Juan M. Restrepo
Department of Mathematics and
College of Earth Oceans and Atmospheric Sciences
Oregon State University
Abstract:
The use of models and data, via data assimilation, is one of the strategies pursued to improve climate and weather predictions and retrodictions. In these application areas the norm is that the number of degrees of freedom in the models is vastly larger than the data available. This is notoriously problematic in the oceans where data gathering is challenging and the dynamics have statio-temporal scales that span large ranges.
The Dynamic Likelihood filter is a data assimilation scheme that is designed specifically for hyperbolic and advection-dominated problems. It aims to improve predictions that combine data and model outcomes in a Bayesian framework by extending the application of the likelihood over longer spatio-temporal filtering frames. When the data has low uncertainty and the data is sparse, the methodology is competitive with other filtering methods with regard to computational complexity, and superior in its estimates.
The talk will describe the filter and show how it performs on simple wave problems. It will then be compared to other methods on problems where it is clear what the optimal outcome should be, demonstrating its efficiency and efficacy.
