Fig. 1: Schematic of the network-based constraint to evaluate Climate Sensitivity (netCS) approach.

First step is the mapping of gridded data into a lower-dimensional representation using Knowledge Discovery and Data mining algorithms. From large and complex Sea Surface Temperature (SST) simulations, it infers SST regions with δ-MAPS and discovers causal structure with the causal discovery algorithm PCMCI. The Tigramite approach allows to estimate the causal effects. Second step is the evaluation of the network with respect to reference networks (reconstructed from HadISST and COBEv2). Two distance metrics are proposed. Weighted Wasserstein Distance (\({WWD}\)) evaluates the patterns and the Distance Average Causal Effect (\({D}_{{ACE}}\)) evaluates the connectivity patterns between the main regions of the climate system. A weighting scheme approach^27,28,29 converts distances into weights, which allows to constrain both Equilibrium Climate Sensitivity (ECS) and Transient Climate Response (TCR) with new weighting distributions.