Advanced Earth model building during iterative imaging updates using machine learning from an East Mediterranean basin case study

This study presents the effective implementation of machine learning (ML) using convolutional neural networks (CNN) to fine-tune the Gulf of America (GOA), historically known as the Gulf of Mexico, regionally trained CNN from Fernandez et al. (2023) and apply transfer learning to iteratively construct the models with the predicted salt body horizons during imaging cycles for a massive (∼35,000 km2) and complex East Mediterranean (EMED) data set. These salt body horizons display improved and consistent tracking of the top of salt (TOS) and base of salt (BOS) events; thereby, capturing the fine details and complexity of the geology. A considerable efficiency gain in model building is achieved by minimizing the labor-intensive task of interpreting salt body horizons. A 2 to 3-time turnaround time improvement record was compared with a fully manual interpretation. ML can directly provide a highly accurate and precise interpretation for the straightforward and well-imaged regions.