Thesis Type:PhD thesis
The first part of this thesis studies the low-frequency ocean-atmosphere coupling relevant to describe tropical meridional modes using an analytical framework and simple sensitivity experiments. We consider a simple Gill-Matsuno atmospheric model coupled to a thermo- dynamic slab ocean model to describe the wind-evaporation-sea surface temperature (WES) feedback. We develop an analytical understanding of the coupled processes responsible for the growth and propagation of meridional mode-like structures by projecting the ocean and steady-state atmosphere equations onto parabolic cylinder functions. By doing this the cou- pling simplifies to the non-normal interaction of di↵erent sea surface temperature (SST) modes mediated by atmospheric Kelvin and Rossby waves. Under a homogeneous coupling the system simplifies to independent, coupled equatorially symmetric and antisymmetric (meridional mode-like) modes, with the following major findings: the non-normal interaction is responsible for a propagation from high latitudes to low latitudes, and this process works similarly for equatorially symmetric and antisymmetric coupled structures. That similarity breaks near the equator where the antisymmetric structure grows through a positive feedback mediated by the first antisymmetric Rossby wave, while the symmetric structure is subjected to a negative WES feedback arising from the atmospheric Kelvin wave interaction. Taken all together this explains why the WES feedback preferentially sustains meridional mode-like structures.
The second part of this thesis deals with the stochastic parameterization of tropical climate variability. Although usually seen as arising from slow non-linear interactions, devi- ations from Gaussianity in the Tropical Pacific may originate through fast variability that may be parameterized as multiplicative noise. We develop and apply a stochastic parameter- ization that accounts for the Tropical Pacific SSTs deviations from Gaussianity. This model, termed CAM-LIM retains the successful elements of the regular Linear Inverse Model (LIM) but it improves in the representation of the higher order statistical moments. The CAM-LIM is used to correctly model the marginal and conditional probability density functions of the different SST principal components. It is found that the model is successful at describing the observed deviations from Gaussianity such as: it rightfully describes the ratio between positive and negative Niño events, and the frequency of extreme Niño events as well.