Towards a GCM Emulator for Exoplanet Applications
Daniel Abdulah, Massachusetts Institute of Technology
The exploration of exoplanetary climates has significantly advanced over the past years owing to the generalization of new climate models and information-rich data from a new generation of observatories (e.g., JWST). While these new atmospheric models have a fidelity adequate to interpret JWST data, their computation time is a key bottleneck to leverage them in standard Bayesian analysis framework. Following recent groundbreaking successes for Earth climate models, we aim to develop an climate emulator for the broader parameter space associated with exoplanets. To this end, we turn to MIT’s SuperCloud to generate a training set of 10,000-climate simulations with the ExoCAM model, varying orbital obliquity, length of day, atmospheric abundance of CO2, and atmospheric abundance of H2. We then train two emulators on this data with the task of generating T-p profiles and a 2D surface temperature field respectively.