Abstract. Accurately quantifying volcanic impacts on climate is a key
requirement for robust attribution of anthropogenic climate
change. Here we use the Unified Model â United
Kingdom Chemistry and Aerosol (UM-UKCA) compositionâclimate model to simulate
the global dispersion of the volcanic aerosol clouds from the three
largest eruptions of the 20th century: 1963 Mt Agung, 1982 El Chichón,
and 1991 Mt Pinatubo. The model has interactive stratospheric chemistry
and aerosol microphysics, with coupled aerosolâradiation interactions
for realistic compositionâdynamics feedbacks. Our simulations align
with the design of the Interactive Stratospheric Aerosol Model
Intercomparison (ISA-MIP) âHistorical Eruption SO2 Emissions
Assessmentâ. For each eruption, we perform three-member ensemble model
experiments for upper, mid-point, and lower estimates of SO2
emission, each re-initialised from a control run to approximately match the
observed transition in the phase of the quasi-biennial oscillation (QBO) in the 6 months after
the eruptions. With this experimental design, we assess how each eruption's emitted SO2
translates into a tropical reservoir of volcanic aerosol and analyse
the subsequent dispersion to mid-latitudes. We compare the simulations to the volcanic forcing
datasets (e.g. Space-based Stratospheric Aerosol Climatology (GloSSAC);
Sato et al., 1993, and Ammann et al., 2003) that are used in historical integrations
for the two most recent Coupled Model Intercomparison Project (CMIP) assessments.
For Pinatubo and El ChichĂłn, we assess the vertical extent of the simulated volcanic clouds by comparing modelled extinction to the Stratospheric Aerosol and Gas Experiment (SAGE-II) v7.0 satellite measurements and to 1964â1965 Northern Hemisphere
ground-based lidar measurements for Agung. As an independent test for
the simulated volcanic forcing after Pinatubo, we also compare simulated
shortwave (SW) and longwave (LW) top-of-the-atmosphere radiative forcings to the flux anomalies
measured by the Earth Radiation Budget Experiment (ERBE) satellite
instrument. For the Pinatubo simulations, an injection of 10 to 14âTg SO2 gives
the best match to the High Resolution Infrared Sounder (HIRS)
satellite-derived global stratospheric sulfur burden, with good
agreement also with SAGE-II mid-visible and near-infra-red extinction
measurements. This 10â14âTg range of emission also generates a
heating of the tropical stratosphere that is consistent with the
temperature anomaly present in the ERA-Interim reanalysis. For El
ChichĂłn, the simulations with 5 and 7âTg SO2 emission give best
agreement with the observations. However, these
simulations predict a much deeper volcanic cloud than represented in
the GloSSAC dataset, which is largely based on an interpolation between
Stratospheric Aerosol Measurements (SAM-II) satellite
and aircraft measurements. In contrast, these simulations show
much better agreement during the SAGE-II period after October 1984.
For 1963 Agung, the 9âTg simulation compares best to the forcing
datasets with the model capturing the lidar-observed signature of the
altitude of peak extinction descending from 20âkm in 1964 to 16âkm in
1965. Overall, our results indicate that the downward adjustment to SO2
emission found to be required by several interactive modelling studies when
simulating Pinatubo is also needed when simulating the Agung and El
ChichĂłn aerosol clouds. This strengthens the hypothesis that
interactive stratospheric aerosol models may be missing an important
removal or re-distribution process (e.g. effects of co-emitted ash)
which changes how the tropical reservoir of volcanic aerosol evolves
in the initial months after an eruption. Our model comparisons also
identify potentially important inhomogeneities in the CMIP6 dataset
for all three eruption periods that are hard to reconcile with variations
predicted in the interactive stratospheric aerosol simulations. We
also highlight large differences between the CMIP5 and CMIP6 volcanic
aerosol datasets for the Agung and El ChichĂłn periods. Future
research should aim to reduce this uncertainty by reconciling the
datasets with additional stratospheric aerosol observations.