Carbon Dioxide

Pilot Top-down CO2 emissions and removals associated with Terrestrial Carbon Stock Changes by nations

Dataset Description

The Committee on Earth Observations (CEOS) Atmospheric Composition – Virtual Constellation (AC-VC) GHG team has generated top-down gridded and country-level estimates of net carbon exchange (NCE) and changes in terrestrial carbon stocks. Processes that add or remove CO2 from the atmosphere are summarized in the diagram below. Two global, top-down products have been developed in these pilot budgets:

  1. Annual net land-atmosphere CO2 fluxes.
  2. Annual changes in terrestrial carbon stocks.
These products are provided over the period 2015-2020 on both a global grid and as country-level totals with error characterization.

Top-down inverse models analyze ground-based, airborne, and space-based measurements of CO2 to produce spatially-resolved estimates of the Net Carbon Exchange (NCE) between surface and atmosphere. NCE can be further subdivided into fossil fuel emissions (FF), lateral carbon fluxes due to rivers, crop, and wood, and changes in land carbon stocks (ΔC).

Processes that emit CO2 into the atmosphere and remove it at the surface and how they relate to stock changes and net surface-atmosphere flux. The largest anthropogenic emissions are from fossil fuel combustion and cement manufacturing. Agriculture, forestry and other land use constitute the second largest emitter. The natural carbon cycle both removes CO2 through photosynthesis and emits it through heterotrophic respiration and wildfires (biomass burning, BB shown here).


These estimates are derived from space-based, airborne, and surface-based measurements of CO2 using an ensemble of state-of-the-art flux inversion systems that use different transport models, meteorology, and inverse methods [e.g. Peiro et al., 2021]. The flux inversion systems provide spatially- and temporally-resolved estimates of surface-atmosphere fluxes for land and oceans from which country level annual land-atmosphere CO2 fluxes can be estimated.

Two estimates are provided based on two different datasets of atmospheric CO2 concentrations:

  1. fluxes estimated from in situ CO2 measurements in the National Oceanic and Atmospheric Administration (NOAA) Observation Package (ObsPack;, and
  2. fluxes estimated from NASA Orbiting Carbon Observatory-2 (OCO-2) column-averaged CO2 dry air mole fraction, XCO2, retrievals over land.
For these pilot products, fossil fuel CO2 emissions are prescribed from a bottom-up emissions inventory and held fixed, while terrestrial carbon fluxes (including those from the AFOLU sector) and ocean carbon fluxes are optimized to match the spatial and temporal fluctuations present in the observations within their uncertainties. This approach was adopted because the available ground-based, airborne and space-based measurements of atmospheric CO2 cannot clearly discriminate emissions from fossil fuel and AFOLU sources, and the fossil fuel CO2 sources are generally much better known. Future updates in the observing system are expected to dramatically improve the resolution and coverage, facilitating inversions that could simultaneously constrain fossil fuel and AFOLU emissions for future GSTs.

The Net carbon exchange (NCE) includes contributions from the Net biospheric exchange (NBE) and the Fossil Fuels. Inverse models derive estimates of NCE. Spatially-resolved estimates of NBE (middle) can be derived by subtracting the fossil fuel contributions (right) from NCE. Here, fossil fuel emissions are prescribed to derive national-scale estimates of NCE and NBE.

Time-dependent, sub-national-scale fossil fuel CO2 emissions from the Open-source Data Inventory for Anthropogenic CO2 (ODIAC) database were adopted for these pilot products. Fire emissions were derived from the Global Fire Emission Database v4.1s. Given these inputs, the net carbon exchange (NCE) between the surface and atmosphere is estimated using the inverse models at 1-degree latitude-longitude resolution.

To facilitate comparisons of these top-down flux estimates to national AFOLU inventories, a mask is applied that only includes “managed lands” within each country. These values are then adjusted to account for changes in land carbon stocks that do not immediately appear as CO2 emissions. These include “lateral” carbon fluxes from the terrestrial biosphere such as land-to-ocean transport of carbon by rivers ( a natural process perturbed by human activities) and import/export of harvested agricultural and wood products ( a human activity). These estimated changes in terrestrial carbon stocks reflect the combined impact of direct anthropogenic activities and changes to managed ecosystems in response to rising CO2, climate change, and disturbance. Regional-scale comparisons of these adjusted results with the original NCE estimates can be used to assess the fraction of the total CO2 emissions from the terrestrial biosphere that are being tracked by the AFOLU inventories.

Spatially-resolved estimates of the net carbon stock change, ΔC, can be derived from estimates of the Net Biospheric exchange by subtracting contributions from lateral carbon transports associated with wood and crops and rivers. National-scale maps of ΔC, NBE, Crop and Wood exports, and River exports are shown here.

Examples of country level emissions for 2015 – 2018 from inverse model experiments using only in situ (IS) or OCO-2 Land Nadir and Land Glint (LNLG) estimates of CO2. Contributions to the Net Carbon Exchange (NCE) is shown (grey) as well as its contributions from Fossil Fuels (red) and the land biosphere (ΔC; green). Land carbon losses associated with lateral exports by rivers (violet) and crops and wood harvest (brown) are shown for comparison. All fluxes are expressed in units of Petagrams (1015 grams = 1012 kg = 109 tons) of Carbon per year, expressed as PgCyr-1. A Pg of carbon corresponds to 3.66 Pg of CO2. Positive values indicate emissions into the atmosphere (sources) while negative values indicate removals (sinks). Uncertainties are expressed by the vertical dimension of the box.

Summary of Results

Country-level estimates of the net carbon exchange (NCE) were derived in experiments that used only in situ (IS) measurements and in experiments that used only OCO-2 XCO2 observations over land (LNLG). CO2 emissions associated with fossil fuels were subtracted to yield estimates of the net biospheric exchange (NBE). These values were adjusted for lateral carbon transports associated with imports or exports of crops or wood or carbon transport by rivers to yield estimates of the net carbon stock change, ΔC. These top-down stock change estimates must be multiplied by 3.66 (the ratio of the molar masses of CO2 and carbon) to yield estimates of CO2 emissions or removals for each country.

On regional scales, we find that the terrestrial carbon stocks increased across the northern extra-tropics (20-90N) but decreased in the northern tropics (0-20N) over 2015-2018, a period strongly impacted by the 2015 El Niño. Country-level flux estimates generally show robust signals for large extratropical countries (e.g., USA, Russia, China). Agreement between the simulations employing only in situ (IS) CO2 measurements and those using estimates of XCO2 derived from OCO-2 Land Nadir and Land Glint (LNLG) observations generally decreases for mid-sized countries (e.g., Turkey), particularly in regions with sparse observational coverage by the surface in situ network (such as the tropics). Large divergences between the IS and LNLG results occur in some regions, particularly North Africa. This could be related to biases in current OCO-2 XCO2 retrievals or poor coverage by the in situ network. The sparsity of independent CO2 measurements in these regions precludes definitive conclusions.

For future stocktakes, these estimates will be refined as new space-based observing systems are deployed to expand observational coverage and resolution of the atmospheric XCO2 distribution. Complimentary expansions of ground-based and aircraft-based CO2 measurements are critically needed in under-sampled regions both to complement the coverage of the space-based data and to assess their accuracy. Improvements to flux inversion systems will further refine results, as reductions to systematic transport errors will be critical for refining carbon flux estimates.

Technical Characteristics

Spatial resolution: 1-degree latitude by 1-degree longitude and national

Geographical coverage: Global

Temporal coverage: 2015 – 2020 (and onwards)

Update frequency: Annual

Format: Spreadsheet (.csv), netCDF, GeoTIFF

Data Policy: Open

Associated Guidance or User Manual

Under development

Points of contact for queries

Brendan Byrne

David Baker
CIRA, Colorado State University

David Crisp