CMIP6 Volcanic Forcing Error Quantified to be Smaller than Internal Variability
CMIP simulation data from the successive phases has underpinned much progress in climate research in the last decades. CMIP6, continues the progress, with improved correspondence between observed and CMIP6-simulated historical climate reconstructions, when contrasted to earlier CMIP simulations. While model fidelity continues to increase, the role of uncertainties in the forcing datasets that underpin the historically forced period remain an unresolved climate research challenge. After the CMIP6 endorsed forcing dataset collection was released for use, a problem was discovered with the volcanic forcing dataset - too large magnitude SAOD loading around the Mt Pinatubo (1992) volcanic eruption. Using two of the CMIP6-contributing models, we investigate the climate impact of the changes to the forcing dataset (v3 -> v4) undertaken to correct the standing issue.
Using two of the CMIP6-contributing models, CanESM5, a coupled ocean-atmosphere configuration that uses a relatively coarse horizontal resolution enabling large ensembles, and EAMv1, an atmosphere-only configuration that provides relatively high horizontal and vertical resolution (high-top) with a smaller ensemble size, we undertook transient historical experiments following the CMIP6 historical methodology. Simulations were run using both the v3 (CMIP6 endorsed) and v4 (revised, corrected) volcanic forcing datasets, and the model output was analyzed to ascertain the impact of the forcing change. Differences in temperature, precipitation, ocean heat content, and radiative forcings were found to be small when compared to the unforced variability captured by comparing intra-ensemble simulations for each model. The single exception to this result was found in tropical atmosphere anomalies at 24 km altitude, where a peak magnitude of 3C was recorded immediately after the eruption.
While considerable uncertainties remain, the utility of the CMIP6-historical simulation archive using the endorsed volcanic forcing dataset (v3) has been validated. Using a two-model suite, which spans a range of CMIP6-contributing model complexities, the forcing dataset error was found to have less impact on simulated fields than the unforced internal variability that occurs across single simulations within a single model simulation ensemble. No known issues with CMIP6 forcing have been identified that undermine the utility of the archive.