Probably the most frequently asked question in reservoir characterisation is “do advanced and time-consuming techniques predict outcomes better than simpler ones?” Often the main stumbling block is that the more advanced our models become, the more remote they get (and are allowed to get) from the data. A good model needs to satisfy the requirement that at least some deterministic rules can be applied to it. At worst, taking the ‘advanced' route is an admission that the framework was never really there - hence the scepticism over some of the results.

To illustrate this particular problem, an example is presented of a North Sea oil field which has undergone a step change in its reservoir description, The field is a submarine fan sandstone development of Upper Palaeocene age. The principal reservoir uncertainty is net sand distribution, which is related to the stacking pattern of the sand bodies and the interactions between them.

The pre-development approach used to characterize the distribution of net sand employed “advanced” geological modelling techniques almost exclusively. These were applied at two scales. Large-scale net to gross variations (the channel framework) were described using a “template” for sand body disposition, derived from well-based sedimentology and from field studies of the disposition of analogue reservoirs. Smaller scale (intra-channel scale) variations were quantified geostatistically; objects representing different facies types were assigned properties taken from sedimontological and petrophysical data, supplemented with statistical data from published analogues, and then distributed in a geostatistical model.

The “advanced” model suffered from the problem that its own framework was model-driven. However at that time the most obvious medium for bridging this gap, the 3D seismic, was considered to be unresolvable except for picking top reservoir.

Disappointing early development drilling results inevitably led to a reappraisal of the “advanced” model. Of a variety of possible approaches, including more advanced modelling, the solution chosen was to go back to the weak point, the deterministic framework, and attempt to apply a more rigorous and revealing seismic interpretation. With time a new interpretation evolved around a range of integrated techniques; these included relative amplitude and AVO reprocessing, AVO/AVA interpretation using partial seismic stacks, seismic amplitude modelling, seismic facies mapping, use of time, horizon and coherency displays, and implementation of an acoustic impedance inversion. Most of the original issues around seismic recognition and resolution became resolved through this effort and a channel architecture was able to be built up. In addition, using the new framework a method for accurately mapping average net to gross variations within the main channels was devised using seismic attributes.

The new description was tested with additional drilling, and resulted in a step change in the outcome of well predictions compared with the original model.

The difference between the approaches lay in the presumptions about the type of “advanced” characterisation required for this field. The solution lay in the more advanced application and integration of relatively traditional deterministic techniques, and making better use of the data. This was opposed to methods such as object modelling and geostatistics which were perceptually more sophisticated and attractive but which were data-limited in this case, and as a result yielded poorer results.

AAPG Search and Discovery Article #90937©1998 AAPG Annual Convention and Exhibition, Salt Lake City, Utah