An atmospheric data assimilation system combining biomass and atmospheric CO₂ data for constraining biosphere carbon fluxes

Quantification of the long-term carbon uptake by the land biosphere is of key importance for climate action. Traditional methods of estimating the carbon sink include atmospheric inversions, which use CO₂  measurements to reduce inherent biases in simulations of the land biosphere. The atmospheric  CO₂ measurements used are informative on different time scales from days to decades, which are often difficult to separate from the data. Additional data sources can be used to separate the decadal change in sink magnitude from the shorter-term impacts of e.g. droughts. An example is the use of remotely-sensed above-ground biomass changes that have recently gained traction to estimate the stock change of carbon at the surface (Δbiomass), caused by months and years of integrated Net Ecosystem Exchange (NEE). We therefore built a Bayesian framework in which we constrain decades of daily NEE with both atmospheric CO2 observations as well as satellite-based Δbiomass products. With this integration we aim to better constrain the magnitude, inter-annual variability and location of land carbon sinks and sources. We focus the initial tests of the system on European carbon fluxes and find that Europe is a small long-term sink of CO₂, albeit with large regional differences. Most notably, vegetation of central European comes out as a net source of CO₂  into the atmosphere in our system, a finding that is supported by both by the Δbiomass product and the atmospheric CO₂ data. In this presentation we further explore the limits of the attempted integration, aiming to pave the way for future syntheses of atmospheric inversions with novel data products.