Epistemic and aleatoric uncertainty quantification in weather and climate models

This paper looks at the uncertainties in parameterisations in weather and climate models, using Bayesian Neural Networks to break uncertainties down into epistemic uncertainty (from uncertainty in the model) and aleatoric uncertainty (from natural variability in the data).

For a short summary of the paper, you can find my blog post here, and the full, open-access, paper is available here