Spatial and temporal variability of column-integrated CO2: identifying drivers and variations from high-resolution model simulations and OCO-2 observations

Abstract:

Isolating the drivers and variations in column-averaged dry air mole fraction of carbon dioxide (X_CO2) is essential for mining information from space-based remote-sensing observations, such as those available from the Orbiting Carbon Observatory-2 (OCO-2). Contrary to the large number of studies analyzing the variability of surface CO₂ concentrations, studies analyzing the spatiotemporal variability of X_CO2 are relatively limited. More importantly, these results are either based on a sparse network of ground-based total column observations (i.e., from the Total Column Carbon Observing Network - TCCON) or derived from low-resolution model simulations. In this study, using the high-resolution (~7 km) GEOS-5 model simulated fields and the high-density observations from OCO-2, we investigate how variability in surface fluxes and/or meteorological drivers impact the observed X_CO2 variability across a range of scales. The study focuses on ~13:30 LT and is designed to highlight the significant contributors to local and regional scale X_CO2 variability from daily to seasonal timescales. In collaboration with the OCO-2 Validation team, the variability information is also being used to identify small geographical areas (<1° or ~100km) where the X_CO2 is expected to be relatively constant. These small areas then serve as target regions for examining the potential of external variables (for e.g., surface reflectance, aerosol) to generate biases (variability) in the X_CO2 retrievals in those regions. We will also show comparison results of the model-based variability analyses with the variability statistics derived from actual OCO-2 retrievals. This comparison serves as an important consistency check for the simulated fields from the GEOS-5 model. Finally, we will review these results in terms of assessing and quantifying representation errors as well as developing and implementing data thinning/‘superobbing’ algorithms for OCO-2 retrievals.