Seismic Facies Classification of Salt Structures and Sediments in the Northern Gulf of Mexico Using Self-Organizing Maps

Proper geologic reservoir characterization is crucial for energy generation and climate change mitigation efforts. While conventional techniques like core analysis and well logs provide limited spatial reservoir information, seismic data can offer valuable 3D insights into fluid and rock properties away from the well. This research focuses on identifying important structural and stratigraphic variations at the Mississippi Canyon Block 118 (MC-118) field, located on the northern slope of the Gulf of Mexico, which is significantly influenced by complex salt tectonics and slope failure. Due to a lack of direct subsurface data like well logs and cores, this area poses challenges in delineating potential reservoirs for carbon storage. The study leveraged seismic multi-attribute analysis and machine learning on 3-D seismic data and well logs to improve reservoir characterization, which could inform field development strategies for hydrogen or carbon storage. Different combinations of geometric, instantaneous, amplitude-based, spectral frequency, and textural attributes were tested using Self-Organizing Maps (SOM) to identify distinct seismic facies. SOM Models 1 and 2, which combined geometric, spectral, and amplitude-based attributes, were shown to delineate potential storage reservoirs, gas hydrates, salt structures, associated radial faults, and areas with poor data quality due to the presence of the salt structures more than SOM Models 3 and 4. The SOM results presented evidence of potential carbon storage reservoirs and were validated by matching reservoir sands in well log information with identified seismic facies using SOM. By automating data integration and property prediction, the proposed workflow leads to a cost-effective and faster understanding of the subsurface than traditional interpretation methods. Additionally, this approach may apply to other locations with sparse direct subsurface information to identify potential reservoirs of interest.