Understanding and Analysing Seismic Images - Insight through Appropriate Modelling

Seismic images are often provided to interpreters as the only element to work with, besides well logs. However, the possible geological complexity of both overburden and reservoir, as well as survey, wavelet, and processing are all parameters affecting seismic images. An ideal seismic modelling would be using full-waveform methods to generate complete synthetic seismic records, then process the latter in the same manner as the observed data. But it is a complex and time-consuming task, impossible at production and routine levels. Interpreters often use the simplest method of all, i.e., “trace modelling” or “1D convolution”, but this corresponds to strong, often forgotten, model assumptions: homogeneous horizontal layers. Using a 3D-spatial ray-based prestack convolutional method allows instead accounting for many of the above-mentioned effects, at a similar cost than 1D convolution, but in a flexible and more complete manner, permitting sensitivity tests on various parameters. This would help interpreters better understanding what seismic can/cannot see and assessing their results in each specific case.