All the data derived from the previously described 13 processing techniques categories (without the laboratory-derived velocities) incorporated into the initial data pool from 0 to 244 mbsf are shown in Figure F7. The data are also given in Table T2. We merged all data categories with a total of 5400 values into a single depth/velocity matrix by offsetting the depth of the individual velocity categories by 5 mm. Looking at the hole section in a compressed representation, it is difficult to detect well-supported trends (Fig. F7). Although only data that passed several quality criteria were included, the data set remains extremely spiky and velocity variations of 1600 to 2800 m/s for the same depth interval are common. Based on the large standard deviation of the data, we rejected a simple smoothing of the values. Our first step toward simplifying the data were carried out in 25-m sections. We used our own method as described in the "Seismic Stratigraphy" section of the "Explanatory Notes" chapter of the Leg 178 Initial Reports volume (Shipboard Scientific Party, 1999a). This method is based on individual polynomial fittings of variable orders and uses selectable confidence intervals around the polynomial fit. An example of a six-step cleanup is given in Figure F8. An interpolated trendline connecting values chosen is shown together with the initial data pool in Figure F7 (blue line). It should be noted that the polynomial fitting and exclusion technique will always favor incorporating regions with high data density. In most cases, this is an improvement over simple averaging, since outlyers are completely removed and do not affect the resulting data. On the other hand, this method by no means ensures the extraction of only good data out of clusters in case the majority of the data for one depth interval is erroneous and the described data separation technique fails.
Subsequently, the data were filtered with a low-pass filter (Fig. F9) designed to filter out short wavelength variations. The frequency range of the pass band is set to reduce the vertical resolution of the filtered velocity log to approximately 2 m. The final representation of our approach is given in Figure F10. All data processing within this study is based on raw unsynchronized data with respect to depth shifts between logging runs and with respect to the different transmitter receiver pairs used for the different data categories. To achieve a comparable profile with regard to other depth-shifted logging data processed by the Lamont-Doherty Earth Observatory Borehole Research Group (LDEO-BRG), the resulting data were graphically fitted with the integrated resistivity (IMPH) data. The program used is AnalySeries 1.2 (Paillard et al., 1996), and the 14 matchpoints and resulting depth shifts are given in Table T3 and Figure F10. According to the LDEO-BRG depth scale, the resulting depth shifts are larger near the drill pipe between 87 and 100 mbsf, very reasonable in the interval 100 to 207 mbsf and 230-244 mbsf, and unrealistic high in the short interval between 207 and 212 mbsf. The higher shift values at the base of the drill pipe may be due to problems encountered during the process of reentering the tools after the logging run (Shipboard Scientific Party, 1999b). Variable shift values between 0 and 3 m can be easily explained with a combination of three effects:
In the final velocity profile, the effect of those three factors is combined to various degrees depending on the varying importance of a single data category for a specific depth interval to the finally chosen data. However, the large shifts in the depth interval 207-212 mbsf are probably unrealistic and the result of a mismatch or incorporation of erratic data into the final velocity data selection. Since the correlation is entirely based on graphical correlation with the same systematics for the whole section (matching regional highs and lows of the reference with regional highs and lows of the filtered data curve), and given that the erroneous interval is very short, we decided to stick to the correlation and accepted its limitations in the mentioned depth interval 207-212 mbsf.
Finally, after compiling the data from all depth sections (0-75, 75-244, and 244-360 mbsf), the contacts were smoothed, resulting in a continuous profile that is displayed together with the laboratory velocity data values (Fig. F11, Table T4).
The computed velocity curve can be compared for validation to the other representative downhole logs obtained at Site 1103 (Shipboard Scientific Party, 1999b): the neutron porosity (APLC), the bulk density (RHOM), the electrical self-focusing resistivity (SFLU), and the magnetic susceptibility (RMGS) (Fig. F12). The chosen logs show reliable values, except the anomalous ones in the RMGS log (~117 mbsf), certainly caused by the APS bow spring lost in the hole.
The velocity curve is general correlated with the RHOM, RMGS, and SFLU logs, and anticorrelated with the APLC log. The velocity curve shows the same features as the other logs that are divided in five units (Fig. F12). The first unit is characterized by low porosity and high resistivity, density, and velocity values. The second unit exhibits porosity values between 25% and 50% and lower susceptibility, density, velocity, and resistivity values. It is interesting to point out that the thin beds (~132 and ~142 mbsf) seen in all the logs (high SFLU and RHOM values; low NPHI and RMGS values) are also found in the velocity curve. In the third unit, the resistivity, the susceptibility, and the velocity logs show the same higher values at the top with a slight tendency to decrease down the hole. The fourth unit is characterized by a sharp reduction of the previous logs and a sharp increase in the porosity values. In the last unit, we note a distinct jump to lower density, resistivity, susceptibility, and velocity values and higher porosity. The high variability of the logs in this part is not seen in the velocity curve, probably because of the smoothing method used to reconstruct it.
Additionally we present a comparison of the original seismic data and a depth-migrated section generated by using the new velocity data. The original time section already published in the Leg 178 Initial Reports volume (Shipboard Scientific Party, 1999b), the new velocity data, and the migrated depth section are shown in Fig. F13. The data presented confirm the location of the major shelf unconformity at 222 ms TWT, or 243 mbsf. The unconformity between seismostratigraphic units S1 and S3 (Shipboard Scientific Party, 1999b) is seismically expressed by a strong negative and subsequent positive reflection around 222 ms TWT below seafloor (Fig. F13A). The decline and rise in acoustic impedance (acoustic impedance = velocity x density) within the depth interval 220 to 245 mbsf seen in the velocity and density data of Figure F12D are most likely the cause of this reflector. The positive reflection around 206 ms TWT on the other hand, is probably still part of the S1 topset package.