The Effects of Assimilation on the Physics of an Ocean Model. Part II: Baroclinic Identical-Twin Experiments

Abstract

Part I has shown that a simple assimilation scheme can have a significant effect on the physics of the model. Part II concentrates on the effects of nudging assimilation in a full primitive equation model, the Free Surface Cox Code, illustrating the value of the previous results in a more realistic scenario and providing guidelines relevant to the assimilation of real data. Using a 1° resolution, midlatitude baroclinic setup and justifying results with simple physical models, the assimilation of different types of data and the effects of various data restrictions are studied. It is shown that some assimilations fail, for example, velocity alone is numerically unstable, and that success depends on the relative importance of barotropic?baroclinic modes, for example, sea surface height alone drives a predominantly barotropic response. The preferred type of data depends on length scale relative to the Rossby radius. Interestingly, all are benefited by topography; for example, assimilation of density corrects not only the baroclinic fields but also the barotropic (cf. inverse techniques). Some assimilations illustrate spurious artefacts of the assimilation, for example, sea surface height and density altering the Kelvin wave structure. Identical-twin experiments allow assessment of final errors and times for convergence. A simplistic approach to the problems of limited data coverage indicates that lack of data in time and horizontal space can be easily overcome at this resolution, whereas lack of depth data is critical. A projection scheme for obtaining subsurface fields from surface data is also discussed.