Bayesian Approach to CO2 Retrievals for the OCO-2 Instrument Using a Surrogate Forward-Model

• Activity Number: 24
• Type: Topic Contributed
• Date/Time: Sunday, August 9, 2015 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics and the Environment
• Author(s): Jenny Brynjarsdottir* and Amy Braverman and Jonathan Hobbs
• Companies: Case Western Reserve University and Jet Propulsion Laboratory and California Institute of Technology
• Keywords: Carbon cycle ; Remote sensing ; Bayesian analysis ; Adaptive MCMC

Abstract

After a successful satellite launch in July 2014, NASA's Orbiting Carbon Observatory-2 (OCO-2) mission is in the process of collecting space-based measurements to quantify variations in atmospheric carbon dioxide (CO2). Atmospheric CO2 columns are inverted from reflected sunlight in three spectral regions that make a single "sounding". For each sounding the inversion is performed with remote sensing retrieval algorithms that give an estimate of the posterior mode and posterior covariance matrix of the CO2 column. Even thought the forward model is not linear, the posterior distribution is usually treated as being Gaussian with these estimated moments. In an effort to obtain the full posterior distribution we perform the inversion with an adaptive MCMC algorithm. Since the forward model used in the remote sensing retrieval algorithms is too computationally expensive for MCMC iterations, we instead use a fast but physically realistic surrogate model.