The Holuhraun volcanic eruption in Iceland in 2014 provided an unprecedented opportunity to examine ACI and how well they are represented in climate models. Malavelle et al. (2017) used Collection 5 data from the MODIS Aqua satellite and provided an assessment of the impact of the large release of sulfur dioxide on cloud effective radius (reff) and cloud liquid water path (LWP), finding a considerable impact on the former, but no impact on the latter. We revisit this eruption with a considerably extended satellite record which includes new Collection 6 data from the Terra and Aqua satellite and additional years of data from 2015-2020. This tripling of satellite data allows an even more rigorous assessment of ACI, including the impacts not just on cloud micro-physical properties (reff and LWP), but also on the macro property cloud coverage, indicated by cloud fraction (CF). As shown in the enclosed figure: changes in cloud properties caused by the Holuhraun volcanic perturbation estimated using machine-learning (ML) and MODIS observations. The spatial distribution and zonal means of the changes in (a) reff and (b) CF are shown in the left panels, while right panels show probability density functions (so that the areas under the curves are equivalent) for MODIS and ML-MODIS. We present analyses both from a novel machine learning approach and from more standard climatological analysis and conclude that machine learning enables us to tease out robust impacts on CF that had previously been too noisy to determine.
Our results show that cloud fraction is significantly increased and appears to surpass cloud brightening and to be the dominant factor in aerosol indirect radiative cooling. Climate models are unable to replicate such strong impacts on CF (Ghan et al., 2016). These results show that the ongoing debate about the cooling impact of aerosols is far from over while climate models continue to inadequately represent the complex macro- and micro-physical impacts of ACI. Our study provides new directions and constraints for improving model representation of ACI.
References:
Malavelle, F. F. et al., Nature 546, 485-491, (2017).
Glassmeier, F. et al., Science 371, 485-489, (2021).

