Least-squares reverse time migration for the Cascadia ocean-bottom dataset

Pre-processing

In our 2D study, we extract a 2D shot line from directly above the OBS line. Following the processing step in Dash et al. (2009), we apply a gapped deconvolution with a 12-ms gap length and 300-ms filter length to suppress the source bubble. Out of the ten OBS receivers, only six were used for final imaging. These receivers were selected according to their data quality. Three receivers were rejected due to large tilt angles, and one receiver was rejected because it was saturated with noisy traces.

Since we only use the down-going wavefield, only acoustic separation above the seafloor is needed. We used the adaptive decomposition method of Schalkwijk et al. (1999).

These data are bandpassed between 5 and 45 Hz to avoid dispersion in the time-domain finite-difference calculation. These data before and after pre-processing are shown in Figure 5.

dataCom
Figure 5. (a) A hydrophone common receiver gather (CRG) before pre-processing and (b) down-going data after pre-processing.

Figure 6 (a) shows the migration velocity used for our RTM and LSRTM. The velocity ranges from 1480m/s in the water column to 1820m/s in the sediment layer. The BSR lies at an approximate depth of 1500-m with a velocity inversion.

Least-squares reverse time migration for the Cascadia ocean-bottom dataset