The potential for ocean prediction and the role of altimeter data

Abstract

A skillful ocean forecasting capability will depend on (1) adequate data input, (2) adequate computing power, and (3) properly designed and adequately validated ocean models for data assimilation and forecasting. Once these conditions are satisfied, which is feasible with existing technology, forecasts of meandering currents and eddies up to a few months appear to be a reasonable expectation. Simulation studies suggest that the prediction of meandering currents and eddies requires high horizontal resolution (∼10 km), but only low vertical resolution. High vertical resolution can be obtained by using the circulation model to provide horizontal and vertical advection to a grid of one‐dimensional mixed‐layer models. Circulation models on subdomains of major ocean basins with extensive open ocean boundaries require accurate boundary information at all levels in the vertical and for the duration of forecasts longer than a few days. Basin‐scale models appear to be the most promising source for this information. To provide useful boundary conditions, the basin‐scale models must also resolve the meandering currents and eddies. Class 7 computers (∼1 gigaflop and 32–128 million words), expected on the market in the middle and late 1980s, are required for eddy‐resolving models of major ocean basins. Currently, satellites and atmospheric models provide the only prospects for oceanographic data and forcing functions with global coverage and resolution adequate for ocean circulation prediction models. For this purpose the most useful fields potentially available from satellites are current‐related variations in the sea‐surface elevation derived from altimeter data and scatterometer‐derived wind stress. Choosing an appropriate altimeter track pattern is critical. Successful ocean prediction appears feasible without extensive subsurface data acquisition by using the circulation model to convert the potentially well‐observed fields at the surface into subsurface information. This looks more promising than attempting to assimilate the extremely sparse subsurface data. Where adequate subsurface data are available (nowhere at present), they should, of course, be used. Subsurface data and major field programs are vital for local and regional subsurface validation of the forecast models, but altimeter data are essential for global prediction and representation of individual current meanders and eddies. Otherwise, in most regions it would be necessary to rely on the simulation skill of the model and the forcing functions, which are only weakly coupled to many of these oceanographic features. Within the next decade the appropriate satellite data, Class 7 computers, and eddy‐resolving basin or global‐scale models should become available, and should be used to form the heart of a global ocean prediction system. Part of this paper outlines a strategy for global ocean prediction based on these elements and on plans at the Naval Ocean Research and Development Activity.