Caltech Hybrid Modeling Tools

• ACTM Performer: Caltech team;

• Author: Costa Christopoulos (cchristo@caltech.edu) and Tapio Schneider (tapio@caltech.edu)

This section provides an overview of the CliMA code, including links to documentation containing tutorials. The modeling and learning framework for this project is modular, and code for each component resides in unique repositories. A hybrid machine-learning model (TurbulenceConvection.jl) is calibrated with Ensemble Kalman Processes in CalibrateEDMF.jl. As training data we employ a library of Large Eddy Simulations (LES) [Data], driven by conditions found in state of the art climate simulations at selected location on the globe.

Code repositories and data libraries associated with the project:

CliMA Packages

Package	Code	Docs	Purpose
CalibrateEDMF.jl	Link	Link	Framework to learn about cloud processes from library of LES Data
EnsembleKalmanProcesses.jl	Link	Link	Implementation of gradient-free optimization techniques
TurbulenceConvection.jl	Link	Link	Implementation of EDMF scheme of turbulence, convection and clouds
OperatorFlux.jl	Link		A machine learning package for Fourier Neural Operators
LES library	Data		LES generated training data at current climate and 4K warming simulations