Page 323 - Acquisition and Processing of Marine Seismic Data
P. 323

314                                  6. DECONVOLUTION


























           FIG. 6.1  Effect of the bandwidth on the resolution of seismic data. (A) Wide band (4–140 Hz), and (B) narrow band (4–60 Hz)
           seismic data.

           because a wide spectral band always corre-   Deconvolution broadens the band-limited spec-
           sponds to a better temporal resolution. As an  trum,andeventriestoconvertitintoawhitespec-
           example, Fig. 6.1 compares small portions of  trum consisting of a balanced amplitude
           wide (4–140 Hz, Fig. 6.1A) and narrow band   distribution for the whole available frequency
           (4–60 Hz, Fig. 6.1B) seismic sections, where the  band. The theoretical basis underlying the use of
           resolution is much better in the case of wide  deconvolution to achieve these goals is basically
           bandwidth. That is, the dominant frequency of  to remove (at least, to compress) the source wave-
           broadband data is relatively higher, which ulti-  let embedded in the seismic data. The time length
           mately results in a smaller wavelength, increas-  of the wavelet before deconvolution is much
           ing the resolution of the seismic signal.    largerthanthatofthewavelet afterdeconvolution
              Deconvolution has a number of significant  (Fig. 6.2).
           effects on seismic data, as shown schematically  Although deconvolution is generally applied
           in Fig. 6.2. In ideal conditions, the deconvolution  to preprocessed shot or common depth point
                                                        (CDP) gathers, it can also be applied to stack sec-
           • improves temporal (vertical) resolution
                                                        tions, which is known as deconvolution after
           • removes or compresses the seismic wavelet
                                                        stack (DAS). If applied to prestack data, the
              from the data
                                                        shots or CDPs must be properly preconditioned
           • removes the ringy character of the reflections
                                                        and the coherent and random noise components
           • eliminates multiple reflections
                                                        must be removed as far as possible, since the
           • whitens the amplitude spectrum to obtain
                                                        quality of deconvolution output is quite sensi-
              broad-band data
                                                        tive to noise in the data. Figs. 6.3 and 6.4
              The recorded seismic data is band limited at  compare the preprocessed shots and stack sec-
           both high- and low-frequency ends of the spec-  tions before and after a spiking deconvolution,
           trum because of the ghost reflection notches and  respectively. The deconvolution is especially
           the high swell noise amplitudes, respectively.  effective on the early arrivals and removes the
   318   319   320   321   322   323   324   325   326   327   328