Accelerating Seismic Modelling and Inversion by Exploiting Wavefield Smoothness

Seismic imaging is at the core of the hydrocarbon upstream market. Given a surge in acquisition datasets with campaigns exceeding hundreds of thousands of shots, the corresponding modelling of wave propagation through complex media poses a formidable computational challenge. In the past, 2D acquisitions and 2D modelling represented a common approximation which decreased the computational cost by orders of magnitude compared to 3D methods. Nowadays, most campaigns and the complexity of structures (e.g. subsalt) are inherently 3D such that the validity of the 2D approximation is questionable. We propose a new method, which propagates waves in a 3D domain, but at the cost of a 2D method. This is achieved by a multipole expansion for axisymmetric (2.5D media). A corresponding code by Tarje Nissen-Meyer is well established in global seismology; this project aims at applying the method for local scales. This will include rewriting parts of the meshing, and adding absorbing boundaries, whereas the bulk of the method will remain unchanged. We will extensively test this new approach against reference solutions, and apply it to synthetic data such as the SEG/EAGE salt and overthrust models. Relations to the seismic industry exist as well, such that we shall eventually validate it against actual data. The code is orders of magnitudes cheaper than 3D modelling, and thus will allow for rapid imaging. A second component of this work will tie the new methodology to true 3D modelling in a hybrid approach. This approach will be based on "exact boundary conditions", which have recently been shown to work excellently by Prof. Johan Robertsson and co-workers at ETH Zurich. The joint framework of axisymmetric and hybrid modelling shall be of wide interest to the general fields of seismic modelling, imaging and inversion.

This project falls within the "Earth resources" research area.