MsTMIP Version 1 Screening


MsTMIP Version 1 contains all model results that have undergone basic screening. During screening every effort was made to semantically map original outputs provided by the modeling teams to the required MsTMIP set of variables. A complete crosswalk from originally uploaded model outputs to required post-processed MsTMIP variables is available in spreadsheet form (https://docs.google.com/spreadsheets/d/17YAXpj1gu0g8Wi2SyNu90bUgy9OFALLQMlnK9DmE-BI/edit). The color coding of the cell denotes presence (green) or absence (red) of required variables. Note that several variables are calculated using mass balance and many variables are reported with non-standard naming conventions. These are noted in the target cell. Original variables not part of the MsTMIP required list are not considered, apart from some instances where they are used for mass balance calculations.


In addition to semantic checks, screening also includes checks on sign convention, limits (e.g., negative values for GPP), as well as whether model output conforms to the spatiotemporal resolution and simulation protocol of MsTMIP. This screening is performed on a test subset of model variables: AutoResp, Evap, GPP, HeteroResp, NEE, NEP, NPP, TotalResp, and TotLivBiom. These are bolded entries in the accompanying spreadsheet.


During this process special emphasis was placed on NEE. The MsTMIP protocol defines NEE as: Net Ecosystem Exchange = HeteroResp + AutoResp + Disturbance_Emissions_(CO2) + Product_Emissions_(CO2) - GPP; where Disturbance_Emissions_(CO2) is assumed to include fire flux and land use change emissions. Most models cannot simulate all NEE sub-processes. Similarly, several models cannot simulate all 5 MsMTIP global simulations. Any deviations from the protocol, e.g., missing NEE sub-process, simulation availability, and/or other issues, are noted below for each model. 


It is important to remember that the model results released are not original model outputs from participating models. Rather, outputs have undergone post-processing to match the MsTMIP protocol as much as possible. Steps in that processing are detailed below for each model and include semantic remapping of model outputs (https://docs.google.com/spreadsheets/d/17YAXpj1gu0g8Wi2SyNu90bUgy9OFALLQMlnK9DmE-BI/edit). MsTMIP Version 1 results are packaged as netCDF v4 files with one variable (for each combination of model and simulation) at a monthly timestep from 1901 to 2010. Finally, inclusion in MsTMIP Version 1 does not imply that the model results themselves are error free; merely that basic thresholds for usability in scientific manuscripts have been met.


BIOME-BGC
1. SG2 and SG3 are not available.
2. For all simulations fixed land cover types are used.
3. For BG1 only constant N deposition is used.
4. For RG1 only 1801 to 1900 are simulated. This time period was mapped to 1901-2010 by using 1801-1810 as 1901-1910 and 1801-1900 as 1911-2010.
5. Net Ecosystem Exchange = HeteroResp+AutoResp+Fire_flux-GPP; land use change and product emissions are not simulated.


CLASS-CTEM-N+
1. Units do not match protocol for Evap and some large outlying values present in carbon fluxes.
2. All carbon fluxes are capped at 1000 gC/m2/month. Values in exceedance of this threshold are set to NA.
3. Barren/vegetated mask excludes high productivity vegetated areas.
4. Net Ecosystem Exchange = -Net Ecosystem Productivity; disturbance emissions (fire flux and land use change emissions) and product emissions are not simulated.


CLM4
1. Original NEP has units gC/m2/s; NEP was converted to kgC/m2/s during processing.
2. Net Ecosystem Exchange = HeteroResp+AutoResp+Disturbance_Emissions_(CO2)+Product_Emissions_(CO2)-GPP where Disturbance_Emissions_(CO2) includes fire flux and land use change emissions. Net Ecosystem Exchange also includes hrv_xsmrpool flux (maintenance respiration deficit).


CLM4VIC
1. Net Ecosystem Exchange = HeteroResp+AutoResp+Disturbance_Emissions_(CO2)+Product_Emissions_(CO2)-GPP where Disturbance_Emissions_(CO2) includes fire flux and land use change emissions. Net Ecosystem Exchange also includes hrv_xsmrpool flux (maintenance respiration deficit).


DLEM
1. DLEM simulates latitudinal bands -88.25 to +88.25 only, the missing bands were infilled as ocean or barren pixels.
2. The Net Ecosystem Exchange reported does not incorporate all reported simulated subcomponents. It is calculated using mass balance:  Net Ecosystem Exchange =  HeteroResp+AutoResp-GPP+Proddec+Lulc where Proddec is Product_Emissions_(CO2) and Lulc is annual land use change emissions. Lulc is first apportioned equally across all 12 months. DLEM does not simulate fire flux.


GTEC
1. BG1 is not available.
2. For RG1 only 1900 to 1999 are simulated. This time period was mapped to 1901-2010 by using 1900-1910 as 1901-1910 and 1900-1999 as 1911-2010.
3. Lat 90.75N and Lon: 81.25E is a bad pixel (e.g., GPP c.1 300 gC/m2/month for a tundra pixel). Pixel set to zero for all screened variables.
4. Net Ecosystem Exchange = HeteroResp+AutoResp+Product_Emissions_(CO2)-GPP; fire flux and land use change emissions are not simulated.


ISAM
1. HeteroResp flux original units in kgC/m2/week; units converted (using an average week/month ratio of 365/12/7) to kgC/m2/s during processing.
2. TotLivBiom (calculated using mass balance: C_leaf+C_rootc+C_rootf+C_stem [the sum of all live carbon pools]) units are kgC/m2/s (!); TotLivBiom is converted to kgC/m2 during processing.
3. The Net Ecosystem Exchange reported does not incorporate all reported simulated subcomponents. It is calculated using mass balance: Net Ecosystem Exchange = HeteroResp+AutoResp+Ld-GPP; where Ld is Disturbance_Emissions_(CO2). Ld original units kgC/m2/month are converted to kgC/m2/s for this calculation. Fire flux and product emissions are not simulated.


LPJ-wsl
1. BG1 is not available.
2. A dynamic global vegetation model is used for natural vegetation (as opposed to the prescribed forcing data for land cover/use change), while the crop grid cell fraction is determined by the forcing data.
3. Net Ecosystem Exchange = HeteroResp+AutoResp+Disturbance_Emissions_(CO2)-GPP; where Disturbance_Emissions_(CO2) includes both fire flux and land use change emissions. Product emissions are not simulated. Net Ecosystem Exchange also includes grazing emissions.


ORCHIDEE-LSCE
1. BG1 is not available.
2. Net Ecosystem Exchange = HeteroResp+AutoResp+Disturbance_Emissions_(CO2)+Product_Emissions_(CO2)-GPP; where Disturbance_Emissions_(CO2) includes only land use change emissions. Fire flux is not simulated.


SiB3
1. BG1 is not available.
2. SiB3 lacks carbon pools.
3. Net Ecosystem Exchange = -Net Ecosystem Productivity; disturbance emissions (fire flux and land use change emissions) and product emissions are not simulated.


SiBCASA
1. BG1 is not available.
2. Net Ecosystem Exchange = HeteroResp+AutoResp+Disturbance_Emissions_(CO2)-GPP; where Disturbance_Emissions_(CO2) includes only land use change emissions. Fire flux and product emissions are not simulated.


TEM
1. RG1/SG1 have a different barren/vegetated mask than BG1/SG2/SG3 in the original model output. To harmonize the barren/vegetated mask for all simulations the BG1 mask is used for RG1/SG1 during processing.
2. Several model outputs contain bands, linear features, on a North-South axis, with missing or abnormally small values located in either North America (Quebec), South America (Colombia-Brazil-Peru), and/or Africa (longitude 15.25E from 31.75N to 25.75S). These were infilled using nearest neighbor interpolation. Banded output by simulation:
   1. BG1, RG1, SG1, SG3: GPP, HeteroResp, NEE, NPP, SWE, TotLivBiom, TotSoilCarb
   2. SG2: All variables
1. Net Ecosystem Exchange = HeteroResp+AutoResp+Disturbance_Emissions_(CO2)+Product_Emissions_(CO2)-GPP where Disturbance_Emissions_(CO2) includes fire flux and land use change emissions.


TRIPLEX-GHG
1. SG1 is not available.
2. For all simulations fixed land cover types are used.
3. Original model output saved as one file per variable per month. e.g., a single file contains every month of June from 1901 to 2010 for a given output field.
4. All carbon flux original units in gC/m2/mon/10000; these are converted to kgC/m2/s during processing.
5. Evap original units in mm/day; Evap is converted to kg/m2/s during processing. 
6. Net Ecosystem Exchange = -Net Ecosystem Productivity; disturbance emissions (fire flux and land use change emissions) and product emissions are not simulated.


VEGAS2.1
1. BG1 is not available.
2. The prescribed forcing data for land cover/use change is not used for unmanaged land. Instead a dynamic global vegetation model is used. For managed land --crop/pasture fraction and wood harvest-- the prescribed forcing data was used.
3. Net Ecosystem Exchange = HeteroResp+AutoResp+Disturbance_Emissions_(CO2)+Product_Emissions_(CO2)-GPP where Disturbance_Emissions_(CO2) includes fire flux and land use change emissions. 


VISIT
1. BG1 is not available.
2. For RG1 and SG1 Net Ecosystem Exchange values are identical.
3. RG1 contains a different set of variables; only GPP, NEE, CarbonPools1-3, and Evap are available.
4. Net Ecosystem Exchange = -Net Ecosystem Productivity; disturbance emissions (fire flux and land use change emissions) and product emissions are not simulated.