The Elo rating system as a new narrative for ranking climate models

Monerie, P.-A. ORCID: https://orcid.org/0000-0002-5304-9559 (2026) The Elo rating system as a new narrative for ranking climate models. Environmental Research Communications. ISSN 2515-7620 doi: 10.1088/2515-7620/ae661c (In Press)

Abstract/Summary

It is challenging to define the models best suited to simulating specific aspects of the climate system in particular regions. Selecting models for focused analysis (e.g., impact studies) requires an in-depth understanding of their ability to simulate processes of interest. Ranking models using multiple metrics and variables is difficult. Furthermore, a root-mean- square error-based approach, for example, can lead to conclusions that can be difficult to interpret. For example, I show that a Model A produces a standardised root-mean-square error that is 0.55 lower than a Model B. While a 0.55 reduction in standardised root-mean- square error confirms a better fit, it is not immediately intuitive. Instead, I propose using the Elo rating system to rank climate models. This framework enables multiple variables and metrics to be easily combined, providing a new probabilistic and comparable narrative. For example, Model A can be assigned an 82.5% probability of outperforming Model B, mapping their performance gap onto an intuitive 0-100% skill scale. I suggest that interdisciplinary collaborations are key to evaluating the climate system's many facets. By pooling expertise and data across domains, we can use the Elo rating framework to build a more comprehensive picture of model performance.

Item Type	Article
URI	https://centaur.reading.ac.uk/id/eprint/129563
Identification Number/DOI	10.1088/2515-7620/ae661c
Refereed	Yes
Divisions	Science > School of Mathematical, Physical and Computational Sciences > NCAS Science > School of Mathematical, Physical and Computational Sciences > Department of Meteorology
Publisher	IOP Science
Download/View statistics	View download statistics for this item