Impact of Volcanic Eruptions on CMIP6 Decadal Predictions: A Multi-Model Analysis
The climate impacts of large volcanic eruptions can be isolated using large ensemble simulations, but traditionally such attribution studies examine the mean forced response across a large sample of internal variability base states. The simulated response cannot easily be compared to an observed benchmark given the small sample of well-observed large eruption events. In this study, a multi-system dataset of contributions to the Decadal Climate Prediction Project (DCPP) of the Coupled Model Intercomparison Project Phase 6 (CMIP6) is analyzed to isolate volcanic aerosol impacts under realistic internal variability base states. Specifically, the component C experiment protocol of DCPP-CMIP6 called for hindcast ensembles initialized just prior to the eruptions of Agung, El Chichón, and Pinatubo that used background volcanic aerosol forcing instead of the historical volcanic aerosol forcing used for component A hindcasts. The difference between DCPP-A and DCPP-C hindcasts then isolates the effect of realistic volcanic aerosol forcing given the same historical initial conditions. Across six different decadal prediction systems, robust responses are found for top-of-atmosphere global radiation as well as global surface and upper ocean temperature, with the latter exhibiting decadal-scale memory of the eruption. Regional impacts vary depending on the hemispheric symmetry of the forcing event and generally show less consistency across systems (particularly in the second pentad of the hindcasts). A direct comparison to observations reveals that, in general, volcanic forcing improves decadal hindcast skill, but a notable exception is that Pinatubo forcing degrades tropical Pacific skill in all prediction systems considered.
The climate system response to large volcanic eruptions and the potential for that response to enhance climate prediction skill at decadal lead times remain uncertain. This research advances understanding on both fronts by considering six different prediction systems that provided parallel with/without volcanic aerosol forcing hindcasts for the three largest eruptions of the late 20th century. The comparison sheds new light on the realism and model structural uncertainty of the response, and it also highlights the distinct responses to different eruptions. The study provides a new benchmark for understanding coupled model response to volcanic aerosol forcing that will be useful for future assessments (e.g., CMIP7).
Despite strong agreement in predicting the large-scale post-eruption radiative effects, including long-lasting surface cooling, other expected impacts (acceleration of NH polar vortex, Eurasian warming, development of El Niño conditions, strengthening of the Atlantic Meridional Overturning Circulation) are found to exhibit considerable dependence on the model and the eruption.