Data compression for 3-D oceanographic fields using empirical orthogonal functions

Naval oceanographic and acoustic information systems need to be capable of handling and manipulating large quantities of environmental data. This is required to support the use of acoustic models in future sonar applications, which require a complete 3D description of the ocean environment. Increasingly, use is being made of 3D numerical oceanographic forecast models to provide this information (e.g. FOAM, a global Forecasting Ocean Atmosphere Model which is being developed for operational use by DRA and the UK Meteorological Office). However, the quantities of data that such models generate are significantly greater than those which can be handled by current naval communications systems. An efficient technique is therefore required for compressing 3D oceanographic data to enable it to be transmitted quickly and with little or no loss of accuracy to an at-sea user. The authors describe a 3D data compression technique that uses empirical orthogonal functions (EOFs) and variable bit length encoding, to achieve high compression. The use of EOFs for data compression is not new. Clancy (1992) and Garther et al. (1991) describe the application of this technique in the US Navy´s Oceanographic Data Distribution System (NODDS). However, the present article gives new results on the accuracy of reconstructed environments and the impact of data compression on derived acoustic fields