Global Multi Resolution Ocean Prediction/Analysis with Scale Recursive Estimation and Multi Sensor Data Fusion

PI: Iskandarani, Mohammed (Rosenstiel School of Marine and Atmospheric Science (RSMAS))
Co-PI(s): Chassignet, Eric (Florida State University), Shrinivasan, Ashwanth (Tendral LLC), Sharma, Neha (Woods Hole Group Inc.)
Start Year: 2019 | Duration: 3 years
Partners: Florida State University, Tendral LLC, Woods Hole Group Inc.

Project Abstract:

This proposal seeks to develop and test a highly efficient multi-resolution data assimilation framework for ocean state estimation and prediction at a hierarchy of scales – from global to local. Present practices in data assimilation inject large scale observations into global models while regional models assimilate information at localized scales. In the absence of domain-wide, two-way information exchange between global and regional models, however, the observational impact is suboptimal and inconsistently spread across scales.

The new data assimilation system will have the following enhancements: 1) adequately address the multi-scale nature of oceanic flows and observing systems, 2) make optimal use of available observations, 3) assimilate highly local and non-standard data streams and 4) scale with emerging computing architectures. Our approach is based on the scale-recursive estimation which can provide state predictions and uncertainties across a range of scales in a rigorous, computationally efficient and statistically optimal manner. A prototype global multi-resolution ocean prediction system has already been developed and is currently used to provide ocean current forecast for the offshore industry worldwide. The present proposal aims to further evaluate and enhance the scale-recursive methodology, to assess the accuracy versus resolution trade-offs and to extend the system by assimilating new types of observations, namely velocity observations and frontal information.

The testing and evaluation will rely on a multi-resolution system consisting of nested HYCOM models: a global 1/4-degree model, a 1/16-degree model of the Caribbean and the Gulf of Mexico and regional models of the Gulf of Mexico at resolutions ranging from 1/32 through 1/64 degrees. Several 3-year hindcast experiments will be performed to gauge the effectiveness of the one and two-way information exchange in the prediction system.

The model will be enhanced to assimilate velocity observations obtained from the global drifter program, rig mounted ADCP and HF-radar. We will assess the utility of these data for reconstructing the surface and subsurface velocity fields, and if they lead to a better prediction of subsurface eddies that are difficult to detect from surface observations alone.

A major concern of ocean assimilation systems is the possibility of misplacing frontal features in space and time. Experience has shown, however, that predictions that assimilate frontal information are generally more accurate than those without it. We will develop and test methods to assimilate this frontal information to further correct estimates obtained by assimilating routine observations.

We expect our efforts to produce a validated scale recursive system for assimilating multi-scale observations. Estimates produced by the system can be used for both short-range forecasting and sub-seasonal forecasting applications. The outcome will directly impact the ability for the Navy to make predictions across scales and will also improve the safety and efficiency of offshore industrial operations.

BAA: N00014-18-S-B007
BAA Topic: Topic 6: New Approaches for Data Assimilation to Improve Operational Ocean Prediction