Analysis of Connected Climate Systems via Deconvolution

Abstract:

Deconvolution is a technique most often used in signal and image processing to remove the effects of a system’s impulse response and recreate the input signal from a given output. In the context of paleoclimate, deconvolution by spectral division has been used to recover the climate system’s impulse response, also known as its transfer function, given the δ¹⁸O time series record of the north pole as the input and the south as the output (or vice versa). The working hypothesis of polar synchronization justifies the use of deconvolution methods. Various regularization approaches and spectral analysis show a clear connection of millennial scale periodicity linking the polar climates over the past 100,000 years. Tests of spectral peak consistency across regularization factors and of peak validity indicate that the connection is a result of the data and is not an artifact of the method used. Deconvolution can be applied to other linearly connected climate systems including teleconnected systems. Sea surface temperature dipoles found in the North Atlantic Ocean basin, for example, also display potentially geographically linked features, and correlation between the dipoles themselves suggests synchronization of adjacent dipoles. Having identified this system of synchronized variations with linear phase relations, deconvolution methods can be used to investigate potential transfer functions across different scales.