On the ability of the LODYC GCM to simulate the thermocline depth in the equatorial Atlantic

Abstract

Using a multivariate model testing procedure that distinguishes between model inadequacies and data uncertainties, we investigate the ability of the LODYC GCM to simulate the evolution of the 20°C isotherm depth during the 1982–1984 FOCAL/SEQUAL experiment in the equatorial Atlantic. Two different versions of the model are considered: the “Ri” version which has a Richardson number dependent parameterization of vertical mixing and the new “TKE” version which uses a local estimation of the turbulent kinetic energy to parameterize vertical mixing. Some effects of the forcing uncertainties are considered by forcing the TKE version with three equally plausible wind stress fields whose differences are consistent with the measurement and sampling errors, and the drag coefficient indeterminacy. The resulting uncertainties in the model response are substantial and can be as large as the differences between simulations with the two GCM versions, which stresses the need to take the forcing uncertainties into account. Although only one Ri run is available, it is shown that the “TKE” parameterization significantly improves the representation of the equatorial upwelling and the simulation of the depth of the thermocline in the eastern Atlantic. However, there remain significant differences with the observations which cannot be explained by the forcing uncertainties that were considered. The two model versions perform better in the equatorial wave guide than in the 12°N-12°S domain, and they are better distinguished over large domains than along sections, which shows that a global multivariate view point must be used in model-reality comparisons. Finally, a comparison with a linear multimode model emphasizes the need for greater model complexity to properly simulate the equatorial upwelling and the thermocline variability in the tropical Atlantic.