doi:10.1111/j.1365-2478.2008.00681.x

Geophysical Prospecting, 2008, 56, 213–227

2D PP/PS-stereotomography: Application to a real 2D-OBC dataset ∗

M. Alerini1, ,† , G. Lambar´e1,‡ , R. Baina2 , P. Podvin1 and S. Le B´egat1, 1 Ecole ´

∗∗

des Mines de Paris, Fontainebleau, France, and 2 OPERA, Pau, France

Received May 2007, revision accepted September 2007

ABSTRACT It has been shown on an ‘ideal’ synthetic dataset that PP/PS-stereotomography can estimate an accurate velocity model without any pairing of PP- and PS-events. The P-wave velocity model is first estimated using PP data and then, fixing this velocity field, the S-wave velocity is estimated using the PS data. This method needed to be evaluated further and we present here the first application of PP/PS-stereotomography to a real dataset: the 2D East-West Mahogany OBC line (Gulf of Mexico). We are here confronted with data which do not fit our working assumptions: coherent noise (due to an approximate separation of PP- and PS-events and some remaining multiples), probably some anisotropy and 3D effects. With a careful selection of the stereotomographic picks, which allows one to decrease the effect of the picked coherent noise by the automatic picker, our application can demonstrate the relevance of our approach in the upper part of the profile, where anisotropy and 3D effects might be low. We can thus estimate, without any pairing of PP- and PS-events, a velocity field which provides not only flat common image gathers, but also PP- and PS-depth migrated images located at the same positions. For the deeper part of the profile, a significant shift in depth appears. In addition to possible anisotropy, 3D effects and a more complex velocity field (‘salt body’), this is due to the quality of the PZ- and X-components profiles: The PZ-component profile where the PP-stereotomographic picking is performed, is polluted by conflicting converted or multiple events and the X-component profile, where the PS-stereotomographic picking is performed, is highly noisy. This study emphasizes the need to develop accurate selection criteria for the stereotomographic picks.

INTRODUCTION The use of converted waves in reflection seismic imaging can be a significant improvement. Indeed P- and S-reflected waves bear complementary information about the elastic properties of the medium. It concerns both the reference velocity model (P- and S-wave velocities) and the short wavelength components of the elastic parameters describing the model (reflectivity). The propagation of S-waves can be advantageously used for structural imaging, as for example in gas cloud contexts

∗

E-mail: [email protected] at: SINTEF Petroleum, Trondheim, Norway ‡ Now at: Compagnie G´en´erale de G´eophysique, Massy, France ∗∗ Now at: Geophysical Consulting, Pau, France † Now

 C

2008 European Association of Geoscientists & Engineers

where the S-wave propagation is far less affected by gas than P-wave propagation (Thomsen et al. 1997; Gaiser et al. 2001). Concerning the short wave-length components of the elastic parameters, in addition to AVA/AVO analysis of converted waves (Garotta and Granger 1987), linearized inversions of PP- and PS-diffracted/reflected wave fields have been proposed (Tarantola 1986; Jin et al. 1992; Nicol´etis et al. 1997). They allow a better lithological identification (Polskov et al. 1980), a better fracture characterization (Kendall and Kendall 1996; Granger et al. 2001) and better rock property analysis (Kendall et al. 1998b). In marine environments, with the development of 4component (4C) receivers laid on the sea bottom, converted waves have received increasing attention from the exploration geophysics community (Garotta, Granger and Dariu 2000). These 4C are one hydrophone recording the pressure and

213

214 M. Alerini et al.

three geophones recording the particle displacement (or velocity). The most commonly used acquisition system is the so-called OBC, for ocean bottom cable (Caldwell 1999). A cable containing 4C receivers is laid on the seabed and a vessel pulls a source at the sea surface. In recent years, the use of S-waves through PS-conversions in marine reflection seismics has shown better quality results than the use of PP-data in some configurations. For example, real improvements have been reported when imaging through a gas cloud (Thomsen et al. 1997; Granli et al. 1999). Processing of converted waves requires a significant adaptation of the sequence such as wavefield separation, converted waves migration or velocity model building (Stewart et al. 2002, 2003; Bo¨elle and Ricarte 2003). An additional difficulty may arise when the analysis needs a matching of PP- and PS-events. In depth imaging it means determining P- and Swave velocity models migrating PP- and PS-data at the same depth and the same horizontal position. Considering the poor conditioning of the estimation of the reference velocity models, the matching of PP- and PS-images is generally insured by pairing a priori PP- and PS-reflected events. In standard traveltime tomography, this seriously complicates the picking step (Gerea, Nicoletis and Rakotoarisoa 2001; Stopin 2001; Broto et al. 2003), even if some practical solution can be found (Foss, Ursin and de Hoop 2005). Alerini et al. (2007) presented an extension of stereotomography to converted PP/PS-waves. A 2D approach was developed for inverting P-wave and S-wave isotropic reference velocity models in a nearly fully automatic way. They showed that for a synthetic data example and optimal preprocessing and picking, PP/PS-stereotomography allowed one to invert, without any pairing of events, both P- and S-wave velocity models, providing PP- and PS-migrated images at the same positions. It appeared, thus, important to further test the approach in a real data context, where coherent noise, 3D effects and anisotropy could alter the results. In the present paper we first recall some basic ideas of PP/PSstereotomography and then present the first application on a real dataset, the East-West 2D-4C OBC Mahogany line (Caldwell et al. 1998). The quality of the inverted velocity model is assessed in terms of flatness of common image gathers and in terms of focusing depths of PP- and PS-reflections. Even though the automatic picking used for stereotomography is much easier than the one used for picking global events, it remains the bottleneck of the method. Indeed, we use here an automatic approach which can be inaccurate in (coherent) noise context. We emphasize here the practical difficulties of the application of PP/PS-stereotomography and in particular,

 C

the difficult selection of the stereotomographic picks after the automatic stereotomographic picking (Billette et al. 2003).

PP/PS-STEREOTOMOGRAPHY Stereotomography is a slope tomography which was introduced, developed and applied by Billette and Lambar´e (1998). Alerini et al. (2007) presented an extention to converted PSwaves. We shall here just briefly recall the basis of PP/PSstereotomography. Stereotomography is based on the use of locally coherent events for estimating velocity macro-models from seismic reflection data. Such locally coherent events are defined by their positions at the surface (source and receiver for example, s, r), their two-way traveltimes, T sr , and their two slopes in the common source and common receiver gathers (ps , pr ), which are the local tangents to the locally coherent events. For PP/PS-stereotomography, those events have to be identified as primary PP- or PS-reflections/diffractions. A PP/ PS-stereotomographic dataset consists then of NPP PP-stereotomographic picks and NPS PS-stereotomographic picks:   NP S NP P , (s, r, ps , pr , Tsr ) j=1 d = (s, r, ps , pr , Tsr )i=1 . (1) This data will allow us to invert a model defined by  NP S NP P MP , (X, βs , βr , Ts , Tr ) j=1 , (C P )k=1 , m = (X, βs , βr , Ts , Tr )i=1  MS (CS )l=1

(2)

M

MP and (CS )l=1S denote respectively the MP and where (C P )k=1 MS parameters describing the P- and S-wave reference velocity models, X the position of the diffraction/reflection point, (β s , β r ) the two shooting angles and T s and T r the two oneway traveltimes, for the rays propagating from X towards the source in the P-wave velocity model and towards the receiver in the S-wave velocity model, respectively. In PP/PS-stereotomography we can first invert the PPstereotomographic dataset into the P-wave reference velocity model. This is done by standard stereotomography. We then fix the P-wave reference velocity model and invert the PSstereotomographic dataset into the S-wave reference velocity model. PP/PS-stereotomographic optimization was described in Alerini et al. (2007). We would like to focus here more on the stereotomographic picking, which is, although easier than the picking in traveltime tomography, definitively the main bottleneck in our later application to real data. We use an automatic picking tool (Billette et al. 2003) the principle of which is to compute the slope of the locally coherent events by local slant stack and the coherency of this events

2008 European Association of Geoscientists & Engineers, Geophysical Prospecting, 56, 213–227

2D PP/PS-stereotomography 215

by semblance of the envelope of the trace  2    |x−x0 |