Understanding and reducing uncertainty in atmospheric and Earth system models requires effective ways to explore model behaviour, compare different models and connect simulations with observations. Key challenges include identifying the processes that cause model uncertainty, distinguishing uncertain parameters from differences in model structure, and establishing whether observational constraints remain robust across models.

This session brings together modelling, observational and statistical communities developing methods and applications that address these challenges. We welcome both single-model and multi-model studies, including perturbed parameter ensembles, coordinated model experiments and model intercomparison projects.

Topics include:

Ensemble design, sensitivity analysis and attribution of uncertainty within and across models.

Statistical and machine-learning emulation of individual models and multi-model ensembles, including approaches that capture shared behaviour and model-specific differences.

Bayesian calibration, history matching, model weighting, and emergent or process-based observational constraints.

Joint use of multiple observations, accounting for observational uncertainty, model dependence and structural differences.

Evaluation of how uncertainties propagate across coupled model components and how coupling alters relationships between model variables and observations.

Diagnosing structural model deficiencies using model–observation comparisons to guide model development.

Designing observations and model experiments to improve constraints on climate uncertainty.

Applications across atmospheric composition, clouds, physical atmospheric processes and their coupling with other Earth system components are encouraged, from regional to global scales. We particularly welcome studies investigating whether constraints derived from one model remain informative across other models, and whether improved agreement with observations leads to more reliable climate projections.