Seismic inversion is routinely used in the delineation of drillable leads and prospects; however the algorithms and techniques that are most widely used today suffer from a number of shortcomings.  In this paper we demonstrate how a new inversion technology, which utilises per-facies compaction trends and rock physics models within a Bayesian pre-stack inversion frame-work, is able to deliver reliable and repeatable estimates of seismic facies and corresponding absolute rock properties from seismic data without the requirement for direct well calibration and model building. The joint impedance and facies inversion (Ji-Fi) technology is the result of a 4 year R&D collaboration between Ikon Science and CSIRO, with financial and technical support from Tullow Oil plc.

Conventional inversion approaches fall broadly into two categories, namely ‘absolute’ and ‘relative’.  Absolute inversions typically require a starting low frequency model, and it is this aspect which introduces the main source of user bias and error, because to construct the ‘correct’ initial model, you must be quite certain of the geology (N:G, porosity, Sw, etc.) at each seismic trace location. Relative inversions are often preferred in exploration settings, precisely for the reason that they do not contain bias from a starting model (although a reasonable estimate of the ultra-low frequency trends is required to ensure reasonable determination of the underlying reflectivities).  Whilst relative inversions provide an unbiased estimate of elastic properties, they only contain information in the seismic bandwidth and therefore cannot easily be used to estimate absolute rock properties such as porosity, etc.

The joint inversion scheme proposed in this paper provides seismic facies consistent with absolute rock property estimates from post or pre-stack seismic data, without the requirement for a conventional low frequency model. Instead, the inversion framework utilises, per-facies, rock property compaction trends and derives the low frequency model during the inversion process.

Upper Palaeocene log data from a regional study of 30 wells located in Central North Sea, covering an area of 35,000 sqkm, were analysed using a clustering analysis to determine what elastic log responses could be differentiated and to what geological facies they relate. Based on this analysis, the log responses were classified into soft shale, hard shale, brine and oil bearing reservoir. These facies categories, in conjunction with the corresponding elastic logs (Vp, Vs and Rhob), were used to generate per-facies compaction trends, referenced to the mudline (figure 1).

These trends were used to build background models per facies by hanging the trends from the seabed horizon at each trace location in the 3D survey area. For each survey statistical wavelets were generated and used in the inversion. The objective of the inversion was to predict both the presence of Palaeocene sands and their corresponding hydrocarbon distributions within the Forties Formation in the Forties field (Rose et al., 2011) and the Palaeocene channel sands from the Upper Balmoral Member of the Montrose Group in the Brenda field (Jones et al., 2004).

Presented at NCS Prospect Fair 2015 by Ana Somoza.