Visualizing the Reservoir Distribution of a West-Central Saskatchewan Heavy Oil Field by Integrating 3D Seismic and Well Log Data

Sabrina Sarzalejo and Bruce Hart
McGill University, Montreal, Quebec, Canada

The Lower Cretaceous Mannville deposits have been described by many authors as very complex due to their fluvial-estuarine origin and the geomorphologic configuration of the Paleozoic terrain over which they were deposited. This complexity makes understanding the reservoir architecture of the Dina-Cummings members (lower Mannville) a difficult task. The heavy oil fields in central Alberta and Saskatchewan are all in these complex reservoirs which need to be understood in order to proceed with the complicated techniques (horizontal drilling, SAGD, VAPEX, etc) required to exploit these deposits.

In an effort to help establish a workflow for characterizing these units, we integrated three-dimensional seismic and well data in order to predict the distribution of facies in our study area. In this poster we show the reservoir model generated by integrating the geological and seismic data as well as the methodology followed.

The Dina-Cummings reservoir is up to 40 meters thick and sits on top the regional Sub-Cretaceous unconformity in direct contact with the Devonian Duperow carbonates. Well logs and core show, from bottom to top, thick amalgamated channel sands (Dina Member) that are overlain by an interval of heterolithic facies which are, in turn, covered in most of the area by a coal approximately one meter thick (Cummings Member). This succession can be interpreted as the product of a transgressive episode that eventually culminated in the deposition of the marine shales of the overlying Lloydminster Member. The thick sands of the Dina –Cummings interval are interpreted as a series of fluvial to estuarine channel sands that are covered by a succession of estuarine point bars, tidal flats and lagoonal sediments that are coved in many cases by a paleosol and the Cummings coal. In the seismic data the reservoir is seen as series of discontinuous reflections that range from chaotic to parallel bedded. A series of channels can be seen cutting the interval at various levels in time slices through the seismic cube. Core data documents the lithofacies that are associated with these channels.

Several wells penetrate the interval of interest but they cannot be correlated easily at the reservoir level because of the complexity of the depositional environments and history of this geological interval. As such, we sought to predict the distribution of lithologies (sand/shale) in the lower Mannville by training a neural network to predict gamma ray logs from combinations of seismic attributes. An interval between the sub-Cretaceous unconformity and a continuous seismic marker within the Rex member (stratigraphically above the Lloydminster member within the Mannville Group) was chosen for this seismic attribute study. Our results allow us to predict the broad-scale 3-D distribution of lithologies, but problems still exist, for example in areas of poor seismic data quality. Another problem is associated with the small contrasts in acoustic impedance between sands and shales that make characterizing the reservoir with conventional p-wave seismic data a challenge. Nevertheless, by integrating our attribute-based lithology predictions with log facies and core descriptions, we have succeeded in generating a better reservoir model than could be produced using logs, core or 3-D seismic data alone.

AAPG Search and Discovery Article #90075©2008 AAPG Hedberg Conference, Banff, Alberta, Canada