Machine learning for parameterization of subgrid processes in the atmosphere
Abstract
The parameterization of subgrid processes such as deep convection contributes to uncertainty in climate modeling and numerical weather prediction. Machine learning can be used to learn new parameterizations directly from high-resolution model output, but the resulting parameterizations do not always lead to stable and accurate simulations when implemented in a coarse-resolution model. Previous work suggests that a machine learning approach based on an ensemble of decision trees (random forest) can be used to robustly emulate a conventional moist convection scheme. Here we use this decision-tree approach to learn a parameterization from coarse-grained output of a quasi-global high-resolution model with hypohydostatic rescaling. We discuss the performance of the parameterization when implemented at coarse resolution in the same model with a focus on statistics of mean and extreme precipitation. In addition, we discuss approaches to improve generalization across climates and to provide insights into the underlying dynamics.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2019
- Bibcode:
- 2019AGUFM.A43E..07O
- Keywords:
-
- 3305 Climate change and variability;
- ATMOSPHERIC PROCESSES;
- 3337 Global climate models;
- ATMOSPHERIC PROCESSES;
- 3365 Subgrid-scale (SGS) parameterization;
- ATMOSPHERIC PROCESSES;
- 1942 Machine learning;
- INFORMATICS