For CMIP6, we are submitting to ESGF CAM's CO2 tracer. For WACCM, this tracer is prognostic and is used in radiative computations, while for CAM it is a diagnostic quantity and is not used in radiative computations. I previously received a related question from someone trying to set up a CESM experiment and was concerned about the high values in the CO2 tracer. My answer to them, which uses CESM/CAM terminology, is below. I'm not sure what is the best way to explain this to people getting our output from ESGF.The 4D CO2 field in the runs that you are looking at is purely a diagnostic tracer. It is not being used in surface biogeochemistry computations or atmospheric radiative computations. In these runs, co2vmr is being used in these computations. So I don't think that these elevated CO2 concentrations indicate a problem.The CO2 value used in surface biogeochemistry computations is determined by the xml settings CLM_CO2_TYPE and OCN_CO2_TYPE. These variables can be one of 'constant', 'diagnostic', 'prognostic', and it is set to 'diagnostic' in these runs. When these variables are set to 'diagnostic', the surface components use co2vmr. When it set to 'prognostic', the surface components use the bottom layer value from the 4D CO2 field.The CO2 value used in atmospheric radiative computations is determined by the CAM namelist variable rad_climate. In these runs, this namelist variable has an entry 'N:CO2:CO2'. The N denotes non-advected, and with this setting radiative computations uses co2vmr. If this were 'A:CO2:CO2', radiative computations would use the advected 4D CO2 field.In our CESM nomenclature, we call this a BDRD configuration. This string is at the end of the compset longname.BD implies that biogeochemistry uses diagnostic CO2RD implies that atmospheric radiative computations use diagnostic CO2Runs with a prognostic carbon cycle are BPRP.The 4D CO2 tracer is forced by surface CO2 fluxes from CLM, POP, and fossil fuel emissions (which is zero in B1850 runs). It is higher than observed because we have (an undesirable) drift in the land and ocean carbon cycles. It just so happens that the drift in both components is a source of CO2 to the atmosphere. This causes the 4D CO2 field to keep increasing. In BDRD runs, there is no feedback to eliminate this drift, since the values are purely diagnositc and are not used in any computations. In BPRP runs, these elevated CO2 values trigger CO2 concentration feedbacks in CLM and POP which reduces the drift.