GEOPARD - Geostatistical Event-based Object-model Predicted from Analogue Reservoir Deposits

The GEOPARD project is a Knowledge-building Project for Industry sponsored by the Research Council of Norway and four industry partners. It is an inter-disciplinary project, collaborating closely with the expertise of academia in geoscience and statistics. Three-dimensional geological models play an important role in predicting underground fluid flow and can be utilized across various fields, including hydrocarbon reservoirs and CO₂-storage. Our aim is to improve model predictions by introducing geological concepts into the core of the statistical modelling algorithms.

The main delivery of the project is a model that not only generates a realistic representation of geology but also captures uncertainty and conditions data, including well data and outcrop analogues. This will meet the industry demands of more direct utilization of geological knowledge and analogue databases in development decisions.

Modeling facies within a reservoir can be challenging due to the trade-off between geological realism and the ability to condition to subsurface data and handle uncertainties within reasonable computational time. Methods range from basic stochastic facies modeling algorithms with data conditioning as main feature, to advanced physical simulations where the geological depositional processes are the main concern. We are developing a new rule-based algorithm for modeling shoreface depositional systems, where the main aim is to bring geological concepts into the core of the statistical modeling framework. The algorithm builds on the object-based facies modeling technique, which provides a Bayesian framework to handle conditioning, assimilate input data and manage uncertainties. The novelty lies in making a prior geological model that generate geometries and place facies objects from a set of geological rules instead of random placement. We translate the dominant geological processes into a set of rules from which we develop our facies modeling algorithm to generate geologically realistic prior models.