A wavefield domain dynamic approach: Application in reverse time migration

Abstract

This paper proposes a novel technique to handle the wavefield domain involved in the procedures of seismic modeling, reverse-time migration (RTM), and full-waveform inversion (FWI). This method considers that the size of the wavefield domain varies with time, in other words, that it expands concomitantly to the propagation. However, in the geophysical literature, this dynamism has always been neglected as the wavefield domain is constantly considered to be fixed, thus, representing what we call a static approach (SA). This assumption may incur unnecessary use of available computational resources, thereby compromising application performance. Herein, we create a so-called dynamic approach (DA), capable of obtaining truly significant gains in terms of memory consumption and computational time. This new methodology is based on the application of an empirical filter that delimits the wavefront. This filter functions as a window and it is applied at each timestep until the wavefront reaches the model's boundaries, selecting the area where the seismic wavefield exists. This approach tries to approximate the computational domain to the propagation domain in order to obtain valuable computational gains, by eliminating unnecessary work, thus reducing the amount of work needed to perform forward and backward propagation. We compare both approaches using the Pluto model. The seismic data generated from the Pluto model is very large and it was not possible to use the static approach with it relying only on the random-access memory (RAM) of the used hardware. In order to perform the conventional RTM, we implement and compare the effective boundary technique for wavefield reconstruction with the RTM using the proposed dynamic approach. With the dynamic approach, it was possible to perform RTM of a 2D seismic data obtained from the Pluto model using only the RAM of the computational nodes and without the need of reconstruction techniques.