GNGTS 2014 - Atti del 33° Convegno Nazionale

60 GNGTS 2014 S essione 3.1 Water-layer multiple separation and imaging. After the subtraction of water layer multiples has been performed according to eq. 2, and subsequent adaptive subtraction, WVSP data D all is split into two different sets: “primary” wavefield D 1 and “multiple” wavefield D 2 : D all = P + nonwlM 1…N + wlM 1…N + IM D 1 = P + nonwlM 1…N + IM D 2 = wlM 1 + wlM 2 + … +wlM N where P stands for the primary reflections, wlM 1…N for the water layer multiples of order 1…N, nonwlM 1…N for the other types of surface-related multiples of order 1…N and IM for all the internal multiples. Primary reflections cannot be completely isolated, as only a subset of surface-related multiples has been predicted. However, as water layer multiples can be the most energetic subset of interfering wavefield, such a partial noise attenuation is still beneficial (as proved by the comparison of results shown in figure 3a and 3b, described in the next section). Then first order only water-layer multiple reflections wlM 1 can be nicely identified by iterating the previously described SRME prediction and subtraction procedure on D 2 subset, as sketched in figure 1a. By eliminating higher order water layer multiple reflections, we obtain a new subset D wlm1 that contains only the recorded multiple events that are reflected only one time over water-bottom (from source side). Thus, multiple estimation iteration allows to isolate a single multiple generator (i.e., water- bottom): migration of D wlm1 naturally avoids any cross-talk which degrades the final image, and no post-imaging cross-talk attenuation needs to be applied. A simple source geometry transform allows to perform the migration of D wlm1 multiple reflections without the need of any modification of standard migration tools, analogously to “receiver mirror migration”. Source geometry is modified accordingly to the ray-paths of the first-order multiples traveling in the water-layer. A two-stage mirroring is performed (Fig. 2b): the first mirror moves the water-bottom from its original position to the reverse position L 1 (in the z-negative half- plane); then second mirror moves the sea surface from its original position to a symmetrical position L 2 in respect of the ghost water bottom L 1 (that is assumed to be planar – constant dip, at least locally around source position), allowing to compute virtual source position S’ (lying on mirrored sea-surface L 2 ). Migration velocity model is then extended for negative depths with constant water velocity, and then any standard migration code can be applied. Source-side double mirror transform can be obviously combined with receiver-side mirrored migration, to widen subsurface illumination. It must be noted that in this case two different velocity models (source-side: v(-z)=V water , receiver- side: v(-z,x,y)=v(z,x,y) ). Fig. 2 – a) receiver mirror migration geometry: a new acquisition geometry is built by reverting receiver depths; b) double mirror migration geometry: in addition to the receiver depths transform, the source location is modified unfolding water-layer ray-path.