Bridging Models and Data - Assimilation, Model Hierarchies, Causal Inference and Digital Twins

Explore advanced data assimilation techniques and their applications across scientific disciplines in this comprehensive seminar by Dr. Nan Chen from the University of Wisconsin-Madison. Learn how data assimilation serves as a crucial bridge between mathematical models and observational data, beginning with traditional methods before advancing to cutting-edge innovations. Discover how to integrate models of varying complexity through reconfigured latent data assimilation approaches, with practical applications demonstrated using equatorial Pacific Ocean systems. Master the principles of assimilative causal inference (ACI), a novel Bayesian framework that traces causes backward from observed effects to study predictability and attribution in climate tipping points, model reduction, and extreme events. Understand how ACI identifies dynamic causal interactions without requiring direct observations of candidate causes, accommodates short datasets, and scales efficiently to high dimensions while providing online tracking of intermittent causal roles. Examine a nonlinear neural differential equation modeling framework that exploits generalized Koopman theory to discover latent representations of state variables, enabling closed-form solutions to nonlinear data assimilation problems. Gain insights into computationally efficient digital twin development and explore applications spanning atmospheric and ocean science, materials science, and data science from a researcher who has received the Kurt O. Friedrichs prize and ONR Young Investigator Award.