Bruce D. Cornuelle
Peter F. Worcester
David B. Chester
Scripps Inst. of Oceanogr., Univ. of California at San Diego, La Jolla, CA 92093
Brian D. Dushaw
Bruce M. Howe
Univ. of Washington, Seattle, WA 98105-6698
The Acoustic Mid-Ocean Dynamics Experiment (AMODE) measured time series of reciprocal acoustic travel time and temperature with a 700-km-diam array of six moorings south of Bermuda during 1991--1992. Time-independent inversions give a series of snapshots of the barotropic and baroclinic eddy field. Observations taken at different times can be combined, using a dynamical ocean model, to generate improved estimates of the evolving eddy field, provided that the model is accurate. The AMODE observations have been used to test a nonlinear, quasigeostrophic, limited-area ocean model by attempting to fit a long-term dataset, and by initializing the model with a `training' dataset and then predicting forward in time for comparison to data not used in the initialization. Using an rms error measure of misfit between the observations and the predicted data, the model prediction has 40% less error than persistence over 12 days, and about 20% less error than persistence over 24 days. The extent to which the data misfit is sensitive to external model parameters, such as the radius of deformation or viscosity, allows the parameters to be optimized to give maximum predictability, and also gives a measure of how well the model has been tested by the dataset. [Work supported by ONR.]