Progress toward Estimating Surface Carbon Dioxide Fluxes at the Regional Scale Using an Augmented Ensemble Kalman Filter

Surface carbon dioxide (CO₂) fluxes are difficult to directly measure at the regional to global scales, but atmospheric CO₂ inversions have shown promise in constraining the fluxes using observed atmospheric CO₂ concentration. With the advent of new observations of CO₂ concentration, including high-resolution satellite measurements, denser operational tower and aircraft networks, and intensive field campaigns such as the NASA Atmospheric Carbon and Transport – America (ACT-America) mission, it has become necessary to develop advanced data assimilation techniques to efficiently assimilate the new observations and properly account for the multiple sources of uncertainties when solving for the CO₂ fluxes. However, our current understanding of how to design data assimilation systems for this application is limited, partly due to lacking knowledge about the underlying dynamics and error propagation.

Here we present the development of a new regional-scale ensemble-based data assimilation system for estimating surface CO₂ fluxes over North America, with a special focus on investigating the error covariance structure and evolution and contribution of different error sources to the posterior CO₂ flux uncertainties. The system is based on the Ensemble Kalman Filter (EnKF) component of The Pennsylvania State University Unified Data Assimilation system and uses the Weather Research and Forecasting model coupled with chemical transport (WRF-Chem) as its dynamical model. Observations of both CO₂ concentrations and atmospheric dynamical and thermodynamic variables will be assimilated simultaneously. We will present recent results from observing system simulation experiments that explore the impact of atmospheric transport error and CO₂ tower observing network on the estimated CO₂ fluxes and their posterior uncertainties. The focus will be on biogenic fluxes in the eastern United States during the summer of 2016, when the first ACT-America flight campaign took place. Finally, we will discuss some specific techniques for this particular data assimilation application, including covariance inflation and localization, and the optimum design of observing network and assimilation techniques for carbon estimation and monitoring.