Deep marine deposits are an important petroleum reservoir type. While existing object and trend based stochastic modelling approaches can capture some of the heterogeneity, the limited quantities of data result in a very broad spread of possible model outcomes. Furthermore, such modelling approaches fail to honour the key factors that control the distribution of reservoir sand, namely accommodation, sediment supply and basin topography.
Existing numerical process based modelling tools are capable of recreating flow and deposition from a single turbidite event. These tools are computationally expensive and while they capture the fine detail of a single event they are unsuitable for re-creating reservoirs comprised of the accumulated result of hundreds of individual flows.
A new simplified process based modelling tool capable of rapidly recreating the flow and deposition of hundreds of flow events is presented. The model accounts for seabed topography, gravity, friction, kinematics, ocean currents, sedimentation and erosion rates. Individual events start in a confined feeder channel and experience a hydraulic jump where the flow stalls and widens into a lobe. This method is combined with stochastic elements for inclusion of reservoir uncertainty and conditioning to well data.
The geometry of a deep marine deposit is modelled by generating many individual events and stacking them on top of each other, thereby mimicking the actual sequence of deposition. Intermediate deposition of clay rich material, from hemipelagic material and numerous small clay rich mass failure events, are also modelled. These intermediate events can either be of constant thickness or filling available accommodation space.
Initial input is a bottom surface (seabed), well data and various physical parameters describing the physical process generating a turbidite. The goal is to generate a geologically sound realisation given the physical properties and the constraints imposed by surfaces and well data.
The basic idea is to have a simplified process model that can condition to well data. This combination of process model and exact conditioning makes a rigorous probabilistic specification very difficult. The lack of rigour allows the inclusion of realistic physical processes, but it requires robust algorithms since no properties are guaranteed from a theoretical model.
