Automated mapping of carbonate build-ups and palaeokarst from the Norwegian Barents Sea using 3D seismic texture attributes

Geological heterogeneity in hydrocarbon reservoirs is normally related to variability in depositional facies, diagenesis and structure. On the Loppa High, Norwegian Barents Sea, reservoir quality in prospective Upper Palaeozoic carbonates is considered to be controlled mainly by primary depositional facies variability linked to active faulting, with a secondary overprint of palaeokarst. The best quality reservoirs are anticipated in build-up facies. Although high-quality 3Dseismic data exist over the prospect, it is a challenge to map the buildups and palaeokarst in three-dimensional space using conventional 3Dseismic interpretation tools. This is mainly due to the internal heterogeneity of the build-ups and palaeokarst, which are characterized by a range of seismic reflection/attribute patterns. A procedure for multi-attribute mapping of seismic facies is described and utilized to provide a truly threedimensional interpretation of build-up and palaeokarst geobodies. A neural network classifier is used to analyse a set of 3D attributes that capture the seismic stratigraphical and structural patterns inherent in the data. These attributes are referred to as 3D texture attributes since they describe the reflector-geometry in a small 3D neighbourhood. Pattern recognition is interpreter-guided, with the desired seismic facies to be mapped input as ‘training data’ and analysed prior to their presentation to the 3D-classification system. Once the classification is made, the interpreter is presented with a 3D classified volume and an uncertainty analysis of each of the mapped facies/classes. This is used to evaluate the results prior to their visualization in 3D space. The method allows the interpreter to identify quickly a volume where the features of interest occur and to focus the interpretation work directly on them. In addition, the 3D visualization of the mapped geobodies brings important new information that might be overlooked when inspecting data in vertical or horizontal data windows. In the Norwegian Barents Sea case study, 3Dseismic facies mapping provides the first truly three-dimensional interpretation of the build-ups and palaeokarst, where their external form, juxtaposition, cross-cutting relationships and structural information are preserved. These data give new insights as to the stratigraphic distribution and internal variability of the build-ups and palaeokarst system; this information is important to estimation of reservoir volume, connectivity and variability.