In a Tertiary basin offshore China, we used "Colored Inversion" (Lancaster and Whitcombe, 2000) to study the distribution of sand and shale in the subsurface. The geological complexity and the consequent borehole and seismic data inconsistencies in the area necessitated the application of colored inversion instead of other, more preferred, inversion methods. Colored inversion was used in a multi-attribute hybrid-inversion workflow to generate relative acoustic and shear impedances and density volumes so reservoir properties could be determined through stratigraphic interpretation. We discuss the results and the lessons we have learned.
Lancaster and Whitcombe (2000) presented colored inversion (CI) in a heuristic approach for inversion of seismic data into relative impedances. In CI, a single operator convolves with the seismic traces to produce relative impedances. The CI operator has amplitude spectrum that maps the mean seismic spectrum to the mean log-impedance spectrum and has a constant phase of -90°. The impedance logs are incorporated in the creation of the CI operator based on the assumption that the gross spectral form of impedance logs from wells in any given field is constant. Although algorithmically crude, the method has the appeal of simplicity and quickness of application. No explicit wavelet extraction or spectral shaping, which are often very complex processes, is needed. We found the method useful in the offshore China study area.
The basin consists of a Tertiary fluvial-lacustrian system draped over Ordovician horst blocks. The depositional system is modeled as an interconnected series of channels, levies, and point bars. Many thin gas- and oil-bearing sand zones constitute the shallow reservoirs. The primary inversion-study objective was to delineate these sands and find their interconnectivities, starting very shallow at the upper Tertiary, to the top Ordovician unconformity.
Although several exploratory wells and some developmental wells exist in the area, the quality of the logs was poor--the log-responses were not consistent with each other, and or with the corresponding seismic data. The wavelets extracted through well ties varied significantly from location to location. Also, the quality of seismic data was inappropriate for a model-based inversion because of the presence of numerous normal faults associated with the horst blocks. Under the circumstances, we concluded that inverting for relative impedances and density rather than for their absolute values, and using a simplified, unconstrained sparse-spike inversion (Oldenburg et al., 1983) would be preferred. We used CI, which was known to work significantly better than the conventional recursive inversion and benchmarked well against the unconstrained sparse-spike inversion (Lancaster and Whitcombe, 2000).
The application provided acceptable relative acoustic and shear impedances for stratigraphic interpretation, delineating reservoir sands in the area. We also carried out a detailed and integrated reservoir description (IRDTM) based wavelet processing and spectral shaping process described by Poggiagliomi and Allred (1994) on the Kirchhoff Prestack Time Migration (PSTM) stack and compared the results with those of CI.
