Data description

Document containing the description of the current CARBONES data (V2.0)

The objective of this document is to give the users of the CARBONES portal a brief and synthetic overview of the available products and services. This document contains five sections:

  • Section 1: gives a rapid overview of the data assimilation system used (CCDAS) in the generation of the data currently available on the portal;
  • Section 2: contains a list of products available on the portal;
  • Section 3: contains a description of the processing available (spatial aggregates and time filtering);
  • Section 4: contains a description of the visualisation tools available;
  • Section 5: contains the list of the other products currently displayed and compared to CARBONES product.

1. Description of the CCDAS version used to generate the data

The description of the CCDAS corresponds to the data currently displayed on the portal. The document will be updated after each new release of data on the portal. The data currently displayed on the portal correspond to version V2 (released in February 2013) and have been obtained with a 4 steps sequential assimilation system in which:

  • Step 1: Ocean fluxes are obtained from Neural Network trained with Takahashi pC02 climatology as well as raw pCO2 data and the corresponding variables that control to first-order the surface ocean pCO2 and issued from: satellite observations (Surface Chlorophyll, Sea Surface Temperature) and model outputs (Temperature, Salinity, Mixed Layer Depth,...);
  • Step 2: MODIS vegetation greenness (NDVI) data are assimilated to optimize the phenology parameters of the ORCHIDEE biosphere model: 3 to 4 parameters per Plant Functional Types. Note that only 10 out of 13 PFTs are optimized;
  • Step 3: Daily mean CO2 and heat fluxes from around 70 FluxNet sites are assimilated, using the posterior parameter estimates and error covariance matrix from Step 2. The parameter set is enlarged compared to step 2 to include all critical parameters controlling daily to seasonal/inter-annual flux variations (20 parameters per PFTs, around 150 in total);
  • Step 4: Assimilation of surface CO2 measurements are performed as an overall correction of prior land fluxes issued from Step 3 (using LMDZ transport) to provide the best 20-year surface flux reanalysis. The hypothesis is that the atmospheric data provide a global constraint and that a "flux correction" is less sensitive to "ecosystem model structural errors" than a parameter adjustment. Note that the complementary approach where the parameters of ORCHIDEE are corrected using prior and covariance estimates from step 3 has been performed and will be displayed as "additional product".

The ocean fluxes from Step 1 are also optimized in the final step with the atmospheric data but with rather small uncertainties. On the other end, the a priori fluxes of fossil fuel emissions, compiled in the framework of CARBONES by IER, and the biomass burning fluxes processed from GFEDv3 (specific process to distinguish permanent deforestation from savannas burning-regrowth cycle), are not optimized in Step 4. A schematic representation of the methodology used to obtain the version 1 of the CARBONES product is given in the Figure 1. Note finally that the land NEE flux correction in step 4 led to a re-adjustment of the gross carbon fluxes (GPP, Resp) from step 3 to match the NEE.

Figure 1: Schematic representation of the step-by-step assimilation used to obtain the V2 version of the 20 year CO2 reanalysis

2. Description of products available on the CARBONES portal

The Carbon Data Assimilation System delivers fluxes and stocks of carbon, as well as latent heat and sensible fluxes over land. These products were generated for a 20 year period (1990-2009) at a resolution of 1° x 1° degrees. The calculations for the current version were done at a coarser resolution of 2.5° x 3.75° for the final assimilation of atmospheric data (step 4 only) and were then interpolated on the 1°x1° degree grid. The CCDAS products available with the current version are listed in Table 1 below:

  Type 1 Type 2 Type 3 Comments Units
Carbon Fluxes Net Fluxes   Natural and Fossil Fuel Fluxes Over land: From step 4 assimilation
Over ocean: from step 1 further corrected in step 4 		kgC / m2 / hour
Net Ocean Fluxes From the step 1 assimilation
Natural Fluxes From the 20 year run of ORCHIDEE model with parameters optimized in step 3
Natural Land Fluxes
Total Land Fluxes
Fossil Fuel Fluxes Prior: From the CARBONES/IER inventory
Land Components Forest Net Ecosystem From the 20 year run of ORCHIDEE model with parameters optimized in step 3
Grass Net Ecosystem
Crop Net Ecosystem
Biomass Burning Prior: From GFEDv3
Land Components   Photosynthetic Flux From the 20 year run of ORCHIDEE model with parameters optimized in step 3
Respiration Flux
Other Fluxes     Latent Heat From the 20 year run of ORCHIDEE model with parameters optimized in step 3 W / m2
Sensible Heat
Carbon Stocks Living Biomass   Total (above and below ground) From the 20 year run of ORCHIDEE model with parameters optimized in step 3 kgC / m2
Total above ground
Total below ground
Leaf Area Index No Unit
Soil Organic   Total Soil Organic Carbon kgC / m2
Table 1: List of products provided by the current version of the CCDAS

3. Description of processing available on the CARBONES portal

The processing available with the current version of the CARBONES portal concerns two types of filtering:

  • Temporal filtering
  • Spatial averaging
3.1 Temporal filtering

The portal allows displaying the products either at their original temporal resolution or for coarser temporal means (Table 2)

  Carbon fluxes Heat fluxes Carbon stocks LAI
3 hourly values Original resolution Original resolution No No
24 hour means Yes Yes Original resolution Original resolution
Monthly means Yes Yes Yes Yes
Yearly means Yes Yes Yes Yes
Decadel means Yes Yes Yes Yes
20 year mean Yes Yes Yes Yes

Table 2: Information on the original resolution of the products and on the temporal means available.

The units depend on the type of temporal averaging and spatial aggregation: for instance flux in kgC/m2/hr is transformed into kgC/m2/day or kgC/m2/month or kgC/m2/year, and the spatial aggregates are transformed into kgC or GtC (see below).

In addition to calculating the means, the time filtering algorithm can be applied to fit the data to a polynomial of degree 2, with 3 harmonics. The residuals are filtered in the frequency space using a low-pass filter of 667 days and a high-pass filter of 80 days. The output variables from the temporal filtering are listed in Table 3.

CARBONES post-processing output parameter Description
Means Algebraic means (Table 2)
Smoothed trend The polynomial fit plus the low-pass filtered residuals using a cut-off value at 667 days
Smoothed curve The fitted function plus the low-pass filtered residuals using a cut-off value at 80 days
Re-centred parameter The original data minus the period mean

Table 3: Types of temporal filtering available

3.2 Spatial Averaging

All variables (fluxes and stocks) can be aggregated onto predefined regions for the visualization of the time series of fluxes and stocks (not for displaying 2D maps). This aggregation is performed based on 2D masks describing each of the preselected regions at 1°x1° resolution.

  • There are 3 general types of regions:
    • Land only,
    • Ocean only,
    • Land + ocean.
All available regions are listed in the Table 4 below.
Table 4: List of available regions for spatial aggregation

Note: For the ocean regions there are no carbon stocks and no estimation of the latent heat and sensible heat fluxes.

4. Description of visualisation tools available on the CARBONES portal

The visualisation tools available on the portal allow generating in an interactive way two types of plots:

  • Flux and carbon stock time series
  • Flux and carbon stock maps

In addition, static (non-interactive) plots, have been generated to show the model fit to the different data streams that were assimilated. The document with these plots is available on the CARBONES portal.

4.1 Flux Time Series Plots

To generate a time series plot, the menus listed in Table 5 are available:

MENU Comment Example
Variable One of the variables listed in Table 1 Net ecosystem flux
Temporal period Allows zooming to any particular period Jan 1, 1990 to Jan 1, 1992
Temporal filtering One or several of the possibilities listed in Table 3. See also Table 6 to see for which temporal resolutions these variables are available Means
Temporal resolution One or several of the possibilities listed in Table 2 Daily means
Spatial aggregation One or several of preselected regions in Table 4, only land regions for carbon stocks Global

Table 5: List of options available for time series plots

Note: Several variables and results for several temporal filters (e.g. monthly means and daily means) can be displayed on the same graph.

In addition:

  • It is possible to compare the net CO2 flux from the MACC 2 project with the corresponding product from CARBONES.
  • The estimated error on the CARBONES products is provided.

Units:

  • For carbon fluxes the units depend on the temporal averaging period: GtC/year for yearly data; GtC/month for monthly data; kgC/day for daily data; and kgC/hour for 3-hourly data.
  • W for the latent and sensible heat fluxes.
  • For carbon stocks the units depend on the temporal averaging period: GtC for yearly and monthly data; and kgC for daily data.
Averaging time C flux and latent heat flux C stocks including LAI
Mean Smooth trend Smooth curve Recentered parameter Mean Smooth trend Smooth curve Recentered parameter
3-hourly ✓              
Daily ✓ ✓ ✓   ✓ ✓ ✓ ✓
Monthly ✓       ✓     ✓
Yearly ✓       ✓     ✓
20 years                

Table 6: Specification of the temporal averages for which different types of temporal filtering results are available (checked boxes) for time series plots.

4.2 Flux maps

MENU Comment Example
Variable One of the variables listed in Table 1 Net ecosystem flux
Regions This option allows zooming to one of the regions listed in Table 2 Global
Temporal period Allows selecting a particular period January 1 2000 (for daily means)
Temporal averaging One or several of the possibilities listed in Table 3. See also Table 8 to see for which temporal resolutions these variables are available Smoothed trend
Range of values displayed Possibility to choose the minimum and maximum value displayed 0 to 5
Colour scale Possibility to select the colour scale Rainbow

Table 7: List of options available for map plots

Units:

  • KgC/m2/time step for carbon fluxes;
  • W/m2 for energy fluxes;
  • KgC/m2 for stocks.
Averaging time C flux and latent heat flux C stocks including LAI
Mean Smooth trend Smooth curve Recentered parameter Mean Smooth trend Smooth curve Recentered parameter
3-hourly ✓              
Daily ✓ ✓ ✓   ✓ ✓ ✓ ✓
Monthly ✓       ✓     ✓
Yearly ✓       ✓     ✓
20 years ✓             ✓

Table 8: Specification of the temporal averages for which different types of temporal filtering results are available (checked boxes) for map plots.

Figure 2. Portal view of flux maps

Figure 3. Portal view of regional time series charts

5. List of additional product compared to CARBONES main product

As part of the evaluation and validation strategy of CARBONES, the portal offers the possibility to compare the standard results of the CARBONES CCDAS to other model or data-oriented products. In this way we are able to provide benchmarks either for CARBONES or for the other products depending on the quantities. The list of products is intended to be flexible and to increase over time and even after the formal end of CARBONES project. So far the following products are available:

  • Fnee_MACC: the latest release of the global net land carbon flux exchanges from MACC-2 GMES core service.
  • Fnee_LPJ: the net land carbon fluxes estimated from the "Lund-Potsdam-Jena" terrestrial ecosystem model.
  • Fgpp_jung: the photosynthetic land carbon uptake estimated from the "FluxNet" database and a statistical model to extrapolate in space and time the GPP, using climate data and satellite information on the vegetation activity.
  • Fgpp_lpj: the GPP land carbon fluxes estimated from the "Lund-Potsdam-Jena" terrestrial ecosystem model.
  • Fresp_jung: the respiration land carbon flux estimated from the "FluxNet" database and a statistical model to extrapolate in space and time the RESP, using climate data and satellite information on the vegetation activity.
  • Fresp_lpj: the RESP land carbon fluxes estimated from the "Lund-Potsdam-Jena" terrestrial ecosystem model.
  • Foce_takahashi: The climatological air sea flux exchange from Takashashi et al. 2005.