Dynamic Plane-wave LSRTM

When high computational efficiency is in demand, LSRTM can be performed with the dynamic encoding approach (Krebs et al., 2009; Schuster et al., 2011), where one plane-wave gather is used for each iteration and the ray parameter $p$ (corresponding to surface shooting angle) is dynamically changed from one iteration to another. Figure (d) shows the LSRTM image with dynamic encoding after 31 iterations. It has resolution comparable to Figure (c) but contains more noticeable artifacts (see Figure (d) and (d)).

Figure 3.14: The migration velocity model for the field data test, obtained by full waveform inversion.

Figure 3.15: The migration images obtained by: (a) conventional shot-domain reverse time migration, (b) plane-wave reverse time migration, (c) plane-wave least-squares reverse time migration and (d) plane-wave LSRTM with dynamic encoding. The blue and red boxes indicate the areas for zoom view.

Figure 3.16: The zoom views of the red boxes: (a) conventional shot-domain RTM, (b) the plane-wave RTM, (c) the plane-wave LSRTM and (d) the plane-wave LSRTM images with dynamic encoding.

Figure 3.17: The zoom views of the blue boxes: (a) conventional shot-domain RTM, (b) the plane-wave RTM, (c) the plane-wave LSRTM and (d) the plane-wave LSRTM images with dynamic encoding.

Figure 3.18: The misfit vs iteration number curve for plane-wave LSRTM shows fast and stable convergence even when the velocity is not completely accurate.

Figure 3.19: The common image gathers extracted from the plane-wave RTM image of the field data.

Figure 3.20: The common image gathers extracted from the plane-wave LSRTM image after 30 iterations for the field data test.