Integrating outcrop and subsurface data to improve the predictability of geobodies distribution using a 3D training image: A case study of a Triassic Channel – Crevasse-splay complex

Fluvial sandstones deposited by high-sinuosity fluvial systems are one of the most complex reservoirs to predict and model with confidence, a reflection of both the geometries and complex distribution of the component geobodies. By integrating both analogue outcrop data and associated subsurface data, as well as new technical advances in the reconstruction of the outcrop in 3D (Digital Outcrop Models, DOM), the geostatistical parameters, which condition the modelling of these reservoirs, can be better determined. In addition, DOMs also allow us to easily extract the necessary georeferenced input data (digitized outcrop interpretations, geometrical parameters, as well as key surfaces) and so create geocellular outcrop models (GOM); a useful tool with which to contrast the results obtained from geostatistical simulations, as well as to quantify the uncertainty associated with the results. In this study, classical field data, digital data derived from outcrop models and subsurface data were combined in order to carry out a geostatistical modelling of a Channel – Crevasse-splay complex outcrop analogue, located in the Triassic Red Beds of Iberian Meseta (TIBEM). Geostatistical modelling results were obtained by combining Object-based (OBM) and MultiPoint Statistics-based (MPS) modelling techniques. A critical element in this study was the design of appropriate modelling workflows with Petrel® which would best reproduce the distribution of heterogeneities at macroscale. The designed modelling workflow was used to construct a 3D Training Image (TI) of a fluvial reservoir comprising both a meandering channel system and its associated overbank sandstone deposits. The resulting TI represents all geobodies described in the studied outcrop example and is exportable to similar fluvial reservoirs. This TI was then used in MPS simulations, in order to establish how it could assist in the prediction of the reservoir geobodies, as well as confirming to what extent this prediction matched the outcrop. •Outcrop/Behind Outcrop methodology is used to both characterize and provide input to geostatistical modelling.•A modelling workflow, using Object-based modelling and rule-based methods, was designed to create a 3D Training Image.•Geocellular Outcrop Models provide a valuable source of information enabling effective quality control of modelling results.•MPS simulation, in agreement with input from the paleogeographic reconstruction, generate a good prediction of the geobodies