Mathematics and Data Science for Improved Physical Modeling and Prediction of Sea Ice
Abstract
Sea ice is a critical component of the Earth s climate, and also plays a vital role in socioeconomic activities and DoD operational,capabilities in the Arctic. Yet, despite its importance, modeling of sea ice dynamics is still in its infancy. For instance, many cu,rrent approaches rely on continuum approximations offering limited skill in capturing the mechanics and thermodynamics of sea ice as, a material undergoing deformation, fracture, and phase transitions. These deficiencies become increasingly apparent at scales of or,the main objective of this proposal is to harness the strength of modern applied mathematics and data science, combined with advance,s in data acquisition and computing capabilities, to develop the next generation of sea ice modeling techniques that are capable of,producing robust, uncertainty-quantified sea ice forecasts and improved physical understanding of processes on these scales. Drawing, upon the interdisciplinary expertise of our team in mathematics, data science, climate dynamics, and Arctic system observation and,processes, we propose to address a carefully chosen hierarchy of problems. Our ultimate goal is to develop and disseminate a global-,scale prediction system that includes novel representations of discrete sea ice dynamics, coupled with data assimilation algorithms,for improved initialization and parameter estimation from high-resolution satellite imagery and in situ observations. At the core of, this framework is a suite of superparameterization approaches, providing physically-realistic, yet computationally-efficient, repre,sentations of small-scale ice dynamics via embedded high-resolution models within the grid elements of coarse global models. Our pro,posed approaches include first-principles Lagrangian ice floe models, developed specifically for this project, as well as statistica,l models based on stochastic modeling and machine learning. Besides offering high dynamical fidelity, this model architecture is com,patible with advanced data assimilation and parameter estimation algorithms using ideas from particle filtering and Bayesian inferen,ce, which we will employ to develop a self-contained prediction framework. This framework will be mathematically rigorous, yet compu,tationally efficient, and will build upon techniques pioneered by our research team in diverse areas. As one of our main deliverable,s, we propose to disseminate the framework as a well documented software product, together with relevant case studies. In addition,,since the judicious use of data underlies many aspects of our approach, another major focus will be to compile and interpret high-re,solution datasets, combining satellite imagery and in situ measurements of sea ice, oceanic, and atmospheric variables. We propose t,o create a valuable resource for researchers by disseminating these datasets through an online repository. For the purposes of the p,roposed research, these datasets will be employed in the training of superparameterization models and parameter estimation schemes,,as well as in process studies that will advance phys, adaptable to DoD forecasting systems, including the Navy Global Environmental Model, thus directly impacting DoD strategic planning, and operational capabilities in the Arctic. More broadly, it is expected that the modeling techniques developed in the project will, find wide applicability in many DoD-relevant contexts where time-evolving complex systems play a role. The project will contribute,to workforce development through multidisciplinary training of postdoctoral researchers and students. Available for public release.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Dec 06, 2022
- Source ID
- N000142312014
Entities
People
- Dimitrios Giannakis
Organizations
- Board of Trustees of Dartmouth College
- Office of Naval Research
- United States Navy