GeoSeminar: Building subsurface models with AI

TU community join the Department of Geosciences for a GeoSeminar over building subsurface models with AI.

A realistic model that delineates the structure, stratigraphy, and rock properties plays a pivotal role in understanding the Earth’s subsurface, and is essential to natural resource exploration, carbon storage, and civil engineering. Traditionally, building such models requires extensive human interaction with multiple data modalities. For example, to build a structural model, one needs to interpret multiple horizons and faults that define the key structures, which can be time-consuming even for experienced seismic interpreters.
We attempt to automate and accelerate the subsurface model building workflow with artificial intelligence (AI), specifically, with deep learning. We use deep learning models in many key steps of the workflow, including seismic and well log data quality check and conditioning, structural and stratigraphic interpretation, generation of attributes, as well as predicting rock properties. We will see the value of AI in building subsurface models with greatly reduced turn around time, while also discussing some lessons learned along the journey.

Brief Bio:
Tao Zhao is the data science manager for interpretation at SLB. Tao joined SLB in 2019 as a senior data scientist, developing deep learning applications for seismic processing and imaging. From 2017 to 2019, Tao was a research geophysicist at Geophysical Insights. Tao has PhD and MS degrees in geophysics from the University of Oklahoma and the University of Tulsa, and BE degree in exploration geophysics from China University of Petroleum (East China). Tao received the J. Clarence Karcher Award from the Society of Exploration Geophysicists (SEG) in 2023, and the best paper award from the 2024 SEG-AAPG IMAGE annual meeting.