Characterizing the errors in AIRS mid-tropospheric CO2 retrievals

Mid- to upper-tropospheric carbon dioxide (CO2) has been retrieved since 2002 from thermal infrared channels of NASA's Atmospheric Infrared Sounder (AIRS) instrument. Unlike retrievals using near-IR data from the Japanese Greenhouse gas Observing SATellite (GOSAT) or NASA's Orbiting Carbon Observatory 2 (OCO2), the sensitivity of AIRS measurements does not peak near the surface; however, the AIRS tropospheric CO2 data cover nearly the entire globe across a decadal time period - they should provide a good constraint on long-term surface CO2 fluxes at broad spatial scales, in the absence of significant biases. In this study, we attempt to characterize the systematic and random errors in AIRS CO2 retrievals by comparing to CO2 fields generated by the PCTM transport model using CarbonTracker-optimized fluxes. We examine both the standard and support products of AIRS Version 5 Release Level 2 CO2, together with other parameters (e.g., cloud top pressure/temperature) retrieved using coincident microwave measurements from the Advanced Microwave Sounding Unit (AMSU). We formulate a bias correction for AIRS CO2 against these parameters, remove the bias, and use the bias-corrected data to solve for weekly flux corrections across 2009-2011 at a 3.0 x 3.75 deg resolution (lat/lon).