Skip to main content Skip to secondary navigation
SCERF research group in ARCO Courtyard
Main content start

SCERF faculty, staff and students 2022. The Stanford Center for Earth Resources Forecasting (SCERF) provides research in the exploration, evaluation & development of Earth Resources, whether Energy, Water or Minerals.

Our motivation

"The challenge of living on a planet with a growing population requiring an increasing need for such resources. Any exploration, evaluation & development will have to rely on smart decision-making processes that address properly large uncertainties due to limited observations, whether from drilling, geophysical or flow dynamic data, and to mitigate environmental impact"

Our mission

"To provide solutions for such problems from data acquisition to decision analysis. We focus on developing state-of-the-art data scientific methods for the integration of spatial data over many scales, the quantification of uncertainty of subsurface systems, the value of information of data sources in the context of decision making purposes."

The SCERF program receives funding from industrial affiliate members from both the mineral and energy industry and funding for groundwater & geothermal resources from government entities and the Stanford Doerr School of Sustainability.

Learn More about SCERF

Generative AI for multi-scale subsurface modelling

At SCERF, we have pioneered generative deep learnings for subsurface modeling across scales. Some of our recent key highlights include:

  • LAMGeo initiative: Large generative AI Model for Geomodelling. A large AI model will be trained based on GANs or diffusion models for fast field-scale subsurface modeling and smart decision making integrating various static and dynamic data as well as experts’ geological knowledge. The pre-trained AI model can be used for energy and environmental applications for climate change mitigation, such as CO2 geological storage, subsurface water resources, natural hydrogen, etc. [GitHub Repository:, to be released in September 2024.]
  • PorousRockGAN: Multiscale Porous Rock Modeling based on Generative Adversarial Networks (GANs). An end-to-end GANs has been developed to integrate multi-source and multi-scale imaging data of porous rock, achieving high-resolution stochastic characterization of porous rocks across multiple scales. [Liu and Mukerji, GRL, 2022:
  • GeoBRAIn: an open-source project for Geoscientific Bayesian Reasoning with Artificial Intelligence. It serves as a dynamic network to leverage advanced computational and AI techniques for data assimilation and inference to enhance decision-making processes in subsurface geosciences. [GitHub Repository:, to be released in September 2024.]
Sunset view from Stanford Campus

Foundational References

Eidsvik J, Mukerji T, Bhattacharjya D (2015) Value of information in the earth sciences. Cambridge University Press, Cambridge.

Scheidt, C., Li, L., Caers, J (2018) Quantifying uncertainty in subsurface systems, American Geophysical Union, Wiley, 288p.

Read about SCERF research in EOS