R. Donner (with A. Witt):
Temporary Dimensions of Multivariate Palaeoclimate Records - Signatures of Climate Change ?


Abstract:
Palaeoclimatic variations are recorded in geological sequences and observed in terms of sets of parameters reflecting
different aspects of climatic history. However, time series arising from corresponding measurements are typically rather
short and insufficiently resolved such that ''standard'' measures in nonlinear data analysis are not appropriate for
characterizing such data sets.
To overcome this problem, we investigate the estimation of fractal dimensions based on the decay of eigenvalues of the
covariance matrix. Our method is demonstrated by studying numerical examples yielding a qualitative robustness of our
approach.
As a particular example, we apply the method to chemical trace element abundances recorded in the marine sediment
offshore the East Antarctic coast. The temporal variability of our dimension estimates is used as a novel measure for
characterizing long-term climate change. The influence of extreme events within the data is briefly discussed.