Nonlinear prediction and analysis of solar-geophysical disturbances within the ionosphere

This thesis presents novel and robust nonlinear techniques that can be used to analyse and predict geophysical time series. The techniques presented can cope with the problems of noise and non-continuity that are typical of solar-terrestrial data sets. Existing nonlinear analysis techniques are prone to given spurious indications of nonlinear behaviour in such circumstances. These new techniques can be applied as either an alternative to established techniques, to determine less fragile and more qualitative nonlinear properties that may be presented, or as a precursor - to determine whether or not it is viable to proceed with the application of standard nonlinear analysis methods. The methods in this thesis have a general applicability that extends to any time series prediction problem, and are not necessarily limited to geophysical applications. Furthermore, these techniques have been adapted to form the basis of a real-time ionospheric forecasting demonstrator.