A Method to Separate Proxy Climate Records into Relevant Temporal and Individual Periodic Signals and Application to the "LR04" Benthic Foraminifera δ18O Stack

Abstract:

The objective of the effort reported here was the development of a method to extract the temporal signature and the individual periodic and quasi-periodic time-varying signals from proxy climate data. The effort was based on the assumption that the explicit delineation of each periodic term might ultimately be useful in helping to identify the mechanism for the 100-kyr glacial cycle in Quaternary proxy data. The method is an iterative application of both fixed and variable period filters scanned across the applicable portions of the data set. In the current application, the window for each filter is a single period wide, allowing for the maximum retention of the time-varying amplitude of the signal. An iterative process, including a time-series term, is required in order to counter the very broadband nature of the filter. Details of the method will be presented and the results of its application to the “LR04” Stack (Lisiecki and Raymo, 2005) will be discussed.  Graphical correlation of the individual signals with Obliquity, Precession, and Eccentricity (Laskar, et al, 2004) will be discussed. After removal of a constant from the data, the first thru third-order time series accounts for 75% of the variability in the data. Seven statistically justifiable periodic terms account for 94% of the remaining variability. The shortest period used was the 22.2 kyr average of the 3 precession period signals identified by spectrum analysis. A very strong obliquity driven term of 41.0 kyr was also observed as expected due to tuning of the data to that term. Averaging of the spectrum analysis data was used to determine a value of 96.05 kyr for the glacial cycle. The four remaining terms had periods between 180 and 2000 kyr and were determined using the variable period filters. Shorter period terms were not used because of their excessive requirements on the degrees of freedom available in the data set.