Activity 2A – Shallow structures & Advanced interpretation
Delimitation of subsurface geologic features from surface seismic data
We aim to develop neural-based new approaches for advanced interpretation of 3D seismic data with a much reduced human intervention at the Artificial Intelligence Center of Excellence for Geosciences (AICEG). We shall pursue such research to the available and/or 3D seismic data that can be procured from premier oil companies for automatic delineation of subsurface geologic features and properties. The expertise built during the last five years on neural networks would be utilized to achieve the goal. The approach is based on Artificial Intelligence consisting of several steps: special filtering and processing of 3D seismic data to improve the signal; computation of several attributes specific to a geologic target; fusing the attributes into a single attribute, called the meta-attribute, through training and testing over a small volume of data; validating the process; and finally running the trained system over the entire volume. This will automatically pick up the geometry of subsurface geologic feature/target based on characteristics of meta-attribute, and aid in quick and advanced interpretation of 3D seismic data. The ANN approach would be extended further based on other neural algorithms, and applied to address issues related to landslides, earthquakes, and glaciers.
An attempt of procuring high-resolution 2D/3D seismic data at the foothills of NW and NE Himalaya would be made from the available sources or industries. These data can be analyzed by utilizing advanced processing and modeling for the delineation of fine-scale structures, geometry and disposition of decollement (i.e. the boundary between underlying Indian plate and overriding Eurasian plate), ramp limiting the ruptures, thrust geometry of the Himalaya, configuration of other faults etc. with a view understanding the seismogenesis and geodynamic evolution of the Himalaya and adjoin regions.