On December 18, 2025, Maëlle Coulon–Decorzens defended her PhD thesis entitled “The Art and Science of Climate Model Tuning: From Theory to Practice”, carried out under the supervision of Frédéric Hourdin and Najda Villefranque.

Climate models are essential tools for understanding the climate system and producing projections. They nevertheless rely on many uncertain parameters, whose adjustment—known as tuning or calibration—has long been performed in an empirical manner. The recent emergence of new statistical methods is transforming this key step in model development, making tuning a genuine object of scientific research.

During her PhD, Maëlle explored the potential of the History Matching with Iterative Refocusing (HMIR) method to improve both the tuning and the understanding of climate models. In particular, she developed a guided–zoomed configuration of the LMDZ model, making it possible to constrain model parameters using in situ daily observations (SIRTA site, Paris region), while limiting biases related to synoptic dynamics in order to focus on biases in physical parameterizations. This work led to the first tuning experiment of LMDZ directly targeted at site-level observations.

The thesis also investigated error compensations between clouds and radiative transfer when tuning global radiative fluxes. Using the HMIR method, Maëlle proposes an original approach to quantify these compensations and shows that taking 3D radiative effects into account helps avoid an artificial overestimation of cloud fractions in the model.

Finally, the feedback, methodological reflections, and critical perspectives offered by Maëlle constitute a valuable contribution to the evolution of tuning practices and, more broadly, to the future development of climate models.

Congratulations to Maëlle on her very rich thesis defense, which highlighted the quality and originality of her work!