Analysis of a bistable climate toy model with physics-based machine learning methods

We propose a comprehensive framework able to address both the predictability of first and second kind for high-dimensional chaotic models. For this purpose, we analyse properties newly introduced multistable climate toy model constructed by coupling Lorenz '96 with zero-dimensional energy balance model. First, attractors system are identified Monte Carlo Basin Bifurcation Analysis. Additionally, detect Melancholia state separating two attractors. Then, Neural Ordinary Differential Equations applied in order predict future