Lagrangian and Coupled Data Assimilation enhanced by Machine Learning to improve Operational Ocean Prediction
Abstract
We design and build an ensemble-based ocean data assimilation (DA) system that can effectively assimilate observation data relevant to constraining a high-resolution analysis of the ocean surface, with computational costs comparable to the existing 3D-Var NCODA system used by the US Navy. Two guiding principles are the use of innovative new approaches to: (1) utilize new observational data types that are typically ignored for operational ocean data assimilation, and (2) reduce the computational costs of the cycled ocean data assimilation process. The research approach is developed under the assumption that time-dependent dynamically defined background error covariance estimates are necessary to assimilate underutilized observationsalong with the coming wave of high resolution observational data, in order to accurately project the impact of these observations to the unobserved areas of the ocean. Ensemble-based background error covariance estimates based on proxy ML models are to be demonstrated as a viable alternative to running an ensemble of full resolution ocean GCMs at short lead times. Using these tools, previous successes from Lagrangian Data Assimilation (LaDA) and CoupledData Assimilation (CDA) experiments that were demonstrated in a variety of simplified models will be translated to high-resolution Navy models.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jul 20, 2020
- Source ID
- N000142012580
Entities
People
- Stephen G Penny
Organizations
- Office of Naval Research
- Regents of the University of Colorado
- United States Navy