Title: Data Fusion and Spatial Inference for Remote Sensing
Presenting Author: Amy Braverman
Organization: Jet Propulsion Laboratory

Abstract:
Remote sensing data are inherently spatial, and a substantial portion of their value for scientific analyses derives from the information they can provide about spatially dependent physical processes. However, no instrument observes everywhere all the time, and even where they do observe, their data have gaps and errors. Scientific inferences about the processes of interest can be improved if gaps can be filled and errors accounted for in the form of probabilistic uncertainties that can be propagated forward through subsequent analyses. In this talk I will discuss Spatio-Temporal Data Fusion (STDF) in this context. That is, the objective of data fusion is to combine data from multiple remote sensing instruments in order to provide a single best-estimate data set, with uncertainties attached to every datum so derived. The input data sets can have different coverages, footprint sizes, shapes, and orientations, and error characteristics. STDF capitalizes on the differing strengths of heterogenous data sources in order to drive uncertainty to a minimum and exploit the synergy of complementary observations. In some cases, this may even allow us to construct new data products that estimate quantities that are not directly observed by any remote sensing instrument. I will illustrate with some results from our fusion of AIRS (Atmospheric Infrared Sounder) and OCO-2 (Orbiting Carbon Observatory-2) carbon dioxide data.