Simula@BI: Data assimilation: Methods, convergence results and challenges

Data assimilation (DA) is an umbrella term for methodologies that combine measurement with mathematical models in a probabilistically reasonable way. In DA, the current state of a system given partial and noisy observations is represented by a distribution in DA, where the assimilation of measurements into the distribution is ideally achieved through Bayesian inference.

For problems involving complexities, such as high-dimensional state space or non-linear models, DA often has to be computed by means of approximation methods. One class of approximation methods that relate closely to Monte Carlo methods is the ensemble-based methods (ensemble Kalman filtering and particle filtering). In this talk, I will present an overview of the main ideas of these ensemble-based filtering methods, cover recent convergence results, and discuss some of the challenges that remain open for these methods, particularly relating to the assimilation of observations into all ensemble members of such methods.

• Visit our website to read more about the research center Simula@BI