M-33 Migration Velocity Analysis in Complex Media: Application to the Marmousi Data Set

We demonstrate a method for 2D velocity estimation in complex media characterized by triplicated ray fields. Due to multipathing, Migration Velocity Analysis (MVA) principle has to be revisited, requiring in particular the use of Common Angle Gathers (CAGs) instead of the classical common offset panels. For the inversion itself, locally coherent events are automatically picked in the prestack depth migrated volume and defined by their positions and two slopes. The velocity is estimated by optimizing the flatness of these local events in the Common Image Gathers (CIGs) via ray tracing. We present the results obtained on the 2D synthetic Marmousi data set and show the possibility of an efficient inversion using an automatic picking on selected CIGs and CAGs and a local optimization algorithm associated to a multi-scale approach. Introduction – The estimation of the velocity model required for prestack depth migration is still a critical step in the depth imaging process, especially in complex media. As MVA iteratively combines velocity estimation and migration, it is important for the velocity model to be consistent with these two dependent steps: MVA based on Kirchhoff migration should be coupled with a method providing smooth velocity models. Such a condition is satisfied in the method developed in [2]. We defined in this paper a locally coherent event everywhere a migrated gather displays lateral continuity over a few traces and pointed out that a single locally coherent event picked in a CIG contains information on the velocity model (value of the local slope). We also explained in 2D how to use this information to update the velocity model. This method has been successfully applied on a 2D real data set, using automatic picking [3]. Lateral variations of the velocity model were retrieved after inversion, providing well focused images. However, media with triplicated ray fields (due to very high lateral variations of the velocity model) can not be directly handled. We first shortly analyze why specific problems arise in complex media and present how to overcome these difficulties. We then test the modified method on the 2D Marmousi data set, known for its high complexity [10]. MVA in complex media – Whatever the MVA process chosen to update the velocity model, it relies on the general idea that CIGs should present flat events if the exact velocity model is used for migration: all illuminated points are indeed repositionned after migration at their exact location in the depth domain, which does not depend on the acquisition parameter such as the offset. However, this well known statement fails in complex media where common offset migration generates spurious effects, in particular kinematic artifacts [9;11]. The reason for failure may be simply understood as follows: the mapping between the data (time) domain and the depth domain is not uniquely defined due to multipathing in the ray field. A convenient way to overcome this difficulty is to modify the definition of CIGs: instead of migration at constant offset, migration at constant diffracting angle [5] should be preferred, as it provides flat events for the exact velocity model [11]. Such panels are obtained by sorting at each position in the target the contribution of traces, considering the diffracting angle. Once the domain has been modified, we still need to adapt MVA as defined by [2]. Tomographic Migration Velocity Analysis – The method developed here works in the prestack depth migrated domain, defined for 2D data by a 3D migrated volume , where denotes the position on surface, the depth and the diffracting angle. The estimation of velocity is based on flattening locally EAGE 63rd Conference & Technical Exhibition — Amsterdam, The Netherlands, 11 15 June 2001 coherent events in this domain, treated as uncorrelated events (i.e. not attached to a particular reflector). For