Fig. 4: 2D metric space <WWD, DACE>.
From: network-based constraint to evaluate climate sensitivity
Each of the 27 CMIP6 model is represented by a marker with its ensemble mean Weighted Wasserstein Distance \(({WWD})\) value on the x-axis and its ensemble mean Distance Average Causal Effect (\({D}_{{ACE}}\)) value on the y-axis. \({WWD}\) and \({D}_{{ACE}}\) values are the mean values of the metrics computed with respect to both HadISST and COBEv2. Number in parenthesis in the legend refers to the number of ensemble members. The ground truth is represented at the bottom left corner (0,0): the closer the marker is to the origin, the more realistic the model. To evaluate the performance of a new model, one can compute its \({WWD}\) and \({D}_{{ACE}}\) values and locate the model in the 2D metric space as a first assessment. \({WWD}\) and \({D}_{{ACE}}\) are moderately correlated, with Pearson correlation coefficient equals to 0.62.
