Paper submitted to EAGE Annual 2026

Summary

The Mad Dog field is one of the largest producing fields in Gulf of America (GoA). Discovered by bp in 1998, the field commenced production in 2005 and is located approximately 190 miles south of New Orleans, near the edge of the Sigsbee Escarpment. As with many subsalt fields in the GoA, seismic imaging at Mad Dog field is particularly challenging due to the presence of complex salt geometries in the overburden, which strongly distort seismic wave propagation and degrade image quality.

In recent years, elastic full waveform inversion (FWI) has been successfully applied to ocean-bottom node (OBN) data, leading to significant improvements in velocity model building and seismic imaging (Liu et al., 2024). At Mad Dog field, the application of elastic FWI to OBN data resulted in a stepchange improvement in subsalt reservoir imaging (Liu et al., 2023). While the rich low-frequency content of OBN data is well suited for recovering long-wavelength background velocity models, the limited high-frequency energy due to two-way propagation and subsurface attenuation, constrains FWI’s ability to resolve fine-scale reflectivity. In contrast, legacy 3D vertical seismic profile (VSP) datasets retain substantially higher-frequency content due to shorter propagation paths and receiver placement in close proximity to the reservoir (Rollins et al., 2015). This contrast raises a key question - can legacy VSP data be revitalized through an elastic FWI workflow to deliver the high-resolution imaging that OBN data alone cannot achieve?

To assist the continued development of the Mad Dog field, two large-scale 3D VSP surveys were acquired in 2015 and 2018 (Li et al., 2016, van Gestel et al., 2019). A representative geological crosssection (Figure 1a) highlights subsalt faulting as a primary risk to reservoir deliverability, underscoring the need for higher resolution imaging to guide well placement and evaluate connectivity. However, the existing 25 Hz OBN elastic FWI-derived reflectivity (FDR) image (Figure 1b) does not provide sufficient detail for this purpose. To overcome this limitation, we apply elastic FWI to the legacy 3D VSP datasets to finetune the OBN-derived velocity model and recover additional high-frequency detail in both the velocity model and FDR image.