Summary
Shell conducted its first dual-well 3D DAS-VSP survey concurrently with an OBS survey in a deep water environment in the Gulf of Mexico in 2012. This survey produced about 40M picks of the first arrival times (FAT) which were used to diagnose and update velocity models for improvement of both borehole and surface seismic images of subsurface structures. We developed a procedure to use the VSP-FAT to diagnose the velocity models derived from surface seismic surveys and monitor the velocity updating process. The method first was used for selecting a suitable initial velocity model. After the traveltime tomography inversion of FAT, this diagnosis approach was applied again to the updated VTI-inversion model to ensure the velocity updating effort is on the right track. We used the Absolute and Relative Misfits (AM & RM) and apparent velocities to quantify the velocity model uncertainties as functions of depth, azimuth, and offset. Both DAS-VSP data at two wells and OBS data were migrated with the initial VTI velocity model and the updated VTI-inversion model. It is found that both borehole and surface seismic images generated with the VTI-inversion model are improved from those obtained with the VTI-initial model, especially for the seismic amplitudes at a target event.
Introduction
Distributed acoustic sensing (DAS) systems using the fiber optical cable have played important roles in the borehole seismic monitoring and imaging. In 2012, Shell simultaneously acquired its first 3D dual-well DAS-VSP (Vertical Seismic Profiling) with an OBS (Ocean Bottom Sensor) survey (Mateeva et al., 2013) in a deep water environment in the GOM (Figure 1). This DAS-VSP dataset, with over 40M traces, provides rich borehole seismic data to diagnose and update the surface seismic velocity models and examine the effects from model updating on improvements of both borehole and surface seismic images of subsurface structures. This study is a complementary to the earlier efforts of 3D OBS-VSP surveys in the GOM (Hornby and Burch, 2008; Wu et al, 2011) with the limited numbers of fiber sensors.
