| Summary: | The north-east Atlantic margin (NEAM) is important for hydrocarbon exploration because of the growing evidence of hydrocarbon reserves in the region. However, seismic exploration of the sub-surface is hampered by large deposits of flood basalts, which cover possible hydrocarbon-bearing reservoirs underneath. There are several hypotheses as to why imaging beneath basalt is a problem. These include: the high impedance contrast between the basalt and the layers above; the thin-layering of the basalt due to the many flows which make up a basalt succession; and the rough interfaces on the top-basalt interface caused by weathering and emplacement mechanisms. I perform forward modelling to assess the relative importance of these factors for imaging of sub-basalt reflections. The boundary element method (BEM) is used for the rough-interface modelling. The method was selected because only the interfaces between layers need to be discretized, in contrast to grid methods such as finite difference for which the whole model needs to be discretized, and so should lead to fast generation of shot gathers for models which have only a few homogeneous layers. I have had to develop criteria for accurate modelling with the boundary element method and have considered the following: source near an interface, two interfaces close together, removal of model edge effects and precise modelling of a transparent interface. I have improved efficiency of my code by: resampling the model so that fewer discretization elements are required at low frequencies, and suppressing wrap-around so that the time window length can be reduced. I introduce a new scheme which combines domain decomposition and a far-field approximation to improve the efficiency of the boundary element code further. I compare performance with a standard finite difference code. I show that the BEM is well suited to seismic modelling in an exploration environment when there are only a few layers in the model and when a seismic profile containing many shot gathers for one model ...
|
|---|