INTERNALLY-GENERATED VARIABILITY IN SOME OCEAN MODELS ON DECADAL TO MILLENNIAL TIMESCALES
Timothy J. Osborn
A thesis submitted in fulfilment of the requirements for the degree of Doctor of Philosophy at the University of East Anglia
Passive variability, generated by the modulation of random weather events by the inertia of the oceans, is studied with a range of simple ocean models of the upwelling-diffusion type. Earlier work is extended by parameterising global-mean vertical heat transport in a number of depth-dependent and time-dependent ways. These are developed from theoretical and empirical considerations, as well as from comparison with a more complex ocean general circulation model (OGCM). They are an attempt to separate the effects of isopycnal and diapycnal advection and diffusion. The spectra of passive variability is shown to be sensitive to these parameterisations. The highest sensitivity is to the replacement of a completely mixed layer by a layer of enhanced mixing, which results in all the combinations tested exhibiting weaker low frequency variability than some previous studies found.
Active variability, generated by instabilities and interactions within the ocean and between ocean and atmosphere, requires more complex ocean models which explicitly model ocean dynamics. Previous work, which found considerable Southern Ocean and Atlantic Ocean variability in the LSG OGCM under mixed boundary conditions, was extended in a number of ways. First, a similar experiment was performed, but with a rather different OGCM. Numerical problems greatly reduced the usefulness of the results. Second, the sensitivity of the internal variability to some of the model's physical and numerical details was investigated. Using an alternative convective adjustment scheme can reduce the magnitude of the internal variability by 70%. Using an improved parameterisation of brine rejection from sea-ice freezing also reduced the magnitude of variability, although stronger stochastic forcing could induce large North Atlantic oscillations. Third, it was shown that the dominant variability was purely a Southern Ocean phenomenon, since the signals which propagate around the Atlantic Ocean play no active role in that mode. A second type of propagation was identified - westward around the Antarctic continent - and was explained as a coupled 'salinity - coastal upwelling' wave motion.
Finally, the active variability of a hybrid coupled model was studied. This model consisted of the same OGCM, but coupled to a statistical atmosphere model rather than to mixed boundary conditions. The atmosphere model was constructed on the basis of results from a 19 year simulation with an atmosphere general circulation model forced by observed sea surface temperatures. It included active air-sea fluxes of fresh-water, momentum and heat. The fresh-water flux model appeared to reduce the magnitude and period of the ocean variability, but this was shown to have little significance. In fact, when the statistical model was improved it acted to slightly strengthen and lengthen the oscillation maxima and minima. The air temperature (i.e., heat flux) model weakened the convective feedback which causes the model's variability, so that the oscillations were weaker. But it was unable to prevent the variability from occurring, and was unable to prevent a partial collapse of the North Atlantic thermohaline circulation under stronger forcing.