Using multi-attribute neural networks classification for seismic carbonate fades mapping: a workflow example from mid-Cretaceous Persian Gulf deposits

Abstract

Abstract Seismic facies mapping in large seismic surveys can be time consuming, even if only a basic overview of the facies distribution is needed. Therefore this study outlines an approach for the use of volume-based seismic attributes from 2D surveys for automated seismic facies mapping within carbonate settings. The study area is located in the central Persian Gulf, offshore Iran. The interval of interest is the mid-Cretaceous Sarvak Formation, which is part of the extensive Cretaceous shallow-water carbonate platform of the eastern Arabian Plate. A set of nine volume-based seismic attributes, calculated from time, amplitude and frequency information of post-stacked 2D seismic data, was chosen to characterize geological information within the interval of interest. The volume-based attributes were supplemented by two grid-based attributes to highlight structural elements. The geological significance of each attribute was evaluated by comparing it with results of seismic sedimentological/geomorphological studies. Furthermore, statistical methods were applied to highlight direct relationships amongst the attributes. The results of these tests were then used to choose a limited set of attributes for neural network-based multi-attribute classifications. The results show that seismic attributes derived from 2D surveys can be used to map basic seismic facies types in carbonate settings and that the outlined, general approach might be applied in other studies.