YIP Multi-model data assimilation and uncertainty quantification
Abstract
Complex dynamical systems are ubiquitous in many areas, including geophysics, climate science, engineering, neuroscience, and material science. These systems involve complicated nonlinear interactions across spatiotemporal scales, strong intermittency, and extreme events. Effective modeling of complex systems advances the understanding of nature and facilitates many essential subsequent tasks, such as data assimilation (DA), prediction, and uncertainty quantification (UQ).This project aims to develop mathematically tractable and computationally efficient nonlinear stochastic models from a multi-model viewpoint. It contains three parts. First, a stochastic nonlinear surrogate multi-model framework is proposed. It starts with a fuzzy dynamical classification for feature extraction and regime detection. With systematic model identification, a set of physically informed surrogate models are developed to capture nonlinear and intermittent dynamics in each regime. The multi-model framework provides a more effective characterization of intermittent and non-Gaussian features than traditional single-model approaches and allows highly efficient multi-model DA and UQ. Second, with a rigorous derivation of a continuum description of particle motions, a new hybrid Eulerian and Lagrangian DA strategy is proposedto improve the DA of both the large-scale and mesoscale features of the ocean. A set of cheap stochastic surrogate models of the ocean field and a randomized selection strategy of observations are further proposed to significantly improve computational efficiency. One crucial and unique feature of the proposed hybrid DA framework is that it facilitates using closed analytic formulae for both the Eulerian and the Lagrangian DA. In addition to developing mathematical theories, the project aims to assist and improve operational systems. Particularly, characterizing coupled ocean and atmosphere in the tropics is one of the most challenging scientific problems with a significant societal impact. In the last part of the project, a hierarchy of stochastic dynamical models is proposed to provide a complete dynamical and statistical depiction of the equatorial phenomena, including the coupled El Nino-Southern Oscillation (ENSO) and the Madden-Julian Oscillation. These models are used to simulate and understand these phenomena with an appropriate UQ, which is beyond the capability of traditional deterministic systems. These new models will also be utilized to advance multi-modelDA and prediction. In addition, developing new information metrics beyond the standard path-wise errors for quantifying the skill of stochastic models, measuring the statistical response, and assessing the predictability is another focus of the work.The methods developed in this project fit many Navy-relevant problems. The effective stochastic nonlinear surrogate multi-model framework aims toimprove the description and understanding of many complex systems with nonlinear multiscale structures, strong intermittency, and extreme events. These are typical features of the problems the Navy is interested in. The hybrid Eulerian and Lagrangian DA strategy can be adaptable to DoD data assimilation and forecasting systems and has the potential to be incorporated into the US Navy Global Prediction Systems. In addition, improving the understanding and forecasting of the coupled ocean and atmosphere, especially the ENSOcomplexity, directly impacts DoD strategic planning and operational capabilities. Together with the proposed information metrics, these statistically accurate dynamical models facilitate the study of the model response, predictability, and UQ in forecasting nature. Finally, the project will contribute to workforce development through multidisciplinary training of students and postdocs.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Mar 15, 2024
- Source ID
- N000142412244
Entities
People
- Nan Chen
Organizations
- Office of Naval Research
- United States Navy
- University of Wisconsin System