An Assessment of Biases in Satellite CO2 Measurements Using Atmospheric Inversion

Abstract:

Column-integrated CO₂ mixing ratio measurements from satellite should provide a new view of the global carbon cycle, thanks to their ability to measure with great coverage in places that are poorly sampled by the in situ network (e.g. the tropics) using a new approach (full-column averages rather than point measurements). For this new insight to be useful, however, systematic errors in these data must first be identified and removed.

Here we use atmospheric transport modeling to perform a global comparison of satellite CO₂ measurements to higher-quality reference data (in situ data from flasks and aircraft, column CO₂ data from the upward-looking spectrometers of the TCCON network) to assess systematic errors in the satellite data. This broad comparison is meant to complement the more direct validation done at specific TCCON sites. A suite of 3-D CO₂ mixing ratio histories are generated across 2009-2014 using combinations of several different a priori fossil fuel, land biospheric, and oceanic CO₂ fluxes run through the PCTM off-line atmospheric transport model driven by MERRA 1°x1.25° winds and vertical mixing parameters. Each member of the suite is forced to agree with in situ CO₂ measurements (flask, tall tower, and routine light aircraft profiles) through use of a variational carbon data assimilation (4Dvar) system. The optimized 3-D CO₂ fields are then compared to ACOS column CO₂ retrievals of GOSAT data, with the differences being fit to different independent variables (aerosol optical depth, atmospheric path length, surface albedo, etc.) to derive a GOSAT bias correction. ACOS-GOSAT CO₂ retrievals, corrected by this scheme, as well as with the “official” ACOS bias correction, will then be assimilated using the same 4Dvar approach. The benefit of the GOSAT data with and without the bias corrections will then be assessed by comparing the optimized CO₂ fields to independent data (TCCON column data, as well as aircraft data left out of the in situ inversions).

Finally, we will compare these GOSAT-optimized CO2 fields to early-mission column CO₂ data from the Orbiting Carbon Observatory (OCO-2) across August-October 2014. We will attempt to characterize the relative GOSAT-to-OCO-2 CO₂ retrieval biases and to identify their causes by plotting them against independent variables.