Entropies and predictability of variability indices of the tropical Pacific

This doctoral thesis is concerned with the problems of the predictability and the temporal structure of indices of the climatic variability in the tropical Pacific, which is known in the scale of decades as El Niño-Southern Oscillation (ENSO). For this purpose, time series of the anomalies and persistences of the Southern Oscillation Index (SOI), Multivariate ENSO Index (MEI) and of the Sea Surface Temperature (SST) were investigated. Methods of the dynamical and conditional shannon entropies were applied for the investigation of the predictability of symbolic sequences derived from the time series. The investigation of the conditional entropies for symbolic sequences shows that the most probable Events of ENSO occur after constant short sequences. Time correlations are found for several events; these determine the predictability of a sequence as a function of its length. The evolutions of short sequences representing transitions between ENSO states are relatively less predictable. The most predictable short sequences have been studied in detail. It was further found that, in most cases, SST is the most reliable information source. The analysis of the wavelet spectra of the time series shows strong periodicities of 2 to 4 years, which appear between 1900 and 1960, and between 1970 and 2000 in ENSO. There is evidence of a non-markovian process being responsible for these frequency components. Furthermore, the anomalies of the SST series show a gradient of frequency components towards smaller periods.