Generative model based geomodelling informed by multiple geophysical observations and geological insights

Minghui Xu
With the increasing computational capacity of computers, the reparameterization of high-dimensional geomodels and the generation of realistic geomodels through latent low-dimensional features have become feasible. Generative deep learning models have demonstrated their impressive performance, not only in computer science but also in geoscience. Despite these advancements, the application of generative modeling to practical geophysical and geological scenarios still requires further exploration. Bridging the gap between synthetic data applications and practical implementation is crucial for unlocking the full potential of generative models in addressing real-world challenges within the geosciences. This research not only contributes to the refinement of generative geomodelling techniques but also holds promise for applications in critical areas such as carbon capture, sequestration, and sustainable development.