Improving the representation of internal waves in the Navy and NOAA data assimilative forecasting systems

PI: Chassignet, Eric (Florida State University)
Start Year: 2022 | Duration: 3 years
Partners: Naval Research Laboratory NOAA/NCEP,  U. of Southern Mississippi, U. of Michigan, Tendral, LLC.

Project Abstract:

The Global Ocean Forecast System (GOFS) is the U.S. Navy’s operational global ocean prediction system that runs daily at US Navy production centers. The system depicts the location of mesoscale features such as oceanic eddies and fronts, i.e., the “ocean weather”, and provides accurate 3- dimensional ocean temperature, salinity, and current structure to the Fleet. In the last decades, GOFS has improved its predictive capabilities for ocean circulation over a wide range of frequencies and wave numbers. The assimilation of observational data using NCODA-3DVAR, a three-dimensional variational data assimilation (DA) technique, has significantly lowered the forecast errors of subtidal fields (Chassignet et al., 2009; Cummings and Smedstad, 2014; Luecke et al, 2017). A major step forward in the HYCOM system was achieved with the introduction of tidal forcing (Arbic et al., 2010, 2012, 2018). The current implementation of NCODA-3DVAR data assimilation data of observational data is, however, not without drawbacks. It causes shocks in the positioning of mesoscale fields and these shocks can result in high-frequency internal gravity waves, which appear as “noise” in the tidal bands and inertial bands in regions with strong mesoscale activity. These spurious internal waves cause an excess of energy when compared to observations (drifters) and/or to simulations without data assimilation. The main objective of this proposal is to have more accurate internal tide predictions in GOFS by reducing the generation of high-frequency noise introduced by the data assimilation and by taking advantage of the new observational datasets. To minimize the noise introduced by the DA, we propose to evaluate several data assimilation techniques and to quantify their impact on the representation of high-frequency motions. The data assimilation methods are NCODA-3DVAR (default configuration; Cummings, 2005), TSIS (Srinivasan et al., 2021), 4DVAR (Ngodock and Carrier, 2014), and LETKF (Penny, 2014). The evaluations will be performed in idealized, regional, and global HYCOM configurations.