Online Program Home
  My Program

All Times EDT

Abstract Details

Activity Number: 271 - Statistical Modeling and Uncertainty Quantification for Atmospheric Remote Sensing Retrievals
Type: Topic Contributed
Date/Time: Tuesday, August 4, 2020 : 1:00 PM to 2:50 PM
Sponsor: Section on Physical and Engineering Sciences
Abstract #313242
Title: Bayesian Uncertainty Quantification for Atmospheric CO2 Retrieval Using Functional Principal Component Based Emulators
Author(s): Anirban Mondal* and Jonathan Hobbs and Pulong Ma and Emily Kang and Bledar Konomi
Companies: Case Western Reserve University and Jet Propulsion Laboratory and The Statistical and Applied Mathematical Sciences Institute and University of Cincinnati and University of Cincinnati
Keywords: Functional Principal Component; Active Subspace; Nearest Neighbor Gaussian Process; Emulator; Remote Sensing; Markov Chain Monte Carlo
Abstract:

NASA's Orbiting Carbon Observatory-2 collects tens of thousands of observations of reflected sunlight daily, and the mission's retrieval algorithm processes these indirect measurements into estimates of atmospheric CO2 and other states. The retrieval is an inverse problem and consists of a physical forward model for the transfer of radiation through the atmosphere. Here we use a Bayesian approach, which casts this inverse solution as the posterior distribution of the state given the observed spectra. The posterior distribution is not available in closed form and MCMC method is used to sample from the posterior, which requires a large number of evaluations of the expensive forward model. So, here we focus on the emulator approach where a statistical representation of the forward model is built based on some simulation runs of the forward model using nearest neighbor Gaussian process. We use functional principal component analysis approach to reduce the dimension of the functional radiance data and active subspace approach to reduce the dimension of the input state vector. The emulator runs instantaneously resulting in a computationally efficient retrieval algorithm based on MCMC.


Authors who are presenting talks have a * after their name.

Back to the full JSM 2020 program