Synthetic Example: Marmousi2 Data with a Fixed Spread Survey

The plane-wave LSRTM is applied to a synthetic dataset for the Marmousi2 model. The original model is modified to be the size of 8 km $\times$ 3.5 km with a 10 m grid interval (Figure (a)). Synthetic data are generated with a fixed spread geometry where 801 shots are excited with a 10 m offset interval at the depth of 10 m. Each shot is recorded with 801 receivers with a 10 m receiver interval. A Ricker wavelet with a 20-Hz peak frequency is used as the source wavelet, and the record length is 8 sec in time. The true data are generated by a pseudo-spectral modeling method (Kosloff and Baysal, 1982). Conventional shot-domain RTM is first applied to the dataset with a smooth migration velocity (Figure (b)) to give the image shown in Figure . In the shallow part, there are strong artifacts associated with the source positions and the ocean bottom even after high-pass filtering. The amplitudes of the deep reflectors are very weak.

Figure 3.2: (a) Modified Marmousi2 model and (b) the smooth migration velocity model. The migration velocity is smoothed by a triangle smoothing filter with a window length of 100 m to get rid of the fine scale structures.

Figure 3.3: The conventional shot-domain RTM image for the Marmousi2 model.