Luthi, Stefan M.1, Israel Rivera Rabelo1, Aernout Schram de
Jong1, Hilbrand Haverkamp1
(1) Department of Applied Earth Sciences, Delft University of Technology, 2628 RX Delft, Netherlands
ABSTRACT: Insights from the Analysis of High-resolution 3-D Seismic and High-density Well Data in a Paleocene Shallow Fluvio-deltaic Reservoir with Low Net Deposition, Suriname (South America)
The Tambaredjo Field in Suriname contains two thin sand reservoirs containing heavy oil in Paleocene fluvio-deltaic deposits. Because of its shallow depth (around 400 meters) and difficult reservoir geometry, much of the field has been developed with 200-meter grid drilling. Recently, high-resolution 3-D seismic has been acquired over 80 km2, which, together with the data from over 500 wells, provides a unique data set to study these deposits. We trained artificial neural networks on reliable well data “ground truth” to segment the seismic volume into facies, either using reflection or acoustic impedance data. The results showed a clear spatial distribution that reflects the overall distribution of depositional environments. We then trained a neural network to recognize individual lithofacies using the acoustic impedance volume, and found a reasonable success rate, although the reservoir sands were sometimes confused with other lithofacies, or were below resolution of the seimic data set. The geometry of the sand bodies was analyzed with a variety of 3-D visualization techniques and using higher-level pattern recognition methods. We found that the knowledge base for such an analysis is still very thin. The sands show poor lateral continuity, mostly because of the relatively high erosion rate in the system, and they were therefore difficult to track over longer distances. Their distribution is also strongly influenced by syn-sedimentary tectonics, mostly consisting of normal faulting along the northern edge of the Guyana shield.
AAPG Search and Discovery Article #90026©2004 AAPG Annual Meeting, Dallas, Texas, April 18-21, 2004.