Assessment of forcing changes across CMIP eras using reduced-complexity climate models

Previous studies showed that historical forcing changes can partly explain differences between climate model simulations of phases 5 and 6 of the Coupled Model Intercomparison Project (CMIP). With the new CMIP7 historical forcings now delivered, we investigate how forcings have changed from the CMIP5 to CMIP7 generations, and use the FaIR reduced-complexity climate model to quantify the impacts for simulated 1850-2100 global mean surface temperature.

First, we perform historical simulations using CMIP5, AR5, CMIP6, AR6, and CMIP7 forcing datasets using FaIR’s IPCC 6^th Assessment Report (AR6) calibration. To quantify the impact of individual forcing change on the simulated temperatures, we run simulation ensembles changing individual forcings (CO₂, CH₄, N₂O, other greenhouse gases, aerosol-radiation interactions, aerosol-cloud interactions, solar, volcanic, ozone, and land use forcings) for each CMIP and AR era. In particular, we show three major changes in CMIP7 relative to CMIP6: i) 1850-1900 is 0.1 K colder, largely driven by changes in stratospheric aerosol forcing; ii) 1960-1980 is 0.07 K warmer, largely driven by changes in stratospheric aerosol forcing; iii) over the 20^th century, the smaller aerosol-cloud interaction forcing translates into a temperature increase of 0.04 K, whereas the solar forcing change drives colder temperatures by 0.02 K.

Second, to qualitatively assess potential impact of forcing changes on climate model tuning, we recalibrate FaIR using the different CMIP era forcings. In particular, we quantify how forcing dataset choice affects the estimates of key climate metrics (e.g. Equilibrium Climate Sensitivity or Transient Climate Response) and the distribution of FaIR parameters (e.g. carbon cycle parameters or shallow and deep ocean heat capacities). We also compare ensembles of future projections produced using the new calibrations. We show that forcing changes result in relatively small impacts on emergent parameters, e.g. ECS and TCR are up to 2.5 % higher in CMIP6 calibration compared to CMIP7. These translate in simulated temperature estimates for 2100 colder by up to 0.2 K across various SSP scenarios when using the CMIP7 calibration instead of the CMIP6 one.

Overall, our results provide a comprehensive assessment of forcing changes across CMIP eras, in particular for the new CMIP7 datasets, and their implications for simulating historical and future climate. We discuss the value of reduced-complexity models for fast sensitivity testing of new forcings datasets and establish a workflow to test future updates of inputs4MIPs forcings.