A Sequential OBS–Streamer Full-Waveform Inversion Workflow for Marine Velocity-Model Building: Synthetic and South China Sea Case Studies

Marine velocity-model building using full-waveform inversion (FWI) is strongly affected by acquisition geometry because different surveys sample different parts of the wavefield. In this study, we investigate a sequential OBS (ocean-bottom seismometer)–streamer FWI workflow that uses the velocity model obtained from OBS FWI as the starting model for subsequent streamer FWI. Synthetic experiments are first performed to compare the spatial sensitivity of OBS and streamer geometries using source–receiver and single-shot sensitivity kernels, followed by comparisons among OBS-only, streamer-only, and sequential inversions. The results indicate that the long-offset OBS data provide velocity constraints over a broader depth range, whereas the dense streamer acquisition mainly improves shallow and intermediate-depth structures. During sequential inversion, the large-scale velocity variations introduced by OBS FWI are largely preserved, while additional refinements are provided by the streamer data in regions with stronger shallow sensitivity. In the synthetic test, sequential OBS–streamer FWI reduces the relative L2 velocity error from 5.782% to 1.600% in the shallow zone and from 12.301% to 10.660% in the intermediate-to-deep zone, relative to the initial model. The workflow is further applied to colocated OBS and streamer data from the South China Sea, and reverse-time migration using the sequential model produces improved reflector continuity and event focusing. These results demonstrate that the sequential workflow improves velocity-model accuracy in the synthetic test and imaging quality in the field example by incorporating complementary constraints from OBS and streamer data.