Technology focus: Multi-layer tomography for seismic velocity model building

Added: 28th May 2020 by CGG

Many regions of the world exhibit complex layered geology with significant seismic wave velocity and anisotropy contrasts that present a major challenge for standard grid based tomography methods. Using seismic reflection tomography to globally update a model that contains velocity contrasts can introduce significant errors into the inverted model near the layer boundaries, such that the positions of the layer boundaries in the model become uncoupled from the corresponding reflector positions in the updated seismic image.

To help overcome these limitations, layer stripping approaches are used to enable accurate repositioning of velocity and anisotropy boundaries. The model is divided into a set of layers defined by the major seismic velocity boundaries (these usually correspond to major geological boundaries but not always), which are updated iteratively in a top?down manner. For the inversion of a particular layer, the starting velocity and anisotropy from the target layer are allowed to “flood” through the boundary position and residual velocity error from imaged (migrated) data in depth (as in depth of the reflectors below the surface) is used only from the layer to be updated. Following the tomographic update of the chosen layer velocities, the data are re-imaged and the correct position of the base of layer boundary reinterpreted prior to final calibration to the actual depth of the layer boundary (horizon) as observed in available borehole (well) data.

For a typical North Sea project, as many as six iterations of layer stripping may be required to update all the major velocity and anisotropy contrasts in the geology. This layer stripping approach is time consuming and is also prone to serious velocity errors. Since the method only treats one layer at a time in a top?down manner, it precludes any intercommunication between connected model layers during the inversion process. In addition, it is common practice to “freeze” a layer once it has been updated. While this avoids corrupting the depth position of carefully interpreted layer boundaries, it also results in the propagation of any unsolved often high-frequency velocity errors into the deeper model layers resulting in instabilities in the deeper update.

Multi-layer tomography

Multi?layer tomography is an extension of non?linear slope tomography (Guillaume et al., 2008; Montel et al., 2009). It uses a new hybrid model format that uniquely defines the velocity and anisotropy values for each model layer as a b-spline mesh while also carrying the precise information for the layer boundaries as horizons. The non?linear inversion scheme performs kinematic de?migration and re?migration of measurements of the residual velocity error and the positioning of the layer boundaries.

This allows the layer boundaries to be repositioned by map migration during the inversion process, preserving their travel-time. This means the traditional layer stripping workflow can be discarded and all layers in the model updated simultaneously with information from all layers contributing to the global inversion scheme, resulting in a significant improvement in overall model stability. Furthermore, the integrated horizon information in the hybrid model allows the update of each model layer to be uniquely parameterised to achieve the best possible inversion result. Layers no longer need to be frozen during the inversion process, as the method allows any layer to be updated during model building without compromising the result. In addition, since the entire initial model is updated during each pass of multi?layer tomography, improvements to the imaging at deeper reservoir levels can be monitored at all stages of model development. A final benefit is that the imaging model used to image the data can be different from the initial model for the inversion. This allows measurements of velocity error and interpretation of horizons to be made on data imaged with a legacy model which gives a good image but which has been generated by global grid-based or layer stripping approaches, contains too much high-frequency detail or is calibrated with out-dated well data so is not suitable as a starting model.

Improvements in model accuracy and stability from using multi-layer tomography naturally translate into improved seismic images with less reflector distortion and greater confidence in the image. Discarding the traditional layer-stripping workflow and returning to a global approach for complex layered geology also yields significant efficiencies in model building and interpretation effort.

The image shows a comparison between the result of seven iterations of model building using layer stripping (left) and three iterations of global multi-layer tomography (right).