A primary enterprise-wide science and technology objective of the NOAA Next-Generation Strategic Plan (NGSP) is an integrated environmental modeling system that properly accounts for physical processes across land – atmosphere – ocean. Modeling at NCEP and other institutions and operational centers has traditionally been focused on individual earth system components (whether it be the atmosphere, land or ocean), with exchange across the different earth systems treated as one way boundary forcings. As part of its Next Generation Strategic Plan, NOAA is moving towards a fully integrated modeling system that actively couples the different modeling systems together, with the aim of improving both short term and long term forecast skill.
The next Climate Forecast System version 3 (CFSv3) will be the prototype for NCEP’s Next Generation Global Prediction System (NGGPS), with a fully integrated environmental modeling system. The CFS version 2 (CFSv2; Saha et al 2010) comprised of the Global Forecast System (GFS) for the atmosphere, the GFDL MOM4 for the ocean, the GFDL SIS for sea-ice, and the Noah land model. For CFSv3 we envision a more tightly coupled modeling system that will simulate the exchange of fluxes across the atmosphere – ocean boundary by properly accounting for ocean wave dynamics and the corresponding physical processes such as wave induced langmuir mixing and growing waves that modulate these fluxes. For this we shall rely on using the latest state-of-the-art wave model, WAVEWATCH III®, coupled with MOM6, the latest version of the ocean model from GFDL
To provide accurate and reliable initialization this work will be tied together with a companion proposal to develop a ‘strongly coupled’ Data Assimilation (PI: Penny), in which the analysis itself is constructed as a fully coupled state using cross-covariances to integrate information across all domains. This approach helps to eliminate coupling shocks, and for poorly observed domains it also draws new observational information into the analysis. ECMWF is exploring this approach for their CERA system (Laloyaux et al., 2015) using an incremental 4D-variational scheme that employs coupling in the outer loop, though the land-surface and wave analyses are still performed separately. While the strongly coupled DA approach may be challenging to implement in a fully-coupled system using 4DVar, an ensemble-based approach makes the implementation tractable. Further, by using a grid-point-based observation-space localization DA method such as the Local Ensemble Transform Kalman Filter (LETKF) (Hunt et al., 2007), implementation of strong coupling is actually quite simple.