Research Projects
My research focuses on developing computationally efficient methods to identify greenhouse gas (GHG) emission sources using machine learning. I am currently using machine learning models to compute atmospheric transport in GHG flux estimation using Bayesian inference. Here are some of the projects that I have been working on:
FootNet is a physics-guided deep learning model that can emulate atmospheric transport. It is trained on half a million examples generated using a physics-based model (STILT). It is 1000x faster than STILT and can compute footprints on-the-fly.
Here we developed a computationally efficient method for scalable GHG flux inversions using a deep learning emulator of atmospheric transport (FootNet).
