Reducing uncertainty in reservoir models through the application of process-based and digital outcrop modelling

Abstract

The deposits created by fluvial processes are complex and heterogeneous as well as challenging to model and characterise, especially with the presence of limited subsurface data (e.g. well and seismic data). This represents a significant source of uncertainty in the modelling of fluvial systems, and in conjunction with limited statistically driven modelling approaches current in use, make the modelling of meandering fluvial systems problematic. By harnessing a new generation of the process-based stochastic modelling and latest advancements in three-dimensional (3D) digital survey technologies, the study presented in this thesis aimed to reduce uncertainty in the fluvial reservoir models by producing realistic meander geometries and facies distributions which will match patterns seen in the well data from the reservoir. Excellent outcrop exposures of the Mid-Jurassic Long Nab Member in the Scalby Formation, North Yorkshire, UK, allowed for detailed field-based interpretations and outcrop characterisations. The exhumed meander-plain of the studied outcrop contains two storeys of channel bar complex, and these storeys are characterised by various depositional patterns and resembled by some modern analogues. The upper-storey is dominated by spatially unconfined meander-belts with comparatively high meandering index. On the contrary, the lower-storey is represented by deposits formed in a confined channel setting with low meandering index. Statistical data are extracted from modern analogues using satellite imagery of meandering rivers that represent both confined and unconfined meander-belts, in addition to 3D digital datasets obtained from outcrop using Unmanned Aerial Vehicles (UAVs). These facilitate the construction of accurate and realistic channel and facies models, which allowed then for comprehensive quantitative and qualitative analysis. The statistical data derived from modern analogues are multi-dimensional in nature, which makes it difficult to analyse. Consequently, applying data mining techniques such as parallel coordinates in order to investigate and identify the key relationships within modern analogues is vital. Development of new methods for integrating surface models with the processes of meander belt formation has successfully enabled 3D reconstruction of multi-storeys of the Long Nab Member. Furthermore, novel techniques of variogram computation, facies classification and upscaling classified-facies from the Digital Outcrop Models (DOMs) provided important conditioning data for building a realistic stochastic facies model (in this case sequential indicator simulation). The quantitative data and 3D visual assessments derived from the resultant models demonstrate the efficiency of using process-based stochastic and digital outcrop modelling approaches to model the geometry of a heterogeneous meandering fluvial reservoir. The high-quality of these models constitute excellent bases from where to extract pseudo wells and training images to be employed in other facies modelling approaches, and also make it a sound input for flow simulation studies.

Date of Award	1 Aug 2020
Original language	English
Awarding Institution	The University of Manchester
Supervisor	Emma Finch (Supervisor) & David Hodgetts (Supervisor)