The objective was the generation of a long-term layered ocean current dataset for the central and southern Atlantic Ocean, of suitable quality to support deep water offshore engineering design, marine operational planning and oil spill simulation. A model scheme was formulated to provide a reference fine resolution offshore grid covering the Atlantic Ocean from 30 degrees north to the Antarctic land mass. Very fine resolution grids were nested to model the continental shelf and slope regions of principal interest for oil and gas exploration and production. Various schemes were evaluated for representation of the riverine discharge, which is of major importance in some parts of west Africa. Additionally, optimal data assimilation scheme formulation was considered in order to best represent ocean surface topography, registration of major dynamic features such as fronts and eddies, and also correct characterisation of ocean density structure. Particular attention has been given to the equatorial region where energetic Kelvin wave propagation is generated by wind forcing; such waves contribute significantly to the most severe ocean current speeds, but can be spuriously misrepresented by inappropriate assimilation of erroneous observational data. The computational runtime required to undertake a long-term ocean reanalysis was large, even when using a major supercomputing facility. Therefore, it was extremely important to ensure that the model configuration was optimised in a structured pilot evaluation in advance of the full production run in order to achieve ‘fitness for purpose’ for the intended applications of the resultant dataset. This paper discusses the detailed process of structured model optimisation, where various permutations of setup had been validated against available in-situ measurements. The final model has now been completed for a 20year run and has been used in a number of regional metocean studies for Oil and Gas clients. The inclusion of location specific processes offers significant advantages over previously existing models.

This content is only available via PDF.
You can access this article if you purchase or spend a download.