Wavelet analysis of nonstationary and chaotic time series with an application to the climate change problem

Our aim here is to show merits of the so-called wavelet transform of time series as an analyzing tool of the simultaneously nonstationary and chaotic processes that usually appear in the Earth sciences. The merits are illustrated on some surrogate time series as well as the natural ones that are observed in the present-day climate change problem. As a result of a wavelet transform of the natural hemispheric averaged surface air temperature time series a crossover scale is recognized that cuts off the intra- and interdecadal temperature oscillations, which are obviously internally induced, from a secular trend-like component of unknown origin.