Machine learning sheds light on how aerosols affect clouds

The Earth’s atmosphere is rich in tiny, suspended particles called aerosols that affect the climate. Aerosols can be formed by natural processes such as volcanic eruptions and forest fires. However, they also result from the fossil fuels burned for us to drive cars, fly aeroplanes and operate factories. This increase in aerosols could have a significant impact on climate change. Aerosols play a role in cloud formation, so they could help to keep the planet cooler since clouds reflect sunlight.

Unfortunately, scientists have struggled to determine precisely how clouds respond to changes in aerosol levels, making climate change projections even more uncertain. This is because the impacts of other meteorological variables have made it difficult to isolate data on aerosol-cloud interactions. However, researchers supported by the EU-funded CONSTRAIN project have found a way to study such interactions through the unique opportunity offered by Iceland’s volcano eruption in 2014. Their findings have been published in the journal ‘Nature Geoscience’.

The Icelandic eruption resulted in a huge aerosol plume. “This massive aerosol plume in an otherwise near-pristine environment provided an ideal natural experiment to quantify cloud responses to aerosol changes, namely the aerosol’s fingerprint on clouds,” explains study lead author Dr Ying Chen of Paul Scherrer Institut, Switzerland, in a news item posted on the website of the University of Exeter, United Kingdom. Dr Chen carried out this work as a research fellow at the British university.“Our analysis shows that aerosols from the eruption increased cloud cover by approximately 10%,” continues the scientist. “Based on these findings, we can see that more than 60% of the climate cooling effect of cloud-aerosol interactions is caused by increased cloud cover.”

Following an air polluting event such as a volcanic eruption, there are much higher concentrations of aerosol particles in the air. Since water droplets form around aerosol particles, the higher aerosol concentrations make it easier for cloud droplets to form. However, these cloud droplets tend to be smaller and more numerous, causing the polluted clouds to look brighter than they would normally be, scatter more light and become more reflective.

Whereas earlier research indicated that this brightening of clouds accounted for most of the cooling caused by cloud-aerosol interactions, the new study’s findings tell a different story: “Volcanic aerosols also brightened clouds by reducing water droplet size, but this had a significantly smaller impact than cloud-cover changes in reflecting solar radiation,” notes Dr Chen.

To reach their conclusions, the researchers used a satellite-based machine learning approach. Prof. Jim Haywood, a research fellow at CONSTRAIN project partner the Met Office, United Kingdom, states: “This work is radically different as it does not rely on models; it uses state-of-the-art machine learning techniques applied to satellite observations to simulate what the cloud would look like in the absence of the aerosols. Clear differences are observed between the predicted and observed cloud properties which can be used to assess aerosol-cloud-climate impacts.”

The study’s results provide substantial observational evidence on the cooling impact of aerosols that will help scientists to improve climate models. Through this, it is furthering the CONSTRAIN (Constraining uncertainty of multi decadal climate projections) project’s goal to improve climate projections for the next 20 to 50 years.

For more information, please see:

CONSTRAIN project website