Effective prediction of regional climate, especially rainfall, at interannual to decadal timescales is of considerable importance to decision makers. To investigate predictions at these timescales, a new set of climate model experiments, called the ‘decadal’ experiments was set up as part of CMIP5. Simulation of rainfall in climate models, however, is often poor and the decadal experiments provide little predictability for rainfall. Previous work has demonstrated that SST has greater predictability and sea surface temperature (SST) has a strong influence on terrestrial rainfall. Thus, SST predictions can be used to derive rainfall ahead of time. Such Indo-Pacific SST-rainfall relationships are used operationally in the seasonal forecasting of Australian rainfall. On this basis, the thesis investigates the possibility of rainfall prediction over Australia at interannual timescales using outputs of SST from these decadal experiments. The prediction skills of nine SST indices that are relevant predictors of Australian rainfall are first quantified. It is found that most indices are not predictable beyond the first year. Two approaches for enhancing their predictability timescale are examined: 1) Investigating the effect of drift on predictability and identifying the drift correction method that leads to the best predictability 2) Identifying other indices that inherently have a higher predictability. The key findings around model drift include: (i) under sampling of initialisation years can lead to spurious estimates of drift and predictability limits over the tropical Pacific, (ii) prediction skill is enhanced with more complicated drift correction methods, and (iii) drift correcting individual models prior to multi-model averaging leads to clear improvements in skill. This thesis also examined in detail a newly identified Pacific-Atlantic transbasin climate mode, TBV, to be significantly related to Australian rainfall. We found that this mode showed predictability timescales that exceeded El Niño Southern Oscillation (ENSO) across multiple models. Most importantly, we showed the co-occurrence of TBV and ENSO to intensify the drying and wetting effects of ENSO over Australia. Using all this information, a simple rainfall prediction model is designed and applied over Australia. The results show that there is indeed merit in decadal predictions of SST for interannual rainfall prediction over Australia.