- Arvind Saibaba (Mathematics, NCSU)
- Julianne Chung (Mathematics, Emory University)
- Scot Miller (Environmental Health Engineering, John Hopkins University)
Prerequisites: Linear algebra, interest in programming.
Outline: Atmospheric inverse modeling is used to estimate greenhouse gas fluxes or air pollution emissions at the Earth’s surface using observations of these gases collected in the atmosphere. The launch of new satellites, the expansion of surface observation networks, and a desire for more detailed maps of surface fluxes have yielded many computational and statistical challenges. Previous work have resulted in efficient iterative methods  for solving the inverse problem and quantifying the reconstruction accuracy; however, the solution process depends crucially on the choice of good hyperparameters.
Objectives: This project will investigate methods, using surrogate models, for selecting good hyperparameters in a computationally efficient manner. Examples of surrogate models under consideration include Gaussian processes, Neural Networks, and Tensor-based function approximation.
Outcomes: Successful code developed as a part of this project will be incorporated into publicly developed code  and potential publication in a peer-reviewed journal.
References: Cho, T., et al., Computationally efficient methods for large-scale atmospheric inverse modeling. Geosci Model Dev, 2022. 15: p. 5547-5565.  Cho, T., Inverse-modeling/genHyBRmean: Efficient methods for large-scale atmospheric inverse modeling. 2021: https://zenodo.org/record/5772660#.Y0XEvuzMKL6 [accessed Oct 2022].