolution limit of individual beds is 1.75-2 m (Badley, 1985). In contrast, individual chert layers are typically from 1 to 10 cm thick and rarely reach thicknesses of 20 cm. Consequently, where MCS reflections correspond to synthetic reflections, and to mapped chert layers, their causal relationship is not obvious, unless interference effects of several closely spaced chert layers create an impedance contrast large enough to cause a reflection. The poor recovery in the Cretaceous section does not allow us to test this hypothesis.