Questo evento è disponibile solo in lingua inglese
Lecturer
Michel Crucifix, Université Catholique de Louvain, Belgium
Abstract
Traditional time-series analysis, such as Fourier transform, often give frustratingly disappointing results when applied to paleoclimate data. For example, a Fourier analysis of the Pleistocene climate stack record essentially indicates a red-noise, while the experimentalist’s trained eye actually ‘sees’ something. Is is just wishful thinking?
Not necessarily. One must realise that most statistical significance algorithms test the data against the null hypothesis that the data contains no other information than red noise. In practice this is not what the experimentalist wants to test. He or she has in mind a much more elaborate model, about the processes that have generated the data, the way they have been archived, and the way they have been measured. These positive bits of information are crucial to analyse the paleoclimate record.
The present talk is a case study. We start form the EPICA CO2 record as an example, and illustrate how accounting for information in a modelling framework allow us to extract information from the data about the response to the astronomical forcing, noise levels, and even about the stability of the climate system. We also illustrate some of the pitfalls, which exist in most time-series analysis approaches, but which can be analysed more explicitly here.
Organizer
Carlo Barbante
Venezia, Italy, DAIS - Università Ca' Foscari di Venezia - San Giobbe - Aula E - DAIS - Università Ca' Foscari di Venezia - San Giobbe - Aula E, Venezia, Italy -
24 Apr 2013
Website