This paper tests the ability of Artificial Intelligence methods, and more specifically Deep Learning, to eliminate the Gaussian noise that disturbs the data of a dynamic system. The authors demonstrate this using a highly chaotic model as a hard test case.

The paper presents a valuable review of state-of-the-art large ensemble methodology, comparing observational-based analogues and model-generated perturbations with the aim of studying the impact of long-term variability on European past, present and future climate. This work is helpful as a theoretical and methodological benchmark for a number of open issues in ensemble modeling, that are currently the object of intense discussions in the research community. Given that this topic very much relates to the work done within the Coupled Model Intercomparison Project, and ultimately to the redaction of the IPCC Assessment Reports, the paper is of major interest for the research community as well as for the broader public.

This study suggests a necessary shift in the paradigm of nonlinear dynamical systems analysis in climate science, ecology, and beyond. The author delves into a relatively unexplored facet of critical transitions, which is of great importance in examples within the Earth and environmental sciences. In particular, she explains mechanisms based on these ideas in examples of systems that describe population dynamics and climate transitions.

This is an interesting contribution to the long-running debate on the anisotropic scaling of atmospheric dynamics, confirming that it is neither 3D nor 2D, but rather 23/9 D. This is of interest to the whole atmospheric community.

This is a new and timely contribution to the blocking phenomena in atmospheric circulation and extremes that result. The latter are stimulating and of fairly general interest to the non-linear geoscience community and beyond.