Accelerating climate models using an AI surrogate
We accelerate Community Atmospheric Model (CAM) simulations by replacing the moist physics and radiation with an AI surrogate using a novel python-fortran bridge (TorchClim) in the Community Earth System Model (CESM). We show that a deep neural network surrogate trained on data from CAM itself can produce a stable model that reproduces the climate and variability of the original model, albeit with some biases.
ACCESS_NRI_Prasad.pdf (4.5 MB)
Note: this topic is part of the 2023 ACCESS Community Workshop Poster session