Pattern-Based Geological Modeling of Deep-Water Channel Deposits in the Molasse Basin, Upper Austria

Stright, Lisa¹, Andre G. Journel² (1) Stanford University, Stanford, CA (2) Stanford University, N/A, CA

The goal of reservoir modeling is to generate numerical representations of reservoir geology that honors both hard (well and core) and soft (seismic and production) data collected from the reservoir. Traditional numerical reservoir modeling approaches are limited in their ability to include interpretive information and their resulting models often lack geologic realism. A shifting modeling paradigm from 2-point variogram-based statistical methods toward training image-based methods is empowering modelers to integrate descriptive geological interpretations with numeric field data. Advances in pattern-based modeling (Strebelle,2000; Arpat, 2005; Zhang, 2006) have made it possible to mimic the geometry of complex geologic structures while conditioning to the diverse suite of data typically encountered in hydrocarbon reservoirs, presenting a greater opportunity to include qualitative geological interpretations into the numerical reservoir model.

A demonstration of a pattern-based modeling study is presented for the Puchkirchen field in the Molasse Basin in Austria. The Puchkirchen field is located in a large, elongate, deep-water channel belt confined within a foreland trough. The seismically mappable channel belt is dominantly filled by turbiditic conglomerate and sub-seismic, thin-bedded sandstone intervals. The channel fill is further complicated by chaotic slump and debris-flow deposits. The complex nature of the channel fill is captured through hand drawn training images which reflect sedimentological studies of well logs, core and seismic data (Hubbard, 2006). The pattern-based algorithms produce multiple alternative numerical models of channel fills, all drawing spatial facies distributions from the training image and all locally constrained to well, core and seismic data.