Data reduction for inverse modeling: an adaptive approach v1.0

Abstract. The number of greenhouse gas (GHG) observing satellites has greatly expanded in recent years, and these new datasets provide an unprecedented constraint on global GHG sources sinks. However, a continuing challenge for inverse models that are used to estimate sinks is the sheer satellite observations, sometimes millions per day. These massive often make it prohibitive implement modeling calculations and/or assimilate observations using many types atmospheric models. Although very large, information content any single observation modest non-exclusive due redundancy with neighboring measurement noise. In this study, we develop adaptive approach reduce size geostatistics. A guiding principle data more regions little variability less high variability. We subsequently tune evaluate synthetic real case studies North America from NASA's Orbiting Carbon Observatory-2 (OCO-2) satellite. proposed reduction yields accurate CO2 flux estimates than commonly method binning averaging data. further metric choosing level reduction; can dataset average one ∼ 80–140 km specific here without substantially compromising estimate, but find reducing quickly degrades accuracy estimated fluxes. Overall, developed could be applied range problems use large trace datasets.