How do CMIP6 models simulate future changes in the terrestrial carbon cycle?

How do CMIP6 models simulate future changes in the terrestrial carbon cycle?

Scientific Motivation

Uncertainties in projections of atmospheric carbon dioxide concentrations are largely driven by uncertainties in the land carbon cycle (e.g., Friedlingstein et al. 2014). The C4MIP simulations used CMIP5 models to compare concentration-driven versus emissions-driven RCP8.5 simulations. The resulting spread for land carbon uptake was large, in part due to uncertainties in model structure and parameters. How has this changed with CMIP6 models?

land-flux-carbon
Figure 4d from Friedlingstein et al., 2014. Annual global air to land carbon flux for 11 emissions-driven CMIP5 models.

Proposed Hacking

A limited number of CMIP6 models have released results from C4MIP emissions driven simulations using the SSP5-8.5 scenario. These results can be compared with the concentration driven SSP5-8.5 runs, with a focus on land carbon fluxes.

Anticipated Data Needs

Monthly land carbon fluxes (e.g., GPP, NPP, NBP) for esm-ssp585 (C4MIP) and ssp585.

Anticipated Software Tools

Standard python analysis tools (xarray, numpy, matplotlib).

Desired Collaborators

Anyone interested in terrestrial climate changes, carbon cycle modeling, and/or uncertainty quantification.

1 Like

Hi Katie - this sounds like a great project. I too am interested in terrestrial components. My first interest would be in comparing historical simulations of the CO2 effect to latitudinal flux distributions coming from atmospheric inversions, and one could also compare historical global CO2 effects to future trajectories across models. Have you thought of looking at historical or hist-bgc output as well? Cheers,
Britt

Thanks, Britt. These are great ideas too. I made a data request for C4MIP output under the future scenario (esm-ssp585), but I think there are only a few models that have completed and published this type of output. See: https://docs.google.com/spreadsheets/d/1AVZTNHcZME5xr0_hMbhvZY6z4twk3o3GHMz2Di3ZUN0/edit?usp=sharing

Perhaps looking at historical simulations (hist-bgc and/or esm-hist) would provide more models to compare with. Certainly the Tier 1 historical simulations would provide more output.