GNGTS 2013 - Atti del 32° Convegno Nazionale

ITERATIVE DECONVOLUTIONS TO COMPENSATE WAVELET STRETCHING ON 4 TH ORDER TRAVELTIME KINEMATIC E. Biondi 1 , E. Stucchi 1 , A. Mazzotti 2 1 Department of Earth Sciences, Geophysics, University of Milan, Italy 2 Department of Earth Sciences, Geophysics, University of Pisa, Italy Introduction. Normal move-out (NMO) correction applied to common-midpoint (CMP) gathers is needed for stacking, AVO analysis and for other seismic processing steps. The result of this operation are CMP gathers where the offset and velocity effects have been removed and thus all the traces, recorded at variable source to receiver offset, simulate a zero-offset kinematic. The NMO corrected CMP gathers are then employed for AVO analysis and for building stacked images. However, it is well known that the traditional sample by sample NMO correction introduces the stretching of the reflected wavelets (Buchholtz, 1972). These distortions are caused by the non-parallelism of the local traveltime of each reflected event, or, in the frequency domain, as the consequence of the nonphysical energy changes introduced by the non-stationary time shifts applied. Additionally, there are other drawbacks of the standard NMO correction, such as the partial duplication of the recorded events and the time inversion of the samples of a reflection (Masoomzadeh et al. , 2010). Typically, in order to avoid such negative effects, which may compromise the quality of the stacked images, a mute function is applied to the distorted part of the corrected CMP gathers. Yet, the application of the mute function limits the stacking process to only the near vertical reflections. For example, in case of long-offset acquisitions the wavelets used in the stacking process are constrained to a limited portion of the recorded data and therefore the muting reduces the exploitable information provided by the wide-angle wavelets. As a matter of fact, there are many instances where long-offset acquisitions are crucial, such as sub-basalt exploration or seismic undershooting. In these cases the far offset reflections on the CMP gathers are necessary for imaging purposes (Colombo, 2005). Since in exploration seismology it is now common to acquire and process data with more than 10 km of offset, a nostretch NMO correction procedure is of interest. Many authors have proposed alternative methods (e.g. Perroud and Tygel, 2004; Masoomzadeh, 2010). However, these techniques are affected by limitations, in particular they do not account for interfering reflections and offset variations of the waveforms, which are commonly present in wide-angle CMP gathers. Moreover, they may introduce horizontal coherent noise that after stacking can mimic true reflected events. In this paper we propose an extension of the algorithm of the normal moveout through iterative partial corrections and deconvolutions which deals with long-offset data and offset varying waveforms (Mazzotti et al. , 2005). We test our method on a synthetic seismic gather which presents long-offset traces, interfering events and amplitude and phase variations with offset of the reflected waveforms. In addition to these features, we add random noise to the gather and we simulate under-shooting acquisitions using offsets greater than 1000 m (Fig. 1). We also apply our nostretch algorithm to a subset of an offshore marine line (Rocchi et al. , 2007) simulating again an under-shooting pattern by muting the short offset traces. We start illustrating the normal moveout through partial corrections (NMOPC) and then we show the synthetic and real applications. Method. The Normal Moveout through partial corrections can be divided into three phases: wavelet estimation, partial NMO correction and shaping deconvolution. We based our algorithm on the 4 th order traveltime approximation proposed by Taner and Koehler (1969). For the wavelet estimation, we employ temporal-offset windows which slide both in time and offset. The short-offset window is projected along the 4 th order curve corresponding to the central time of the window. These windows enable to select portions of the wavefield where the reflected wavelets can be considered as stationary. The time samples and the traces within each window constitute a matrix which contains reflected and interfering events plus random noise. 17 GNGTS 2013 S essione 3.1