This applied axis addresses the urgent need for actionable, high-resolution climate information to support mitigation, adaptation, and policy decisions. Climate models have become essential predictive tools, helping stakeholders assess climate risks and implement adaptation strategies. The axis focuses on developing advanced modeling capabilities, improving physical understanding, reducing uncertainties, and exploring hybrid methods that combine physical modeling with artificial intelligence (AI).
Four main challenges guide this axis:
-
Providing actionable climate information: CERFACS aims to co-develop high-resolution, physically-grounded climate projections with stakeholders to support adaptation planning. This includes using kilometer-scale simulations to assess extreme events (e.g., heatwaves, storms) and developing hybrid downscaling and event-based approaches using AI. The use of SMILEs (Single Model Initial-condition Large Ensembles) and decadal predictions will improve our understanding of near-term climate variability (1–30 years), a critical time horizon for adaptation.
-
Improving model accuracy and reducing uncertainty: Enhancing model components—particularly ocean circulation (AMOC, ENSO), Arctic dynamics, and sea-ice representation using advanced rheology—will improve reliability. Coupled biogeochemical processes (e.g., oxygen minimum zones) and improved model calibration via perturbed parameter ensembles and AI will also be prioritized. Additionally, observational uncertainties, particularly in extreme precipitation, will be better quantified.
-
Understanding aviation–climate interactions: Given aviation’s rising contribution to radiative forcing, CERFACS will refine the representation of contrails in climate models, assess non-CO₂ effects using simplified models, and evaluate aviation-related hazards (icing, turbulence) using kilometer-scale simulations and surrogate models. This research directly supports the decarbonization goals of shareholders like Airbus, SAFRAN, and TotalEnergies.
-
Exploring hybrid and data-driven approaches: Deep learning will be used to emulate complex models, improve signal-to-noise interpretation in climate simulations, and support model calibration and uncertainty quantification. These approaches enable faster, cost-effective diagnostics, especially for extremes and rare events.
Strategically, this axis is grounded in the following Stratetic Axes : Data-Driven Modeling (machine learning, surrogate modeling, uncertainty quantification) and Sustainable Programming (efficient HPC usage, FAIR data principles, and digital twins). These efforts align with major programs such as TRACCS (France) and European projects like Impetus4Change and MOSCITO.
CERFACS works closely with key shareholders (Météo-France, CNES, EDF) and new partners (IRD, Global South institutions), reinforcing its national and international leadership in climate science. The axis aims to produce robust, scalable tools and knowledge to meet society's climate challenges from global scales down to actionable local insights.
