Special Session 191: Stochastic Dynamical Systems Under Levy Noise: Theory and Applications

Levy noise versus Gaussian-noise-induced transitions in the Ghil-Sellers energy balance model
Larissa Serdukova
University of Leicester, Department of Computing and Mathematical Science
England
Co-Author(s):    Valerio Lucarini, University of Leicester, UK. Georgios Margazoglou, Moody`s Corporation, UK
Abstract:
We investigate the effects of stochastic forcing on the Ghil-Sellers energy balance climate model by introducing fluctuations in solar irradiance. Using numerical simulations, we analyse noise-induced transitions between coexisting warm and snowball climate states. We consider multiplicative stochastic forcing driven by Gaussian noise and $\alpha$-stable L\'{e}vynoise with $\alpha \in (0,2)$, focusing on the statistics of transition times and the structure of the most probable transition paths. While Gaussian noise-induced transitions in metastable systems are well understood and serve here as a reference case, much less is known about the corresponding behaviour under L\'{e}vynoise, particularly in high- or infinite-dimensional systems. In the weak-noise regime, we find fundamentally different scaling laws for residence times in metastable states. For Gaussian noise, the classical Kramers-type exponential scaling is recovered, whereas L\'{e}vy noise leads to power-law scaling with exponent $-\alpha$, consistent with rigorous results for related systems with additive L\'{e}vy forcing. This behaviourcan be interpreted by modelling the L\'{e}vy noise as a compound Poisson process. We also examine transition paths in a reduced state-space projection. Gaussian-driven transitions pass through the same unstable edge state, while L\'{e}vy-driven transitions follow distinct paths through the closest basin boundary to the originating attractor.