Quantifying uncertainty and improving reduced-order predictions of partially observed turbulent dynamical systems
Seminar Room 1, Newton Institute
AbstractCo-author: A. J. Majda (Courant Institute, NYU) The issue of mitigating model error in reduced-order prediction of high-dimensional dynamics is particularly important when dealing with turbulent geophysical systems with rough energy spectra and intermittency near the resolution cut-off of the corresponding numerical models. I will discuss a new framework which allows for information-theoretic quantification of uncertainty and mitigation of model error in imperfect stochastic/statistical predictions of non-Gaussian, multi-scale dynamics. In particular, I will outline the utility of this framework in derivation of a sufficient condition for improving imperfect predictions via a popular but heuristic Multi Model Ensemble approach. Time permitting, the role and validity of 'fluctuation-dissipation' arguments for improving imperfect predictions of externally perturbed non-autonomous turbulent systems will also be addressed.
If it doesn't, something may have gone wrong with our embedded player.
We'll get it fixed as soon as possible.