Global Carbon Project: Tracking the fate of man-made CO2

After the global average of fossil carbon dioxide emissions dropped significantly in 2020, this year they are again approaching levels before the COVID-19 pandemic. This is the finding of the Global Carbon Project (GCP). Every year, scientists take stock of how much CO2 has been anthropogenically emitted around the world and is absorbed again by natural sinks.

According to the preliminary report, the global share of CO2 that remains in the atmosphere will continue to rise this year, by 2.0 ppm to an expected 415 ppm (parts per million, a unit of measure for the composition of gases). Taken together, CO2 sinks on land and in the oceans continued to absorb about half (54 percent over the past decade) of carbon dioxide emitted. The rise in atmospheric CO2 is driving sinks on land and in the ocean, while climate change is reducing carbon sinks (by about 15 percent on land, and by about 5 percent in the ocean).

As every year, scientists from the Max Planck Institute for Meteorology (MPI-M) are involved in the calculations of the Global Carbon Budget (GCB). This year, Prof. Julia Pongratz, Dr. Julia Nabel, Dr. Tatiana Ilyina and Dr. Peter Landschützer have again made some novel contributions.


One novel contribution to this year’s GCB was provide by MPI-M in the form of the first ocean CO2 sink (Socean) and land CO2 sink (Sland) predictions, i.e., the annual mean air-sea and air-land CO2 fluxes, for the year 2021 using a neural network-based regression model. Similar to the historical reconstruction using the MPI-SOMFFN method developed within the “Observations, Analysis and Synthesis” group at MPI-M, the new prediction uses the historical (1959-2020) Sland and Socean fluxes and reconstructs those using historical atmospheric CO2 concentrations, fossil fuel emissions and the ONI ENSO index. Once the historical relationship is established, this relationship can be used to estimate Socean and Sland for the year 2021, using the same set of predictors. “This new estimate is exciting” says Peter Landschützer. “On the one hand we now have, for the first time a full prediction of all main components of the carbon cycle in 2021, but on the other hand this is also a test, whether we can – in a very simple way – show that we understand what drives year-to-year variability in the carbon sinks”.

Additionally, this year’s budget for the first time includes observation-based estimates in the Socean calculation. Therefore, CO2 data collected onboard of the yacht “Seaexplorer”, scientifically accompanied by Peter Landschützer, directly contribute to the best air-sea CO2 flux estimate. “I am happy that the Vendée Globe data from Boris Herrmann found their way in the estimation of Socean”, adds Peter Landschützer. “The CO2 data collected from the remote Southern Ocean help to increase the confidence in the observation-based estimates of the air-sea CO2 flux”.

Furthermore, MPI-M provides an Socean estimate computed with the ocean biogeochemistry model HAMOCC. This year, the contribution includes an important novel feature. For the first time, the computation of the ocean CO2 sink is enhanced by adding riverine inputs of carbon and nutrients to the ocean. This new model development, based on recent work by Fabrice Lacroix, consolidated in the model code by Fatemeh Chegini, improves the representation of carbon cycling in the coastal ocean. As a result, model evaluation shows an improvement in Socean estimate in comparison to observations.

Tatiana Ilyina: “This is an important step forward to reduce uncertainties in our understanding of the role of the coastal ocean in the global carbon sink and to constrain regional outgassing of riverine carbon.”


The CO2 fluxes from land use activities have been estimated this year based on updated datasets on global changes in agricultural areas and wood harvest. This lets the different models’ estimates converge. However, the reduced range of estimates does not necessarily imply reduced uncertainty. “Important processes such as degradation of tropical forests, which have gained strongly in importance, and changes from year to year are not well captured in current land use data.” says Julia Pongratz, who coordinates the estimates of land use emissions in the GCP budget. “Accurate, high-resolution observations of land-use dynamics are now more important than ever to reduce the uncertainty around land-use emissions and their trends."

Based on the updated estimates, land use activities released about 2.9 GtCO2 net in the atmosphere in 2021, compared to 3.2 GtCO2 in 2020. The comparatively small net flux comprises substantial emissions, predominantly from deforestation, and CO2 removals, for example from forest regrowth after afforestation or wood harvest. The emission terms remained high during the pandemic and amounted to about 13.3 GtCO2 in 2020. Reduced monitoring capacities and legal enforcement of measures to reduce tropical deforestation in the wake of the pandemic have been observed in multiple countries and may impair emissions reductions.

For the first time, the GCP links in its budget the national greenhouse gas inventories with the independent data of the global carbon cycle models. JSBACH, the land surface model of the MPI-ESM, is one of the models providing this data. “While global carbon cycle models separate natural from anthropogenic effects based on drivers, the country reported data do so based on area” says Julia Nabel, who conducted the JSBACH simulations at MPI-M. The country reported data thus attributes part of the natural terrestrial sink on managed land to the land-use sector, which, summed over all countries, thereby becomes a sink. Accounting for this redistribution of natural sinks, the two approaches are shown to be consistent with each other. “Linking the independent estimates of the global carbon cycle models to the national greenhouse gas inventories is an important step in view of the Global Stocktake, with which the implementation of the Paris Agreement and the common progress towards the agreed goals can be evaluated.” says Julia Pongratz.

The natural land CO2 sink continued to increase during the 2011-2020 period in response to increased atmospheric CO2, albeit with large interannual variability.  The natural land CO2 sink was 11.4 GtCO2 per year on average during the 2011-2020 decade (28% of total CO2 emissions). The large year to year variability in the natural land sink (about 4 GtCO2 yr-1) makes smaller annual changes in anthropogenic emissions hard to detect in the global atmospheric CO2 concentration.

Original publication:

Friedlingstein et al. (2021) Global Carbon Budget 2021. Earth System Science Data (preprint).


Prof. Dr. Julia Pongratz
Max Planck Institute for Meteorology
and LMU Munich, Department of Geography
Email: julia.pongratz@we dont want

Dr. Julia Nabel
Max Planck Institute for Meteorology
and Max Planck Institute for Biogeochemistry
Email: julia.nabel@we dont want

Dr. Tatiana Ilyina
Max Planck Institute for Meteorology
Email: tatiana.ilyina@we dont want

Dr. Peter Landschützer
Max Planck Institute for Meteorology
Email: peter.landschuetzer@we dont want