The goal of this project is to add �??upper air data�?� into CarbonTracker (CT) to solve for net surface CO2 fluxes at regional spatial scales (~500 km2) as well as to improve CarbonTracker�??s long-term flux estimates. Due to its short (5-week) assimilation window, CarbonTracker cannot currently make good use of such upper air data: information from the upper part of the atmospheric column is mis-attributed to local sources rather than to the proper, wider-spread distribution. For it to be used properly, three key changes to CarbonTracker will be required: 1. The 5-week assimilation window currently used in the ensemble Kalman smoother (EnKS) will be lengthened, to permit fluxes to be linked with upper-column concentrations over most of the globe, 2. A separate inversion outside of the current EnKS will be added to estimate coarse- scale fluxes at seasonal and longer timescales. This will provide a less-biased prior flux and a more realistic prior covariance to the EnKS, and 3. The number of flux parameters estimated, currently156 per week, will be increased by a factor of ten to get to the regional spatial scales of interest. Simulation experiments will be run to quantify the improvement of the new setup compared to the existing system and to help design the new EnKS flux parameterization. To implement this simulation capability, a new set of carbon flux models will be added to CarbonTracker to represent a realistic pseudo-�??truth�?�. Using the best new flux parameterization identified with these simulations, the PIs will then estimate regional CO2 fluxes in CarbonTracker using GOSAT, OCO-2, TCCON, AirCore, and aircraft data across 2009-2015, with the aim of identifying the key processes driving flux variability across that span. Finally, to assess how large transport model errors are in comparison to the other errors quantified by our simulation experiments, we will add a second atmospheric transport model, PCTM, into CarbonTracker, to compare to TM5.