Uncertainty quantification for remote sensing Earth system data records
Conference
64th ISI World Statistics Congress - Ottawa, Canada
Format: IPS Abstract
Keywords: remote sensing, uncertainty quantification
Session: IPS 448 - New Statistical Methods for Surrogate Modeling and Inverse Problems
Tuesday 18 July 2 p.m. - 3:40 p.m. (Canada/Eastern)
Abstract
Remote sensing data have provided unique observations of the Earth system for multiple decades, now spanning multiple generations of satellite technology and dedicated missions from multiple international agencies. Integration of data from multiple existing and future platforms can enable unique science investigations and construction of climate data records. Combined use of these remote sensing data records can be aided by a unified treatment of uncertainty quantification (UQ) for the data processing pipeline. This work illustrates the application of a common framework for UQ for the satellite retrieval inverse problem. For many atmospheric remote sensing missions, the retrieval is a mathematical and computational method for inferring geophysical quantities of interest from observed satellite spectra. We will highlight a simulation-based UQ framework and its application to atmospheric carbon dioxide from the NASA’s Orbiting Carbon Observatory-2 and -3 (OCO-2/3) missions. The framework incorporates multiple sources of epistemic and aleatoric uncertainty and provides a statistical model that can be applied to operational data products from the missions.