Semi‐Automatic Tuning of Coupled Climate Models With Multiple Intrinsic Timescales: Lessons Learned From the Lorenz96 Model.

The objective of this study is to evaluate the potential for History Matching (HM) to tune a climate system with multi‐scale dynamics. By considering a toy climate model, namely, the two‐scale Lorenz96 model and producing experiments in perfect‐model setting, we explore in detail how several built‐in choices need to be carefully tested. We also demonstrate the importance of introducing physical expertise in the range of parameters, a priori to running HM. Finally we revisit a classical procedure in climate model tuning, that consists of tuning the slow and fast components separately. By doing so in the Lorenz96 model, we illustrate the non‐uniqueness of plausible parameters and highlight the specificity of metrics emerging from the coupling. This paper contributes also to bridging the communities of uncertainty quantification, machine learning and climate modeling, by making connections between the terms used by each community for the same concept and presenting promising collaboration avenues that would benefit climate modeling research. Plain Language Summary: Climate models are computer simulation codes that incorporate centuries of human knowledge of the physics of planet Earth. They are used to understand the past, the present and make projections about the future of our climate. To validate a climate model, scientists tune a number of its parameters so that it yields a simulated climate resembling real‐life observations as much as possible. The main challenge in this tuning task is the extreme cost of climate models which limits a lot the number of tuning experiments scientists can run. In this paper we are interested in a technique that uses artificial intelligence in order to replace the expensive climate model with a cheaper surrogate. We experiment on a simplified model to assess the strengths and weaknesses of this semi‐automatic technique, and show that it can be more efficient when combined with human expertise. Key Points: The History Matching method is explained in detail then used for tuning a toy coupled model: the Lorenz 96 modelThe importance of several design choices is demonstrated, especially when considering forced experiments such as Atmospheric Model Intercomparison Protocol and Ocean Model Intercomparison ProjectWe argue that this tuning method is semi‐automatic & highlight the importance of human expertise when considering it for real coupled models [ABSTRACT FROM AUTHOR]