Selecting CMIP6 Models for Future Arctic Storylines Using a Novel Performance Score

Storylines are physically plausible scenarios of future climate change, statistically derived from an ensemble of climate model projections and organized according to the magnitude of projected changes in two or more remote drivers that strongly influence the spatial pattern of the climate response. Here, we provide novel insights into the Arctic storylines identified by Levine et al. (2024), where Barents-Kara Sea warming and lower-tropospheric Arctic warming during the extended summer season (May–October) were remote drivers, as we identify a set of models from the Coupled Model Intercomparison Project phase 6 to represent the storylines. We do this by first identifying models that are similar to these storylines in terms of each remote driver response and quantifying this similarity. Second, we evaluate the model’s performance in terms of a simple performance score based on the mean normalized root-mean-square error for multiple climate variables of importance for the storylines. The normalized values vary between 0 and 1 for all variables, allowing them to exert a comparable influence on the score. The advantage of the score is that it provides an easily implementable and interpretable way of identifying models that are characterized by large errors relative to the rest of the ensemble. Finally, we combine the similarity estimate and the score to select models to represent the storylines. We focus on the Arctic during the extended summer season for which the storylines were designed, but also consider other seasons and regions. Through this exercise, we also document the methodology, benefits, and limitations of the score.