Stationary determinism in Observed Time Series: the earth's surface temperature

Abstract

In this work we address the feasibility of estimating and isolating the stationary and deterministic content of observational time series, {\bf Ots}, which in general have very limited characteristics. In particular, we study the valuable earth's surface mean temperature time series, {\bf Tts}, by applying several treatments intended to isolate the stationary and deterministic content. We give particular attention to the sensitivity of results on the different parameters involved. The effects of such treatments were assessed by means of several methods designed to estimate the stationarity of time series. In order to strengthen the significance of the results obtained we have created a comparative framework with seven test time series of well-know origin and characteristics with a similar small number of data points. We have obtained a greater understanding of the potential and limitations of the different methods when applied to real world time series. The study of the stationarity and deterministic content of the {\bf Tts} gives useful information about the particular complexity of global climatic evolution and the general important problem of the isolation of a real system from its surroundings by measuring and treating the obtained observations without any other additional information about the system.