Effects of additive noise on the stability of glacial cycles

TL;DR

This study examines how additive noise affects the stability and predictability of glacial cycles modeled by low-order dynamical systems, revealing that noise can cause trajectories to cluster around stable paths and exhibit sensitive dependence.

Contribution

It demonstrates how stochastic fluctuations influence the stability of glacial cycle models with various attractors, highlighting the role of noise in climate trajectory divergence.

Findings

Noise induces clustering of trajectories around stable paths.

Chaotic models show significant trajectory leakage due to noise.

Strange nonchaotic attractors exhibit intermediate sensitivity.

Abstract

It is well acknowledged that the sequence of glacial-interglacial cycles is paced by the astronomical forcing. However, how much is the sequence robust against natural fluctuations associated, for example, with the chaotic motions of atmosphere and oceans? In this article, the stability of the glacial-interglacial cycles is investigated on the basis of simple conceptual models. Specifically, we study the influence of additive white Gaussian noise on the sequence of the glacial cycles generated by stochastic versions of several low-order dynamical system models proposed in the literature. In the original deterministic case, the models exhibit different types of attractors: a quasiperiodic attractor, a piecewise continuous attractor, strange nonchaotic attractors, and a chaotic attractor. We show that the combination of the quasiperiodic astronomical forcing and additive fluctuations induce a form of temporarily quantised instability. More precisely, climate trajectories corresponding to different noise realizations generally cluster around a small number of stable or transiently stable trajectories present in the deterministic system. Furthermore, these stochastic trajectories may show sensitive dependence on very small amounts of perturbations at key times. Consistently with the complexity of each attractor, the number of trajectories leaking from the clusters may range from almost zero (the model with a quasiperiodic attractor) to a significant fraction of the total (the model with a chaotic attractor), the models with strange nonchaotic attractors being intermediate. Finally, we discuss the implications of this investigation for research programmes based on numerical simulators. }