NEMO working group on Machine Learning and Model Uncertainties
co-chairs : Andrea Storto and Julien Le Sommer
- Organize and plan the development of NEMO for modeling uncertainty (stochastic physics and ensembles)
- Organize and plan the development of NEMO in connection with machine learning
- Ensure liaison between projects focusing on NEMO-based ML applications and the NEMO DevCom and organize discussions on associated topics if relevant.
- Ensure liaison between projects focusing on using NEMO in connection with DA frameworks and the NEMO DevCom and organize discussion on associated topics if relevant.
Scientific discussion topics :
- Uncertainty quantification: learning representations of model errors and tuning model parameters.
- Differentiable emulation (of individual code components or entire models)
- ML-based subgrid parameterization
- Explicit representation of uncertainties (ensemble generation, stochastic modelling, multi-physics simulations)