Machine Learning-Driven Seismic Facies Analysis for Channel Prospect Identification in Offshore Abu Dhabi

This paper presents an integrated approach to seismic facies classification for characterizing channel development in the H carbonate formations of offshore Abu Dhabi. Focusing specifically on waveform classification and multi-attribute analysis methods, we demonstrate how these techniques significantly enhance the identification and interpretation of subtle channel features that conventional methods often fail to detect. Our workflow combines unsupervised classification using SelfOrganizing Maps (SOM), supervised neural network classification, and hierarchical waveform classification to analyze the complex channel systems within these challenging carbonate reservoirs. Results show that the integration of these methods provides robust verification of channel features and deeper insights into their genetic mechanisms, including tectonic influences, sea-level fluctuations, karstification processes, and depositional controls. Validation against RGB color blending results by decomposed seismic data demonstrates much better classification accuracy and much more obtained details, significantly outperforming conventional amplitude-based interpretation. This approach offers valuable insights for optimizing development strategies in similar carbonate channel systems throughout the region.