Previous   |    Next

doi:10.2204/iodp.proc.341.204.2017

Results

Site U1420

In Matlab R2015a we fitted four splines to the bulk density data for Site U1420, and based on the size of the data gap the splines were selectively sampled to create a continuous, partially interpolated record consistent with the sampling rate of typical logging (Fig. F3). We fit the upper ~170 m with a spline achieving R² = 0.800 and root mean squared error (rmse) = 0.0746 with the moving average of the data compilation in this region (Fig. F4). The data gap between ~170 and 450 m CSF-A was interpolated with a minimum curvature spline, which was nearly equivalent to a linear fit between samples on either side of the gap. We fit the lower interval with an aggressive spline, which reached R² = 0.8032 and rmse = 0.0449 deeper than ~450 m CSF-A. For gaps near 500–550 and 700–750 m CSF-A, we interpolated with a smoother spline with a lower R² = 0.38 and rmse = 0.0842 compared to the data compilation. This selective sampling strategy resulted in conservative estimates of variability in poorly constrained intervals.

We created the continuous P-wave velocity curve from a compilation of logged and core-derived measurements where CSF-A and WMSF depth scales were used together directly. This may introduce error because in the CSF-A scale drilling effects such as core expansion caused by overburden release, compression during coring, and coring method are not compensated; these effects are typically on the order of 10%–20% (see “Stratigraphic correlation” in the “Methods” chapter [Jaeger et al., 2014a]). In sum, these sampling biases are difficult to accurately quantify and often affect a difference between core-and log-derived measurements. The velocity curve presented here below the logged interval (i.e., deeper than 288 m WMSF) likely represents minimum estimates because the apparent average velocity decreases with depth as opposed to a typical compaction trend and also because these measurements were collected on the WRMSL, which tends to be negatively affected by incompletely full core liners (Walczak et al., 2015). These measurements were fitted with one of three splines depending on the availability of data in the depth interval (Fig. F5). In the upper 270 m, we sampled from a spline achieving R² = 0.9287 and rmse = 0.0454 compared to the full data compilation. Deeper than 270 m on the combined CSF-A/WMSF scale, we fit the data using two splines. The more aggressive fit had R² = 0.7774 and rmse = 0.1020. Except for data gaps near 270–550 and 700–780 m on the combined depth scale, where the aggressive fit predicts unusually large variability, we sample from this spline. The spline that we used for interpolation in the poorly constrained intervals reached R² = 0.6698 and rmse = 0.1165 with data from the compilation deeper than ~270 m on the combined depth scale.

Figure F14 summarizes the cross-correlation before and after the tie. The cross-correlation is a statistic commonly used to quantify the similarity of signals based on a lag time or offset. In this case we desire a maximum cross-correlation value between our synthetic and seismic signals near zero lag time close to the site. Before visual matching, the maximum cross- correlation within five traces of the site, nearest Trace 2622 on Line GOA 2505, was 0.367 with Trace 2619 with a lag of 13 ms in the 340–1410 ms window. At the site trace, the cross-correlation was calculated as –0.118 at 0 ms lag. After a bulk shift and visual matching, we calculate that the maximum cross-correlation occurred at Trace 2624 and 3 ms lag. At Trace 2622, the maximum cross-correlation was calculated to be 0.188 at 0 ms lag, whereas the maximum cross-correlation for this trace, 0.350, occurred with 3 ms lag. To avoid overinterpretation of the data, we made minimal matches and shifts (n = 4) to achieve visually acceptable results between the synthetic and seismic traces (Fig. F15). Although the correlation is low compared to sites with more complete drilling records (e.g., the well to seismic tie in White and Simm [2003] reported an R² of ~0.87) the overall character of the seismic and synthetic traces agrees well visually among their major trends and reflectors. High-amplitude reflectors in seismic Line GOA 2505 near 475 and 895 ms TWT are well matched by reflectors in the synthetic seismic at ~390 and 820 m seismic depth below sea level (SSL). Low core recovery, variable data quality and source, data compilation handling, and a short logged interval relative to the length of the well-to-seismic tie are likely responsible for the low correlation. Figure F15, shown with the final TDR (Table T4), better illustrates the quality of the tie.

The preliminary TDR created shipboard (see Fig. F11 in the “Site U1420” chapter [Jaeger et al., 2014b]) may represent a maximum depth relationship by using linear approximation through WRMSL velocity measurements near 680 and 900 m CSF-A. We observe that values here are notably higher than the mean of the WRMSL velocity measurements overall but are low compared to the expected in situ logging trend. Unlike a typical compaction trend that increases with depth, the apparent average velocity decreases with depth at this site. Only a single WRMSL density measurement above the logged interval was available, and this measurement value is low among the data set, which results in the calculation of a smaller than expected reflection coefficient at the seafloor. We directly used bulk density values from the WRMSL, which are likely minimum values compared to ideal recovery with perfectly full core liners or an equivalently wireline logged section measured in situ (see the “Site U1420” chapter [Jaeger et al., 2014b]). Although this effect may be particularly important below the end of the logged interval (i.e., below 288 m WMSF), the variability of the core-derived measurements provides changes in the synthetic trace which we correlate to events in the seismic traces near the well location.

Site U1421

At Site U1421, we derived a bulk density compilation used for data fitting and interpolation from STMSL and WRMSL measurements from Holes U1421A–U1421C on the CCSF-B scale. We fitted two splines to the data and selectively sampled from them to create a continuous record. Using the more aggressive fit, we achieved R² = 0.9309 and rmse = 0.0599 with the compilation moving average for the full interval (Fig. F6). Although we found that it was possible to achieve higher R² values, the fit presented here is the best fit constrained such that no density <1 g/cm³ was predicted. At ~150–220, near 410, and 500–575 m CCSF-B, we sampled a smoother spline fit with R² = 0.6980 and rmse = 0.1404.

The full velocity compilation used for spline fitting was created by compiling the logged measurements with the core compilation measurements. In the upper 92 m, we sampled the spline-fitted data from the core compilation with R² = 0.7908 and rmse = 0.0674. The bottom interval consisted of the fitted log-derived compilation from a spline with R² = 0.9718 and rmse= 0.03833. Although both log and core data were available for a short interval, in the area of overlap we sampled from the spline fitted to the wireline-logged data.

We made matches between the synthetic and seismic traces, which were minimal in number (n = 14) and scale. Before matching, the maximum cross-correlation within five traces of the site, nearest Trace 411 on Line GOA 2503, was 0.417 with Trace 414 at –1 ms lag when correlating the 980–1650 ms window. At the site trace we calculate the cross-correlation with the synthetic at 0 ms lag to be 0.365, which was also the maximum in the trace window at any lag. After matching, the maximum cross-correlation, 0.500, occurred at Trace 413 with 0 ms lag. At Trace 411, the nearest to the site, the zero-lag cross-correlation was 0.413 and the trace maximum cross-correlation, 0.448, occurred at 1 ms lag (Fig. F16). Again, despite the low correlation compared to ties based on more complete data, we find that the character of the seismic and synthetic traces agrees well visually in overlay (Fig. F15). High-amplitude reflectors on seismic Line GOA 2503 near 1279, and 1620 ms TWT match well with the synthetic at 1020 and 1395 m SSL. The final TDR is shown in Table T1 with reference to both the SSL and meters seismic depth below seafloor (SSF) datums. Additionally, we note the high agreement between our final TDR and the check shot time-depth constraints, which suggests a well-calibrated model.

Application to previous work

We apply the TDR to estimate depths of several reflectors in seismic Lines GOA 2503 and GOA 2505. These sections and horizons are discussed and interpreted in Worthington et al. (2010) and the “Site U1420” and “Site U1421” chapters (Jaeger et al., 2014b, 2014c). Here, we simply provide a summary table of the depths of these horizons as calculated by our TDR (Fig. F17). At Site U1420, we were able to provide depth estimates for Horizons H1A, H1B, H1, H2A, H2B, H2C, H2D, and H2, which had TWTs less than ~1550 ms. At Site U1421 we provided depth estimates for Horizons H1B, H1, and H2A. For each location we estimated the TWT from the appropriate seismic section near the well and interpolated from our final TDR to generate depth estimates.

Top of page   |    Previous   |    Next