Exploring the structure of the mid-Pleistocene revolution with advanced methods of time series analysis

The mid-Pleistocene climate transition is a complex global change leading to Late Pleistocene ice ages with increased mean ice volume and dominant 100-ka cycle. A thorough understanding of this transition demands quantitative investigations in the time and frequency domains, and in the ''stochastic domain''. Three methods of time-series analysis are presented which have been adapted for this purpose. They are tested by means of predefined, artificial time series and applied to benthic oxygen isotope (delta(18)O) records which serve as ice volume indicator. Results are as follows: (a) The time-dependent mean shows an increase of 0.35 parts per thousand, vs PDB from 942 to 892 ka. (b) Evolutionary spectral analysis reveals an abrupt increase of 100-ka cycle amplitude at approximately 650 ka. (c) Probability density function exhibits a bifurcation behavior at approximately 725 ka. These findings point to a multiple transition from a more linear climate system to a strong nonlinear system. The significant lead of the transition in mean, in relation to the 100-ka amplitude change and bifurcation is left open for explanation.