Exploiting Low-Dimensional Structure in Bayesian Ice Sheet Inverse Problems

Watch a 53-minute webinar presentation where Noemi Petra from the University of California Merced explores low-dimensional structures in Bayesian ice sheet inverse problems. Dive into the complexities of model-based projections used for predicting ice sheet contributions to sea level rise, focusing on identifying and leveraging low-dimensional structures in inverse problems. Learn how these structures emerge from local sensitivity of data to parameters, diffusive models, and sparse data. Understand the significance of the Hessian operator in inverse problems, its low rank properties, and how its approximations serve as preconditioners for Newton-Krylov systems or inform MCMC proposal distributions. Explore the challenges of high-dimensional inversion parameters and computationally intensive model evaluations in the context of Bayesian inverse problems, while discovering how these methods help integrate measurement data to estimate and quantify uncertainties in model parameters.