A source-oriented evaluation of CO2 inverse flux estimates in North America using ACT-America

Abstract:

Rigorous quantification of terrestrial CO2 fluxes is critical to our understanding of the earth’s carbon cycle. Both in situ-based and satellite-based atmospheric CO2 measurements have been used to quantify net ecosystem exchange (NEE) of CO2. We present an objective evaluation of three inverse estimates of CO2 NEE from the state-of-the-science CarbonTracker global inversion system, which assimilates the traditional in situ measurements, Orbiting Carbon Observatory-2 (OCO-2) v9 column retrievals, and a combination of the two platforms, respectively. Independent observations from the Atmospheric Carbon and Transport (ACT) - America aircraft campaign are used to evaluate the three inversions in temperate North America. In the study, we evaluate the three inversion estimates against ACT-America measurements on the three-dimensional adjoint sensitivity space, based on a suite of influence functions. We establish three metrics to evaluate and diagnose the errors in the magnitude of fluxes and the systematic biases, including summed model-obs deviation by space and time (J), summed gridded model-obs deviation (J_grid), and gridded model-obs mean bias (J_bias). We find the in situ-based inversion has the best performance (minimum J) in this region. The OCO-2 based inversion shows the best performance (minimal J_grid) in the Mid-Atlantic region, while the in situ-based inversion shows the best performance (minimal J_grid) in the central and western domain. The multiplatform inverse flux estimates have the poorest overall performance. Summer NEE of CO2 was overestimated in the eastern and southern US and underestimated in the Midwest US and western Canada, from the three inversions. The OCO-2 driven flux inversion shows a pronounced overestimation of NEE in the Central US during the wintertime period. Smaller systematic biases are found in the transition periods (spring and fall) for three inversions. Overall, this objective evaluation framework can be applied to distinguish different flux inversions or ecosystem flux models.