Stochastic Paleoclimatology: Modeling the EPICA Ice Core Climate Records

TL;DR

This paper models and analyzes the stochastic behavior of paleoclimate data from the EPICA ice core over 800 kyr, revealing insights into climate variable coupling and stability during glacial cycles.

Contribution

It introduces a stochastic modeling framework for paleoclimate time series, incorporating multifractal analysis and coupled dynamical systems to understand climate variable interactions.

Findings

Methane and nitrous oxide destabilize climate systems.

Carbon dioxide and temperature have stabilizing effects.

The model reveals causal relationships during glacial transitions.

Abstract

We analyze and model the stochastic behavior of paleoclimate time series and assess the implications for the coupling of climate variables during the Pleistocene glacial cycles. We examine 800 kyr of carbon dioxide, methane, nitrous oxide, and temperature proxy data from the EPICA Dome-C ice core, which are characterized by 100~kyr glacial cycles overlain by fluctuations across a wide range of time scales. We quantify this behavior through multifractal time-weighted detrended fluctuation analysis, which distinguishes near red-noise and white-noise behavior below and above the 100~kyr glacial cycle respectively in all records. This allows us to model each time series as a one-dimensional periodic non-autonomous stochastic dynamical system, and assess the stability of physical processes and the fidelity of model-simulated time series. We extend this approach to a four-variable model with linear coupling terms, which we interpret in terms of the interrelationships between the time series. Methane and nitrous oxide are found to have significant destabilizing influences, while carbon dioxide and temperature have smaller stabilizing influences. We draw conclusions about causal relationships in glacial transitions and the climate processes that may have facilitated these couplings, and highlight opportunities to further develop stochastic modeling approaches.