A Simple Emulator That Enables Interpretation of Parameter‐Output Relationships, Applied to Two Climate Model PPEs DOI Creative Commons
Qingyuan Yang, Gregory S. Elsaesser, Marcus van Lier‐Walqui

et al.

Journal of Advances in Modeling Earth Systems, Journal Year: 2025, Volume and Issue: 17(5)

Published: May 1, 2025

Abstract We present a new additive method, referred to as sage for Simplified Additive Gaussian processes Emulator, emulating climate model Perturbed Parameter Ensembles (PPEs). estimates the value of output sum terms. Each term is mean Process, and corresponds impact parameter or group on variable interest. This design caters sparsity PPEs, which are characterized by limited ensemble members high dimensionality space raise issue sensitivity in emulator design. quantifies variability explained different parameters groups, providing additional insights parameter‐climate relationship. apply two PPEs compare it fully connected Neural Network. The methods have comparable performance with both but provides importance well diagnostics useful optimizing PPE Insights gained from applying method comparing its Network pointed out not been previously addressed. Our work highlights that analyzing used train an data generated trained PPE, former more structure could help inform also proposes questions optimal way working PPEs.

Language: Английский

A Simple Emulator That Enables Interpretation of Parameter‐Output Relationships, Applied to Two Climate Model PPEs DOI Creative Commons
Qingyuan Yang, Gregory S. Elsaesser, Marcus van Lier‐Walqui

et al.

Journal of Advances in Modeling Earth Systems, Journal Year: 2025, Volume and Issue: 17(5)

Published: May 1, 2025

Abstract We present a new additive method, referred to as sage for Simplified Additive Gaussian processes Emulator, emulating climate model Perturbed Parameter Ensembles (PPEs). estimates the value of output sum terms. Each term is mean Process, and corresponds impact parameter or group on variable interest. This design caters sparsity PPEs, which are characterized by limited ensemble members high dimensionality space raise issue sensitivity in emulator design. quantifies variability explained different parameters groups, providing additional insights parameter‐climate relationship. apply two PPEs compare it fully connected Neural Network. The methods have comparable performance with both but provides importance well diagnostics useful optimizing PPE Insights gained from applying method comparing its Network pointed out not been previously addressed. Our work highlights that analyzing used train an data generated trained PPE, former more structure could help inform also proposes questions optimal way working PPEs.

Language: Английский

Citations

0