SEISMIC DATA AND ANALYSIS

A survey of the Hawaii-2 cable between 140° and 143°W was carried out in August 1997 on the Revelle (Stephen et al., 1997), during which single-channel seismic (SCS) data were acquired along the cable track starting at 140°W and heading southwest. Site 1224 is to the southwest of a well-surveyed block but is bracketed by two parallel SCS lines (Line 14a in Figure F1B and a second SCS line not shown in the figure). The seismic source from the generator injector (GI) gun from Seismic Systems, Inc. was run in "pure GI mode." The generator chamber is 45 in³, and the injector chamber is 105 in³. It was estimated that the GI gun was towed at a depth of 8 m or less (Stephen et al., 1997). The streamer consists of two active sections towed at a depth of <20 m. The digital data were acquired on an analog-to-digital acquisition system from the School of Ocean and Earth Science and Technology (SOEST) (Hawaii, USA) (Stephen et al., 1997) and were recorded with a sampling interval of 1 ms and a time delay of 5 s. The spacing between shots was 40 m. A portion of the seismic section along Line 14a (the North line) shown in Figure F4 consists of 200 shotpoints (SPs) around SP 16002. SP 16002 is the closest point of approach to Site 1224.

The amplitude spectrum of the original seismic data recorded at SP 16002 is shown in Figure F5A. The amplitude spectrum of recorded seismic data depends on the source signal, hydrophone response, survey design specifications, Earth structures that the seismic signal passes through, and environmental noise. The GI gun outputs a single large spike with essentially no bubble pulse. The width of the single large spike on the blast phone at the gun was measured to be ~10 ms. Because of the source and receiver depth in the water and the existence of the air/water surface ("ghost effect"), the resultant wavelet recorded is a composite waveform of a three-peaked pulse with a central negative peak twice as large in amplitude as the positive first and third pulses. This situation is common to nearly all marine seismic surveys. The resultant wavelet of the Hawaii-2 observatory (H2O) seismic site survey was thus estimated to have a dominant frequency of 33 Hz (Stephen et al., 1997). The original spectrum for SP 16002 in Figure F5A indeed shows strong seismic energy distributed in the frequency components between 10 and 50 Hz. Moreover, the spectrum also reveals that there are two more energy bands in the original seismic signal with dominant frequencies of 70 and 160 Hz, respectively. These two higher-frequency energy bands shown in Figure F5 for SP 16002 are typical for the recorded seismic data of the survey, and they are seismic signals rather than environmental noise. The three-banded amplitude spectrum of the original record shown in Figure F5 agrees very well with those typically recorded with an air gun at a depth of 10 m and the streamer at a depth of 15 m (Evans, 1997, p. 154). The direct current component due to the ghost effect is also evident in the original spectrum.

Downhole logging data indicate that within the 140-m basement drilled at Site 1224 there are layered structures on the log scale. Synthetic seismograms with different wavelets of variable dominant frequencies could help determine in what seismic frequency bands such layered structures can be identified. The process of synthetic seismogram generation is illustrated in Figures F6 and F7. Using downhole density and compressional wave velocity data, an impedance profile, which is a product of density and velocity (Z = x V), is obtained, as shown vs. depth in Figure F6 and vs. two-way traveltime in Figure F7. The density and wave velocity of seawater is assumed to be 1.024 g/cm³ and 1.5 km/s, respectively. An average density and velocity for the 28-m sediment section of 1.52 g/cm³ and 1.5 km/s, respectively, is estimated from shipboard core measurements. The density and velocity for the basement rock between 28 and 35 mbsf are assumed to be a single-value extrapolation of the logging data at 35 mbsf. A reflection coefficient (RC) series can then be calculated using the formula

r_i = (Z_i+1 – Z_i)/(Z_i + Z_i+1),

from the impedance at a logging depth sample i and the impedance at the adjacent logging depth sample i + 1. The calculated RC series is termed as primary RC without considering multiple reflections, which are shown in depth in Figure F6 and in time in Figure F7. A synthetic seismogram is usually generated using the RC series convolved with a source wavelet. To demonstrate the convolution process, a Ricker wavelet with a dominant frequency of 75 Hz, labeled as Ricker (75Hz) in Figure F7, is used to calculate the seismogram, shown as RickerSeis in depth in Figure F6 and time in Figure F7. By reversing its polarity, the Ricker wavelet is the ideal presentation of the composite air gun-hydrophone signature. A dominant frequency of 75 Hz is used instead of 33 Hz, as indicated from the seismic data, to show possible effects of structures on the synthetic seismogram. From the synthetic seismogram (RickerSeis) shown in Figures F6 and F7 and its comparison with unprocessed SCS data at Site 1224 in Figure F8, the traveltime for reflections from the water/sediment and sediment/basement interfaces agrees well with the field SCS data. However, the seismic energy in the 10- to 50-Hz frequency band in both the synthetic seismogram and field SCS data masks the detailed basement structures revealed by the downhole logs.

It is evident that the resolution of the SCS data needs to be increased with proper seismic data processing algorithms. One simple way is to apply a bandpass (BP) filter to the seismic data in order to eliminate the seismic energy in the frequency band of 10–50 Hz. By analyzing the original spectrum of the SCS data in Figure F5A, we use a bandpass filter of BP (10, 50, 250, 300) to keep all the seismic energy in the frequency band of 50–250 Hz as shown in Figure F5B. We apply the same bandpass filter to the RC series to retain only the structural information in the frequency band of 50–250 Hz so that the filtered RC series is effectively a synthetic seismogram that can be compared with filtered seismic data. The BP-filtered RC series is shown as BPSeis vs. depth in Figure F6 and vs. two-way traveltime in Figure F7. It is enlarged in Figure F7 to show detailed variations. The boundaries between sediment/basement interfaces and between different basement units are shown in two-way travel time in Figure F7. These boundaries are identified by comparison of the impedance and RC signatures on the time profile (Fig. F7) with the depth profile (Fig. F6). Figure F9 shows a comparison between the BP-filtered RC synthetic seismogram and unprocessed SCS data. A portion of BP-filtered SCS data is shown in Figure F10 in comparison with the BP-filtered RC synthetic seismogram. The amplitude spectrum of the BP-filtered SCS data recorded at SP 16002 is displayed in Figure F5B. The agreement between the synthetic seismogram and seismic data in Figure F10 is better than that shown in Figure F9 for the unfiltered SCS data.

In order to better reveal the shallow seismic structure around Site 1224, the advanced high-resolution enhancement technique (Highres) was also used to improve the seismic signal/noise ratio and seismic resolution of the SCS data. Details of the Highres method are given by Sun (2004) in the attempt to resolve 0.5-m dolomite layers in modern marine sediments near the Japan Trench. The amplitude spectrum of the Highres-processed seismic trace at SP 16002 is shown in Figure F5C. In comparison with the amplitude spectrum after simple BP filtering shown in Figure F5B, it shows that Highres enhances the seismic energy in the higher-frequency components within the valid seismic energy band, especially the energy centered at 160 Hz. Comparison of the Highres SCS data with the BP-filtered RC synthetic seismogram in Figure F11 demonstrates that both the seismic resolution and seismic quality are much improved relative to the simple filtered SCS data. The basement units are identified following the seismic signatures of the BP-filtered RC synthetic seismogram in Figures F6 and F7 as will be explained later. Better seismic imaging and higher quality are also evident on the BP-filtered SCS data shown in Figure F12 in comparison with the unprocessed seismic section shown in Figure F4. This helps to demonstrate that even a simple filtering could improve seismic resolution to resolve the detailed crustal structures (Sun and Goldberg, 2000).