GC006
Advances in Emulating Earth System Models
Advances in Emulating Earth System Models
Session ID#: 251950
Session Description:
Running Earth system models (ESMs) requires a heavy computational cost. As a result, ESMs are often only run for a few emission scenarios with a limited number of realizations per scenario. Emulators or reduced-complexity models are used to interpolate in between the simulated scenarios or reduce internal variability uncertainty. But, existing emulators, such as linear pattern scaling, can break down for emulating irreversible changes, aerosol forcings, and compound and rarest extreme events. Further, emulating high spatiotemporal outputs that capture the internal variability of the Earth system remains challenging.
With autoregressive machine learning emulators on the rise, we also welcome submissions from the reduced complexity modeling communities to encourage an interdisciplinary exchange including, but not only:-Benchmarking/validation protocols
-Applications of emulators
-Theoretical insights for interpretability, consistency, etc.
-Reduced-complexity models
-Probabilistic emulation techniques, such as Weather generators, diffusion, GPs, EVT
-(Hybrid) deep learning approaches, including operator, autoregressive, foundation models-Atmosphere-ocean coupled emulators
With autoregressive machine learning emulators on the rise, we also welcome submissions from the reduced complexity modeling communities to encourage an interdisciplinary exchange including, but not only:-Benchmarking/validation protocols
-Applications of emulators
-Theoretical insights for interpretability, consistency, etc.
-Reduced-complexity models
-Probabilistic emulation techniques, such as Weather generators, diffusion, GPs, EVT
-(Hybrid) deep learning approaches, including operator, autoregressive, foundation models-Atmosphere-ocean coupled emulators
Co-Sponsor(s):
- A - Atmospheric Sciences
- IN - Informatics
- NG - Nonlinear Geophysics
- OS - Ocean Sciences
Index Terms:
1616 Climate variability [GLOBAL CHANGE]
1622 Earth system modeling [GLOBAL CHANGE]
1942 Machine learning [INFORMATICS]
4468 Probability distributions, heavy and fat-tailed [NONLINEAR GEOPHYSICS]
Primary Convener: Björn Lütjens, IBM Research, Yorktown Heights, United States
Conveners: Kalyn Dorheim, St. Olaf College, Northfield, United States, Paolo Giani, Massachusetts Institute of Technology; Department of Earth, Atmospheric and Planetary Sciences, Cambridge, United States and Pedram Hassanzadeh, Rice University, Department of Mechanical Engineering and Department of Earth, Environmental and Planetary Sciences, Houston, United States
Student/Early Career Convener: Björn Lütjens, IBM Research, Yorktown Heights, United States
See more of: Global Environmental Change