Evaluation of the Consistency among In Situ and Remote Sensing Measurements of CO2 over North America using the CarbonTracker-Lagrange Regional Inverse Modeling Framework

CarbonTracker-Lagrange (CT-L) is a regional inverse modeling system for estimating CO2 fluxes with rigorous uncertainty quantification. CT-L uses footprints from the Stochastic Time-Inverted Lagrangian Transport (STILT) model driven by high-resolution (10 to 30 km) meteorological fields from the Weather Research and Forecasting (WRF) model. We have computed a library of footprints corresponding to in situ and remote sensing measurements of CO2 over North America for 2007-2015. GOSAT and OCO-2 XCO2 retrievals are simulated using a suite of CT-L terrestrial ecosystem flux estimates that have been optimized with respect to in situ atmospheric CO2 measurements along with fossil fuel fluxes from emissions inventories. A vertical profile of STILT-WRF footprints was constructed corresponding to each simulated satellite retrieval, and CO2 profiles are generated by convolving the footprints with fluxes and attaching initial values advected from the domain boundaries. The stratospheric contribution to XCO2 has been estimated using 4-dimensional CO2 fields from the NOAA CarbonTracker model (version CT2016) and from the Chemical Lagrangian Model of the Stratosphere (CLaMS), after scaling the model fields to match data from the NOAA AirCore surface-to-stratosphere air sampling system. Tropospheric lateral boundary conditions are from CT2016 and from an empirical boundary value product derived from aircraft and marine boundary layer data. The averaging kernel and a priori CO2 profile are taken into account for direct comparisons with retrievals. We have focused on North America due to the relatively dense in situ measurements available with the aim of developing strategies for combined assimilation of in situ and remote sensing data. We will consider the extent to which interannual variability in terrestrial fluxes is manifest in the real and simulated satellite retrievals, and we will investigate possible systematic biases in the satellite retrievals and in the model.