The internal variability generated by FaIR in a stochastic run is constrained as part of the overall RMSE and present-day temperature constraints. However, it appears that this may bias against high variability runs. The below plot, which is part of the fair-calibrate draft manuscript, shows the 10 highest RMSE runs that pass the constraint (red) and 10 lowest (blue). The low RMSE runs all show low variability, but they are also not showing large spreads in future climate projections.
No measure currently exists to determine whether the internal variability generated by FaIR is realistic: at the moment, it is just taken from the CMIP6 ensemble parameter sampling. An idea is to take detrended residuals from the HadCRUT5 ensemble, look at the internal variability in these, and compare the internal varibility in FaIR to these runs (i.e. impose an additional observational constraint).
The internal variability generated by FaIR in a stochastic run is constrained as part of the overall RMSE and present-day temperature constraints. However, it appears that this may bias against high variability runs. The below plot, which is part of the fair-calibrate draft manuscript, shows the 10 highest RMSE runs that pass the constraint (red) and 10 lowest (blue). The low RMSE runs all show low variability, but they are also not showing large spreads in future climate projections.
No measure currently exists to determine whether the internal variability generated by FaIR is realistic: at the moment, it is just taken from the CMIP6 ensemble parameter sampling. An idea is to take detrended residuals from the HadCRUT5 ensemble, look at the internal variability in these, and compare the internal varibility in FaIR to these runs (i.e. impose an additional observational constraint).