The scientific-practical сonference “Seismic exploration in Siberia and beyond” was hold from November 08 to 11, 2022.
The conference was scheduled in a hybrid format (offline and online).
On November 08, the following report was presented at the "Field work and data processing technologies" section of the "Seismic Exploration in Siberia and Beyond" conference:
How and why to model wave fields in seismic acquisition.
The speaker: Ovchinnikov Kirill, Geophysicist, Seismotech Ltd.
The authors are: Finikov Dmitry, Ovchinnikov Kirill, Silaenkov Oleg, Shalashnikov Andrey
The wave fields recorded in seismic surveys include waves of a variety of nature. Wave propagation in a medium is described by tensor wave equations of the most general form (and in inelastic media with absorption, the type of equation is still under discussion). Strictly speaking, the most information a researcher can extract from the observed data is the parameters of these equations relating to the interior points of the medium. This is what is called the inverse problem of seismic acquisition. Finding out the relationship between these parameters and the real properties of the substance of which the Earth's strata are composed is the task of interpretation.
Solving the inverse problem in such a general formulation is not possible not only because of its complexity, incorrectness and structural instability, but also because of the limitations of the system of real observations, the volume and quality of the input data.
In addition to establishing links between the estimated elastic parameters and the material composition of rocks, the task of the interpretation stage of seismic data analysis is to obtain a detailed description of the medium in terms of the parameters of the wave equation. Here the important role of modeling seems obvious - one cannot solve the inverse problem without knowing how to solve the forward task.
Mapping the set of estimated parameters into the original wave field is a necessary, although insufficient condition for the correctness of the solution. We must admit that this criterion is not always checked, although its importance is generally recognized.
Traditionally, the processing results in obtaining of various images of the medium, which are transmitted to the interpretation. It is important that these images contain the correct dynamic parameters of the waves and are not complicated by noise. It turns out that this can also be achieved through the use of model wavefields. The paper will demonstrate the use of modeling to compute reference gathers and the application of these gathers to both correction of post-migration amplitude distortions, static shift estimates, and noise suppression. In addition, the processing use special noise-wave modeling algorithms used to suppress them.
There are many different ways to model wavefields. The paper will focus on the modeling method of layer-by-layer recalculation by integral field transfer operators, on the method of inverse migration transformation. The application of grid schemes to solve direct problems is discussed in the context of hybrid solutions obtaining.
Finally, the prospects for modeling in machine learning are quite evident, although such applications of direct problems seem to be too expensive for the time being. However, the discussed algorithms are suitable for performing computations in a limited target domain, which allows us to hope for a successful application of the methods in this class of problems as well.