You must have a javascript-enabled browser and javacript and stylesheets must be enabled to use some of the functions on this site.
 
   

 

HYCOM Ocean Prediction and Altimeter Data Assimilation

Eric Chassignet(1)

(1) U. of Miami/RSMAS, 4600 Rickenbacker Cway, Miami, FL, 33149, United States

Abstract

Co-authors: H.E. Hurlburt, O.M. Smedstad, J. Cummings, H. Kang, C. Thacker, L. Parent, P. Brasseur, A. Srinivasan, T. Chin, and R. Baraille

A broad partnership of institutions is presently collaborating in developing and demonstrating the performance and application of eddy-resolving, real-time global and basin-scale ocean hindcast, nowcast, and prediction systems using the HYbrid Coordinate Ocean Model (HYCOM). The plan is to transition these systems for operational use by the U.S. Navy at the Naval Oceanographic Office (NAVOCEANO), Stennis Space Center, MS, and the Fleet Numerical Meteorology and Oceanography Center (FNMOC), Monterey, CA; and by NOAA at the National Centers for Environmental Prediction (NCEP), Washington, D.C. The partnership is also the eddy-resolving global ocean data assimilative system development effort that is sponsored by the U.S. component of the Global Ocean Data Assimilation Experiment (GODAE). The systems not only need to run efficiently on a variety of massively parallel computers, but they also need to include sophisticated, but relatively inexpensive, data assimilation techniques for assimilation of satellite altimeter and in-situ data. We will report on the performance of (1) a multi-variate optimum interpolation (MVOI) that use either the Cooper and Haines (1996) technique or synthetic T & S profiles for downward projection of SSH and SST, (2) the Singular Evolutive Extended Kalman (SEEK) filter, and (3) the Reduced Order Information Filter (ROIF). Both the SEEK filter and ROIF are especially well suited for large dimensional problems: In the SEEK filter, the dominant eigenvectors describing the model variability are used to specify the initial background error covariance matrix in decomposed form and this leads to fully three-dimensional, multivariate dynamically consistent corrections. The ROIF method factors the covariance functions into horizontal and vertical components and represents the correction field implicitly, using techniques transplanted from statistical mechanics (Gaussian Markov Random Field). The reduced order aspect of ROIF refers to the fact that the information matrix is approximated as a banded matrix. The SEEK filter uses the non-linear model to propagate the error statistics forward in time while the ROIF assumes a tangent linear approximation to the system dynamics.

 

Workshop presentation

 

                 Last modified: 07.10.03