Emulation of climate model output using machine learning or other data driven approaches. Chair: Vassili Kitsios (CSIRO)
Link: S2 B2 - Google Docs
Description: Reduced-order models of the climate, or climate emulators, approximate the dynamics of the climate in a computationally cheap fashion as compared to a full complexity general circulation model (GCM). These reduced-order models can be developed from fundamental principles by instead representing a lower dimensional system. They can also be developed using machine learning, statistical learning or a variety of other data-driven methods, which exploit existing GCM output, reanalysis data, or real-world observations. Computationally cheap climate emulators will enable: a more complete assessment of climate risk for a broader range of climate emissions scenarios; and exploration of science questions and methods development via a hierarchy of representative models.
Conclusions/Actions from Breakout Session
Please use this thread for further discussion on this talk.