A Road Map for Improving the Treatment of Uncertainties in High‐Resolution Regional Carbon Flux Inverse Estimates.

Atmospheric inversions allow us to estimate the terrestrial carbon sink by combining atmospheric observations with atmospheric transport models. However, these inverse estimates remain highly uncertain. Here we quantify uncertainties in simulations of North American atmospheric CO2 concentrations using a probabilistic approach. We demonstrate that uncertainty in fossil fuel emissions is a key factor in the uncertainty surrounding biospheric flux estimates. We show that atmospheric transport uncertainties in state‐of‐the‐art numerical weather models diminish when averaged over time, while uncertainties in large‐scale CO2 boundary inflow considerably impair our ability to quantify regional fluxes. Current estimates of the North America land sink that neglect the uncertainties in CO2 boundary inflow and fossil fuel emissions are likely overconfident. Our findings suggest that targeted use of new atmospheric observations and improved quantification of uncertainty components are a promising avenue to improve atmospheric inversions with the goal to refine estimates of biospheric CO2 fluxes on regional and continental scales. Plain Language Summary: The uncertainty in biospheric carbon dioxide (CO2) flux estimates drives divergent projections of future climate and uncertainty in prescriptions for climate mitigation. The terrestrial carbon sink can be inferred from atmospheric CO2 observations with transport models via inversion methods. Regional CO2 flux estimates remain uncertain due to the mixture of uncertainties caused by transport models, prior estimates of biospheric fluxes, large‐scale CO2 boundary inflow, the assumptions in the inversion process, and the limited density of atmospheric CO2 observations. Understanding the characteristics of these uncertainties in space and time is essential for accurate biospheric CO2 flux estimates. Here we identify the terms that most confound biospheric flux estimates. Our results show that, over North America, (i) biospheric fluxes dominate the model uncertainty over all timescales. (ii) Contrary to expectation, fossil fuel emissions are the second largest source of uncertainty at all timescales. (iii) Transport uncertainties are large at short timescales but act like random errors decreasing with time averaging. (ix) Continental boundary inflow uncertainties are large near the boundaries and become significant at seasonal to annual timescales. We propose sampling and analysis strategies that can better quantify and reduce uncertainties in both fossil fuel emission and biospheric flux estimates. Key Points: We quantify biospheric flux, fossil fuel emission, atmospheric transport, and boundary inflow uncertainties in modeled atmospheric CO2Biospheric fluxes and fossil fuel emissions are the largest contributors to atmospheric CO2 uncertainty over North AmericaTransport uncertainties can be approximated by random errors, while boundary inflow uncertainties are persistent and can be sizeable [ABSTRACT FROM AUTHOR]