GODAE is sponsored by
3.1 Global High Resolution Analysis and Forecast at the Mesocale
Lead-author: Harley Hurlburt (NRL, Stennis)
Author/co-authors: H. Hurlburt1, E. Dombrowsky2, G. Brassington3, E. Chassignet4, J. Cummings5, M. Drevillon6, Y. Drillet2, E.Greiner7, E. J. Metzger1, P. Oke8, T. Pugh3, A. Schiller8, J. F. Shriver1, O. M. Smedstad9, C.-E. Testut10, B. Tranchant6, A. Wallcraft1, G. Warren11
1Naval Research Laboratory, Stennis Space Center, MS, USA
2Mercator-Ocean, Ramonville Saint Agne, France
3Centre for Australian Weather and Climate Research, BoM, Melbourne, Australia
4Florida State University, COAPS, Tallahassee, FL, USA
5Naval Research Laboratory, Monterey, CA, USA
6CERFACS, Toulouse, France
7CLS, Ramonville Saint Agne, France
8Centre for Australian Weather and Climate Research, CSIRO, Hobart, Australia
9Planning Systems, Inc., Stennis Space Center, MS, USA
10MGC, Blagnac, France
11Bureau of Meteorology, Melbourne, Australia
Abstract
The feasibility of global ocean weather prediction was just emerging as GODAE began in 1997. Ocean weather includes phenomena such as meandering currents and fronts, eddies, the surface mixed layer and SST, equatorial and coastally trapped waves, upwelling of cold water, and Rossby waves, all influencing ocean variables such as temperature (T), salinity (S), currents, and sea surface height (SSH). Adequate realtime data input, computing power, numerical ocean models, data assimilation capabilities, atmospheric forcing, and bathymetric/boundary constraints are essential to make such prediction possible. The key observing systems and real-time data inputs are SSH from satellite altimetry, satellite and in situ SST, T or T&S profiles (e.g. ARGO, TAO, PIRATA, BTs), and atmospheric forcing. The ocean models dynamically interpolate the data in conjunction with the data assimilation, convert atmospheric forcing into oceanic responses, and forecast the ocean weather, applying the bathymetric/boundary constraints in the process. The results are substantially influenced by ocean model simulation skill and it is advantageous to use an ocean model that is eddy-resolving (typically < 8 km grid increments), not just eddy-permitting. Since the most abundant ocean observations are satellite surface data and subsurface data are very sparse in relation to the space scales of the mesoscale ocean features that dominate the ocean interior, downward projection of surface data is a key challenge in ocean data assimilation. The need for accurate prediction of ocean features that are inadequately observed, such as the mixed layer depth, places a major burden on the ocean model, the data assimilation, and the atmospheric forcing. The sensitivity of ocean phenomena to the atmospheric forcing and the time scale for response affect the time scale for oceanic predictive skill, sensitivity to the initial state vs the atmospheric forcing as a function of forecast length, and thus oceanic data requirements and prediction system design. Outside surface boundary layers and shallow regions, forecast skill is ~1 month globally and over many subregions and is only modestly reduced by reverting toward climatological forcing after the end of the atmospheric forecasts vs using analysis-quality forcing for the duration. Finally, global ocean prediction systems must demonstrate the ability to provide initial and boundary conditions to nested regional and coastal models that enhance their predictive skill. Demonstrations of feasibility in relation to the preceding phenomena, requirements, and challenges will be drawn from the following global ocean prediction systems: BLUElink (Australia), HYCOM (USA), Mercator-Ocean (France), NCOM (USA), and NLOM (USA).
(Last Updated: 13-10-2008)